Technical Design Report
The BABAR Collaboration
March, 1995
_
This document and the material and data contained therein, was developed under
sponsorship of the United States Government. Neither the United States nor
the Department of Energy, nor the Leland Stanford Junior University, nor their
employees, nor their respective contractors, subcontractors, or their employees,
makes any warranty, express or implied, or assumes any liability or responsibility
for accuracy, completeness or usefulness of any information, apparatus, product or
process disclosed, or represents that its use will not infringe privately owned rights.
Mention of any product, its manufacturer, or suppliers shall not, nor is intended
to, imply approval, disapproval, or tness for any particular use. A royalty-free,
nonexclusive right to use and disseminate same for any purpose whatsoever, is
expressly reserved to the United States and the University.
i
TheBABAR Collaboration
LAPP Annecy, Annecy-le-Vieux, France
D. Boutigny, Y. Karyotakis, S. Lees-Rosier, P. Petitpas
INFN, Sezione di Bari and Universita di Bari, Bari, Italy
C. Evangelista, A. Palano
Beijing Glass Research Institute, Beijing, China
G. Chen, Y.T. Wang, O. Wen
Institute of High Energy Physics, Beijing, China
Y.N. Guo, H.B. Lan, H.S. Mao, N.D. Qi, W.G. Yan, C.C. Zhang, W.R. Zhao, Y.S. Zhu
University of Bristol, Bristol, UK
N. Dyce, B. Foster, R.S. Gilmore, C.J.S. Morgado
University of Bergen, Bergen, Norway
G. Eigen
University of British Columbia, Vancouver, British Columbia, Canada
C. Goodenough, C. Hearty, J. Heise, J.A. McKenna
Brunel University, London, UK
T. Champion, A. Hasan, A.K. McKemey
Budker Institute of Nuclear Physics, Novosibirsk, Russia
A.R. Buzykaev, V.N. Golubev, V.N. Ivanchenko, S.G. Klimenko, E.A. Kravchenko, G.M. Kolachev,
A.P. Onuchin, V.S. Panin, S.I. Serednyakov, A.G. Shamov, Ya.I. Skovpen, V.I. Telnov
California Institute of Technology, Pasadena, California, USA
D.G. Hitlin, J. Oyang, F.C. Porter, M. Weaver, A.J. Weinstein, R. Zhu
University of California, Davis, Davis, California, USA
F. Rouse
University of California, IIRPA, La Jolla, California, USA
A.M. Eisner, M. Sullivan, W. Vernon, Y.-X. Wang
Technical Design Report for the BABAR Detector
ii
University of California, Irvine, Irvine, California, USA
K. Gollwitzer, A. Lankford, M. Mandelkern, G. McGrath, J. Schultz,
D. Stoker, G. Zioulas
University of California, Los Angeles, Los Angeles, California, USA
K. Arisaka, C. Buchanan, J. Kubic, W. Slater
University of California, San Diego La Jolla, California, USA
V. Sharma
University of California, Santa Barbara, Santa Barbara, California, USA
D. Bauer, D. Caldwell, A. Lu, H. Nelson, J. Richman, D. Roberts, M. Witherell, S. Yellin
University of California, Santa Cruz, Santa Cruz, California, USA
J. DeWitt, D. Dorfan, A.A. Grillo, C. Heusch, R.P. Johnson, E. Kashigin, S. Kashigin, W. Kroeger,
W. Lockman, K. O'Shaughnessy, H. Sadrozinski, A. Seiden, E. Spencer
Carleton University and CRPPy, Ottawa, Ontario, Canada
K. Edwards, D. Karlen, M. O'Neilly
University of Cincinnati, Cincinnati, Ohio, USA
S. Devmal, B.T. Meadows, A.K.S. Santha, M.D. Sokolo
University of Colorado, Boulder, Colorado, USA
A. Barker, B. Broomer, E. Erdos, W. Ford, U. Nauenberg, H. Park, P. Rankin, J. Roy, J.G. Smith
Colorado State University, Fort Collins, Colorado, USA
J. Harton, R. Malchow, M. Smy, H. Staengle, W. Toki, D. Warner, R. Wilson
Technische Universitat Dresden, Institut fur Kern- und Teilchenphysik,
Dresden, Germany
J. Brose, G. Dahlinger, P. Eckstein, K.R. Schubert, R. Schwierz, R. Seitz, R. Waldi
Joint Institute for Nuclear Research, Dubna, Russia
A. Bannikov, S. Baranov, I. Boyko, G. Chelkov, V. Dodonov, Yu. Gornushkin, M. Ignatenko,
N. Khovansky, Z. Krumstein, V. Malyshev, M. Nikolenko, A. Nozdrin, Yu. Sedykh, A. Sissakian,
Z. Silagadze, V. Tokmenin, Yu. Yatsunenko
Technical Design Report for the BABAR Detector
iii
University of Edinburgh, Edinburgh, UK
K. Peach, A. Walker
INFN, Sezione di Ferrara, Ferrara, Italy
L. Piemontese
Laboratori Nazionali di Frascati dell' INFN, Frascati, Italy
R. Baldini, A. Calcaterra, R. De Sangro, I. Peruzzi (also Univ. Perugia), M. Piccolo, A. Zallo
INFN, Sezione di Genova and Universita di Genova, Genova, Italy
A. Buzzo, R. Contri, G. Crosetti, P. Fabbricatore, S. Farinon, R. Monge, M. Olcese, R. Parodi,
S. Passaggio, C. Patrignani, M.G. Pia, C. Salvo, A. Santroni
University of Iowa, Iowa City, Iowa, USA
U. Mallik, E. McCliment, M.-Z. Wang
Iowa State University, Ames, Iowa, USA
H.B. Crawley, A. Firestone, J.W. Lamsa, R. McKay, W.T. Meyer, E.I. Rosenberg
Northern Kentucky University, Highland Heights, Kentucky, USA
M. Falbo-Kenkel
University of Lancaster, Lancaster, UK
C.K. Bowdery, A.J. Finch, F. Foster
Lawrence Berkeley Laboratory, Berkeley, California, USA
G.S. Abrams, D. Brown, T. Collins, C.T. Day, S.F. Dow, F. Goozen, R. Jacobsen, R.C. Jared,
J. Kadyk, L.T. Kerth, I. Kipnis, J.F. Kral, R. Lafever, R. Lee, M. Levi, L. Luo, G.R. Lynch,
M. Momayezi, M. Nyman, P.J. Oddone, W.L. Pope, M. Pripstein, D.R. Quarrie, J. Rasson,
N.A. Roe, M.T. Ronan, W.A. Wenzel, S. Wunduke
Lawrence Livermore National Laboratory, Livermore, California, USA
O. Alford, J. Berg, R.M. Bionta, A. Brooks, F.S. Dietrich, O.D. Fackler, M.N. Kreisler,
M.A. Libkind, M.J. Mugge, T. O'Connor, L. Pedrotti, X. Shi, W. Stoe, K. van Bibber,
T.J. Wenaus, D.M. Wright, C.R. Wuest, R.M. Yamamoto
University of Liverpool, Liverpool, UK
J.R. Fry, E. Gabathuler, R. Gamet, A. Muir, P. Sanders
Technical Design Report for the BABAR Detector
iv
University of London, Imperial College of Science, Technology and Medicine,
London, UK
P. Dornan, A. Duane, L. Moneta, J. Nash, D. Price
University of London, Queen Mary & West eld College, London, UK
D.V. Bugg, P.F. Harrison, I. Scott, B. Zou
University of London, Royal Holloway & Bedford New College, Egham, Surrey, UK
Y. Gao, M.G. Green, D.L. Johnson, E. Tetteh-Lartey
University of Louisville, Louisville, Kentucky, USA
C.L. Davis
McGill University, Montreal, Quebec, Canada
D. Britton, R. Fernholz, D. MacFarlane, P. Patel, C. Smith, B. Spaan, J. Trischuk
University of Manchester, Manchester, UK
J. Allison, R. Barlow, G. La erty, K. Stephens
University of Maryland, College Park, Maryland, USA
C. Dallapiccola, M. Foucher, H. Jawahery, A. Skuja
University of Massachusetts, Amherst, Massachusetts, USA
J. Button-Shafer, J.-J. Gomez-Cadenas, S.S. Hertzbach, R.R. Ko er, M.G. Strauss
Massachusetts Institute of Technology, Cambridge, Massachusetts, USA
R.F. Cowan, M.J. Fero, R.K. Yamamoto
INFN, Sezione di Milano and Universita di Milano, Milano, Italy
M. Calvi, C. Cattadori, R. Diaferia, F. Lanni, C. Matteuzzi, F. Palombo, A. Sala, T. Tabarelli
University of Mississippi, Oxford, Mississippi, USA
M. Booke, S. Bracker, L. Cremaldi, K. Gounder, R. Kroeger, J. Reidy, D. Summers
Universite de Montreal, Montreal, Quebec, Canada
G. Beaudoin, M. Beaulieu, B. Lorazo, J.P. Martin, P. Taras, V. Zacek
Mount Holyoke College, South Hadley, Massachusetts, USA
H. Nicholson, C.S. Sutton
INFN, Sezione di Napoli and Universita di Napoli, Napoli, Italy
N. Cavallo, L. Lista, S. Mele, P. Parascandolo, C. Sciacca
Technical Design Report for the BABAR Detector
v
University of Notre Dame, Notre Dame, Indiana, USA
J.M. Bishop, N.N. Biswas, N.M. Cason, J.M. LoSecco, A.H. Sanjari, W.D. Shephard
Oak Ridge National Laboratory/Y-12, Oak Ridge, Tennessee, USA
F.S. Alsmiller, R.G. Alsmiller, Jr., T.A. Gabriel, J.L. Heck
LAL Orsay, Orsay, France
D. Breton, R. Cizeron, S. Du, A.-M. Lutz, J.M. Noppe, S. Plaszczynski, M.-H. Schune,
E. Torassa, K. Truong, G. Wormser
INFN, Sezione di Padova and Universita di Padova, Padova, Italy
F. Dal Corso, M. Morandin, M. Posocco, R. Stroili, C. Voci
Ecole Polytechnique Palaiseau, LPNHE, Palaiseau, France
L. Behr, G. Bonneaud, P. Matricon, G. Vasileiadis, M. Verderi
LPNHE des Universites Paris 6 et Paris 7, Paris, France
M. Benayoun, H. Briand, J. Chauveau, P. David, C. De La Vaissiere, L. Del Buono, J.F. Genat,
O. Hamon, P. Leruste, J. Lory, J.-L. Narjoux, B. Zhang
INFN, Sezione di Milano and Universita di Pavia, Pavia, Italy
P.F. Manfredi, V. Re, V. Speziali, F. Svelto
University of Pennsylvania, Philadelphia, Pennsylvania, USA
L. Gladney
INFN, Sezione di Pisa, Universita di Pisay and Scuola Normale Superiorez,
Pisa, Italy
G. Batignaniy, S. Bettarini, F. Bosi, U. Bottigliy, M. Carpinelli, F. Costantiniy,
F. Forti, D. Gambino, M. Giorgiy, A. Lusianiz, P.S. Marrocchesi, M. Morgantiy ,
G. Rizzo, G. Triggianiy, J. Walsh
Prairie View A&M University, Prairie View, Texas, USA
M. Gui, D.J. Judd, K. Paick, D.E. Wagoner
Princeton University, Princeton, New Jersey, USA
C. Bula, C. Lu, K.T. McDonald
Technical Design Report for the BABAR Detector
vi
INFN, Instituto Superiore di Sanita, Roma, Italy
C. Bosio
INFN, Sezione di Roma and Universita \La Sapienza," Roma, Italy
F. Ferroni, E. Lamanna, M.A. Mazzoni, S. Morganti, G. Piredda, R. Santacesaria
Rutgers University, Rutgers, New Jersey, USA
P. Jacques, M. Kalelkar, R. Plano, P. Stamer
Rutherford Appleton Laboratory, Chilton, Didcot, UK
P.D. Dauncey, J. Dowdell, B. Franek, N.I. Geddes, G.P. Gopal, R. Halsall,
J.A. Lidbury, V.J. Perera
CEA, DAPNIA, CE-Saclay,1 Gif-sur-Yvette, France
R. Aleksan, P. Besson, T. Bolognese, P. Bourgeois, A. de Lesquen, A. Gaidot, L. Gosset,
G. Hamel de Monchenault, P. Jarry, G. London, M. Turluer, G. Vasseur, C. Yeche, M. Zito
Shanghai Institute of Ceramics (SICCAS), Shanghai, China
J.R. Jing, P.J. Li, D.S. Yan, Z.W. Yin
University of South Carolina, Columbia, South Carolina, USA
M.V. Purohit, J. Wilson
Stanford Linear Accelerator Center, Stanford, California, USA
D. Aston, R. Becker-Szendy, R. Bell, E. Bloom, C. Boeheim, A. Boyarski, R.F. Boyce,
D. Briggs, F. Bulos, W. Burgess, R.L.A. Cottrell, D.H. Coward, D.P. Coupal, W. Craddock,
H. DeStaebler, J.M. Dorfan, W. Dunwoodie, T. Fieguth, D. Freytag, R. Gearhart, T. Glanzman,
G. Godfrey, G. Haller, J. Hewett, T. Himel, J. Hoe ich, W. Innes, C.P. Jessop, W.B. Johnson,
H. Kawahara, L. Keller,2 M.E. King, J. Krebs, P. Kunz, W. Langeveld, E. Lee, D.W.G.S. Leith,
V.G. Luth, H. Lynch, H. Marsiske, T. Mattison, R. Melen, K. Mo eit, L. Moss, D. Muller,
M. Perl, G. Oxoby, M. Pertsova, H. Quinn, B.N. Ratcli , S.F. Scha ner, R.H. Schindler,
S. Shapiro, C. Simopolous, A.E. Snyder, E.J. Soderstrom, J. Vav'ra, S. Wagner, D. Walz,
R. Wang, J.L. White, W. Wisniewski, N. Yu
Stanford University, Stanford, California, USA
P. Burchat, R. Zaliznyak
1 Sub ject
2 Retired
to approval of funding agency.
Technical Design Report for the BABAR Detector
vii
Academia Sinica, Taipei, Taiwan
H.-Y. Chau, M.-L. Chu, S.-C. Lee
University of Texas at Dallas, Richardson, Texas, USA
J.M. Izen, X. Lou
INFN, Sezione di Torino and Universita di Torino, Torino, Italy
F. Bianchi, D. Gamba, G. Giraudo, A. Romero
INFN, Sezione di Trieste and Universita di Trieste, Trieste, Italy
L. Bosisio, R. Della Marina, G. Della Ricca, B. Gobbo, L. Lanceri, P. Poropat
TRIUMF, Vancouver, British Columbia, Canada
R. Henderson, A. Trudel
Tsinghua University, Beijing, China
Y.P. Kuang, R.C. Shang, B.B. Shao, J.J. Wang
Vanderbilt University, Nashville, Tennessee, USA
R.S. Panvini, T.W. Reeves, P.D. Sheldon, M.S. Webster
University of Victoria, Victoria, British Columbia, Canada
M. McDougald, D. Pitman
University of Wisconsin, Madison, Wisconsin, USA
H.R. Band, J.R. Johnson, R. Prepost, G.H. Zapalac
York University, Toronto, Ontario, Canada
W. Frisken
Babar and the distinctive likeness are trademarks of
Laurent de Brunho and are used with his permission.
Copyright c Laurent de Brunho
All Rights Reserved
Technical Design Report for the BABAR Detector
viii
Technical Design Report for the BABAR Detector
Contents
1 Introduction
1
2 Detector Overview
7
2.1
2.2
Introduction
General Design Considerations
2.2.1 Acceptance
2.2.2 Charged Track Resolution and Multiple Scattering
2.2.3 Photon Eciency and Resolution
2.2.4 Identi cation of Hadrons
2.2.5 Interaction with the Accelerator
2.2.6 Considerations of Cost and Schedule
The Experimental Design
Detector Optimization
2.4.1 Impact of Particle Identi cation
2.4.2 Optimization of Performance versus Cost
Detector Performance
2.5.1 Vertex Detector
2.5.2 Drift Chamber
2.5.3 Particle Identi cation
2.5.4 Electromagnetic Calorimeter
2.5.5 Muon and Neutral Hadron Detector
2.5.6 Electronics, Trigger, and Data Acquisition
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : :
2.3
2.4
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : :
2.5
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : :
7
8
8
9
11
12
13
14
14
20
20
21
25
25
28
30
32
37
39
x
2.6
2.7
Physics Performance : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
2.6.1 Acceptance and Mass Resolution for Decays to Charged Particles
2.6.2 Separation between B Decay Vertices : : : : : : : : : : : : : : : :
2.6.3 0 Eciency and Resolution : : : : : : : : : : : : : : : : : : : : :
2.6.4 Lepton Identi cation : : : : : : : : : : : : : : : : : : : : : : : : :
2.6.5 Charged Hadron Identi cation : : : : : : : : : : : : : : : : : : : :
Performance for Non-CP Physics : : : : : : : : : : : : : : : : : : : : : : :
2.7.1 Other B Physics : : : : : : : : : : : : : : : : : : : : : : : : : : :
2.7.2 Detector Issues for Charm Physics : : : : : : : : : : : : : : : : : :
2.7.3 Detector Issues for Tau Physics : : : : : : : : : : : : : : : : : : :
2.7.4 Detector Issues for Two-Photon Physics : : : : : : : : : : : : : : :
3 Physics with BABAR
3.1
3.2
3.3
3.4
3.5
Physics Context : : : : : : : : : : : : :
Simulation Tools : : : : : : : : : : : :
3.2.1 ASLUND : : : : : : : : : : :
3.2.2 GEANT Simulation|BBSIM
Studies of B 0 ! J= Modes : : : : : :
3.3.1 B 0 ! J= KS0 : : : : : : : : :
3.3.2 B 0 ! J= KL0 : : : : : : : : :
3.3.3 B 0 ! J= K 0 : : : : : : : : :
Studies of B 0 to Double Charm Modes
3.4.1 B 0 ! D+D, : : : : : : : : :
3.4.2 B 0 ! D+D, : : : : : : : : :
3.4.3 B ! DD : : : : : : : : : : :
Studies of B ! Modes : : : : : : :
3.5.1 B 0 ! +, : : : : : : : : : :
39
40
41
42
43
43
44
45
46
46
48
51
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
Technical Design Report for the BABAR Detector
51
55
55
56
57
57
58
60
61
61
63
64
64
65
xi
3.5.2
3.5.3
3.6
3.7
3.8
3.9
3.10
3.11
3.12
3.13
3.14
B0
! 00 : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
B + ! + 0 Decays : : : : : : : : : : : : : : : : : : : : : : : : : :
B0
!
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Tagging Modes : : : : : : : : : : : : : :
3.7.1 Kaon Tags : : : : : : : : : : : :
3.7.2 Lepton Tags : : : : : : : : : : :
3.7.3 Other Tags : : : : : : : : : : :
3.7.4 Combined Tagging : : : : : : :
Estimate of CP -Angle Measurement : : :
3.8.1 Method of Calculation : : : : :
3.8.2 Tagging Modes : : : : : : : : :
3.8.3 CP Modes : : : : : : : : : : : :
CKM Matrix Determination : : : : : : :
3.9.1 Vcb : : : : : : : : : : : : : : : :
3.9.2 Vub : : : : : : : : : : : : : : : :
3.9.3 Vtd : : : : : : : : : : : : : : : :
Rare B Decays : : : : : : : : : : : : : :
3.10.1 B ! : : : : : : : : : : : : :
3.10.2 B ! Xs`+`, : : : : : : : : : : :
3.10.3 B ! Xs : : : : : : : : : : : :
CP Asymmetries in Charged B Decays :
Charm Physics : : : : : : : : : : : : : :
3.12.1 D0D0 Mixing : : : : : : : : : :
3.12.2 CP Violation in Charm Decays
Tau Physics : : : : : : : : : : : : : : : :
Two-Photon Physics : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
67
68
68
72
73
74
76
76
77
77
78
79
81
82
84
85
85
86
87
87
87
88
88
89
90
91
Technical Design Report for the BABAR Detector
xii
3.15 Summary
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
4 Vertex Detector
4.1
97
Vertex Detector Requirements
4.1.1 Resolution
4.1.2 Acceptance
4.1.3 Eciency
4.1.4 Radiation Tolerance
4.1.5 Reliability
Vertex Detector Overview
4.2.1 Choice of Technology
4.2.2 Detector Layout
4.2.3 Electronic Readout
4.2.4 Mechanical Support
Detector Performance Studies
4.3.1 Resolution
4.3.2 Pattern Recognition
4.3.3 Solid Angle Coverage
Silicon Detectors
4.4.1 Requirements
4.4.2 Silicon Detector Design
4.4.3 Fanout Circuit Design
4.4.4 R&D on Detectors and Fanouts
Electronic Readout
4.5.1 Introduction
4.5.2 Readout Chip
4.5.3 Hybrid Design
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
4.2
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
4.3
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
4.4
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
4.5
92
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Technical Design Report for the BABAR Detector
97
97
98
98
99
99
100
100
100
102
104
105
106
112
114
115
115
117
121
124
124
124
125
130
xiii
4.5.4 Data Transmission
4.5.5 Baseline Design
4.5.6 Power Supplies
4.5.7 Electronics R&D
Mechanical Support and Assembly
4.6.1 IR Constraints
4.6.2 Module Assembly
4.6.3 Detector Assembly and Installation
4.6.4 Detector Placement and Survey
4.6.5 Detector Monitoring
4.6.6 R&D Program
Services, Utilities, and ES&H Issues
4.7.1 Services and Utilities
4.7.2 ES&H Issues
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
4.6
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
4.7
: : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
5 Drift Chamber
5.1
5.2
5.3
5.4
132
133
135
135
136
137
137
139
142
144
146
147
147
147
151
Physics Requirements and Performance Goals
Tracking Chamber Overview
Projected Performance
Drift System Design
5.4.1 Cell Design
5.4.2 Layer Arrangement
5.4.3 Total Channel Count
5.4.4 Cell Studies
5.4.5 Gain Variations
5.4.6 Electrostatic Forces and Stability
5.4.7 Pattern Recognition Studies
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : :
151
153
155
158
158
159
162
166
170
171
171
Technical Design Report for the BABAR Detector
xiv
5.5
5.6
Gas Choice and Properties
Mechanical Design
5.6.1 Endplates
5.6.2 Inner Wall
5.6.3 Outer Wall
5.6.4 Joints
5.6.5 R&D Program on Structural Components
5.6.6 Wires
5.6.7 Feedthroughs
5.6.8 Stringing
5.6.9 Endplate Connections
5.7 Front-End Electronics
5.8 High Voltage System
5.9 Calibration and Monitoring
5.9.1 Calibration
5.9.2 Slow Controls
5.9.3 Monitoring
5.10 Integration
5.10.1 Overall Geometry and Mechanical Support
5.10.2 Cable Plant and Utilities Routing
5.10.3 Access
5.10.4 Integration Aspects of the Gas System
5.10.5 Installation and Alignment
5.11 R&D Using Prototype Drift Chambers
5.11.1 Prototype I Chamber
5.11.2 Prototype II
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Technical Design Report for the BABAR Detector
172
175
175
176
177
178
179
180
181
183
185
187
188
189
189
189
190
191
191
191
192
193
193
194
194
195
xv
5.11.3
Other R&D E orts
: : : : : : : : : : : : : : : : : : : : : : : : : :
6 Particle Identi cation
6.1
197
Physics Requirement and Performance Goals
6.1.1 Introduction
6.1.2
Flavor Tagging
6.1.3 Exclusive decays
6.1.4 Summary of Requirements
Particle Identi cation Overview
6.2.1 The DIRC
6.2.2 The ATC
Projected Performance
6.3.1 Simulation Based on Prototype Results
6.3.2 Pattern Recognition in the DIRC
6.3.3 Particle Identi cation in the DIRC
6.3.4 Particle Identi cation in the ATC
6.3.5 Performance Requirements for Track Reconstruction
6.3.6 E ects of Backgrounds on PID Detector Performance
The DIRC Detector
6.4.1 DIRC Mechanical Design
6.4.2 Photodetectors and Readouts
6.4.3 Laser Flasher Monitoring and Calibration System
6.4.4 Integration Issues
6.4.5 Research and Development Program
The ATC Detector
6.5.1 ATC Mechanical Design
6.5.2 Photodetectors and Readouts
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
B
: : : : : : : : : : : : : : : : : : : : : : : : : :
B
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
6.2
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
6.3
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : :
6.4
: : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : :
6.5
195
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
197
197
197
199
200
200
201
203
204
204
207
210
212
212
214
216
216
223
225
226
228
232
232
234
Technical Design Report for the BABAR Detector
xvi
6.5.3
6.5.4
6.5.5
Monitoring and Calibration System : : : : : : : : : : : : : : : : : 237
Integration Issues : : : : : : : : : : : : : : : : : : : : : : : : : : : 237
Research and Development Program : : : : : : : : : : : : : : : : 238
7 Electromagnetic Calorimeter
7.1
7.2
7.3
7.4
7.5
Physics Requirements and Performance Goals : : : : : : : :
7.1.1 Physics Processes In uencing Performance Goals : :
7.1.2 Summary of Performance Targets : : : : : : : : : :
Calorimeter Overview : : : : : : : : : : : : : : : : : : : : : :
7.2.1 Technology Choice : : : : : : : : : : : : : : : : : :
7.2.2 Description of the Calorimeter : : : : : : : : : : : :
7.2.3 Readout Chain and Trigger : : : : : : : : : : : : :
7.2.4 Review of Options : : : : : : : : : : : : : : : : : :
Projected Calorimeter Performance : : : : : : : : : : : : : :
7.3.1 Contributions to Photon Resolution and Eciency :
7.3.2 Modeling : : : : : : : : : : : : : : : : : : : : : : : :
7.3.3 Expected Performance for Photons : : : : : : : : :
7.3.4 Expected Performance for 0s : : : : : : : : : : : :
7.3.5 e= Separation : : : : : : : : : : : : : : : : : : : :
7.3.6 Performance and Cost Optimization : : : : : : : : :
Crystal Subassemblies and Readout : : : : : : : : : : : : : :
7.4.1 Photodiode Readout : : : : : : : : : : : : : : : : :
7.4.2 Light Collection : : : : : : : : : : : : : : : : : : : :
7.4.3 Light Yield Measurements : : : : : : : : : : : : : :
7.4.4 Reliability of Inaccessible Readout Components : :
Calibration : : : : : : : : : : : : : : : : : : : : : : : : : : : :
7.5.1 Requirements and Ingredients : : : : : : : : : : : :
Technical Design Report for the BABAR Detector
245
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
: : : : : : : :
245
245
248
249
249
250
253
254
256
256
259
260
261
264
266
269
270
274
276
278
280
280
xvii
7.5.2 Energy Calibration with Beam Events
7.5.3 Source Calibration
Mechanical Support Structure
7.6.1 Design Considerations
7.6.2 Barrel Fabrication, Assembly, and Installation
7.6.3 Endcap Fabrication, Assembly, and Installation
Optimization and Prototype Studies
Crystal Procurement Issues
7.8.1 Radiation Hardness
7.8.2 Quality Control and Testing
: : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
7.6
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
7.7
7.8
: : : : : : : : : : :
: : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : :
8 Muon and Neutral Hadron Detector
8.1
8.2
307
Physics Requirements and Performance Goals
Detector Overview
8.2.1 The Iron Structure
8.2.2 The Active Detector Choice
Projected Performance
8.3.1 Muon Identi cation
8.3.2 Muon Tagging
0
8.3.3
L Detection
Detector Design and R&D
8.4.1 Chamber Construction and Assembly
8.4.2 System Layout
Gas System
8.5.1 Gas Composition and Flow Rates
8.5.2 Mixer
8.5.3 Distribution
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : :
8.3
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
K
8.4
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
8.5
281
282
283
283
288
296
299
301
301
303
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
307
312
312
315
316
316
322
324
328
328
330
335
335
336
336
Technical Design Report for the BABAR Detector
xviii
8.6
8.7
Front-End Electronics and High Voltage
Final Assembly, Installation, and Monitoring
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : :
9 Magnet Coil and Flux Return
9.1
9.2
341
Physics Requirements and Performance Goals
Overview
9.2.1 Description of Key Interfaces
Summary of Projected Magnet Performance
9.3.1 Central Field Magnitude and Coil Performance
9.3.2 Shielding of Forward Q2
9.3.3 Flux Return
Superconducting Solenoid
9.4.1 Magnetic Design
9.4.2 Cold Mass Design
9.4.3 Quench Protection and Stability
9.4.4 Cold Mass Cooling
9.4.5 Cryostat Design
9.4.6 Coil Assembly and Transportation
Cryogenic Supply System and Instrumentation
Flux Return
9.6.1 Overview
9.6.2 Barrel Flux Return Description
9.6.3 End Door Description
9.6.4 Options and Detailed Design Issues
9.6.5 Procurement, Fabrication, Assembly, and Schedule
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : :
9.3
: : : : : : : : : : : : : : : : :
: : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
9.4
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
9.5
9.6
336
337
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : :
10 Electronics
Technical Design Report for the BABAR Detector
: : : : : : : : :
341
342
344
346
346
347
347
350
350
352
354
358
359
361
362
364
364
364
368
371
371
375
xix
10.1 Overview
10.1.1 Introduction
10.1.2 Front-End
10.1.3 Trigger
10.1.4 Data Acquisition
10.2 Silicon Vertex Detector
10.2.1 Requirements
10.2.2 Overview
10.2.3 Data Format
10.2.4 Description of the Readout Module
10.3 Drift Chamber
10.3.1 Requirements
10.3.2 Preampli ers
10.3.3 Digitizer Options
10.3.4 Implementing the All-FADC Scheme
10.3.5 Calibration
10.3.6 Research and Development
10.4 DIRC
10.4.1 Requirements
10.4.2 Baseline Design
10.4.3 Research and Development
10.4.4 High Voltage System
10.4.5 Low Voltage Power Supplies and Control Systems
10.5 Aerogel Threshold Cherenkov Counter (ATC)
10.5.1 Requirements and Overview
10.5.2 Front-End Electronics
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : :
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
375
375
377
378
380
382
382
382
384
384
385
385
386
387
388
391
391
394
394
395
398
400
400
401
401
401
Technical Design Report for the BABAR Detector
xx
10.5.3 Readout
Calorimeter
10.6.1 Requirements
10.6.2 Overview
10.6.3 The Preampli er Card
10.6.4 Digitizing Board
10.6.5 Readout Module
10.6.6 Electronic Calibration
10.6.7 Monitoring and Control
10.6.8 Research and Development
Muon System Electronics
10.7.1 Electronics Requirements
10.7.2 General Architecture
10.7.3 Implementation
10.7.4 Time Measurements
10.7.5 Time Calibration
10.7.6 Monitoring
10.7.7 Diagnostics
Trigger Requirements and Background Rates
10.8.1 Introduction
10.8.2 Requirements
10.8.3 Backgrounds and Trigger Rates
10.8.4 De nitions of Trigger Filter Concepts
10.8.5 Performance of Some Simple Triggers
Level 1 Trigger
10.9.1 Overview
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
10.6
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
10.7
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
10.8
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
10.9
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Technical Design Report for the BABAR Detector
402
402
402
403
406
407
411
413
413
414
414
414
415
416
419
419
420
420
420
420
421
423
424
426
428
428
xxi
10.9.2 Drift Chamber Trigger
10.9.3 Calorimeter Trigger
10.9.4 Global Trigger
10.9.5 Simulation
10.9.6 Trigger System
10.10 Data Acquisition
10.10.1 Introduction
10.10.2 Requirements
10.10.3 Architectural Overview
10.10.4 Readout Crates
10.10.5 Event-Data Flow Control
10.10.6 Event Assembly
10.10.7 Event Distribution
10.10.8 Data Integrity
10.10.9 Research and Development
10.11 Level 2 Trigger
10.11.1 Filter
10.11.2 Implementation
10.11.3 Conclusion
10.12 Global Support Electronics
10.12.1 Detector Monitoring and Control
10.12.2 Data Monitoring Support
10.12.3 Diagnostic Support
10.12.4 Calibration Support
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
11 Computing
11.1 Requirements
432
435
436
437
438
441
441
442
445
449
452
456
460
462
462
463
463
464
465
466
466
468
468
469
475
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
475
Technical Design Report for the BABAR Detector
xxii
11.1.1 Operational Requirements
11.1.2 Technical Requirements
Overview
11.2.1 Chosen Technologies
11.2.2 Overall Description
Computing Model
11.3.1 Architectural Model
11.3.2 Operational Model
11.3.3 Baseline Cost Model
Software Environment
11.4.1 Software Methodologies and Languages
11.4.2 Development Environment
11.4.3 Data Model
11.4.4 Code Management and Distribution
11.4.5 Databases
11.4.6 Graphics
Online System
11.5.1 Introduction
11.5.2 Logical View of the Online System
11.5.3 Subsystem View
11.5.4 Operator View
11.5.5 Implementation View
Reconstruction and Analysis Framework
11.6.1 Scope
11.6.2 Conceptual Overview
11.6.3 Toolkits
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
11.2
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
11.3
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
11.4
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
11.5
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
11.6
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Technical Design Report for the BABAR Detector
476
477
482
482
483
485
485
488
493
498
498
500
501
503
506
508
510
510
512
514
515
516
517
517
518
519
xxiii
11.6.4 Framework
11.6.5 Application Builder
11.6.6 Standard Applications
11.6.7 Job Submission
11.6.8 Bulk Production
11.7 Computing Support
11.7.1 Introduction
11.7.2 Collaboration Support
11.7.3 Infrastructure Support
11.8 Integration Issues
11.9 System Responsibilities and Management
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
12 Interaction Region and Backgrounds
519
519
520
520
520
521
521
521
522
523
524
529
12.1 PEP-II Design
12.1.1 Parameters
12.1.2 Interaction Region Components
12.1.3 Background Implications
12.1.4 Operating Modes
12.2 Tools
12.2.1 QSRAD
12.2.2 Decay TURTLE
12.2.3 OBJEGS
12.2.4 GEANT
12.2.5 GELHAD
12.3 Background Sources
12.3.1 Synchrotron Radiation
12.3.2 Lost Particles
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
530
530
531
532
534
535
535
535
536
536
538
538
538
541
Technical Design Report for the BABAR Detector
xxiv
12.3.3 Hadrons
12.3.4 Luminosity Backgrounds
12.3.5 Other Sources
12.4 Background Rates and Detector Responses
12.4.1 Vertex Detector
12.4.2 Drift Chamber
12.4.3 Particle ID
12.4.4 Calorimeter
12.4.5 Muons
12.5 Physics Impact
12.6 Summary
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
13 Safety
13.1
13.2
13.3
13.4
13.5
13.6
567
Introduction
Detector Safety Overview
Beam Pipe and Support Barrel
Vertex Detector
Drift Chamber
Particle Identi cation
13.6.1 DIRC
13.6.2 Aerogel
Electromagnetic Calorimeter
Muon and Neutral Hadron Detector
Magnet Coil and Flux Return
Generic Hazards
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
13.7
13.8
13.9
13.10
544
547
550
552
553
558
560
561
562
562
563
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
14 Facilities, Assembly, Access and Integration
Technical Design Report for the BABAR Detector
567
567
568
569
569
570
570
571
571
572
573
574
577
xxv
14.1
14.2
14.3
14.4
14.5
Facilities
Detector Coordinate System
Structural Support of Systems
Installation Overview
Detector Component Installations
14.5.1 IR-2 Detector Hall Preparation
14.5.2 Coil and IFR Installation
14.5.3 Barrel Calorimeter Installation
14.5.4 DIRC Tube Assembly Installation
14.5.5 Drift Chamber Installation
14.5.6 Aerogel/Forward Calorimeter Installation
14.5.7 Vertex Assembly Installation
14.5.8 Backward End Plug Installation
14.5.9 Electronics House
14.5.10 Detector Transport System
14.5.11 Final Position of Detector in Beam Line
14.5.12 Service Space
14.6 Detector Maintenance Access
14.7 Detector Integration
14.7.1 Assembly Clearances
14.7.2 Good Neighbor Policy
14.7.3 Common Services
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : :
15 Collaboration Issues and Project Management
15.1 Membership
15.2 Collaboration Council
15.3 Spokesperson
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
577
579
579
580
580
580
582
585
586
587
587
588
589
589
589
589
590
590
591
592
593
593
597
597
599
599
Technical Design Report for the BABAR Detector
xxvi
15.4
15.5
15.6
15.7
15.8
Executive Board
Technical Board
Finance Review Committee
Communications
Construction Responsibilities
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
16 Cost and Schedule
16.1
16.2
16.3
16.4
599
600
600
600
601
605
Introduction
Project Cost
Schedule
Detector Systems
16.4.1 Vertex Detector
16.4.2 Tracking Chamber
16.4.3 Particle Identi cation
16.4.4 CsI(Tl) Calorimeter
16.4.5 Flux Return Instrumentation
16.4.6 Magnet
16.4.7 Electronics
16.4.8 Computing
16.4.9 Management, Installation and Integration
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
Technical Design Report for the BABAR Detector
: : : : : : : : : : : : :
605
605
608
609
609
609
610
611
612
612
612
613
615
1
Introduction
n the thirty years that have elapsed since the surprising discovery of CP violation in KL0
I meson decay, the Standard Model of elementary particle physics has evolved considerably.
The Standard Model with three quark generations accommodates CP violation in quark
decays in an elegant and economical manner. Through one imaginary parameter and three
real numbers, the Standard Model completely characterizes the weak couplings of the six
quarks as expressed in the Cabibbo-Kobayashi-Maskawa matrix. We have little evidence,
however, that this explanation is correct.
Indeed, there are hints from cosmological arguments that, in fact, it is not. CP violation is
a key piece in the puzzle of the matter dominance in the universe. The excess of matter over
antimatter is not an easily explained property of the universe. In any theory with baryonnumber-violating processes (for example, most grand uni ed theories) the universe at very
early times contains equal (thermal) populations of quarks and antiquarks. How and when
the currently observed imbalance develops is one of the key puzzles of cosmology. Attempts to
explain this with only Standard Model CP violation have so far not been successful, leading
to the suggestion that perhaps CP violation physics is the Achilles heel of the Standard
Model. Certainly it merits detailed exploration.
The study of CP -violating asymmetries in the decays of B 0 mesons to CP eigenstates
promises, at last, to provide a test of the Standard Model explanation of CP violation.
More generally, the measurement of these CP asymmetries can provide us with a series of
unique, and uniquely stringent, consistency tests of the quark sector of the Standard Model.
These tests, which will be carried out by the BABAR detector at PEP-II, will provide us with
perhaps the best chance to challenge the Standard Model in new and quantitatively precise
ways. This challenge has come about through the interplay of a series of experimental and
theoretical events.
While it had been understood for several years that the measurement of CP -violating
asymmetries in B 0 decay could lead to important tests of the CKM matrix [Car81], the
experiments seemed beyond reach. The discovery of a surprisingly long b-quark lifetime
by MAC [Fer83] and Mark II [Loc83] at PEP in 1983, together with an unexpectedly
large B 0-B 0 mixing by UA1 [Alb87b] and ARGUS [Alb87a] in 1987 made it possible to
contemplate making these measurements. It soon became clear [SNO89] that the richest
and most straightforward approach involved experiments at a variety of e+ e, machines,
2
Introduction
either in the (4S ) region, in the PEP/PETRA continuum region, or at the Z 0 pole.
Experiments at hadron machines can also address this physics. While hadronic experiments
potentially have larger samples of B mesons, they face substantial trigger and combinatoric
problems [HER94]. The most favorable e+ e, experimental situation, that is, the one producing the smallest statistical error with the least integrated luminosity, is the asymmetric
storage ring rst proposed by Oddone [Odd87]. This machine would boost the decaying B 0
mesons in the laboratory frame, allowing existing vertex measuring technology to measure
the time order of B 0 , B0 decay pairs, even with the short B meson ight distance. This,
however, would require event samples that are obtainable only with signi cant advances in
storage ring luminosity.
PEP-II promises to provide the required luminosity [SLA91], initially 3 1033 cm,2s,1 ,
ultimately 1034, with asymmetric (4S) production at a
of 0.56 (9 GeV electrons on
3.1 GeV positrons). The BB production rate will be 3 Hzat the initial luminosity, ultimately
rising to 10 Hz. The experimental challenge is then to provide high eciency, high resolution
exclusive state reconstruction in a situation new to the e+e, world: a center-of-mass in
motion in the laboratory. The boost is not extreme: at PEP-II, 90 in the center-of-mass
is at 65 in the laboratory. Nonetheless, the challenges this boost poses for the detector are
novel, and not easily met.
The measurement of CP -violating asymmetries in B 0 meson decay will be a major breakthrough. It will, however, be only the rst installment in a rich program that will ultimately
provide us with very stringent tests of the Standard Model. The key to these tests is the
measurement of the unitarity triangle relations between the CKM matrix elements. The
ultimate precision of the unitarity triangle tests depends not only on the CP -violating asymmetries, which measure the angles of the unitarity triangle, but on improved measurements
of the triangle sides as well. The series of measurements that is uniquely possible at an
asymmetric e+e, storage ring operated at the (4S ) resonance will both measure the angles
for the rst time and substantially improve the measurement of the triangle sides. They will,
further, provide unique tests of underlying theoretical assumptions and measurements of experimental systematic uncertainties. Should discrepancies with Standard Model predictions
emerge, the program can also pinpoint particular extensions of the Standard Model, be they
supersymmetry, left-right models, or other speci c models.
While the search for CP violation and for possible contradictions with the Standard Model
CKM picture of its origin is the most exciting physics challenge for BABAR, it is by no
means the only physics opportunity to be exploited. The asymmetric con guration and the
resultant separation of B vertices makes possible searches for rare B decays, such as b ! u
transitions, in a very clean environment. The large production rate for charm, either directly
in continuum cc events or indirectly from B decays, provides opportunities for charm physics.
Similarly, aspects of tau physics can readily be studied here with large data samples in a
clean environment. These opportunities are discussed in some greater detail in Chapter 3.
Technical Design Report for the BABAR Detector
Introduction
3
A detector designed to carry out this ambitious experimental program must be capable
of high eciency reconstruction of extremely rare exclusive nal states in the presence of
combinatoric and accelerator-generated backgrounds. This requirement places a premium on
solid angle coverage, charged particle momentum resolution and species identi cation, and
photon resolution and detection eciency. The design of the BABAR detector attempts to
balance these requirements against one another and against cost and schedule imperatives.
Soon after the approval of PEP-II in October of 1993, e orts began to form a major
international collaboration to build and operate a new detector at PEP-II's single interaction
region. Time to build a detector is short, and there has been a clear convergence of
thinking [SLA93, BEL94, HEL92] about new detector designs. The SLAC directorate,
with the approval of the Scienti c Policy Committee, invited a group of experimentalists
to submit a Letter of Intent and a Technical Design Report. The primary aim of this group
is the detailed study of CP -violating asymmetries in B 0 meson decay, with a wide variety
of other B , charm, , and two-photon physics that will also be accessible at new levels of
precision. An inaugural meeting was held in December of 1993. The Collaboration drafted a
governance document, chose a name and its leadership, and entered into the detailed design
of the BABAR detector.
A detailed Letter of Intent [Bab94] was submitted to the SLAC Experimental Advisory
Committee (EPAC) in June of 1994. The EPAC recommended that the BABAR Collaboration
be approved to proceed to a Technical Design Report (TDR). This recommendation was
accepted by the SLAC Director, who asked for a TDR by the end of 1994. This document
is that report.
The detector described in the Letter of Intent was a conceptual design, which in some
instances contained alternative technology choices for several detector subsystems. The
design discussed herein is substantially more detailed. It contains speci c choices for the
subsystem technologies, and the design of these subsystems has been optimized to provide
a balance between cost and adequate physics performance.
The planning of the PEP-II project has been done in sucient detail that the Laboratory is
con dent that it can be constructed in ve years, starting in October of 1993. In designing
BABAR, we have taken as an important constraint that a detector capable of measuring CP violating asymmetries in B 0 meson decay be operational on a timescale matched to that for
building PEP-II.
The BABAR Collaboration management team is functioning, with a Spokesman, Deputy
Spokesman, Technical Coordinator, and a Project Engineer. System managers are in place,
as well. Should the EPAC recommend approval on the basis of this document, the TDR,
together with a detailed Work Breakdown Structure, cost estimate, critical path schedule,
and Management Plan, will be submitted for approval by the DOE and the funding agencies
of the non-US institutions. With the establishment of Memoranda of Understanding (MOUs)
Technical Design Report for the BABAR Detector
4
Introduction
detailing the speci c responsibilities of each collaborating group, the construction of the
BABAR detector will commence in earnest.
The following Chapter describes the proposed detector, discusses some of the design decisions
and their physics basis and reviews the performance expected component by component.
Chapter 3 presents a physics overview, including results of simulation studies for the key
CP violation study channels as well as an overview of other physics opportunities. Detailed
descriptions of each detector subsystem are then provided in individual chapters. Finally
there are chapters that summarize general detector issues, such as accelerator-induced backgrounds, safety, installation and integration, management, budget and schedule.
Technical Design Report for the BABAR Detector
REFERENCES
5
References
[Alb87a] H. Albrecht et al. (ARGUS Collaboration),
192B, 245 (1987).
[Alb87b] C. Albejar et al. (UA1 Collaboration),
186B, 247 (1987).
[BEL94] BELLE Collaboration, \Letter of Intent for A Study of CP Violation in B Meson
Decays," submitted to the TRISTAN Program Advisory Committee (1994).
[Car81] A. Carter and A.I. Sanda,
45, 952 (1980);
D23, 1567
(1981);
I. Bigi and A.I. Sanda,
B193, 85 (1981);
B281, 41 (1987).
[Fer83] E. Fernandez et al. (MAC Collaboration),
51, 1022 (1983).
[HEL92] \HELENA, A Beauty Factory at Hamburg," DESY 92/41 (1992).
[HER94] HERA-B Proposal, DESY{PRC 94/02 (1994).
[Loc83] N. Lockyer, et al. (Mark II Collaboration),
51, 1316 (1983).
[Odd87] P. Oddone, in Proceedings of the UCLA Workshop: Linear Collider B B Factory
Conceptual Design, edited by D. Stork (World Scienti c, 1987) p. 243.
[SLA91] \An Asymmetric B Factory Based on PEP: Conceptual Design Report," edited
by M. Zisman, SLAC{372, LBL PUB{5303 (1991).
[SLA93] \Status Report on the Design of a Detector for the Study of CP Violation at
PEP-II at SLAC," SLAC{419 (1993).
[SLA94] \Letter of Intent for the Study of CP Violation and Heavy Flavor Physics at
PEP-II," SLAC{443 (1994).
[SNO89] \Report of the B Physics Group: 1. Physics and Techniques," G. Feldman et
al.,in Proceedings of the 1988 DPF Summer Study (Snowmass, Colorado) p. 561
(1989).
Phys. Lett.
Phys. Lett.
Phys. Rev. Lett.
Nucl. Phys.
Phys. Rev.
Nucl. Phys.
Phys. Rev. Lett.
Phys. Rev. Lett.
Technical Design Report for the
A AR
B B
Detector
6
Technical Design Report for the BABAR Detector
REFERENCES
2
Detector Overview
2.1
Introduction
T
he primary physics goal of the BABAR experiment is the systematic study of CP asymmetries in decays of the B 0 and B 0 into CP eigenstates, as discussed in Chapter 3. The
secondary goals are to explore the wide range of other B physics, charm physics, physics,
two-photon physics, and physics that becomes accessible with the high luminosity of
PEP-II. The design of the detector is optimized for the CP studies, but also serves well for
the other physics opportunities.
The critical experimental objectives to achieve the required sensitivity for CP measurements
are:
To reconstruct the decays of B 0 mesons into a wide range of exclusive nal states with
high eciency and low background;
To tag the avor of the other B meson in the event with high eciency and purity;
and
To measure the relative decay time of the two B mesons.
Studies carried out over the last few years at SLAC [SLA91, SLA93] and for B factory
proposals from other sites [CLE93, KEK92, HEL92] have led to a fairly complete picture of
the performance requirements that the detector systems must meet. In the Letter of Intent
[SLA94], we outlined a conceptual detector design that would satisfy these requirements. In
this document, we describe the technical design of the BABAR detector and report the results
of simulation studies.
In this chapter, we list the general design considerations, describe the detector design, give
details of the detector performance, and discuss the capability of the detector to do the
important physics available at PEP-II.
8
2.2
Detector Overview
General Design Considerations
2.2.1 Acceptance
The crucial CP asymmetry measurements at PEP-II are made using events in which one
B 0 , usually decaying into a CP eigenstate, is completely reconstructed, and the other B
in the event is tagged as a B 0 or B 0, using either a lepton or charged kaon. The CP
decay modes of interest generally have branching ratios below 10,4, and reconstructing
them requires observing anywhere from two to six charged particles and often one or more
0 s. The charged particles must be detected with very good momentum resolution, precise
vertex information, and clean particle identi cation. To achieve good eciency for these
rare events, it is important that the full performance of the detector should extend over a
very large solid angle. This is made slightly more dicult by the energy boost, which folds
one-half of the particles into the region cos lab > 0:5. Figure 2-1 shows the relation between
center-of-mass and laboratory polar angles for photons with the designed boost of = 0:56.
It should be noted, however, that the boost causes a separation of the two B vertices, which
provides an additional reduction in background from continuum events.
Figure 2-2 shows the detection eciency for the decay B 0 ! J= KS0 ! e+e,+, as a
function of forward polar angle coverage, which is one critical parameter in achieving good
acceptance. The acceptance for charged particles extends to cos 0:96 in the forward
direction and cos ,0:87 in the backward direction. In both cases, the limit is set by how
close the active coverage of the vertex detector can come to the edge of the rst machine
dipole magnet (B1). These angular limits correspond to a geometric acceptance of about
57% for observing all four charged tracks. For the detectors outside the vertex detector,
the acceptance is designed to extend at least to these limits, so there is no further loss of
geometric acceptance.
Another critical parameter is the minimum momentum cuto in the acceptance for charged
and neutral particles. The momentum distribution for the particles from B decays at rest
is quite soft: the average momentum for a pion from the process studied in Figure 2-2 is
about 1.8 GeV=c, but the momentum range for tagging kaons extends below 0.3 GeV=c. The
pion from the process B 0 ! D+D,; D+ ! D0+ has a typical momentum of around
130 MeV=c. The energy spectrum for photons is even softer, since each photon receives on
average half of the energy of the neutral pion which created it. It is, therefore, necessary to
maintain good eciency and resolution for charged tracks and for photons at low momentum.
The tracking eciency cuts o at pt 60 MeV=c, and the eciency for photons at E 20 MeV.
It is particularly important to identify electrons and muons cleanly over as wide a range
of angle and momentum as possible. The primary leptons from B decays, which are used
Technical Design Report for the BABAR Detector
2.2 General Design Considerations
9
γ
BACKWARD
POLAR ANGLES
80
70
60
50
20
o
o
.7
.8
o
.6
.5
.4
.3 .2
o
50
.1
0
o
40 o
.1
.2
.3
.4
.92
.94
30 o
.5
.6
.7
.8
.85
.9 .92
.94 .96
.98
.96
.99
cos θ c m
0
FORWARD
POLAR ANGLES
o
60
.98
10
80
70
.9
o
o
o
.85
30
90
o
40 o
o
o
o
20 o
10
o
0
9 GeV on 3.109 GeV
Υ(4 ) βγ = 0.56
o
θlab
Protractor showing the relationship between center-of-mass and laboratory
polar angles for photons at = 0:56.
Figure 2-1.
to tag the avor of the B , have momentum as low as about 1 GeV=c. Secondary leptons
coming from the sequence B ! DX , D ! `Y provide additional tagging information, if it
is possible to identify them cleanly. Thus, the eciency of the lepton tag is limited by the
minimum momentum at which electrons and muons can be cleanly identi ed, which is less
than 500 MeV=c for the BABAR detector.
2.2.2
Charged Track Resolution and Multiple Scattering
As a result of the characteristically low momentum of particles produced by B mesons
decaying at rest, the errors on charged particle track parameters are usually dominated by
multiple Coulomb scattering rather than by the intrinsic spatial resolution of the tracking
chamber. For example, the resolution in measuring the longitudinal (z) position of the
B decay vertex, and therefore the decay time, is determined primarily by the amount of
material before the rst measurement in the vertex detector and by the radius of the beam
Technical Design Report for the BABAR Detector
10
Detector Overview
1.2
1.0
B°
ψ Ks
π +π +
+
+
0.8
ε
0.6
0.4
ptmin (MeV/c)
0
100
200
0.2
0
0.7
2-95
0.8
0.9
cosθlab
1.0
7857A15doc
Detection eciency for the decay B 0 ! J= KS0 ! e+ e, + , as a function
of forward polar angle coverage for di erent cuts on the minimum transverse momentum
accepted. The dotted line shows the eciency as a function of backward polar angle
coverage.
Figure 2-2.
pipe. This leads to a strategy of making the rst z measurement as close as possible to
the beam pipe. It is also important to put the second layer as close as possible with little
intervening material for the fraction of tracks in which there is no useful hit in the rst layer.
Multiple Coulomb scattering also dominates the angle and momentum resolutions for most
tracks. Good momentum resolution requires a continuous tracking volume lled with a
mixture of gas and wires that has a long track length and long average radiation length.
Another way to improve momentum resolution is to increase the magnetic eld, although
one must be careful not to compromise the acceptance for low momentum particles. For
most of the momentum range useful for B meson studies, the angle measurements are made
primarily in the vertex detector, before the original angle information is degraded by multiple
scattering.
Technical Design Report for the BABAR Detector
Entries / 0.025 GeV
2.2 General Design Considerations
11
600
400
200
0
0
1
2
3
4
Photon Energy (GeV)
Figure 2-3.
Energy spectrum for photons from the process B 0 ! , ! 0 .
2.2.3 Photon Eciency and Resolution
The low momentum range determines the detector requirements for photon detection even
more than for charged particles. Very good energy and angular resolution in the electromagnetic calorimeter for photons in the energy range from 20 MeV to 5 GeV are important.
Figure 2-3 shows the energy spectrum of s from the process B 0 ! , ! 0,
which is a very important CP mode. The 0 spectrum from the modes D+D,, J= K 0 ,
and J= KS0 is softer than that from , while the spectrum from 00 is harder. This leads
to the choice of a CsI calorimeter with fairly ne segmentation. Good resolution at high
energy requires that the crystals be suciently long so that the rear leakage is negligible.
Low-energy resolution requires that the electronic noise be kept low.
To achieve the required physics performance, one must also keep the number of radiation
lengths of material before the calorimeter to a minimum. The major sources of such material
are the particle identi cation device, and, in the endcap region, the drift chamber endplate
and associated hardware. In the barrel region, much of the material is close to the front face
of the calorimeter. Material in this location is the least troublesome, since both electrons
from a photon conversion often reach the calorimeter.
Technical Design Report for the BABAR Detector
12
Detector Overview
dN/dp [Entries per 50 MeV/c]
8000
6000
4000
2000
0: 0
1:0
2:0
3:0
p [GeV/c]
Momentum spectrum for kaons from B decays, which are used for tagging
the avor of the B .
Figure 2-4.
2.2.4 Identi cation of Hadrons
An essential requirement for the CP physics is to identify kaons for tagging with high
eciency and low probability of wrong sign tags. As can be seen in Figure 2-4, most of
the kaons from the process B ! DX , D ! KY have momentum less than 1 GeV=c. There
is a tail at higher momentum, however, which occurs at forward angles.
The other requirement is to separate pions from kaons in decays such as B 0 ! +,(K +,),
as well as in charmed meson and decays. At the extreme decay angle accepted by the
experiment, one pion from B 0 ! +, decay has a momentum of 4.2 GeV=c at cos lab =
0:95, and the other has a momentum of 1.5 GeV=c at cos lab = ,0:69. Although kinematic
tting provides some help in discriminating between these two-body decay modes, achieving
complete separation requires a Cherenkov detector for particle identi cation.
These physics requirements lead to a choice of particle identi cation system which combines
good dE/dx information from the drift chamber to identify kaons with momenta below
0.7 GeV=c and a Cherenkov device in the barrel and forward endcap regions for higher
Technical Design Report for the BABAR Detector
2.2 General Design Considerations
13
momenta, giving good K= separation up to about 4 GeV=c. There are no two-body decay
pions in the backward endcap region, so there is no reason to supplement dE/dx there.
2.2.5 Interaction with the Accelerator
PEP-II represents a new type of e+e, collider and is, therefore, an ambitious project. Highest
priority is being given to making the machine reliable and easy to operate with consistently
high luminosity and low background. Machine-design choices for the interaction region
impact the detector design. A support tube containing the interaction region magnets,
the masks to reduce beam-related backgrounds, the beam pipe, and the vertex detector,
is envisioned to run through the interaction region at a radius of 20 cm. This is the best
method for achieving precise and reproducible relative alignments of the machine components
within the detector. While the support tube introduces 0.005X0 of extra material into
the ducial volume of the detector, the e ect on resolution is rather small. The angles
and impact parameters are measured in the vertex detector, inside the support tube. The
transverse momentum is measured primarily in the drift chamber, in which there is very
little material for multiple scattering. Over the momentum range of charged particles from
B decay, the degradation in momentum resolution due to the support tube is always less
than 5%. Clearly, the prospect of better machine performance and reproducible alignment
of the vertex detector outweighs these consequences for tracking resolution.
The accelerator magnets closest to the interaction point also have consequences for the
detector design. The need to separate the two beams before the rst parasitic crossing
necessitates dipole magnets 21 cm from the collision point; the need to focus the beams to
a very small size necessitates having quadrupoles, which are also placed completely within
the detector volume. The dipole magnets complicate the vertex detector mechanical design
because they occupy much of the region in which the readout electronics and mechanical
support would be most conveniently located. The length of the detector solenoid magnet
is also limited, because if it were to become too large, it would be necessary to shield a
machine quadrupole (Q2) that contains iron, causing severe complications for the detector
and accelerator, and making access to inner detector components more dicult. There has
been a continuing joint e ort of physicists and engineers working on the accelerator and
detector to understand these con icts and to arrive at solutions which serve the common
goal of obtaining the best physics possible.
Because of the unprecedented beam currents expected during PEP-II operation, there has
been a great deal of attention given to backgrounds in the various detectors. Detailed
simulations of the synchrotron radiation and lost beam-particle backgrounds have been
re ned continuously since the earliest days of the machine design and have had substantial
impact on the design of the interaction region. The anticipated background levels, in ated
Technical Design Report for the BABAR Detector
14
Detector Overview
by a reasonable safety factor, are used to set detector parameters such as radiation hardness
of silicon electronics and bandwidth of the data acquisition system.
2.2.6
Considerations of Cost and Schedule
Although the primary considerations in designing the detector are physics requirements
and detector capabilities, the process should and does involve consideration of cost. Often,
one must choose among di erent ways to obtain comparable performance that have very
di erent costs. As in most such detectors, the calorimeter is the most expensive single
system. Because the calorimeter cost increases with volume, every e ort has been made to
minimize calorimeter volume as much as possible within the constraints imposed by physics
performance.
The schedule for PEP-II construction encompasses a six-month commissioning run starting
in the fall of 1998. Detector commissioning o the beam line should commence at the same
time, so that the detector is ready for installation into the interaction region at the end of
the machine commissioning period. This represents an extremely tight schedule on which to
design, engineer, construct, and assemble an experimental apparatus as large and complex
as BABAR. In considering detector technologies and alternate experimental designs, schedule
constraints have been an important criterion.
2.3 The Experimental Design
The BABAR design is shown in Figures 2-5 and 2-6. It consists of a silicon vertex detector,
a drift chamber, a particle identi cation system, a CsI electromagnetic calorimeter, and a
magnet with an instrumented ux return. The superconducting solenoid is designed for a
eld of 1.5 T, and the ux return is instrumented for muon identi cation and coarse hadron
calorimetry. All of these detectors operate with good performance for laboratory polar angles
between 17 to 150, corresponding to the range ,0:95 < cos cm < 0:87. The vertex detector
is mounted inside the support tube along with the beam pipe, the rst accelerator dipole,
and quadrupole magnets. In this section, we give a brief description of the detector; in
Section 2.5, we summarize the performance of the detector subsystems.
The detector coordinate system is de ned with +z in the boost (high-energy beam) direction
and +y in the vertical direction. The high-energy beam travels clockwise around PEP-II,
so the +x direction is away from the ring center. The coordinate system origin is the
nominal collision point, which is o set in the ,z direction from the geometrical center of the
detector magnet. Although the beams collide with each other head-on, they are separated
Technical Design Report for the BABAR Detector
2.3 The Experimental Design
15
6290
3750
1120
1120
2395
1655
150
150
0.653
100
100
Coil Cryostat
330
460
Calorimeter
1730
1400
1360
1113
1657
36
800
.249 ref.
36
810
10
Q4
Q2
Q1
900
890
Q2
Q1
225
500
3-95
7857A21
Figure 2-5.
Cross-sectional view of the detector.
while still inside the detector magnet eld. The detector is rotated 20 mr relative to the beam
direction (around the y axis) to minimize the resulting orbit distortions. The z direction thus
corresponds to the magnetic eld direction, and deviates slightly from the boost direction.
The tracking system in BABAR consists of the vertex detector and a drift chamber. The
vertex detector is the only tracking device inside the 20 cm radius of the support tube. It is
used to measure precisely both impact parameters for charged tracks (z and r , ); these
measurements are used to determine the di erence in decay times of the two B 0 mesons. The
vertex detector also provides the measurements of production angles, given the momentum
information from the drift chamber. Finally, charged particles with pt between 40 MeV=c
and 100 MeV=c are tracked only with the vertex detector, which must therefore provide
good pattern recognition. To serve these various functions, the vertex detector requires
excellent spatial resolution, low multiple scattering, small segmentation, and reasonably
good resistance to radiation.
Technical Design Report for the BABAR Detector
16
Detector Overview
Figure 2-6.
Three-dimensional view of the detector.
Technical Design Report for the BABAR Detector
2.3 The Experimental Design
17
The vertex detector consists of ve layers of double-sided silicon strip detectors. The inner
three layers are in a barrel geometry with detectors parallel to the beam pipe. The outer
two layers combine barrel detectors in the central region with wedge detectors forward and
backward. The entire vertex detector contains about 0.94 m2 of silicon and about 150,000
readout channels. This is somewhat larger than silicon vertex detectors built to date,
although it is substantially smaller than those designed for LHC. The size is determined
by its dual function as a vertex detector and as the only tracking device within a radius of
20 cm.
The second component of the tracking system is the drift chamber, which is used primarily
to achieve excellent momentum resolution and pattern recognition for charged particles with
pt > 100 MeV=c. It also supplies information for the charged track trigger and a measurement
of dE/dx for particle identi cation. The chamber extends in radius from 22.5 cm, just outside
the support tube, to 80 cm.
For most particles of interest at PEP-II, the optimum momentum resolution is achieved by
having a continuous tracking volume with a minimum amount of material to cause multiple
scattering. By using a helium-based gas mixture with low-mass wires and a magnetic eld
of 1.5 T, very good momentum resolution can be obtained. The forward edge of the chamber
is situated 1.66 m from the interaction point, which makes it possible to obtain reasonable
momentum resolution down to the limit of forward acceptance, 300 mr.
Two design options were considered for the drift chamber: a conventional layout with
four axial and six stereo superlayers, each consisting of four individual layers; and an allstereo arrangement of forty layers with alternating U and V stereo angles. The momentum
and angular resolutions for the entire tracking system were found to be comparable for
the two chambers. The axial/stereo chamber makes it more straightforward to design a
charged particle trigger in hardware which has a fairly sharp cuto in pt . The axial/stereo
arrangement has been adopted as the baseline design for the drift chamber.
The chamber is designed to minimize the amount of material in front of the particle identi cation and calorimeter systems in the heavily populated forward direction. The readout
electronics are mounted only on the backward end of the chamber, and the endplates are
designed as truncated cones probably to be made of carbon ber.
As stated above, there are two primary goals for the particle identi cation system. One is to
identify kaons for tagging beyond the momentum range well-separated by dE/dx. The other
is to identify pions from few-body decays, such as B 0 ! +, or B 0 ! . A new detector
technology is required to meet these goals, and in the barrel region, a DIRC (Detector of
Internally Re ected Cherenkov radiation) is used. Cherenkov light produced in 1:75 3:5 cm2
quartz bars is transferred by total internal re ection, while preserving the angle, to a large
water tank outside the backward end of the magnet. The light is observed by an array of
photomultiplier tubes at the outside of the tank, where images governed by the Cherenkov
Technical Design Report for the BABAR Detector
18
Detector Overview
angle are formed. A mirror at the forward end of the bars re ects the forward-going light,
preserving the angle information. This arrangement provides at least 4 standard deviation
=K separation up to almost the kinematic limit for particles from B decays.
In the forward region, for cos > 0:90, there are two layers of aerogel threshold Cherenkov
(ATC) counters, one of higher density (n = 1:055) with a kaon threshold at 1.6 GeV=c, and
the other a lower density (n = 1:0065) for high-momentum identi cation. This provides good
tagging eciency in the forward region, as well as allowing particle identi cation for pions
from B 0 ! +,. In the backward region, no particle identi cation is needed other than
dE/dx. The tagging kaons are quite slow there, and for any B 0 ! + , event with a pion
in the backward endcap region, the other pion is outside the limiting forward acceptance.
The electromagnetic calorimeter must have superb energy resolution down to very low
photon energies. This is provided by a fully projective CsI(Tl) crystal calorimeter, which
has excellent energy and angular resolution and retains high detection eciency at the
lowest relevant photon energies. The calorimeter consists of a cylindrical barrel section
with inner radius of 90.5 cm and a conical forward endcap. The barrel calorimeter contains
5880 trapezoidal crystals; the forward endcap calorimeter contains 900 crystals. The crystal
length varies from 17.5X0 in the forward endcap to 16X0 in the backward part of the barrel;
the typical transverse size is 4.8 4.8 cm2 at the front face.
Each crystal is read out by two independent silicon photodiodes. Electronic noise and beamrelated backgrounds dominate the resolution at low photon energies, while shower leakage
from the rear of the crystals dominates at higher energies. As mentioned above, minimizing
the calorimeter volume is an important consideration in choosing the detector geometry; in
the present design, it is 5.9 m3 .
To achieve very good momentum resolution without increasing the tracking volume and
therefore the calorimeter cost, it is necessary to have a eld of 1.5 T. The magnet is therefore
of superconducting design, with an inner radius of 1.40 m for the coil dewar and a cryostat
length of 3.85 m. The total thickness of the cryostat and coil is 0.25{0.4 interaction lengths,
which is thin enough for reasonable eciency in detecting neutral hadrons that traverse
it. The magnet is similar to many operating detector magnets, so the engineering and
fabrication should be straightforward. The nonstandard features are segmentation of the
iron for an Instrumented Flux Return (IFR), and the complications caused by the DIRC
readout in the backward region.
The IFR is designed to identify muons with minimum momentum around 0.5 GeV=c and
to detect neutral hadrons. The magnet ux return iron is divided into 20 layers, and the
thickness increases from 2{5 cm. Between most of the iron absorber layers are 3 cm-wide
gaps with Resistive Plate Chambers (RPCs), which serve as the active detectors. A system
with 17 layers of RPCs in the barrel and 16 in the endcap has been costed. With this
con guration, it is possible to identify muons down to 0.5 GeV. As an additional bene t,
Technical Design Report for the BABAR Detector
2.3 The Experimental Design
19
the IFR serves as a coarse hadron calorimeter, making it possible to detect KL0 s. The RPCs
represent a proven technology which adapts well to the BABAR geometry.
The high data rate at PEP-II requires a data acquisition system which is more advanced
than those used at present e+e, experiments. The bunch crossing period of 4.2 ns is so
short that the interactions are e ectively continuous, as in xed-target experiments. The
rate of all processes to be recorded at the design luminosity of 3 1033 cm,2s,1 is about
100 Hz. Simulations of machine backgrounds show hit rates of about 100 kHz per layer in
the drift chamber and about 140 MHz in the rst silicon layer. The goal is to operate with
negligible deadtime even if the backgrounds are 10 times higher than present estimates, an
environment which might develop, especially early in the life of the experiment.
In the electronics architecture adapted for BABAR, the experiment can take data in parallel
with processing and transporting the information from previous events. In addition, a uni ed
architecture for all detector systems minimizes the overall cost while enhancing the reliability
of the system. The memory bu ers that traditionally exist at the event-building stage are
replaced with large storage capacity immediately after digitization. This allows long latencies
in the data collection, which in turn decouples data acquisition from readout of the data.
The Level 1 trigger uses trigger primitives provided by special front-end interfaces from
the drift chamber and electromagnetic calorimeter. The trigger decisions are generated
with a xed 9:5 s latency at a maximum rate of 10 kHz at 10 times the nominal machine
backgrounds. The Level 1 trigger is required for readout of the vertex detector because of
the high hit rates from machine backgrounds. An optional Level 2 trigger could be used to
reduce the number of triggers before data are extracted and transmitted. A nal software
Level 3 trigger running in commercial processors reduces the number of events to a number
that can be written to archival storage.
The computing loads are also larger than those for previous experiments at e+e, colliders,
although they are within what is handled by existing hadron experiments. The wide geographical distribution of the collaboration adds an additional consideration in the design of
the computing system. Although much of the computing system will necessarily be installed
at the SLAC site, there will also be regional centers to provide more convenient access to the
data for the widely distributed institutes that comprise the BABAR Collaboration. Finally,
there will be signi cant computing installations at home institutions with network links to
the regional centers and the SLAC site.
The online requirement for CPU is 3000 MIPS, assuming a 3 kHz rate input to the Level 3
farm. The total CPU power needed for reconstruction, Monte Carlo, DST analysis, and
interactive analysis is about 17,000 MIPS. The storage requirements for BABAR are approximately 100 Tbytes/yr of tape storage. In addition, about 2 Tbytes of disk storage is needed
for a staging space for physics analysis. Aggregate network capacity for oine tasks of more
Technical Design Report for the BABAR Detector
20
Detector Overview
than 200 Mbytes/s is required, dominated by the reconstruction passes and multiple reads
through the data for DST creation and analysis.
A common computing environment based on clusters of distributed UNIX workstations and
industry-standard networks is being developed to handle this massive computing demand.
Flexible solutions are being chosen to take advantage of the evolution that the technologies
will inevitably undergo during the BABAR construction schedule.
2.4 Detector Optimization
In the process of arriving at the BABAR design described above, the collaboration considered
a wide range of detector technologies and geometry options. The motivations for some of
the important choices made early in the design process are described in the Letter of Intent.
Since that time, two important decisions were made which have a large impact on the design
of the entire experiment. These were the choices of hadron identi cation technology and
the optimization of the various detector dimensions, especially the outer radius of the main
tracking chamber.
2.4.1 Impact of Particle Identi cation
At the time of the Letter of Intent, BABAR had under consideration three types of Cherenkov
detectors to extend hadron identi cation beyond the momentum region covered by dE/dx
information from the drift chamber. The design and engineering of the entire detector
depended on choice of technology for particle identi cation. The DIRC occupied less radial
space than either a fast RICH or ATC counters, but the DIRC required penetration of
the backward endcap region for its optical readout. Thus integration engineering for the
experiment could not proceed before this decision was made.
A comprehensive review of the di erent technologies looked at performance obtained by prototypes, simulation of expected physics performance, impact on the rest of the experiment,
and prospects for meeting the tight construction schedule. The collaboration selected the
DIRC solution. Because it uses the smallest radial space of the three candidate solutions,
about 10 cm including mechanical supports and clearance space, it minimizes the cost of the
CsI calorimeter. It does, however, complicate the engineering of the backward endcap and
the access to the detector. Detailed studies indicate that access is possible to the end of the
drift chamber in approximately one eight-hour shift. This is essential to avoid long downtimes
which could compromise the factory-like performance of the machine and detector.
Technical Design Report for the BABAR Detector
2.4 Detector Optimization
21
Since the DIRC is made of long barrel staves, it does not provide coverage in the very forward
direction. This hole in the acceptance is lled by the two layers of aerogel threshold counters.
The layers have indices of refraction which allow K= separation at momenta up to about
4.3 GeV=c. The DIRC plus aerogel, after allowing for small gaps between quartz modules,
cover about 95% of the acceptance of the tracking devices.
2.4.2 Optimization of Performance versus Cost
The optimization of the BABAR detector geometry is tightly constrained by the accelerator
design and by cost considerations. The con guration of the interaction region, in particular the dimensions and locations of the B 1 magnets, determine the angular acceptance
in both forward and backward directions. The resulting center of mass acceptance of
,0:95 < cos cm < 0:87 is reasonably close to the preferred symmetric acceptance.
The forward length of the detector is determined primarily by the need to maintain good
momentum resolution and eciency for charged tracks over this entire acceptance. In
conventional solenoidal detectors, the resolution in the forward direction is signi cantly
degraded because the particles exit the drift chamber at a relatively small radial distance
from the beam. This is unacceptable at PEP-II, where the fraction of particles going forward
is greater and their momentum is higher. In order for tracks at 300 mr to pass through 20
of the 40 drift chamber layers, corresponding to a radius of 52 cm, the active volume of the
drift chamber extends to a distance of 166 cm from the interaction point.
Figure 2-7 shows the mass resolution for B 0 ! +, as a function of the cos cm of the
faster pion. There is a clear break at cos cm = 0:7, where the pions begin to exit the
chamber without traversing all 40 layers. The fraction of events with degraded resolution
and the average degradation would increase if the forward length of the drift chamber were
shortened.
This e ect can be seen even when the resolution is averaged over all decay angles. Figure 2-8
shows the mass resolutions for B 0 ! +, and D+ ! K ,++ as a function of the forward
length of the drift chamber. The B 0 ! +, resolution in particular increases signi cantly
as the chamber is shortened. Probably more important is the fact that the track- nding
eciency would su er for forward tracks traversing fewer than 20 layers, especially with
high rates in the inner layers from machine backgrounds. The solid angle coverage of the
DIRC in the center of mass also depends on the forward length of the detector. In the
backward direction, the drift chamber endplate is located a distance of 111 cm from the
interaction point, re ecting the lower momentum of the tracks.
The detector length beyond the forward drift chamber endplate is strongly constrained by
the necessity to minimize the fringe eld of the detector solenoid at the location of the
Technical Design Report for the BABAR Detector
22
Detector Overview
50
2
2
σM (MeV)
40
2
30
2
1
1
2
1
2
2 2
1 2
1 2
1 1
2 2 1 1
1 2
2 1
1 2
1 2
1 2
1 2
1 2
1 1 1
20
10
0
0
0.2
0.4
0.6
0.8
1
cosθcm
Mass resolution for B 0 ! + , as a function of the cos cm of the forwardgoing pion. The lower curve is for the chosen design, and the upper curve is for a
forward length of the drift chamber which is shortened by 44 cm, corresponding to only
four superlayers at 300 mr.
Figure 2-7.
quadrupole Q2 beam line magnet. There is just enough room for two layers of ATC counters
and the forward calorimeter. The resulting detector length, combined with the calorimeter
depth, then determines the barrel-endcap calorimeter interface.
The largest single element in the detector cost is the CsI calorimeter, and much e ort has
gone into minimizing its cost. The length has been reduced to the minimum compatible with
the demands on forward tracking just discussed. The other primary degree of freedom is the
choice of the outer radius of the drift chamber. The present radius of 80 cm is the minimum
compatible with good performance of the tracking system. Combined with the inner radius
of 22.5 cm, this small outer radius puts signi cant demands on the charged particle tracking
detectors to achieve adequate performance, both in momentum resolution and in pattern
recognition, within the limited space available.
Figure 2-9 shows the mass resolutions for B 0 ! +, and D+ ! K ,++, plotted as
a function of the outer radius. The B 0 ! +, mass resolution degrades by about 10%
for each 4 cm change in the outer radius; there are signi cant e ects in the D+ resolution
as well. The e ects on the robustness of pattern recognition may be greater, especially if
Technical Design Report for the BABAR Detector
2.4 Detector Optimization
23
D+Dπ+π-
B
B
7
28
24
5
20
B
6
π+π- RMS (MeV/c 2)
32
D+
K-π+π+ RMS (MeV/c 2)
8
4
2
3
4
5
6
7
16
SuperLayers
Mass resolutions for B 0 ! + , and D+ ! K , + + as functions of the
forward length of the drift chamber. A superlayer corresponds to four layers, and the present
design has tracks at 300 mr crossing ve superlayers.
Figure 2-8.
machine backgrounds are high, although they are harder to quantify. Experience indicates
that pattern recognition with high background rates requires at least 40 layers in the drift
chamber; if radial space were not at such a premium, more layers would be desired. Given
this minimum number of layers, the only way to reduce the drift chamber radius further is to
reduce the cell size. The cell height of 12.5 mm in the present design is about the minimum
practical size both for mechanical integrity of the endplate and for uniformity of the drift
cell.
The second major reduction in calorimeter cost came from removing about one radiation
length from the depth of the calorimeter given in the Letter of Intent design, averaged over
the full calorimeter. This does a ect the resolution for high energy showers, and some e ort
was made to leave enough length in the forward endcap calorimeter to reduce the serious
impact of shower leakage there. The mass resolution for the benchmark mode B 0 ! 0 0
changes by about 10% for every 0.5 radiation length variation in this length. Signi cant but
smaller e ects are observed through the many physics analyses using s and 0s, such as
B+ ! +0.
Additional savings on the calorimeter cost were achieved at some cost in physics performance.
The total number of crystals has been reduced from about 10,000 in the LOI to the present
6780. This reduces costs for both crystal production and electronics. A crystal in the center
Technical Design Report for the BABAR Detector
24
Detector Overview
Mass resolutions for B 0 ! + , and D+ ! K , + + , plotted as functions
of the outer radius of the drift chamber. Also shown is the corresponding cell size.
Figure 2-9.
of the barrel subtends an angle of about 5 mr. The angular resolution is approximately
proportional to the width of the crystal, and the 0 mass resolution is proportional to the
angular resolution. Further reduction of the segmentation would also increase the fraction
of showers which contain energy from nearby particles.
A backward endcap calorimeter was designed to sit behind the drift chamber endcap and
inside the DIRC bars. It would lie outside the angular acceptance of the vertex detector, and
would subtend about 1.5% of full solid angle in the center of mass. The physics motivations
were two-photon physics and a slight improvement of the hermeticity of the detector. As
part of the cost optimization, the backward endcap calorimeter was removed from the initial
detector design. Space has been left for addition of this item at a later date to enhance the
physics capability of the detector.
The e orts to minimize the drift chamber radius and length also minimize the cost of the
solenoidal magnet. Although the magnet ux return design has 20 iron layers, a minimal
system of 16{17 layers could be instrumented with RPCs at the start of the experiment to
save production costs. The magnet cost would not be reduced if there were fewer iron layers.
Technical Design Report for the BABAR Detector
2.5 Detector Performance
25
In the vertex detector, the one saving that was seriously considered was removal of the middle
layer, reducing the total number of layers to four. Because much of the cost for any vertex
detector lies in detector development, electronics development, mechanical engineering, and
prototypes, the cost savings turned out to be rather small. Given this fact, the special need
in this experiment for excellent eciency of the vertex detector, and the severe limits on
the access to the detector, led to the choice of a ve-layer design. The extra layer provides
redundancy that will be needed if channels or sections of the detector fail over the expected
long-running periods without access for repairs.
The BABAR detector should be capable of carrying out the rich physics program available at
PEP-II for a period of ten years or more. The detector design described in this document is
the result of a continuing process of balancing this requirement with a hard realism about
the need to match funds that are available to the collaborators over a rather short period of
time. As a result of this process, the physics performance has su ered measurably, but not
yet enough to severely limit the future of the experiment. Other steps were considered to
reduce the cost further, but they would have severely compromised the physics performance
for the duration of BABAR.
2.5 Detector Performance
In this section, we describe the performance that is projected for each of the detector
subsystems. We give only the most important results; more detailed discussions appear
in the individual system chapters. Where possible, we show the polar angle dependence
in the center-of-mass frame. The primary performance characteristics are eciency and
resolution. The luminosity required to obtain a given sensitivity is inversely proportional to
the eciency and directly proportional to (1+B/S), where B/S is the ratio of background
events to signal events. The ratio B/S varies widely for di erent decay processes, but is
proportional to the resolution in the invariant B mass. The background also increases with
the mass resolution for intermediate states, such as a D+ or 0. A summary of the major
parameters of all the detector subsections is given in Table 2-2.
2.5.1
Vertex Detector
The vertex detector information dominates the measurement of the track impact parameters,
both along and perpendicular to the beam direction. Background rejection using vertex
information improves with better impact parameter resolution. Figure 2-10 shows the
resolution for both impact parameters as a function of pt . These resolutions are close to
the lower limits that are imposed by multiple scattering in the beam pipe and the active
Technical Design Report for the BABAR Detector
10
3
10
2
σz (µm)
Detector Overview
σxy (µm)
26
(a)
0
1
2
10
3
10
2
3
(b)
0
1
2
600
(c)
500
pT (GeV/c)
σz (µm)
σxy (µm)
pT (GeV/c)
400
600
400
300
200
200
100
100
-1
-0.5
0
0.5
1
(d)
500
300
0
3
0
-1
cos θcm
-0.5
0
0.5
1
cos θcm
Resolutions for impact parameters at the vertex: (a) xy as a function of
at cos lab = 0; (b) z as a function of pt at cos lab = 0; (c) xy as a function of cos cm
for pcm = 1 GeV=c ; and (d) z as a function of cos cm for pcm = 1 GeV=c. For (a) and (b),
the dashed curve represents 50 m=pt .
Figure 2-10.
pt
silicon. At cos lab = 0, which is cos cm = ,0:5, the resolution for a single track is about
50 m=p 15 m in both dimensions. (All such momenta are assumed to be in GeV=c.)
The second set of plots in Figure 2-10 shows the dependence of resolution on cos cm at
constant pcm = 1 GeV=c. One can use 1=p scaling to determine the error for other momenta.
The xy resolution is below 100 m over the range ,0:85 < cos cm < 0:70. The angular
dependence is more symmetric in the center of mass than one might expect; although the
tracks traverse more material at forward laboratory angles, they also have higher momentum
in the laboratory frame. The z resolution is similar to the xy resolution for most of the
angular range but becomes larger as the track angle approaches the z axis. The two
resolutions are approximately related by the expression z xy = sin lab . Determining
the resolution in vertex separation between two B decays requires a full physics simulation;
results from this study are discussed in Section 2.6.2.
t
Technical Design Report for the BABAR Detector
27
σθ (mrad)
σφ (mrad)
2.5 Detector Performance
(a)
10
1
(b)
10
1
0
1
2
3
0
1
2
8
7
6
5
4
3
2
1
0
(c)
-1
-0.5
0
0.5
pT (GeV/c)
σθ (mrad)
σφ (mrad)
pT (GeV/c)
1
cos(θcm)
3
8
7
6
5
4
3
2
1
0
(d)
-1
-0.5
0
0.5
1
cos(θcm)
Resolutions for angle measurements: (a) as a function of pt at cos lab =
(b) as a function of pt at cos lab = 0; (c) as a function of cos cm for p = 1 GeV=c;
and (d) as a function of cos cm for p = 1 GeV=c. For (a) and (b), the dashed curve
corresponds to 1:6 mr=pt , and the open circles correspond to the resolutions using a vertex
constraint, as described in the text.
Figure 2-11.
0;
The two angles for charged tracks are also determined primarily in the vertex detector,
assuming that the curvature is measured in the drift chamber. Figure 2-11 shows the
resolution in both angles as a function of pt and polar angle. At cos lab = 0, the resolution
is about 1.6 mr/p in both and . The angular resolutions are quite good over most of
the acceptance of the experiment, good enough that the momentum error dominates the
mass resolution in almost all cases. For momenta less than about 0:5 GeV=c, the angles
are determined better by using a vertex constraint constructed using the faster tracks.
Figure 2-11 shows the e ects of such a constraint, assuming a reasonable error on the
reconstructed vertex of 100 m. In particular, for slow pions coming from the process
D + ! + D 0 , this gives an angular resolution of about 4 mr.
The angular acceptance for the entire experiment is determined by the vertex detector, which
is limited by machine components. In the forward region, great care has been taken to reduce
Technical Design Report for the BABAR Detector
28
Detector Overview
the space between the outer surface of the B1 magnet and the region covered by active silicon
to about 1 cm. This results in coverage down to about 17, with a slight variation due to the
hexagonal structure. To obtain such tight clearance in the forward direction, the services
for the beam pipe are concentrated in the backward region. In addition, the mechanical
support for the silicon is somewhat larger there, to make installation and precise alignment
more tractable. As a result, the acceptance extends backward to a polar angle of about
150. The resulting acceptance extends from ,0:87 < cos lab < 0:96, which transforms into
,0:95 < cos cm < 0:87.
Because the vertex detector provides crucial tracking information for all charged tracks, it
is important to achieve almost perfect eciency within the acceptance of the experiment.
For tracks with transverse momentum greater than about 100 MeV=c, the tracking eciency
in the drift chamber alone is very high, as it is in existing experiments. The eciency of
yielding a good vertex detector track matching a track already identi ed in the drift chamber
is at least 96%, even under conditions of high machine background. Tracks with transverse
momentum between 40 MeV=c and about 100 MeV=c are reconstructed primarily in the silicon
tracker. Detailed studies of pattern recognition, including levels of machine background and
silicon ineciency higher than expected, show a track- nding eciency of about 88% for
these soft tracks.
Access to the vertex detector will be dicult. It will require removal of the support tube and
the machine components in the interaction region, a process which will occur rarely under
normal running conditions. For this reason, the vertex detector is being designed to be
extremely reliable and robust. This includes having redundant readout paths, for example.
A more important consequence of the robustness requirement is the decision to include ve
layers of double-sided silicon detectors. This provides some insurance that in case of steady
loss of channels over a long run, or an unlikely loss of an entire readout module, the track
eciency stays high even for low momentum tracks.
2.5.2 Drift Chamber
To achieve the required sensitivity for the CP measurements with BABAR, it is necessary to
measure all charged tracks within the acceptance of the detector with excellent precision
and high eciency. The angular acceptance is within 17 < < 150; the interesting tracks
for B physics are in the range 60 MeV=c < p < 2:5 GeV=c, while for charm and tau physics,
some tracks are of higher momentum.
The main tracking chamber provides the precise measurement of p needed for good mass
resolution on exclusive B decays. The full tracking system, consisting of vertex detector
and drift chamber, provides very good pattern recognition capability for charged tracks
even at 10 times the nominal machine background. The pulse height information from drift
t
t
Technical Design Report for the BABAR Detector
2.5 Detector Performance
29
pt resolution at (a) lab = 90 as a function of pt , and (b) pcm = 1:0 GeV=c
as a function of cos cm .
Figure 2-12.
chamber signals is used to measure the mean ionization loss (dE/dx), which can be used
to separate kaons from pions well at low momentum, and to provide some discrimination
at high momentum. Finally, the drift chamber information is used to construct a tracking
trigger, one of the two major triggers for the experiment, over 92% of the full solid angle in
the center of mass.
The primary performance parameter for the tracking chamber is, therefore, the momentum
resolution for charged particles with p between 100 MeV=c and 2.5 GeV=c. The value of
100 MeV=c is set by the the inner radius of the chamber, 22.5 cm, and the need to get enough
measurements inside the chamber for good determination of the momentum. The upper
limit is the kinematic limit for particles resulting from B decays.
Figure 2-12(a) shows the p resolution as a function of p for charged particles at cos lab = 0.
The resolution can be parameterized by (p )=p [0:21% + 0:14% p ] from 0.2 GeV=c
to the kinematic limit for B decays, 2.6 GeV=c. The constant term is the contribution of
multiple scattering, and its low value is due to the use of a helium-based gas in the chamber.
The values here are expected to be slightly optimistic because they assume perfect eciency
and pattern recognition within the chamber. The resolution degrades below a momentum
of 180 MeV/c because the path length is shortened in the chamber, but it is still quite good
down to about 100 MeV/c.
The tracking chamber covers the complete polar angle range allowed by the beam-line
components, ,0:87 < cos lab < 0:96. The o set of the chamber center from the interaction
point in the boost direction results in a p resolution which is fairly independent of cos cm
except at very forward angles, as shown in Figure 2-12(b). The resolution increases for
tracks with cos cm > 0:72 which exit the end of the tracking chamber before they reach the
t
t
t
t
t
t
t
Technical Design Report for the BABAR Detector
30
Detector Overview
maximum radius. For tracks with pt less than 0.2 GeV=c and at forward laboratory angles,
p =p is signi cantly greater than pt =pt because the contribution from the error on the
polar angle measurement becomes large. The acceptance and mass resolution for important
exclusive B decays are discussed in Section 2.6.1.
The dE/dx measurements with the designed chamber provide at least 3 K= separation at
momenta up to 0:7 GeV=c, which covers a substantial fraction of the kaons used for tagging,
as one can see from Figure 2-4. This includes most of the small number of tagging kaons
in the backward endcap region, where no other hadron identi er is used. For pions with
a momentum of 3 GeV=c from the decay B 0 ! +,, the ionization measurements may
provide K= separation at about the 2 level.
2.5.3 Particle Identi cation
There are two important benchmarks of performance for the particle identi cation system.
One is the ability to separate B 0 ! +, from B 0 ! K +, and the other is the e ective
eciency for kaon tagging. Because the momentum range for each of these depends on polar
angle, one must consider their performances separately at di erent angles.
Most of the pions from B 0 ! +, fall in the barrel region, in which the DIRC is providing
the hadron identi cation. The number of photoelectrons observed by the DIRC for pions from
this source is 40 at the backward end of the DIRC, reducing to about 20 at cos lab = 0:0,
and increasing again to 50 at the forward end of the DIRC. Of course, the momentum of the
forward pions is greater, which makes the K= separation more dicult. Figure 2-13 shows
the number of standard deviations of K= separation for pions from this decay mode as a
function of cos cm . It drops from complete separation for backward pions to a separation
of about 4 standard deviations at cos cm = 0:0. It is then constant for forward angles up
to cos cm = 0:7, which corresponds to the forward edge of the DIRC. Two estimates of
K= separation from dE =dx are also shown for reference: they correspond to optimistic and
pessimistic estimates of performance.
For B 0 ! +, events with both pions in the acceptance of the tracking system, the fraction
of pions which miss the DIRC is about 11%; 4% due to cracks between the quartz bars in
azimuth, and about 7% in the forward endcap region. All of the B 0 ! +, events with
a pion in the backward endcap region have their other pions striking the B1 magnet, so
there is no acceptance lost by not having high momentum hadron identi cation there. In
the forward endcap, the low-density ATC layer achieves K= separation to 4.3 GeV=c. In
both barrel and forward endcap regions, the information from the Cherenkov detectors is
combined with dE/dx and kinematics to separate B 0 ! +,. The performance using all
three techniques is discussed in Section 2.6.5.
Technical Design Report for the BABAR Detector
Number of Standard Deviations
2.5 Detector Performance
31
16
14
12
10
8
ATC
DIRC
6
4
dE/dx
2
0
-0.8
-0.4
0
cosθcm
0.4
0.8
separation for pions from B 0 ! + , as a function of cos cm, with
full particle identi cation (DIRC and ATC), and with dE =dx only. The shaded region
delineates two dE =dx curves. The upper curve represents the separation obtained using a
modi ed version of a dE =dx program from Va'vra et al., as described in Chapter 5. The
lower curve includes a degradation of the resolution by 50% to better account for observed
experiences with some large drift chambers in magnetic detectors.
Figure 2-13.
K=
Kaon tagging is important for all of the CP studies to be done with BABAR. The gure of
merit is the e ective tagging eciency, "(1 , 2w)2, where " is the fraction of events tagged
and w is the fraction of those in which the tag is of the wrong sign. Table 2-1 shows the
e ect of the DIRC on the kaon tagging eciency. With perfect hadron identi cation and
the BABAR acceptance, assuming the loss of kaons decaying in ight, the e ective tagging
eciency using charged kaons alone would be about 22%. Much of the kaon spectrum shown
in Figure 2-4 is beyond the momentum region in which dE/dx information is useful. The
e ective tagging eciency for events with charged kaons is 14%, if one uses optimistic dE/dx
information for kaon tagging, combining it with information from lepton tagging. Adding
the information from the DIRC and ATC detectors, the e ective tagging eciency rises to
21%, close to the limit of perfect kaon tagging. The main uncertainties in this number are
due to the uncertainties in the kaon yield from B decays.
In the forward region, the refractive index of the ATC counters is chosen speci cally to
optimize the tagging eciency. Two layers of ATC counters are used, one with n = 1:055
corresponding to a kaon threshold of 1:8 GeV=c, and the other with n = 1:0065 for momenta
Technical Design Report for the BABAR Detector
32
Detector Overview
Information Available
Perfect Identi cation at Production
Perfect Identi cation, with Decays
dE/dx Only
DIRC, ATC, and dE/dx
(1
, 2w)2(%)
27
22
14
21
The e ective kaon tagging eciency. The top line corresponds to every hadron
being identi ed correctly at production. For the line below, every hadron is identi ed
correctly except those which decay before the DIRC. For the last two lines, the eciency
includes the e ect of using lepton tagging information in events with kaons.
Table 2-1.
up to 4.3 GeV=c. Pions with momentum greater than about 0.7 GeV=c are detected with
high eciency, which means reasonable overlap with the dE/dx measurements in the drift
chamber.
Traditionally in ring-imaging Cherenkov systems, pattern recognition or association of photoelectrons with charged tracks can be a dicult problem. It is a measure of the robustness of
the DIRC that good performance is observed in full GEANT simulations with e ectively no
pattern recognition. Every photoelectron with proper timing is used as a candidate for each
track, and all possible values of the Cherenkov angle (c ) are calculated. When all of these
many values of c are plotted in a histogram for a given track, a peak at the true Cherenkov
angle stands above the at background, even in busy events. The ability to measure the
mean value hci is reduced very little from the case with no background. If the number of
photoelectrons Npe were reduced due to lower-than-expected transmission or photocathode
q
eciency, the main e ect on the measurement would be the fact that (hci) / 1= Npe.
Another important use of the particle identi cation detectors is to extend the region of good
lepton tagging to lower momentum. The DIRC gives e= separation of at least 4 standard
deviations for plab < 0:7 GeV=c, giving a second contribution to electron identi cation in a
region in which the calorimeter information is somewhat less clear. This is a useful region
for tagging electrons, especially at backward angles in the center of mass. There is also some
= separation which complements the IFR muon identi cation, again having the greatest
e ect for muons with cos cm < 0.
2.5.4 Electromagnetic Calorimeter
Achieving the desired sensitivity for measuring asymmetries in CP decay modes such as
J= KS0 , J= K 0 , D + D , , and requires observing 0 s with very high eciency and
good resolution. This is the physics goal which drives the performance requirements of the
Technical Design Report for the BABAR Detector
2.5 Detector Performance
2.5
33
9
σE / E (%)
9
2.0
9
9
9
1.5
9
1.0
0.02
0.1
0.2
1
2
Photon Energy (GeV)
Energy resolution at cos cm = 0 as a function of photon energy. The
resolution is de ned as FWHM/2.36. The error bar on the rst point indicates a typical
uncertainty in determining the resolution. The solid line shows the target energy resolution.
Figure 2-14.
electromagnetic calorimeter. The calorimeter is also used for electron identi cation and
to supplement the IFR information in identifying muons and KL0 s, as well as providing
information for the neutral trigger.
For decay modes such as B 0 ! ; ! 0, the 0 momentum ranges up to about
2.5 GeV=c in the center of mass, leading to the energy spectrum for photons in the laboratory
shown in Figure 2-3. From this distribution and others for additional modes, one sees that
the photon energy region of interest for B physics is from about 20 MeV to about 5 GeV in the
laboratory. The B mass resolution for this and other modes using 0s is dominated by the
photon energy resolution. Figure 2-14 shows the energy resolution as a function of photon
energy at cos cm = 0:0. The energy resolution (FWHM/2.36) is about 2.1% at 100 MeV
and about 1.6% at 1 GeV. Included in these numbers is the e ect of material before the
calorimeter, which at a photon energy of 100 MeV degrades the resolution by 10{15% and
reduces the eciency by 12{22%. Figure 2-16 shows the amount of material between the IP
and the various detector subsystems, up to the calorimeter. The amount of that material is
about 0:23X0= sin for the barrel, of which 75% is in the DIRC, and about 0.35X0 for the
forward endcap. The resolution
meets the target energy resolution for photons at this angle,
1
E =E = 1%=E ( GeV) 4 1:2%. The constant term arises from leakage, inter-calibration
errors, and nonuniformity of light collection.
Technical Design Report for the BABAR Detector
34
Detector Overview
3.5
2
σ E / E (%)
3.0
2
2.5
100 MeV
2.0
1.5
2
3
2
2
3
3
2
3
3
3
1 GeV
2
2
2 33
3
1.0
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
cos( θcm )
Energy resolution as a function of cos cm for E lab = 0:1 GeV and 1:0 GeV.
The error bars on the rst points indicate typical uncertainty in determining the resolution.
Figure 2-15.
Figure 2-15 shows the energy resolution as a function of cos cm for photons with energies of
0.1 and 1.0 GeV in the laboratory. The resolution at cos cm = 0:5 is 2.5%(2.0%) for photons
of energy 100 MeV(1 GeV). The resolution at angles more forward than that increases because
of leakage of the shower from the uncovered sides of crystals in the staggered arrangement.
The left edge of the plot is cos cm < ,0:5, which corresponds to 90 in the laboratory; the
resolution remains approximately at at more backward angles.
Another important performance parameter is the eciency for 0s; however, this depends
subtly on details of the data treatment in speci c analyses. This eciency is determined by
a cut on the invariant mass, the value of which depends on the background level for a
particular physics process. Figure 2-17 shows the mass spectrum for 0s from the process
B 0 ! . The resolution function is a Gaussian with 5 MeV, plus an additional tail
at low m( ). The Gaussian part of the m( ) resolution function is dominated by the
angular resolution of the calorimeter for 0s with momentum greater than about 1 GeV=c.
Material in front of the calorimeter is responsible for the low mass tail. In Section 2.6.3, the
0 eciency from full physics simulations of a couple of physics processes is discussed.
The angularqresolution for photons in the barrel region is given by the approximate formula
= 3 mr= E (GeV) 2 mr, which is consistent with scaling from experience with the
CLEO-II calorimeter. The angular resolution is determined by the transverse crystal size
and the average distance to the interaction point.
Technical Design Report for the BABAR Detector
35
X0
2.5 Detector Performance
0.6
0.4
0.2
0
-1
-0.5
0
0.5
1
cos(theta)
Figure 2-16. Amount of material between the IP and the various detector subsystems,
measured in radiation lengths, as a function of cos lab . The ve curves, from bottom to top,
give the material up to and including the beam pipe, the vertex detector, the drift chamber,
the particle-identi cation system, and the aluminum shield in front of the calorimeter. The
aerogel geometry used in making this gure was in an early stage of development and does
not correspond to that described in Chapter 6.
Technical Design Report for the BABAR Detector
36
Detector Overview
140
Entries / 0.001 GeV
120
100
80
60
40
20
0
0
0.05
0.1
0.15
0.2
γγ Invariant Mass (GeV)
Figure 2-17.
invariant mass spectrum from the process B 0 ! .
The barrel calorimeter covers the range ,0:80 < cos lab < 0:89; the forward endcap
calorimeter extends the forward coverage to cos lab = 0:97. Showers at least two counters in
from the edge are contained well enough for good measurement. In the center of mass, this
corresponds to containment of any shower in the range ,0:92 < cos cm < 0:87.
Another parameter that determines the eciency of the calorimeter is the minimum detectable energy for photon showers. The incoherent electronic noise per crystal is expected
to be about 150 keV, which adds up to 750 keV for a typical cluster of 25 crystals. This is low
enough that the minimum detectable energy should be determined primarily by backgrounds
from the event and from beam interactions, and is expected to be 10{20 MeV.
The nal measure of calorimeter performance is the ability to identify electrons. Simulations
and experience with CLEO-II indicate a very high eciency for identifying electrons down to
a momentum of 500 MeV=c. The probability for a pion to fake an electron is about 1 10,3
at high momentum.
Technical Design Report for the BABAR Detector
2.5 Detector Performance
37
50000
b
Muons
40000
c
30000
uds
20000
10000
0
0
1
2
3
4
p (GeV/c)
Momentum spectrum of muons in the laboratory. The three sources are
primary, secondary, and continuum muons.
Figure 2-18.
2.5.5
Muon and Neutral Hadron Detector
The primary experimental goal of the Instrumented Flux Return (IFR) is to reduce the lower
momentum limit for cleanly identifying muons from the value of 1:4 GeV=c csc lab that can
be obtained with an unsegmented iron absorber to about 0:6 GeV=c csc lab . This increases
the eciency for tagging the avor of a B meson substantially, and it also increases the size
of the lepton sample for studies of semileptonic decays. Figure 2-18 shows the momentum
spectrum of muons, both primary and secondary.
Figure 2-19 shows the eciency for identifying muons produced at cos cm = 0 as a function
of laboratory momentum. The eciency is about 50% for muons of 650 MeV=c momentum
and rises to an eciency of 90% for muons of 800 MeV=c momentum. The fraction of
pions decaying before the IFR at a momentum of 1 GeV=c is about 4%, but by comparing
measured momentum with range in the iron, one can reduce this background substantially. In
the region below about 600 MeV=c, there is supplemental information from the DIRC which
helps identify muons. The eciency cuto as a function of plab is shown in Figure 2-19 and
remains approximately the same as the angle varies.
The thickness of the iron layers is graded from 2 cm on the inside to 5 cm on the outside. The
ner segmentation of the innermost layers is chosen to identify muons of the lowest possible
momentum. Below 1.0 GeV=c in the center of mass, the muons are predominantly secondary,
Technical Design Report for the BABAR Detector
Detector Overview
µ efficiency
38
1
0.9
0.8
0.7
0.6
0.5
0.4
θcm = 90
0.3
o
0.2
0.1
0
Figure 2-19.
0
0.5
1
1.5
2
2.5
p (GeV/c)
Eciency for identifying muons with cos cm = 0 as a function of plab .
from semileptonic decays of the charm product of a B decay. In events in which one
B decay is completely reconstructed, it is possible to use information about the isolation of
the lepton to help separate primary from secondary muons.
Although the IFR has been designed and optimized with muons as the primary goal, the
importance of detecting the KL0 from B 0 ! J= KL0 has been kept in mind. The goal is
not to obtain an accurate measurement of the KL0 energy, but simply to identify the KL0
and to measure its angle reasonably well. For one set of cuts, the eciency for detecting the
KL0 from B 0 ! J= KL0 is 19%, with a signal/background ratio of 3.5. The most important
background process is B 0 ! J= K , K ! KL0 in which the K cannot be vetoed.
A third use of the IFR is to detect whether there is any hadronic energy which has escaped
the electromagnetic calorimeter. This is useful, for example, in the study of exclusive b ! u
semileptonic decays, for which a serious background comes from continuum charm events
which contain leptons. In suppressing these backgrounds, it is helpful to know if there is
substantial missing energy, and if so, in what direction it lies. Although this has not been
studied speci cally for BABAR, it is known to be helpful in such studies now done at CLEO.
i.e.,
Technical Design Report for the BABAR Detector
2.6 Physics Performance
39
2.5.6 Electronics, Trigger, and Data Acquisition
The primary performance measure of a trigger system is its eciency for benchmark physics
processes. The charged track trigger requires at least two tracks in the drift chamber, one
with pt > 0:21 GeV=c and another with pt > 0:13 GeV=c. The neutral trigger requires two
showers in the electromagnetic calorimeter, both with reconstructed energy deposits above
a threshold that is ecient for muons. The orthogonality of the requirements allows good
cross calibration of trigger eciency. With no further restrictions, the total trigger rate is
simulated to be about 8 kHz at 10 times nominal background, which is greater than the
speci cation of 2 kHz. The rate can be reduced further rather simply by introducing a pt cut
on the track with the largest transverse momentum and an energy cut on the largest energy
deposit.
The eciency for events such as B 0 ! +,, B 0 ! X is simulated to be close to 100%
for any reasonable variation of the proposed trigger. There are so many charged tracks and
photons with high momentum in such events that the eciency is very robust.
Tau events with 1+1 prongs and low-mass two-photon production are more demanding of the
trigger, and therefore drive the detailed trigger logic. Eciencies larger than about 70% are
expected for events with + ! e+ , , ! , . The crucial performance parameter for
such events is not the eciency itself, but the systematic error in determining the eciency.
Since the eciencies for events are dominated by the angular acceptance of the experiment,
they should be measurable with great precision.
Reasonable thresholds for the trigger described above are pt > 0:6 GeV=c for at least one track
in the charged particle trigger and E > 0:6 GeV for at least one cluster in the neutral trigger.
With these thresholds, the rate is about 400 Hz at nominal backgrounds and 1400 Hz at 10
times nominal. Thus, the proposed trigger architecture with Level 1 and Level 3 meets the
design requirements and has very high eciency. The optional Level 2 ensures the robustness
of the design. When more information is available on such parameters as the Level 3 latency
and event-building rate, it will be known whether the Level 2 trigger is needed.
2.6 Physics Performance
The previous section contains the results of a series of simulation studies of particular
detector systems, often based on simulations of a single track or shower. In this section,
we complete the picture by discussing aspects of the detector performance which require
full physics simulations to study properly, or require combining information from multiple
detector systems. The physics modes studied are those benchmarks described in Chapter 3.
Technical Design Report for the BABAR Detector
40
Detector Overview
Detector
SVT
Technology
Double-sided
Silicon Strip
DC
Small Cell
Drift Chamber
PID
PID
CAL
MAG
IFR
Dimensions
5 Layers
r = 3:2 , 14:4 cm
,0:87 < cos < 0:96
40 Layers
r = 22:5 , 80:0 cm
,111 < z < 166 cm
1:75 3:5 cm2 quartz
,0:84 < cos < 0:90
Performance
z = xy = 50 m=pt 15 m
= = 1:6 mr=pt
( )
pt =pt
= [0:21% + 0:14% pt ]
= 20 , 50
4
separation for
all B decay products
ATC
n=1.0065, 1.055
Npe = 10
0:916 < cos < 0:955 =K separation up to 4.3 GeV=c
1
CsI(Tl)
16 , 17:5 X0
E =E = 1%=E ( GeV) 4 1:2%
q
4:8 4:8 cm crystals = 3 mr= E (GeV) 2 mr
Superconducting
IR=1.40 m
Segmented Iron
L = 3.85 m
B = 1:5 T
RPC
16{17 Layers
> 90%
for p > 0:8 GeV=c
DIRC
Table 2-2.
Npe
> K=
Parameter summary (all angles are in the laboratory).
2.6.1 Acceptance and Mass Resolution for Decays to Charged
Particles
The best measures of the overall acceptance of BABAR are the eciencies for multiparticle CP
nal states. The eciency found for the process B 0 ! J= KS0 , J= ! `+`,, KS0 ! +, is
57%, which corresponds to about 87% per track, including geometric acceptance. The polar
angle acceptance set at the vertex detector by the B1 accelerator magnets is about 91%, so
this represents rather little additional loss for lepton identi cation, KS0 nding, and so forth.
Another benchmark channel is B 0 ! D+D, ! +(K ,+),(K +,), which emphasizes
acceptance for charged particles with very low momentum. The eciency for measuring each
slow bachelor pion is about 87%, including the geometric acceptance of 91%. The ability to
track slow pions in the vertex detector is crucial in maintaining this high an eciency. The
total eciency for reconstructing a B 0 in this mode is about 40%.
Good mass resolution is needed to reduce the combinatoric backgrounds, and for all-charged
decay modes, this is dominated by the momentum resolution. The primary physics benchmark for momentum resolution is the mass resolution for the mode B 0 ! +,, in which the
Technical Design Report for the BABAR Detector
2.6 Physics Performance
41
pions are at the kinematic limit. It has the added feature of a large background peak from
B 0 ! K +, separated by only about 43 MeV=c2 in mass, making the mass resolution even
more critical than for other modes. Simulation shows a mass resolution of about 21 MeV=c2
where this value does not yet take into account the e ect of missed hits and fake hits in the
drift chamber. This resolution is determined by the tracking length in the drift chamber
(80 , 22 = 58 cm), the magnetic eld, the intrinsic resolution of each measurement, and the
number of radiation lengths within the chamber volume.
The key to reducing backgrounds for B decay modes such as B 0 ! D+D, is good resolution
on the mass di erence m = m(D) , m(D). Because the pion is slow in the D rest frame,
the resolution on this mass di erence is typically dominated by the angular resolution of
the slow pion. Simulation shows a resolution of about 0:4 MeV in m, which compares
very well with existing experiments. This is achievable by using the production point and
a precise vertex detector measurement near the beam pipe, which reduces the e ect of the
multiple scattering on the angle measurement.
2.6.2
Separation between
B Decay Vertices
The most important role of the vertex detector is to determine the separation between the
two B decay vertices along the z axis, which is needed to measure the CP asymmetries. For
CP studies in modes with little background, this does not place a very stringent requirement
on intrinsic position resolution. The average separation of the two B vertices is 250 m,
and the asymmetry peaks at z 550 m. The degradation in the measurement of the CP
asymmetries due to imperfect vertex resolution is less than 10% if the separation is measured
with an error of 125 m. Figure 2-20 shows the resolution in z zCP , ztag for the case
of B 0 ! +, with a primary lepton tag. The expected distribution for z can best be
characterized as a narrow Gaussian of about 70 m which contains 80% of the events and a
wider gaussian of about 220 m which contains the rest. The resolution in z for this and
other modes is determined primarily by the error on the z position of the tagging vertex.
For measurements of CP decay modes with small backgrounds, the vertex resolution is
suciently good that there is little degradation of the asymmetry measurement compared
to the case of perfect resolution. For other modes, especially the modes used to measure
sin 2 , the vertex separation of the two B decays has an additional importance, since it is
used to suppress the dominant backgrounds. The z distribution of the background is a
gaussian centered at zero with 100 m, where the width is determined by the resolution.
For all of these modes, the e ective background level at z 250 m is very sensitive to
the resolution (z), and the CP measurement improves as this resolution decreases. The
precision achievable is limited by multiple scattering, which sets a natural scale for the point
resolution in the inner layers.
Technical Design Report for the BABAR Detector
Detector Overview
Events / 20µm
42
400
350
300
250
200
150
100
50
0
-1000 -750
Figure 2-20.
lepton tag.
2.6.3
0
-500
-250
0
250
500
750
1000
∆Zmeas - ∆Zgen (µm)
Resolution in z zCP , ztag for the case of B 0 ! + , with a primary
Eciency and Resolution
Full physics simulations were done to measure the eciency for nding 0s embedded in
full events. For the process B 0 ! J= KS0, J= ! `+`,, KS0 ! 00, the eciency is 25%,
or about 0.44 times the eciency obtained for the same process with KS0 ! +,. This
corresponds to an eciency for each 0 of about 60% each, including geometric acceptance.
Since the typical 0 momentum is slightly less than 1 GeV=c, this result demonstrates the
importance of high detection eciency for photons of modest energy.
The material in front of the barrel calorimeter is dominated by the DIRC quartz bar and
associated mechanical supports. The low-mass tail shown in Figure 2-17 is mostly due to
pairs produced in the material before the calorimeter, which are not included in the shower
energy. The net e ect of all such material is to reduce the eciency for 0s in the
mode by about 15% and to degrade the mass resolution by 10%.
For most B decay modes with at least one neutral pion, the energy resolution of the
calorimeter dominates the mass resolution. The benchmark decay analogous to B 0 ! +,
Technical Design Report for the BABAR Detector
2.6 Physics Performance
43
is B 0 ! 0 0; not only is it an important physics mode, but the resolution is closely related
to that for such modes as B + ! +0. The energy resolution averaged over the entire
detector leads to a mass resolution for B 0 ! 00 of about 64 MeV(FWHM/2.36). The
resolution function is, of course, asymmetric with a low-mass tail. The background from the
continuum is larger than the signal for reasonable estimates of the branching ratio, so the
success of this analysis depends critically on maintaining the quoted energy resolution.
2.6.4 Lepton Identi cation
The identi cation of high-energy leptons from B 0 decays such as those coming from B 0 !
J= KS0 is straightforward, and the eciency is very high. Tagging leptons are more challenging, however, since the spectrum of useful leptons extends to rather low momentum, as seen
in Figure 2-18. A full study shows an e ective tag eciency for primary leptons of 13.2%,
assuming a semileptonic branching fraction of 10.5%. Besides the branching fraction, the
factors which determine the tagging eciency are the primary and secondary lepton spectra,
the polar angle acceptance, the eciency at low momentum, and the probability of pions
decaying. Good identi cation eciency is obtained for low momentum electrons using the
calorimeter and the DIRC or ATC counters; for low momentum muons with the IFR and
the DIRC. This contributes to additional tagging eciency using secondary leptons.
2.6.5 Charged Hadron Identi cation
The bulk of charged hadron identi cation is done with the particle identi cation detectors,
the DIRC and ATC counters. In Figure 2-13, we showed the =K separation obtained for
the important decay mode B 0 ! +, using the DIRC and dE/dx information. To gain full
separation from the K + , decay, both of these and kinematics will be used.
Figure 2-21 shows the eciency for accepting a B 0 ! +, event as a function of 2 , the
di erence in 2 between a +, assignment and a K +, assignment, for two di erent assumptions about drift chamber performance. The 2 represents the consistency with a given
hypothesis using either kinematics and dE/dx alone or including the particle identi cation
detectors. A reasonable measure of good suppression of the competing mode is 2 > 9.
Using optimistic calculations of drift chamber performance (right-hand plot), and requiring
that 2 > 9 from dE/dx and kinematics alone, one gets an eciency of about 32%. On the
other hand, adding the information from the DIRC and ATC counters, the corresponding
eciency is about 86%. The left-hand plot shows the same information for an estimate of
resolution in momentum and dE/dx which is based on operating experience with CLEO. In
this case, the eciencies for 2 > 9 are about 7% and 75%, depending on whether the PID
Technical Design Report for the BABAR Detector
44
Detector Overview
1.0
All PID
Kin. + dE/dx
Efficiency
0.8
0.6
0.4
0.2
0
3–95
0
16
8
24 0
16
8
∆χ2
∆χ2
24
7894A3
The eciency for accepting a B 0 ! + , event as a function of 2 =
2
+
,
2
( ) , (K + , ), using pessimistic (left) and optimistic (right) estimates of drift
chamber performance. In calculating the 2 for the two hypotheses, kinematics and dE/dx
Figure 2-21.
are used for the dashed histograms, while for the solid histograms all information is used.
detectors are being used. The DIRC and ATC are even more important if the drift chamber
performance does not quite meet its design goals.
The other important analysis measuring the CP asymmetry parameter sin 2 , that of the
modes B 0 ! , is somewhat less demanding in hadron identi cation. The penguin decay
B 0 ! K + , is expected to have a small branching ratio, even compared to B 0 ! . The
branching ratio for B 0 ! K + , is not that small, but the separation, both by kinematics
and by identi cation of the charged particles, is somewhat easier, due to the lower average
particle momenta. The continuum background, which is relatively more important for this
mode, is also signi cantly suppressed by good K= separation.
2.7 Performance for Non-CP Physics
While operating at the (4S ), the BABAR experiment will be used for a broad program of B
physics, charm physics, tau physics, and two-photon physics. The high luminosity of PEP-II
ensures that, in all of these areas, it will be possible to do physics well beyond present
experiments. Although the success of the CP program is the primary goal in the design of
Technical Design Report for the BABAR Detector
2.7 Performance for Non-CP Physics
45
BABAR, the needs of the other parts of the physics program have been kept in mind. In most
cases, the performance of the detector can be estimated reliably by modest extrapolation
from the sensitivity of present CLEO-II analyses.
Fortunately, the main detector requirements for all of these physics goals are the same as
for the CP program. There are only a couple cases in which additional constraints were
imposed by these other physics areas. Most notable of these is in the trigger design, which
is more demanding for tau and two-photon physics than it is for bottom or charm physics.
Most importantly, none of the other physics is seriously compromised by the choices we have
made for B physics.
2.7.1
Other
B Physics
The BABAR detector is optimally designed for all B physics, not just the measurement of CP
asymmetries. The detector requirements for other B physics are included in the requirements
for the CP physics. The events have the same characteristics, and most of the interesting B
physics involves complete reconstruction of either one or both B decays.
An example is the decay B + ! 0 `+ , which is a prime candidate for measurement of
jVubj. The basic detector requirements are good lepton identi cation and good eciency
for detecting all particles in the event, which are also important for all CP studies. The
dominant backgrounds in the existing study of this mode are from cc events in which a
fast lepton from one charm particle is combined with two pions from the charm particle in
the opposite hemisphere. At PEP-II, these two charm vertices are separated by 600 m
transversely and 600 m longitudinally, so it should be rather straightforward to separate
them from the signal events in which all three tracks come from the same vertex. In addition,
one uses the missing momentum in the event to \reconstruct" the neutrino. Missing neutral
hadrons dominate the resolution for this reconstruction, so the ability to detect such hadrons
should improve the information and help to reduce backgrounds.
A more challenging task is to measure the decay B + ! + , which is the best way to
measure fB . The analysis requires completely reconstructing a large sample of B , decays,
and then looking to see if the remaining particles are consistent with B + ! + ; + !
`+ ` . An obvious background for this decay is a semileptonic decay in which the hadronic
particles are not detected. Again, a very high eciency for charged particles, photons,
and neutral hadrons is required to obtain the optimal suppression of backgrounds. The
largest gap in the detector acceptance is the 300 mr cone in the forward direction. Detailed
studies are needed to determine whether that loss of acceptance is an important source of
background, and if so, whether some form of veto detector could reduce the loss. (See the
discussion below concerning the similar decay Ds+ ! + .)
Technical Design Report for the BABAR Detector
46
Detector Overview
2.7.2 Detector Issues for Charm Physics
The detector requirements for doing charm physics are almost identical to those for B physics.
The clearest indication of this is the success that CLEO-II has had in charm physics. The
main di erences at PEP-II will be the increased luminosity and the boost along the beam
direction. As shown in the previous section, the acceptance of BABAR is very good, and the
performance is fairly uniform even when studied as a function of center-of-mass polar angle.
The BABAR vertex detector (and the one soon to be installed in CLEO) represents a major
improvement for charm physics. For example, the study of the important semileptonic decay
0
D+ ! K `+ is presently limited in CLEO because of combinatoric backgrounds. Because
of the long D+ lifetime, the D+ travels an average of 700 m at PEP-II, which will make
it possible to remove most combinations which include particles from the production vertex.
The vertex detector will be essential in separating doubly Cabibbo-suppressed decays from
mixing in the analysis of such decays as D0 ! K +,; K +,. The proper time distribution
for the mixed events is proportional to t2 e,t= , which makes it possible to t the time
distribution and extract the mixing component.
One of the charm physics topics sure to be still interesting at the time that BABAR starts taking data is the precise measurement of the Ds+ decay constant in the decay mode Ds+ ! + .
This decay is barely seen above background in present CLEO data. With greater integrated
luminosity and vertex information to help reduce backgrounds, a much better measurement
should be possible. One background that is dicult to remove comes from the semileptonic
decay D+ ! KL0 +, in which the KL0 is undetected. The IFR should be useful in reducing
such a background.
2.7.3 Detector Issues for Tau Physics
The luminosity of PEP-II will make it possible to reach unprecedented sensitivity for rare tau
decay modes, especially neutrinoless modes, in which kinematic constraints keep backgrounds
low. In addition, the enhanced capabilities of BABAR relative to CLEO-II (and ARGUS) will
make it possible to perform precision measurements of tau decay properties in common decay
modes. The design of the experiment, although optimized for B physics, is very well-suited
to the important tau physics. Most of the general design considerations for tau physics are
the same as those for B physics: high eciency for charged particles and photons of fairly low
momentum, lepton and hadron identi cation, good geometric acceptance, etc. It is primarily
in the trigger that the tau physics is more demanding, and the trigger is therefore designed
with that in mind.
Technical Design Report for the BABAR Detector
2.7 Performance for Non-CP Physics
47
In contrast to the case for bb and qq events, which typically contain many tracks, tau pairs
have low multiplicity, and thus an ecient trigger becomes a challenge. High eciency
is not the problem for these measurements, however; precise knowledge of the eciency
is. Because of the redundant, orthogonal, low-multiplicity triggers used in BABAR, these
eciencies should be high and precisely measured. It is therefore unlikely that the trigger
eciency will limit the systematic error on measurements of common tau decays.
The typical separation between the two tau decays is 400 m transverse to the z axis and
300 m along the z axis. The precise vertex resolution of BABAR will make it possible to
exploit this separation on an event-by-event basis to separate multiprong tau pairs from qq
backgrounds.
The tau neutrino mass measurement is an analysis which is particularly sensitive to backgrounds which fake multiprong tau decays. To suppress these backgrounds, one usually tags
the event with a lepton from the other tau. In that case, the major background is from
cc events, where, e.g., one D decays to KL0 `+ ` , and there are no fragmentation particles.
Because charmed particles are also long-lived, ight paths are not so useful to suppress this
background. The IFR permits the detection of KL0 s, however, making this background easier
to veto.
The excellent lepton and hadron identi cation of BABAR is also of great help for tau physics.
The IFR provides ecient = separation at momenta down to 0.5 GeV=c. Improved
discrimination between the and decay modes of the tau is of great help in many
important tau analyses. Ecient =K separation over almost all of the kinematic range is
important for studies of the K and KK systems.
In making precise tau branching ratio measurements, typically both tau decays are fully
reconstructed in each event. The error obtainable depends on the ability to model the
eciency of the detector for full reconstruction of many particles. This in turn relies on
the ability of BABAR to detect, with very high eciency, all tracks and photons within the
geometric acceptance, especially at low momentum. The emphasis placed on high eciency
for low energy photons and low momentum pions for B physics serves to optimize the detector
for this tau physics as well.
To suppress backgrounds from tau decay modes containing an extra 0 or KL0 , one typically
rejects events containing extra showers. Because of the soft energy spectrum for photons
from tau decays, it is important to detect photons cleanly with energies as low as 50 MeV.
Thus, it is important to remove sources of spurious low-energy showers, including extra
showers from hadronic interactions in the calorimeter, which are the hardest to model. The
IFR, operating as a \tail-catcher," may be useful to help in identifying such showers. In
addition, the transverse segmentation of the calorimeter crystals can help a great deal in
distinguishing such showers from true photons.
Technical Design Report for the BABAR Detector
48
2.7.4
Detector Overview
Detector Issues for Two-Photon Physics
The high luminosity of PEP-II will extend the study of exclusive two-photon physics (i.e.,
the reaction e+e, ! e+e, ! e+e,X ) from the present 2 GeV up to at least 5 GeV in
mass. In addition to the gains caused by the much higher luminosity, the BABAR detector will
improve on the existing, very successful CLEO-II design in several ways, including superior
calorimetry at low energies, better particle identi cation at high momenta, precise vertex
measurement, and a very exible data acquisition and triggering system. This will greatly
improve both resonance searches and detailed QCD tests. Most of this two-photon physics
can be done with a detector optimized for the CP -violation physics goals. However, some
of the interesting processes have signi cant impacts on the trigger and on the backward
endcap coverage.
Several two-photon reactions result in nal states with only two charged hadrons, sometimes
with rather small transverse momenta. The challenge for the trigger is to accept these
events without introducing a large rate from beam backgrounds or cosmic rays. Redundant
triggers using either two calorimeter clusters or two drift chamber tracks (or combinations
of these) with variable thresholds should suce for this purpose. In addition, there are
interesting all-neutral nal states (e.g., 00 ) that can also be accepted using the calorimeter
trigger. High eciency can be maintained for nal state masses greater than 2 GeV=c2 , while
modest eciency at lower mass preserves the ability to overlap with present two-photon
measurements.
There are two single-tagged reactions which are quite sensitive to the minimum backward
detection angle. The rst of these, the production of spin-1 resonances, gives a prime
tool for identifying hybrid mesons (1,+) and studying charmonium (1++) states, since
one can uniquely distinguish between spin-1 and even-spin production at low Q2 [where
Q2 = 4Ebeam Ee sin2 2 , and Ee () is the energy (polar angle) of the detected positron]. The
angular distribution of the scattered leptons is strongly peaked at small polar angles, and
the beam asymmetry folds some of the scattered positrons out to detectable regions in
the backward endcap. The second of these reactions, the study of single-tagged exclusive
hadron production at high Q2, allows testing of some very basic QCD predictions in a
regime (Q2 > 3 GeV2 ) which has never been studied before. In both of these cases, the
nal-state hadrons are concentrated primarily in the barrel and forward regions so the only
issue is detection of the scattered positrons with energies in the range 1 < E < 3 GeV.
This is a primary motivation for the backward endcap calorimeter. Although for nancial
considerations it is not included in the present design, we hope to be able to add it later to
upgrade the capability of the detector for two-photon physics.
Technical Design Report for the BABAR Detector
REFERENCES
49
References
[CLE93] CLEO Collaboration, \Detector for a B Factory," Cornell Report CLNS{91{
1047{REV (1993).
[HEL92] \HELENA, A Beauty Factory at Hamburg," DESY 92/41 (1992).
[KEK92] B Physics Task Force, \Progress Report on Physics and Detector of KEK
Asymmetric B Factory," KEK Report KEK{92{3 (1992).
[SLA91] \Workshop on Physics and Detector Issues for a High-Luminosity Asymmetric B
Factory at SLAC," SLAC{373 (1991).
[SLA93] \Status Report on the Design of a Detector for the Study of CP Violation at
PEP-II at SLAC," SLAC{419 (1993).
[SLA94] \Letter of Intent for the Study of CP Violation and Heavy Flavor Physics at
PEP-II," SLAC{443 (1994).
Technical Design Report for the BABAR Detector
50
Technical Design Report for the BABAR Detector
REFERENCES
3
Physics with BABAR
3.1
Physics Context
T
he study of CP violation in B decays is the primary physics objective of BABAR [SLA89].
The aim is to achieve a sucient number of independent determinations of the parameters of the Cabibbo-Kobayashi-Maskawa (CKM) matrix to overdetermine those quantities.
This then provides a test of the Standard Model interpretation of CP violation and, perhaps,
a window into physics beyond the Standard Model [Nir93]. This chapter summarizes the
essential physics points of this endeavor and brie y discusses other physics that can be done
with BABAR. The measurement of CP violation asymmetries puts stringent requirements on
the detector, so optimizing it for this purpose means that it is also well suited to study most
of the other physics of interest in this energy regime.
The two mass eigenstates of the neutral B meson system can be written as:
jBLi = p jB 0i + q jB 0i;
jBH i = p jB 0i , q jB 0 i;
(3.1)
where H and L stand for Heavy and Light, respectively. De ning M (MH +ML)=2; M
MH , ML , and neglecting the tiny di erence in width between BH and BL, the decay widths
satisfy ,H = ,L ,. In this approximation, mixing in the Bd0 system is governed by a single
phase:
!
q = VtbVtd = e2iM :
(3.2)
p
V V
Bd0
tb td
The amplitudes for decays into a CP eigenstate f are
Let us de ne
A hf jHjB 0i; A hf jHjB0 i:
(3.3)
rCP (f ) pq AA :
(3.4)
52
Physics with BABAR
The time-dependent rates for initially pure B 0 or B 0 states to decay into a nal CP eigenstate
at time t can then be written as
2
0 (t) ! f ) = jAj2 e,,t 1 + jrCP (f )j
(3.5)
,(Bphys
2
!
2
1
,
j
r
CP (f )j
+
cos(Mt) , Im rCP (f ) sin(Mt) ;
2
(f )j2
,(B 0phys (t) ! f ) = jAj2e,,t 1 + jrCP
2
!
2
1
,
j
r
CP (f )j
cos(Mt) + Im rCP (f ) sin(Mt) :
,
2
(3.6)
The time-dependent CP asymmetry
0
0
af (t) = ,(B 0(t) ! f ) , ,(B 0 (t) ! f )
,(B (t) ! f ) + ,(B (t) ! f )
(3.7)
is given by
2
Mt) , 2 Im rCP (f ) sin(Mt) :
af (t) = (1 , jrCP (f )j ) cos(
(3.8)
1 + jrCP (f )j2
This result holds for a CP -even nal state, while for CP -odd states there is an additional
minus sign in rCP (f ). When only a single amplitude with a given weak decay phase D
dominates the decay, one has
A = e,2iD :
A
(3.9)
af (t) = , sin 2(M , D ) sin(Mt):
(3.10)
Since Im rCP (f ) = sin 2(M , D ), Equation 3.8 simpli es to
In an e+e, B Factory, the initial B 0 and B 0 are produced in a coherent B 0 B 0 state and
remain in this state until one of the particles decays. If one B decays to a avor-tagging
mode and the other decays to a CP -study mode, the event can be used to reconstruct the
time dependence of the asymmetry. In this case, the time t in the equations above is the
time between the tagging decay and the CP -study-mode decay. The tagging decay may be
the later decay, in which case t < 0. Thus, the time-integrated CP asymmetry vanishes if
jrCP (f )j = 1, which makes essential a measurement of the time dependence. The asymmetric
machine con guration and an accurate vertex determination make this possible.
The Standard Model predictions for CP violation are most often presented in terms of the
unitarity constraints for the three-generation CKM matrix. If one assumes the Standard
Technical Design Report for the BABAR Detector
3.1 Physics Context
53
A=(ρ,η)
α
γ
β
C=(0,0)
B=(1,0)
9-89
6466A16
Figure 3-1.
The Unitarity Triangle.
Model, one can derive relationships between the mixing phases and the weak phases of
various contributions to the decay amplitudes and the angles of a triangle, generally called
the Unitarity Triangle. Unitarity of the CKM matrix requires, among others, the relationship
Vtb Vtd + Vcb Vcd + Vub Vud = 0:
(3.11)
The three complex quantities Vid Vib can be represented as a triangle in the complex plane.
The three angles of this triangle are labeled
!
Vtd Vtb
arg , V V ;
(3.12)
ud ub
!
Vcd Vcb
arg , V V ;
(3.13)
td tb
!
Vud Vub
(3.14)
arg , :
Vcd Vcb
Figure 3-1 shows the unitarity triangle, as it is usually drawn, rescaled by the side VcdVcb .
This makes the base of the triangle real and of unit length. The apex of the triangle is then
the point (; ) in the complex plane, in the notation introduced by Wolfenstein [Wol85] for
the parameters of the CKM matrix.
To overdetermine the unitarity triangle, experiments must x as many of the parameters as
possible that give the sides, jVtbVtd j=jVcbVcd j and jVubVud j=jVcbVcd j, and the angles, , , and
. De ning = Vus = sinCabibbo , the quantities Vtd = , Vud = cos Cabibbo , and Vtb = 1 are
already well-determined up to terms of O(4). The value of Vcb will be well-determined by the
study of semileptonic B decays at CLEO II, with possible further improvement from BABAR.
Thus, the primary aim of the B Factory must be to measure accurately the magnitudes of
Vub , Vtd , and the CP -violating asymmetries in the decay modes Bd ! J= KS0 , Bd ! J= KL0 ,
Bd ! J= K , Bd ! D+D, , Bd ! D+ D, , and Bd ! + , , Bd ! . Determination
Technical Design Report for the BABAR Detector
54
Physics with BABAR
Quark Process Bd Mode M , D
b ! ccs
b ! ccd
b ! uud; ddd
Table 3-1.
J= KS0
J= KL0
J= K 0
D+ D,
D+D,
DD
+,
a1
CP modes and Standard Model asymmetry predictions for Bd decays.
of CKM elements will be discussed below. First, we concentrate on the time-dependent CP
asymmetries.
Table 3-1 shows the Standard Model relationship between the angles of the unitarity triangle
and the predicted arg(rCP (f )) = 2(M , D ), assuming a single weak decay amplitude
dominates in each of these cases. In addition to pure CP eigenstates, other channels with CP self-conjugate quark content can be analyzed for CP -violating contributions. These include
mixed spin states such as J= K and vector-scalar mixtures such as and DD where the
various possible charge assignments have both CP and isospin relationships [Ale91, Ale93].
For the b ! ccs decays, which give channels that measure the angle , we are in a fortunate
situation. Up to corrections of order A4, the penguin diagrams have the same weak phase
as the tree diagrams. Thus, these channels have an unambiguous relationship between
the measured asymmetry and the angle . For channels that measure the angle , the
situation is not quite as good, although still satisfactory. In this case, QCD penguin diagrams
are expected to give contributions on the order of perhaps 5{20% of the tree amplitude,
and they have a di erent weak phase. In addition, the Z 0 -mediated penguin diagram also
contributes at the level of perhaps 5% of the tree amplitude, with the same weak phase
as the QCD penguins [Des95]. Isospin-based analysis can select an isospin channel that
has no QCD-penguin contribution [Gro90]. The uncertainty in sin 2 extracted from the
asymmetry in this channel due to the Z 0 -penguin contributions is then not more than 0.05%.
This is less than the expected experimental uncertainty. Furthermore, an isospin-based
analysis will give a measure of the total penguin contributions, and hence, can be used to
tighten the bounds on the possible Z 0-penguin e ects. In addition, measurement of the
time dependence of the asymmetry will allow us to t the cos(Mt) behavior as well as
the sin(Mt) in Equation 3.8. This then gives a further constraint on the magnitude of
penguin contributions.
Technical Design Report for the BABAR Detector
3.2 Simulation Tools
55
Isospin analyses can probably be carried out most e ectively for the channels, since all
three channels decay to +,0 and so have similar detection eciencies. This channel
has the additional advantage that the interference regions between the di erent charge
combinations of give further information, allowing a determination of sin 2 and cos 2
which resolves the ambiguity between and 90 , [Qui93]. For the case, the isospin
analysis requires a measurement of the rates for B 0 ! 0 0 and B + ! +0 .
The simulation estimates for the accuracy of measurements given below, assume that
penguin contributions are suciently small that the measured asymmetries can be used
directly to determine without isospin analysis. The isospin analysis can be used to check
this assumption and to obtain a penguin-corrected value. The degradation in accuracy due
to this multichannel analysis is not expected to be large.
3.2 Simulation Tools
Detailed simulation work has been performed to determine the eciency of the proposed
detector to make the necessary measurements of the channels of interest and to ascertain
that backgrounds can be simultaneously reduced to an acceptable level. Studies of tagging
modes have also been carried out. Two di erent Monte Carlo simulation packages, ASLUND
and BBSIM, were used for this purpose. ASLUND is a fast parametric Monte Carlo, while
BBSIM is a detailed simulation based on the GEANT package [CER93]. Both are designed
to be exible, so that variations in detector design can be investigated. Brief descriptions of
each are given below. Where not otherwise noted, branching ratio information is taken from
the 1994 PDG compilation [PDG94].
3.2.1 ASLUND
ASLUND consists of two major components: the JETSET 7.3 [Sjo93] event generator
con gured for the PEP-II environment and a parametric simulation of the detector. Other
generators (e.g., KORALB [Jad90]) are available using the interface routine BEGET [Wri94].
The parametric simulation includes charged tracking, particle identi cation, and calorimetry.
Charged particle tracking is simulated using the TRACKERR package [Inn93] to estimate
the fully correlated error matrix, which is then used to smear the track parameters. The main
drift chamber, the silicon vertex detector, and any intervening material are modeled. The
TRACKERR input format is used to specify detector geometry, measurement precision, and
material distribution. The combination of ASLUND and TRACKERR has been compared
to BBSIM for the mode B 0 ! +, and found to agree within 15% for the predicted B 0
mass width.
Technical Design Report for the BABAR Detector
56
Physics with BABAR
Subdetector
Beam Line, Beam Pipe, and Support Tube
Silicon Vertex Detector
Drift Chamber
DIRC
ATC
CsI Calorimeter
Coil and Instrumented Flux Return
Table 3-2.
Simulation Level
Geometry
Digitizations
Hits
Digitizations
Geometry
Hits
Digitizations
BABAR subdetectors modeled in BBSIM.
Simulation of particle identi cation (=K=p) information is available for dE=dx, DIRC, and
aerogel threshold counters (ATCs). The geometrical con guration of these devices can also
be speci ed using the TRACKERR format, and performance parameters can be varied to
assess the sensitivity to detector capabilities.
The geometry of the calorimeter, and any material in front of it can also be speci ed in
the TRACKERR format. Photons converted into e+e, pairs in the material in front of
the calorimeter are tracked to determine whether or not they will deposit any energy in the
calorimeter. Energy loss in material and acceptance losses due to the curl-up in the magnetic
eld are taken into account in determining the calorimeter signal. A detailed parameterization of the calorimeter response based on GEANT simulations has been developed for some
analyses, e.g., and DD. Analogous to the other devices discussed above, energy and
angular resolutions can be varied by modifying simple parameters.
3.2.2 GEANT Simulation|BBSIM
Simulation packages for event generation (BEGET), detector response, and subsequent
analysis (BBSIM) have been written to aid in the design and optimization of the BABAR
detector.
The generation of an event from an e+e, collision and the subsequent loading of decay
products into GEANT banks is handled by the BEGET [Wri94] package. The generators
currently available include (4S ) decays and hadronic continuum events via JETSET [Sjo93]
and + , decays via KORALB [Jad90]. BEGET is capable of overlaying background hits
originating from beam-gas collisions on simulated events.
The BBSIM package is based on the CERN detector description and simulation tool, GEANT
[CER93]. The latter consists of packages to construct the detector geometry using a set
of 15 di erent volume types; to step charged and neutral particles through the detector,
Technical Design Report for the BABAR Detector
3.3 Studies of
B0
! J=
Modes
57
Cut
Geometry and Lepton ID Ineciency
J= Invariant Mass From 0.14 to 0.44 MeV/c
Combined J= Eciency
Geometry, Eciency and Track/Photon Cuts
0 Association Cuts
0 Mass Cuts
KS0 Momentum 0.9 p 2.9 MeV/c
KS0 Invariant Mass
Combined KS0 Eciency
B 0 Momentum From 0.14 to 0.44 MeV/c
B 0 Candidate Mass
Overall B 0 Reconstruction Eciency
Table 3-3.
+, 00
0.82
1.00
0.82
0.78
0.94
0.73
0.99
1.00
0.59
0.82
1.00
0.82
0.55
0.89
0.93
1.00
1.00
0.45
0.99
0.96
0.35
Summary of eciencies for B 0 reconstruction in B 0 ! J= KS0 .
simulating the full variety of interactions with the detector; to de ne, register, and digitize
the Monte Carlo track hits (typically track positions and directions); and to display the
detector components, particle trajectories, and track hits. The BBSIM framework consists
of a driver routine to invoke GEANT; a database facility, DBIN, to de ne detector geometry,
materials and media; and a set of subpackages, one per detector subsystem, to de ne the
subdetector geometry, register track hits, simulate the detector response, store the results,
and perform subsequent analysis. The subsystems included in BBSIM, and the corresponding
levels of simulation, are shown in Table 3-2.
3.3
Studies of
3.3.1
B 0 ! J=
Modes
B 0 ! J= KS0
The analysis of this channel and a study of the background were performed using the
ASLUND Monte Carlo. This work is described in more detail in Reference [Har95]. The
results are based on an assumed branching ratio of 0:5 10,3, calculated using isospin from
the measured B + ! J= K + rate [PDG94].
The eciencies obtained are summarized in Table 3-3. The analysis proceeded as follows.
Candidate J= ! `+`, events were selected. Lepton identi cation eciencies (calculated using BBSIM) and a J= mass cut were applied, and KS0s were reconstructed
Technical Design Report for the BABAR Detector
58
Physics with BABAR
in both the charged and neutral decay modes. In the charged mode, the invariant
mass was required to be within 25 MeV=c2 of the nominal KS0 mass. To help suppress
the combinatorial background in the neutral mode, various cuts were made on the
consistency of the four photons having a common point of origin while also being
consistent with the two 0 sub-masses, and the KS0 mass and lifetime. The invariant
mass was required to be within 40 MeV=c2 of the nominal KS0 mass. Note that
reconstruction of the neutral KS0 decays requires the calorimeter to reconstruct the
0 s.
Candidate J= and KS0 events were combined to form B 0 candidates. A cut was applied
on the B 0 momentum in the (4S ) rest frame (kinematically, this is 340 MeV=c). The
invariant mass of the B 0 candidate was required to be between 5.20 and 5.36 GeV=c2 .
Potential backgrounds to this decay mode include those from B 0 ! J= X , cascade semileptonic decays (b ! cl, ! sl+l, ), and continuum production of quark-antiquark pairs.
Large numbers of each were generated and passed through the analysis described above. The
fractions of events surviving all the cuts are summarized in Table 3-4. Also shown are the
relative production rates compared to signal events, given by
BG B (BG)
normalization =
:
bb B (B 0 ! channel)
The predicted upper limit on the number of background events per reconstructed signal B 0
is given for each source. No background events survived the cuts in these simulations. All
backgrounds together contribute no more than 6%. These levels are considered negligible
for the extraction of sin 2 from the decay asymmetry.
3.3.2
B 0 ! J= KL0
The branching ratio for this channel is the same as that for J= KS0 , but the expected
asymmetry has the opposite sign. Thus, high-eciency detection of KL0s can provide an
alternate measurement of and improve the overall accuracy with which this parameter is
determined.
A study of the mode B 0 ! J= KL0 has been carried out [Wri94b]. Since the KL0 momentum
is not well-measured, the KL0 direction and the J= momentum are combined, and a B -mass
constraint is applied to determine prest, the momentum of the B in the (4S ) rest frame.
The resulting distribution for prest is shown in Figure 3-2, in which only events from the
(4S ) are shown; continuum backgrounds in this channel are negligible, as with other J=
channels. A cut prest 0:42 GeV=c is made to eliminate false combinations.
Technical Design Report for the BABAR Detector
3.3 Studies of
B0
! J=
Modes
Channel
Charged Mode search:
59
Eciency Normalization # per observed
signal event
J= KS0 ! l+ l, + ,
J= X
0.59
7 10,4
8 10,7
1:3 10,7
1.00
3:9 101
7:8 102
8:1 104
1.00
4:6 10,2
1 10,3
1:8 10,2
J= KS0 ! l+ l, 0 0
J= X
0.35
5 10,5
1 10,7
1 10,7
1.00
7:8 101
1:6 103
1:6 105
1.00
1:1 10,2
4:7 10,4
4:7 10,2
Semileptonic
qq continuum
Neutral Mode search:
Semileptonic
qq continuum
Table 3-4.
Background contributions for the J= KS0 channel.
425
_
B0→any, B0→J/ψ+any
400
375
_0
0
0
B →any, B →J/ψKL
350
325
_0
0
*
B →any, B →J/ψK
300
275
250
225
200
175
150
125
100
75
50
25
0
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
p (GeV/c)
Momentum distribution of B 0 (prest ) from (4S ) ! B B in the (4S ) rest
frame, including the e ect of the nominal e+ e, beam smearing and KL0 angular resolution.
Figure 3-2.
Technical Design Report for the BABAR Detector
60
Physics with BABAR
Sample
Reconstruction
#
Background
Other
Eciency
Reconstructed From K
Background
Low K
0.25
544
31
74
High K
0.08
168
65
117
Combined
0.33
712
96
191
Table 3-5.
Eciencies for B 0 reconstruction and sensitivity to for B 0 ! J=
KL0 .
One sees that a substantial background which survives this cut comes from B 0 ! J= K
with K ! KL0 0. Since the asymmetry for J= K is opposite in sign to that for J= KL0 ,
these events contribute to a dilution of the measured asymmetry in the same fashion as do
wrong-sign tags. In order to obtain an accurate estimate of the error on the asymmetry,
the data is divided into two samples, one with a low K background and the other with
a higher background. The two samples are then treated independently in determining
sensitivity. The high K background sample is de ned as follows: any pion in the event is
combined with the KL0 to de ne K candidates. The KL0 momentum is recalculated under
this assumption along with a B mass constraint. Any event in which a candidate K has
a mass within 30 MeV=c2 of the true K mass is assigned to the high background sample.
Events with no K candidates are assigned to the low background sample. Table 3-5 shows
the results for the separate samples and the combined sensitivity for this channel.
3.3.3
B 0 ! J= K 0
Only the neutral decay mode of the K 0 is useful for CP -violation measurements. The
analysis of this channel and a study of the background were performed using the ASLUND
Monte Carlo and the published B 0 ! J= K 0 branching ratio of 1:6 10,3 [PDG94]. The
analysis proceeded in a manner similar to that for B ! J= KS0 described above. First, J=
and KS0 candidates were formed as above, where only the charged decay mode of the KS0 was
used in this case. Candidate 0s were obtained from pairs of calorimeter clusters, assuming
that the photons originated at the nominal interaction position of the experiment. All
combinations of 0 and KS0 candidates were made, and a mass cut applied to the resulting K 0
candidates. All mass cuts and the various subeciencies of each stage of the reconstruction
are summarized in Table 3-6.
Technical Design Report for the BABAR Detector
3.4 Studies of B 0 to Double Charm Modes
61
Cut
Eciency
Geometry and Lepton ID Ineciency
0.82
2
3.06 < J= Mass < 3.14 GeV/c
1.00
Combined J= Eciency
0.82
2
Geometry and E min = 25 MeV/c
0.73
0
2
0.120 < Mass < 0.150 GeV/c
0.95
Combined 0 Eciency
0.69
Geometry and Track Cuts
0.77
0
2
0.468 < KS Mass < 0.528 GeV/c
0.92
Combined KS0 Eciency
0.71
0
Combined K Sub-eciencies
0.50
0
2
0.73 < K Mass < 1.05 GeV/c
0.98
Overall K 0 Eciency
0.50
0
Combined B Sub-eciencies
0.40
0
B Momentum in Rest Frame
0.99
B 0 Candidate Mass
0.99
0
Overall B Reconstruction Eciency
0.39
Table 3-6.
Cuts and Eciencies for B 0 Reconstruction in B 0 ! J=
K 0 .
3.4 Studies of B 0 to Double Charm Modes
3.4.1
B 0 ! D+ D,
A branching fraction of 6 10,4 is used for this channel and is obtained by averaging the
measured branching fractions for B + and B 0 to decay to DDS and multiplying by sin2 C .
Only those D+ modes with no more than one 0 and a branching fraction greater than about
1% are considered. Only those KS0 s in the +, decay mode are included. The channels
used are shown in Table 3-7 with branching ratios taken from Reference [PDG94]. The mass
resolutions quoted are vertex constrained. The eciency for each mode in Table 3-7 includes
cuts on particle identi cation for each track and a cut on 2 probability for the reconstructed
D vertices at 2%. Mass cuts of 3mD are made on each D meson candidate.
Pairs of D meson candidates are combined to form B candidates. Each B candidate is
required to have a total momentum between 180 and 440 MeV=c in the (4S ) rest frame. A
vertex and kinematic tting package is then used to t the entire B event, constraining the
intermediate D masses and vertices and the two Ds to a single vertex. The 2 probability
Technical Design Report for the BABAR Detector
62
Physics with BABAR
Branching Single D
Mode
Fraction Eciency mD (MeV)
,
+
+
K
0.091
0.67
4.3
,
+
+
0
K
0.064
0.42
6.9
KS0 +
0.014
0.59
5.0
0
+
0
KS
0.049
0.38
9.1
0
,
+
+
KS
0.035
0.41
3.7
KS0 , + + 0
0.027
0.26
5.5
Table 3-7.
B ! D+ D, !
D+ ! K , + +
D, ! K + , ,
D+ ! K , + +
D, ! K + , , 0
D+ ! KS0 +
D, ! K + , ,
Eciency and mass resolutions for D+ reconstruction.
Branching
Eciency mB zB # of Events
Fraction (10,3)
(MeV) (m) in 30 fb,1 S/B
8.3
0.45
5.0 47.0
80.0
1.8
11.6
0.24
6.5
54.0
61.0
0.7
2.5
0.38
5.3
58.0
14.0
> 3:0
Eciency, mass and vertex resolution, signal, and background for three of the
D decay mode combinations used in the B ! D+ D, reconstruction.
Table 3-8.
for this t is required to be greater than 2%. The reconstructed mass is required to lie within
20 MeV=c2 of the B 0 mass. For events passing the above cuts, a B mass constraint is added
to determine the B decay point. There are many combinations of the D decay modes listed
in Table 3-7, each with di erent eciencies and resolutions. Table 3-8 gives the eciencies
and resolutions for three of the 21 combinations. The complete set of numbers can be found
in Reference [Cou95].
Potential sources of background are from other decays of the (4S ) and continuum quarkantiquark production. Table 3-8 includes an estimate of signal-to-background ratios (S=B )
for the modes discussed. Channels with two or more 0s in the nal state tended to have large
backgrounds and were, therefore, not included in this analysis. Although each combination
of D decay modes is analyzed separately, one can sum over all modes to obtain a total
reconstructable branching fraction of 0.044, an average acceptance of 0.27, and a nal event
sample of 259 events with a background of 351 events.
Technical Design Report for the BABAR Detector
3.4 Studies of B 0 to Double Charm Modes
63
Branching Single D0 mD0
Mode
Fraction Eciency (MeV)
,
+
K
0.040
0.77
5.2
,
+
0
K
0.138
0.48
8.8
0
+
,
KS
0.026
0.48
4.0
KS0 + , 0
0.049
0.29
6.8
,
+
+
,
K
0.081
0.55
3.7
K ,++,0
0.043
0.34
5.5
Table 3-9.
3.4.2
Eciency for reconstruction of D0s produced in B ! D+D, decays.
B 0 ! D+D,
A branching fraction of 1:5 10,3 is used for this mode which is 2.5 times larger than that
for D+D,. This factor is the typical increase for vector-vector decay modes over comparable
scalar-scalar channels.
The only decay chain considered is that in which each D decays to D0 with a branching
fraction of 68 1:6%. The eciency for nding the slow pions has been studied in more
detailed simulations of the vertex detector and found to be better than 95% over the
momentum range of interest above pt = 60 MeV=c for > 300 mr. The branching fraction,
eciency, and mass resolution for the D0 modes are given in Table 3-9.
In reconstructing D mesons, all D0 candidates are combined with charged pions and a cut
made at 3 on the D{D0 mass di erence. Pairs of D candidates are then t for the B
mass, constraining the vertex and mass of the intermediate D and D0. Table 3-10 gives the
eciency, resolutions, and background levels for three of the 21 combinations. The complete
set of results can be found in Reference [Cou95]. Although each combination of D0 decay
modes is analyzed separately, one can sum over the modes to obtain a total reconstructable
branching fraction of 0.125, an average acceptance of 0.15, and a nal event sample of 473
events with a background of 21 events.
In computing CP reach, D+D, channels are assumed to be dominated by a single CP
state. If this turns out not to be the case, a partial-wave analysis will need to be performed
to disentangle the CP = + and CP = , states. This will decrease the e ectiveness of these
channels by a factor that depends on the fraction of each CP state present.
Technical Design Report for the BABAR Detector
64
Physics with BABAR
B ! D+ D, !
D0 ! K , +
D0 ! K + ,
D0 ! K , +
D0 ! K + , 0
D0 ! KS0 + ,
D0 ! K + ,
Branching
mB
zB # of Events
,3
Fraction (10 ) Eciency (MeV) (m) in 30 fb,1 S/B
1.6
0.41
5.6 53.0.
16.0
> 2:0
11.0
0.21
5.8
65.0
58.0
> 7:0
2.1
0.29
4.3
55.0
10.0
> 1:2
Table 3-10. Eciency, mass and vertex resolution, signal, and background for three of
the D decay mode combinations used in the B ! D+ D, reconstruction.
3.4.3
B ! DD
While this mode is not a CP eigenstate, CP -violating asymmetries can be measured using
both D+D, and D,D+ data [Ale91]. A full simulation of these modes has not been carried
out, but eciencies and backgrounds can be estimated using the two previous studies. The
decays B ! D,Ds+ and B ! D,Ds+ have been observed with branching ratios of 1:20:6%
and 2:1 1:5%, respectively [PDG94]. Rescaling by sin2 C gives reasonable estimates for
the DD rates. Using the acceptances determined in the two previous sections we predict a
total reconstructable branching fraction of 0.085, an average acceptance of 0.168, and a nal
event sample of 557 events.
3.5
Studies of
B
!
Modes
The measurement of sin 2 from a CP -violating asymmetry in the B 0 ! +, mode requires
the selection of rather clean samples of +, as well as 0 and 00 events [Ale92]. The
latter are needed for an isospin analysis to estimate the size of the penguin contribution to the
total decay amplitude and to extract sin 2 from the CP asymmetry for the case in which the
QCD-penguin amplitudes are large [Gro90]. Unfortunately, the branching ratios for charmless two-body B decays into either pions or kaons are expected to be very small, on the order
of 10,5 or less [Bro93]. Therefore, these channels su er from a potentially large background
due to continuum quark-antiquark pair production, despite a clean signature. Simple cuts
strongly reduce the continuum background and allow one to enrich the sample in two-body B 0
decays. In addition, the branching ratio of B ! K modes may be comparable to B ! .
In order to disentangle the +, component of the signal from K events, a very good
particle identi cation system operating up to high momentum (4{4.5 GeV=c) is needed. This
Technical Design Report for the BABAR Detector
3.5 Studies of B ! Modes
65
is one of the main features of the BABAR detector. Studies for all of these channels have been
performed using both ASLUND and BBSIM. Because of the large sample of Monte-Carlo
events needed, the continuum background was studied using ASLUND only. The branching
ratio for B ! +, is taken to be 1:2 10,5, which is based on the measurement from
CLEO [Bat93a].
3.5.1
B0
! +,
Despite apparent simplicity, analysis of this mode, dealing with rejection factors at the 10,5
to 10,6 level, requires a precise knowledge of detector systematics. Strong track quality cuts
will be needed; these are not included in the present study. Software tools (full detector
simulation and pattern recognition packages) are being developed and will be used to re ne
the selection cuts and to give more accurate estimates of signal eciencies and background
rejection.
Reduction of Continuum Background
The reduction of the continuum background is achieved by cuts applied in the (4S ) centerof-mass frame. The rst two cuts exploit the fact that the B mesons are produced almost at
rest, and thus the pions are almost monoenergetic, in the (4S ) rest frame. The third uses
the fact that continuum background events are mostly jet-like, whereas B 0B 0 decays are
not. A cut on particle identi cation (2 ) gives a signi cant background reduction. This
occurs because the sample of background tracks hard enough to masquerade as a B !
event is enriched in kaons and protons.
Charged particles of momentum p? in the (4S ) center of mass are selected in a very narrow
range: 2:35 p?() 2:95 GeV=c. This momentum window contains virtually 100% of
reconstructed pions (or kaons) from two-body decays. The probability that a continuum
charged particle lies in this momentum window is less than 2%. The probability that a
continuum event contains two such particles with opposite electric charge is 3:3 10,3.
Any selected pair of opposite sign particles forms a B 0 candidate. The B 0 candidate
momentum p?(B ) in the (4S ) center of mass frame is required to lie in the 150{500 MeV=c
range. This cut selects pairs with large opening angles. To be conservative, we have used a
rather large window for the B 0 mass cut. The signal-to-background ratio (S/N) is a strong
function of the angle sph between the B-decay axis and the sphericity axis formed by the
remaining tracks in the event. The background is strongly peaked at cossph = 1 while
the distribution of the signal in this variable is at. We divide the data into 3 bins (< 0:7,
0.7{0.9, and 0.9{1.0) in jcossphj in order to optimize the CP -sensitivity in this channel. The
Technical Design Report for the BABAR Detector
66
Physics with BABAR
Cut
+ , Acceptance and p? ( ) in [2:35; 2:95]
0 : Preselected Events (Normalization)
1 : p?(B ) in [0:150; 0:500]
2 : 2 0
1 + 2 + jcossphj < 0:7
1 + 2 + 0:7 < jcossphj < 0:9
1 + 2 + 0:9 < jcossphj < 1:0
Overall Eciency per Event Generated
Table 3-11.
B0 ! +,
0.862
1.000
0.994
0.972
0.538
0.166
0.094
0.798
qq
3.3 10,3
1.000
0.246
0.275
5:0 10,6
6:0 10,6
3:9 10,5
5.0 10,6
Eciencies of background rejection cuts for the B 0 ! + , mode.
particle identi cation requirement on 2 is explained in the following section. The results
are summarized in Table 3-11.
Particle Identi cation
The rejection of K events from the sample is a crucial step in measuring the correct
asymmetry. The criterion used combines information from kinematics (mass resolution),
dE=dx, and the DIRC particle identi cation system. The quality of all three pieces of
information is essential to obtaining good background rejection without signi cant loss of
signal.
Combining all kinematic and PID information leads to a 2 test with up to ve degrees
of freedom. In reality, within the acceptance of BABAR, one particle lies outside of the
DIRC acceptance 10% of the time. Thus, the number of measurable degrees of freedom
di ers from event to event. We de ne a discriminating variable i 2 as the average 2
di erence between the null hypothesis () and any one of the alternate hypotheses. The
ve possible terms are a kinematic term for the pair, 2 kin, and two possible particle
identi cation terms for each member of the pair, 2j dE=dx and 2j DIRC . We retain only
the hypothesis giving 2 0. Simulation samples of B 0 ! and B 0 ! K are used
to determine , the probability of rejecting a true event, and , the probability of
accepting an event with a K , as functions of the minimum 2 needed to reject the alternate
hypotheses. The results obtained with ASLUND are listed in Table 3-12. One can see from
this table that for a signi cance = 5%, obtained for 2crit ' 4, can be maintained
below 1%. The simulation's results for the dE =dx and momentum resolutions are based on
ideal performance of the drift chamber and are probably optimistic. They do not include
the single cell eciencies, track overlaps, background rate e ects, polar angle dependence of
the dE/dx response, or a myriad of other factors which need to be addressed by an operating
Technical Design Report for the BABAR Detector
3.5 Studies of
B ! Modes
67
min 2
0
Kinematics
dE=dx
Kin. + dE=dx
DIRC
Kin. + dE=dx + DIRC
86
75
92
74
98
4
1,
51
36
80
69
95
9
0
4
9
12
6
53
58
87
9.7
12
4.7
1.8
2.4
1.8
1.5
1.1
|
0.8
|
|
|
|
|
Probabilities for selecting a + , event (1 , ) and a K +, event ( ), for
several minimum values 2 , based on an ASLUND simulation.
Table 3-12.
experiment. We note that neither CLEO nor ARGUS have been able to achieve reliable K=
separation for high momentum tracks with this method.
Results
The +, events are separated into three classes for analysis of CP sensitivity [Sny95].
Class 1 contains events with cos sph < 0:7, class 2 contains those with 0:7 cos sph 0:9,
and class 3 is all other events. Using the results listed in Table 3-11 and assuming a branching
ratio of 1:2 10,5 [Bat93a], one can reach a background to signal ratio (B=S ) of around
2 with a large eciency = 54%. While the backgrounds in classes 2 and 3 are higher,
inclusion of these events still improves the overall sensitivity.
The e ect of the di erence between the signal and the background in z the axial separation
of the two B decays, further reduces the impact of continuum backgrounds. This is discussed
below.
3.5.2
B0
!
0 0
The reconstruction of 0 s from calorimeter data is discussed in Chapter 7. The resolution
obtained on the 0 mass using BBSIM is 8.3 MeV=c2 . This leads to an eciency of 84% for
0 reconstruction, restricting 0 candidates to the [100 , 170 MeV=c2 ] mass window. The
same mass window is chosen for background studies with ASLUND.
The results obtained with ASLUND for this mode are summarized in Table 3-13. We assume
a branching ratio for B ! 00 of 5 10,6. After selection of the 00 pair, the same
kinematic cuts as in the +, modes are applied. Using BBSIM, the resolution on the
reconstructed B 0 mass in 00 is found to be 49.7 MeV=c2 , dominated by the errors on the
Technical Design Report for the BABAR Detector
68
Physics with BABAR
Cut
B0 ! 00
qq
0 0 Acceptance and p?(0) in [2:350; 2:950] 0.849 (0.772) 0.6 10,3
0 : Preselected Events (Normalization)
1.000
1.000
?
1 : p (B ) in [0:150; 0:500]
0.989
0.263
2 : Cosine Angle w.r.t. Sphericity Axis 0.8
0.770
0.115
3 : Cosine Angle w.r.t. Sphericity Axis 0.7
0.661
0.071
1+2
0.765
0.032
1 + 2 + m() in [5:120; 5:440]
0.754
0.015
1+3
0.656
0.019
1 + 3 + m() in [5:120; 5:440]
0.648
6.5 10,3
Overall Eciency per Event Generated
0.550 (0.518) 4.0 10,6
Eciencies of background rejection cuts for the B 0 ! 0 0 mode. The
numbers quoted in parentheses are obtained when combinatorial background in signal events
is suppressed.
Table 3-13.
measurement of photon energies. The cut on the 0 masses, together with a 3 cut on the
B 0 mass (5:120 < m < 5:440 GeV=c2), yields an eciency of 55% with a B/S of less than 4
to 1. These simulations have been made using a somewhat di erent detector geometry than
given in this Technical Design Report, but the conclusions are still valid.
3.5.3
B+
! +0
Decays
In addition to the 0 0 mode, the isospin analysis requires results from B + ! +0 decays.
A treatment similar to that described above has been made for this mode, and the results
are presented in Table 3-14. An eciency of 53% is achieved with a B/S of less than 3.5 to 1
using an assumed branching ratio of 1:2 10,5.
3.6
B0
!
The branching ratio for B 0 ! +, is taken to be (f=f )2 BR(B 0 ! +,) = 2:9 10,5 .
The value (f =f )2 = 2:4 is determined from ! and ! data [PDG94]. This
number is higher than that used in the Letter of Intent because recent LEP data on !
has caused a signi cant increase in the best t value. The corresponding rate for B ! +,
is more model dependent; we use R = ,(B 0 ! +,)=,(B 0 ! + ,) = 1. Estimates for
Technical Design Report for the BABAR Detector
3.6
B 0 !
69
Cut
B+ ! +0
+0 Acceptance and p?() in [2:35; 2:95]
0.863 (0.803)
0 : Preselected Events (Normalization)
1.000
?
1 : p (B ) in [0:150; 0:500]
0.987
2 : jCosine Angle w.r.t. Sphericity Axisj 0:8
0.756
3 : jCosine Angle w.r.t. Sphericity Axisj 0:7
0.650
2
4 : 0
0.942
1+2+4
0.717
1 + 2 + 4 + m() in [5:150; 5:410]
0.713
1+3+4
0.617
1 + 3 + 4 + m() in [5:150; 5:410]
0.614
Overall Eciency per Event Generated
0.529 (0.506)
qq
3.8 10,3
1.000
0.252
0.120
0.069
0.487
0.013
4.5 10,3
6.2 10,3
2.3 10,3
8.5 10,6
Eciencies of background rejection cuts for the B + ! + 0 mode. The
numbers quoted in parentheses are obtained when combinatorial background in signal events
is suppressed.
Table 3-14.
the ratio range from a high
q of 15 [Ale91] to a low of about 1/12 [Dea93]. The sensitivity of
to this ratio varies as (R + 1)=2R.
An analysis of this channel and a study of the background were performed using the ASLUND
Monte Carlo. The calorimeter performance was determined using BBSIM and was parameterized for the study. Only the backgrounds due to production of quark-antiquark pairs
in the continuum have been considered, as they are expected to be the dominant source.
Other channels decay to the same nal state, +,0 , and will be subject to similar
backgrounds. Eventually, all three possible nal states will need to be studied to allow a
full isospin analysis. However, it is sucient to show that the detector is well-suited to this
study, which has the potential to provide the best measurement of sin2 . Both calorimetry
and particle identi cation are important in the reconstruction of and the reduction of
the continuum backgrounds. This study assumes small penguin contributions and neglects
the interference between the channels.
The mass distributions of photon pairs obtained using the ASLUND Monte Carlo are shown
in Figures 3-3(a) and (b) for the signal and light-quark continuum events, respectively. The
low-mass tail of the 0 peak is caused mainly by incomplete containment of showers as
de ned by the clustering algorithm, and possibly by longitudinal leakage from the back
of the calorimeter. The peak of the signal distribution is o set slightly to the right of the
nominal 0 mass as a consequence of the combination of the low-energy tail and the resolution
parameterization in ASLUND, which maintains the mean energy of clusters at the correct
value.
Technical Design Report for the BABAR Detector
70
Physics with BABAR
(a)
π ο Candidate Mass (Signal)
Figure 3-3.
events.
(b)
π ο Candidate Mass (Background)
Reconstructed two-photon mass for (a) signal and (b) light-quark background
Candidate B 0 s were reconstructed by combining candidates and all charged tracks of
the opposite sign. Further cuts on the B 0 -candidate momentum in the rest frame of the
(4S ), and on its reconstructed mass, were made. These cuts have a dramatic e ect on the
backgrounds from all sources. The e ects of all the above cuts on the signal are summarized
in Table 3-15.
The fraction of continuum events that contribute a B 0 candidate to the background at this
stage is 5:4 10,3. Assuming a branching ratio of 5:8 10,5 for B 0 ! , we nd that
the continuum background dominates the signal by a factor of 140:1. A number of further
cuts were made to reduce continuum backgrounds:
All B 0 candidates were required to have a vertex within 5 cm of the nominal interaction
point in the z direction. This excluded tracks from conversions and nonprompt decays
with no appreciable e ect on the signal.
The two charged tracks from a B 0 decay to were required to form a good vertex
in three dimensions. In charm background events, the two charged tracks involved may
arise from di erent charm decays and hence be well-separated in space. Candidates
were therefore required to have a vertex 2 per degree of freedom less than 5.
In the decay B ! , the is 100% polarized, so that the angle of the decay
axis with respect to its line of ight has a cos2 decay distribution. Background events,
Technical Design Report for the BABAR Detector
3.6
B 0 !
71
Cut
Signal Eciency
Geometry and Emin ( ) = 25 MeV
0.76
111 m(0 ) 160 MeV=c2
0.98
0
Combined Eciency
0.74
Geometry and pt;min() = 60 MeV
0.91
0:37 m() 1:17 GeV=c2
1.00
+,
Combined Eciency
0.91
Geometry and pt;min = 60 MeV
0.96
B 0 Momentum in Rest Frame
1.00
0
B Candidate Mass
1.00
0
Overall B Reconstruction Eciency
0.65
Table 3-15.
Eciencies of B 0 reconstruction for modes.
however, should show no such correlation. A cut was placed at j cos decay j 0:25 of
the angle between the two directions in the rest frame.
More severe cuts were made on the B 0 -candidate momentum in the (4S ) rest frame,
and on its invariant mass.
As in the +, case, the light-quark background events have a pronounced jet structure, with the identi ed B -decay tracks lying within the jets. The thrust (or sphericity)
axis of the remaining particles in the event is strongly correlated with the B -decay
direction in background events and uncorrelated in signal events. Data are divided
into six bins in j cos thrustj and all data is used to achieve maximum CP -sensitivity.
Results for three of these bins are shown in the Table 3-16. For the best region,
jcosthrust j 0:5 a B=S of less than 3/2 is achieved with kaon tags, while with lepton
tags this ratio is less than 0.15.
The eciencies (and the relative tagging eciencies) of the above cuts for signal and background are summarized in Table 3-16 along with the eciencies for initial reconstruction,
the branching ratios and cross sections, and nally, the overall background contamination
per reconstructed B meson. Unbracketed numbers correspond to kaon tags, bracketed ones
to lepton tags. A multivariate analysis has also been performed, following the lines of
Reference [Ale91], and leads to similar results.
Technical Design Report for the BABAR Detector
72
Physics with BABAR
Cut
Signal
Light Quark
Charm
,
3
Reconstructions and Preliminary Cuts 0.65
6.70 10
3.20 10,3
jZvertj < 5 cm
1.00
1.00
1.00
Chi-squared per d.o.f. of Vertex
0.98
0.92
0.68
jcosdecay j > 0:25
0.98
0.85
0.87
0:14 p(B 0 ) 0:44 GeV=c
0.98
0.42
0.42
0
2
5:20 m(B ) 5:36 GeV=c
0.95
0.31
0.30
Tagging Eciency w.r.t. Signal
1.0(1.0)
0.47(0.049)
1.4(0.065)
One candidate per event
1.0
0.77
0.68
,
5
Eciency for 0:0 < jcosthrustj < 0:5
0.29 1.13(0.117) 10 0.69(0.031) 10,5
Eciency for 0:7 < jcosthrustj < 0:8
0.058 1.65(0.170) 10,5 1.98(0.089) 10,5
Eciency for 0:9 < jcosthrustj < 1:0
0.058 16.0(1.65) 10,5 13.5(0.611) 10,5
Example of background rejection cuts for modes. Unbracketed numbers
correspond to kaon tags (class b), bracketed ones to lepton tags (class a). Data samples in
higher cos(thrust ) bins have higher backgrounds.
Table 3-16.
j
3.7
j
Tagging Modes
Measurements of CP asymmetries and B 0 or Bs mixing require the determination of the B
avor. At BABAR, events are studied in which a coherent B 0B 0 state produces one B meson
decaying into a mode under investigation for CP violation, while the avor of the second
B is reconstructed (or tagged) from its decay products. Any particle with a distinct avor
quantum number provides a possible tag. Typically this information is in the form of a
lepton charge, a D charge, or a K charge.
The observed asymmetry aobs(t) can be related to the expected asymmetry, atheory , by
aobs(t) = D atheory (t)
(3.15)
with a dilution factor jDj < 1, where D = (1 , 2w), and w is the wrong-tag probability.
(The further dilution e ect of smearing due to limited time resolution is discussed in the
following section.) Di erent types of tagging information in a single event are combined,
leading to an event-by-event de nition of the appropriate dilution factor.
Details of the most recent tagging studies are discussed in References [Wal95] and [Pla95].
The results are similar. The numbers reported here are taken from the rst of these studies,
which uses a Monte Carlo simulation based on a version of the event generator JETSET 7.4
[Sjo93] tuned to reproduce ARGUS data [Alb94b]. This study builds on earlier work which
identi ed a number of e ective tagging strategies [Jaf94, Wal94a, Pia94]. The strategy is to
Technical Design Report for the BABAR Detector
3.7 Tagging Modes
73
use all possible information in each event to assign the appropriate avor and dilution factor,
and then to extract asymmetries from a two-dimensional t of the data against time and
e ective dilution. The detector acceptance, momentum and energy resolution, and particle
identi cation are simulated with ASLUND.
A avor estimator that allows us to combine all tagging information in an event is assigned
as follows. For each event, nd all possible tagging identi ers. For each identi er, calculate
the quantities Li, Qi, and Di and de ne Bi = Li Di Qi . The likelihood ratio Li , which
estimates whether the tagging track is correctly identi ed as a particle of type i, is given by
Li =
e,2 =2
Pj=e;;;K;p e,2=2 ;
(3.16)
i
j
where 2i is the sum of the 2 from each available source of information for particle type i.
The avor charge Qi for a track of type i is de ned to be +1 for a B 0 right-sign tag and ,1
for the opposite sign. Right sign is de ned below for each class of decays. The appropriate
dilution factor for Di for this particular tag type and event characteristics is then calculated
from the simulation. Eventually, these numbers will also be checked by the study of events in
which both B mesons decay to tagging modes. For each event a combined avor estimator is
de ned as follows: for events with more than one possible tagging identi er, the Bi are added
like relativistic velocities (v=c) to give a B^ , such that ,1 B^ 1; for events with only one
identi er, B^ = B1 . If the single estimators were statistically independent, B^ would be equal
to the e ective dilution Deff for that event. In fact, they are not fully independent, but one
can measure Deff (B^ ) using doubly tagged modes [Wal95]. A cut on B^ provides a simple way
to select against poor tags using all the information in the event. A two-dimensional t to
the asymmetry against time and B^ does not require such a cut and thus has slightly higher
tagging performance. Other studies [Pla95] use a slightly di erent approach to combining all
possible tagging information, directly de ning an e ective wrong tag fraction xeff for each
event rather than introducing B^ .
3.7.1
Kaon Tags
The sign of a kaon's charge is a signature of its strangeness. The principal sources of K s
from b-quark decay are:
b ! cW , ; c ! W + s; s ) K , ;
b ! XW , ; W , ! cs; s ) K , and/or c ! W + s; s ) K + ;
b ! cW , ; W , ! us; s ) K , and/or c ! XW + ; W + ! us; s ) K + ;
b ! Xss; s ) K , and/or s ) K + ;
b ! sq q (penguin); s ) K , :
(a)
(b)
(c)
(d)
(e)
Technical Design Report for the BABAR Detector
74
Physics with BABAR
We call the K from the dominant process (a) the right-sign kaon. Thus, QK = +1 for
a K + and ,1 for K ,. Processes (b), (c), and (d) yield both right-sign and wrong-sign
kaons, typically with a second strange particle in the event. Doubly Cabibbo-suppressed
and penguin processes (e) give tiny contributions. Events with multiple like-sign kaons
can occur since the processes listed above are not mutually exclusive. Wrong-sign particles
contribute to the wrong-tag probability, but as the list above shows, other characteristics of
the event can be used to reduce the wrong-tag fraction. Aside from the presence or absence
of a second K in the event and its charge, the wrong-tag fraction also depends somewhat on
the kaon momentum in the (4S ) rest frame.
Kaons are identi ed via dE/dx by the DIRC in the barrel, and by aerogel Cherenkov counters
in the forward endcap. This information is combined into a likelihood. Tracks with LK > 0:5
are counted as charged kaons in assigning event classes. All K s with LK > 0:1 contribute to
the determination of B^ . In this simulation, kaons which decay before reaching the DIRC do
not contribute. This is a conservative assumption, since dE/dx information or the geometric
signature of a kaon decay in ight may in fact lead to the recovery of good kaon identi cation
in some cases.
On the basis of the simulations described above, the performance of all kaon tags (with
2 = 0:21. Here, K is the kaon tagging eciency, that is, the
or without leptons) is K Deff
0
fraction of generic B decays for which a kaon tag is available.
3.7.2
Lepton Tags
Electrons and muons in b decays come from several sources:
b ! X W , ; W , ) l, ; (0:105e + 0:105); W , ! , ) l, ( 0:005e + 0:005); (a)
b ! X c; c ! X W + ; W + ) l+ ( 0:09e + 0:09);
(b)
,
,
,
(c)
b ! ccs; c ! X W W ) l ( 0:01e + 0:01);
0
+
,
b ! ccs; cc ! J= orJ= ; J= ) l l ( 0:003e + 0:003):
(d)
The lepton from a semileptonic B decay, process (a), is de ned as a right-sign lepton. Process
(d) is a tiny source of spurious tags when one or the other lepton is missed. Each decay
path leads to a distinct distribution in the lepton momentum, the number and invariant
mass of the other accompanying particles, and the distribution of the remaining energy with
respect to the lepton direction [Jaf94, Pia94]. All such information is used, when available,
to determine the wrong-tag fraction. For example, Figure 3-4 shows the distributions of
right- and wrong-sign leptons as a function of momentum, p, in the (4S ) rest frame. At
low momentum, wrong-sign leptons dominate and can contribute usefully to tagging.
Electrons are identi ed in the electromagnetic calorimeter and at low momenta by the DIRC
and dE/dx measurements. Muons are identi ed in the IFR supplemented by DIRC at low
Technical Design Report for the BABAR Detector
3.7 Tagging Modes
75
Figure 3-4. Distribution of momentum in the (4S ) rest frame for leptons. Right-sign
leptons are l+ for B 0 decays and l, for B 0 .
momenta. In low momentum regions, where the separation from hadrons is not perfect, each
candidate is weighted by its likelihood, and cuts at Le > 0:7 and L > 0:7 are applied.
We take the semileptonic branching fraction of the B 0 meson to be 10:5%. Simulations of
2 = 0:13. Here, l is the lepton tagging eciency,
the lepton tagging performance give l Deff
that is, the fraction of signal events with an identi ed lepton tag. The performance number
includes a cut jB^ j > 0:4. A two-dimensional t to the asymmetry, binning the data in time
2 = 0:16.
and in B^ , gives slightly higher tagging performance, l Deff
Technical Design Report for the BABAR Detector
76
Physics with BABAR
Class Selection Eciency Eciency Performance
( )
(B^ 0:1)
(Def2 f )
A
0.20
.19
0.13
B
0.80
.66
0.21
All
1.0
0.86
0.34
Table 3-17.
all events.
Performance gures for lepton and kaon tags and for combined tagging using
3.7.3 Other Tags
In addition to lepton and kaon tags, charged pions have a signi cant avor correlation at
the high end of the momentum spectrum due to two-body decays with a charged or
meson, and at the low end due to pions from D+ ! +D0 [Sny90]. Additional D tags can
be formed from neutral K s. Strange baryons, though rare, provide a small contribution to
tagging. These e ects are included in the analysis presented here. Ongoing studies continue
to identify further ways of reconstructing events that can lead to a avor assignment with a
non-negligible Def f . All such information will be used to maximize tagging performance.
3.7.4 Combined Tagging
All available information is used to calculate a combined tagging performance factor B^ for
each event, thus creating a two-dimensional distribution in time and the e ective dilution
^ ) (or equivalently xef f ) to be used in tting for the asymmetry. We divide tagged
Def f (B
events into two classes, class A being those events with a lepton which has a high probability
of being a primary lepton from semileptonic B decay, and class B being all other events. This
separation is necessary because the vertex resolution for the tagging decay is di erent for the
two cases, and hence the contribution to a CP -asymmetry measurement must be determined
separately for each case. More precisely, class A includes events containing either: (1) a
lepton with p > 1:4 GeV=c; or (2) a lepton with p > 0:5 GeV=c, and no other charged tracks
or a total energy of less than 0.5 GeV within the same hemisphere in the (4S ) rest frame.
The results are given in Table 3-17. The eciency values include misidenti ed particles, but
the e ective dilution factors correct for this. The quantity labeled \performance" is de ned
to be the sum over bins with B^ 0:1 of Def2 f , where is the fraction of signal events in
the bin. If the lepton information alone were used for class A events, the performance would
be only 0.125; the higher combined performance listed in the table shows the impact of the
additional information on these events.
Technical Design Report for the BABAR Detector
3.8 Estimate of CP -Angle Measurement
77
3.8 Estimate of CP -Angle Measurement
3.8.1
Method of Calculation
Using the approximation that both B s have a boost equal to , namely the boost of the
(4S ), the quantity z = zcp , ztag is related to the time di erence between the two B
decays by:
z :
t=
Based on the simulations described above, each CP analysis channel (loosely called a CP
mode in the following) and each tagging class, are characterized by eciency, background (or
dilution for tags), and z resolution. For each CP mode and tagging class, the distribution in
time and b avor is regenerated from these parameters. The program CPEXTRACT [Sny94]
is then used to perform a binned maximum likelihood t to this data and to make an estimate
of the error on the asymmetry for a sample corresponding to 30fb,1 of (4S ) data. This t
includes the dependence on time, tagging type, and background. Errors are quoted for zero
CP asymmetry, since our sensitivity to small asymmetries is the quantity of interest. More
details of the calculation are given in Reference [Sny95].
The z resolution of background events (such as those that occur in the +, and
analyses) is treated separately in CPEXTRACT. This enhances our ability to discriminate
against these backgrounds since the background events are produced at z = 0, and the
resolution on z is better for background events than for signal plus tag events. The
resolutions on z for backgrounds can be parameterized by two Gaussian distributions,
with 90% of the events in a narrow component (narrow = 80 m) and the remaining events in
the wide component (wide = 280 m). Furthermore, since the expected asymmetry behaves
as sin(K z), where K = M= c, the impact of background on the tted asymmetry
parameters is minimized by correct treatment of the z dependence. This is illustrated by
Figure 3-5 which shows the z distribution for B ! with a class B (mostly kaon) tag, an
assumed asymmetry of 0.866 ( = 30), and an assumed background-to-signal ratio of 9. As
can be seen, the background drops signi cantly by the time the asymmetry develops. The
improvement in CP -resolution obtained by explicitly tting for the z dependence of the
background is shown in Figure 3-6.
Here, wrong-sign tags are assumed to have the same z distribution as right-sign tags.
This is a pessimistic assumption because, in general, the wrong-sign z distribution will
be signi cantly broader, which will tend to reduce the adverse e ects that they have on
the time-dependent asymmetry. In the nal analysis, they must be handled correctly, or a
signi cant systematic error will be introduced.
Technical Design Report for the BABAR Detector
Physics with BABAR
Number/25 µm
78
104
103
102
101
100
-2000
-1000
0
1000
2000
∆z × flavor
Distribution of z for the CP -state using class B tag and an assumed
background-to-signal ratio of 9.
Figure 3-5.
For lepton tagged events, there is a strong dependence of the z resolution on polar angle.
This is accounted for by the simple expedient of dividing the direct lepton tag into two polar
angle bins and treating each bin as an independent tag. Further re nement of this treatment
can no doubt be used to improve accuracy in the actual data analysis.
3.8.2
Tagging Modes
For tagging eciencies, we use the numbers given in Table 3-17 above. For both class A
and class B tags, we estimate the z coordinate of the tag by tting a common vertex to all
tracks except those associated with the CP -state and those with transverse impact parameter
greater than 1 mm. This last e ectively excludes the decay products of kaons, lambdas, etc.
Since the z resolution depends strongly on the direction of the tagging track [Cha95], the
tags are divided into seven bins based on the tangent of the dip angle (jtandip j) in order to
optimize our CP -sensitivity. The numbers used in this document were obtained from a high
statistics ASLUND simulation. Details can be found in [Wea95]. BBSIM simulations [Cha95]
con rm these results for lepton tags.
Technical Design Report for the BABAR Detector
3.8 Estimate of CP -Angle Measurement
79
5
σ/σpure
4
3
2
1
0
10-2
10-1
100
101
B/S
Figure 3-6. Ratio of cp in the presence of background to cp obtained without background as function of the background-to-signal (B=S ) ratio. The solid curve corresponds to
tting for the z dependencies of the background and the signal. The dashed curve results
if z information is not exploited.
3.8.3
CP
Modes
The parameters required to determine sensitivity (branching fractions, number of events
expected, backgrounds, and zCP resolutions) for each CP -mode under consideration are
compiled in Table 3-18. These results are collected from the studies described above. The
expected number of events before tagging is based on 30fb,1 collected on the (4S ). The
z resolutions for each mode are characterized by three numbers: narrow , wide , and the
ratio of the area of the narrow Gaussian to the wide one. For D+D,, many decay modes
contribute, and events are grouped into two classes, four-prong and six-prong decays. A
single Gaussian then gives an adequate parameterization of the z resolution of D+D, events
in each class. Similarly, for D+D, decays, events are classi ed by the number of 0s in the
nal state, and then a single Gaussian suces to characterize the z resolution of each class.
For J= KL the background consist of two components: a CP symmetric part (the rst
number) and a CP anti-symmetric part given in parenthesis. For all modes with background
the number in square brackets gives the degradation in cp that results from the background.
Our procedure of estimating the CP -violating asymmetry by dividing the data into samples
with varying background levels and tting explicitly to the z distribution of the background
Technical Design Report for the BABAR Detector
80
Physics with BABAR
Mode
Branching Usable Eciency Reconstructed Background z Resolution
Fraction Sample
Number
0
,3
J= KS 0:5 10
2160
0.41
1106
negligible 36=111=16
0
,3
J= KL 0:5 10
2160
0.33
712
191(96)[1.3] 36/111/16
J= K 1:6 10,3 788
0.39
307
negligible 37/155/10
+
,
,4
D D
6 10
993
0.25
248
109 [1.02] 52(6-prongs)
75(4-prongs)
D+ D, 7 10,4 3130
0.155
485
62 [1.005]
43 (0 0)
51 (1 0)
69 (2 0)
+,
1:2 10,5 432
0.80
346
1957 [1.11] 34/161/4.4
,5
5:8 10
2088
0.56
1162
33420 [1.18] 36/240/24
Table 3-18. Simulated eciencies, backgrounds, and vertex resolutions for CP modes.
The usable sample re ects the visible branching ratio and the 30 fb,1 sample size.
Reconstructed number of events does not include tagging eciencies. The background
number given is for the total sample; in the analysis the data are divided into bins of
varying background to minimize the adverse impact on CP resolution. The factor by which
background actually degrades the resolution is given in square brackets.
when estimating the CP -violating asymmetry q
gives a substantial improvement in sensitivity
over a naive treatment which uses a factor of 1 + B=S .
The resolution on the CP -violating phase (M , D ) for each CP state is given in Table 3-19.
The combined sensitivity for sin 2 and sin 2 is estimated to be 0:059 and 0:085,
respectively, for 30 fb,1, or one year's data at nominal luminosity. The channels J= K ,
D D, D D , , and a1 , which are not pure CP eigenstates, are assumed to be dominated
by a single CP eigenstate. If this is not the case, the resolution obtained with these modes
will be reduced. An accuracy on the CP phase comparable to the DD modes can be expected.
The results in this table di er in many ways from those given in the Letter of Intent. The
di erences arise from two major sources. First, our knowledge of the branching ratio to the
modes of interest has improved somewhat, resulting in changes in these key input numbers.
Second, more complete simulation work, including backgrounds, has been carried out for
most of the channels included in Table 3-19, resulting in changed eciencies and signal-tobackground ratios. The number given for the a1 channel is an estimate based on the work
of Reference [Ale91]. The number obtained in that paper has been rescaled, re ecting the
integrated luminosity assumed here, using the tagging eciency expected for this detector,
and correcting the assumed branching fraction for B ! . For the DD channel, the
estimate is based on the simulations of the DD and DD channels and the measured
B ! Ds D branching ratios.
Technical Design Report for the BABAR Detector
3.9 CKM Matrix Determination
CP State
J= KS0
J= KL0
J= K
D+ D,
D+D,
DD
Combined
+,
a1
Combined
81
Br
M , D A Tags B Tags Combined
0:5 10,3
0.15
0.13
0.098
0:5 10,3
0.25
0.21
0.16
,3
1:6 10
0.29
0.25
0.19
,4
6 10
0.32
0.27
0.21
7 10,4
0.24
0.20
0.15
,4
8 10
0.15
0.059
1:2 10,5
0.29
0.27
0.20
,5
5:8 10
0.16
0.16
0.11
6 10,5
0.24
0.085
Error on the measurement of (M , D ) from various channels for 30 fb,1 .
The A-tag sample is chie y direct leptons, and the B sample is all other tags. The numbers
marked * (a1 and D D) are estimates; all others are based on simulations.
Table 3-19.
The accuracy achievable for the channel depends on the ratio R = ,(B 0 ! + ,)=,(B 0 !
,+) [Ale91, Ale93]. The results given in Table 3-19 assume R = 1. The dependence of
the sensitivity on this assumption is shown in Figure 3-7.
3.9 CKM Matrix Determination
The values of the CKM matrix elements determine the lengths of the sides of the unitarity
triangle. Accurate measurements of these elements are thus an important ingredient in
overconstraining the triangle and probing the source of CP violation. Furthermore, the values
of the CKM elements are fundamental input parameters within the Standard Model and
hence, must be measured as precisely as possible; better knowledge of these parameters may
provide some insight into their origin. The present status of these elements is summarized in
Figure 3-8 in the Wolfenstein parameterization, where the shaded area is that allowed in the
Standard Model for the apex of the unitarity triangle in the (; ) plane. The parameters
used for the plot of the allowed region for the unitarity triangle are taken from the draft of a
Review of Modern Physics article by P. Burchat and J. Richman [Ric95]. The two directlymeasured CKM parameters are jVcbj = 0:040 0:003 and jVub=Vcbj = 0:076 0:026. The
value of the parameter that describes Bd0-B 0d mixing is m(B 0 ) = xd =B0 = 0:43 0:06 ps,1.
The top quark mass evaluated at the relevant scale for this mixing isp mt = 165 15 GeV.
The hadronic matrix elements obtained in lattice calculations are fB BB = 200 40 MeV,
and BK = 0:75 0:05. The resulting errors are added in quadrature, and the lines shown
Technical Design Report for the BABAR Detector
82
Physics with BABAR
4
σcp(R)/σcp(1)
3
2
1
0
0
0.25
0.5
0.75
1
1.25
1.5
R
Dependence of the error on the CP asymmetry angle on the assumed value
for the ratio of decay widths R = ,(B 0 ! + , )=,(B 0 ! , + ), normalized to the case
R = 1.
Figure 3-7.
in the plot correspond to 1:5 limits. The overlapping limits result in the following set of
allowed ranges for the unitary angles:
,0:59 sin 2 1:0
0:29 sin 2 0:88
,1:0 sin 2 1:0 :
(3.17)
It is important to remember that this picture represents the Standard Model and can be
dramatically altered if new physics is present. Examples include new contributions to or
Bd0 -B 0d mixing, even if no new sources of CP violation are present. It is equally important
to note that, since the range of values of parameters such as BK and BB is constrained
by theory rather than experiment, the overlap region is a best estimate rather than a 90%
con dence level result.
3.9.1
Vcb
Values for Vcb are extracted from the investigation of inclusive and exclusive semileptonic B
decays. CLEO and LEP have determined Vcb at the 7.5% level [Ric95], depending on the
Technical Design Report for the BABAR Detector
3.9 CKM Matrix Determination
83
εK, |Vcb|
η
xd
|Vub|
ρ
Figure 3-8.
Constraints in the Standard Model in the
plane.
The shaded area
corresponds to that allowed for the apex of the rescaled unitarity triangle.
extraction technique, and BABAR can expect to reduce these errors to the few percent level
due to the expected large data sample and ecient reconstruction techniques. The inclusive
semileptonic branching fraction BSL can be determined from:
1. Measurement of the inclusive single-lepton momentum spectrum. This technique
yields signi cant data samples, but the procedure used to t the observed spectrum
to the expected shape for primary and secondary leptons from B and charm decay,
respectively, introduces a large model dependence.
2. Charge and angular correlations in dilepton events. This o ers less model dependence,
as the measured correlations can be used to separate the primary and secondary lepton
spectra, instead of relying on theory.
3. Separate measurement of BSL for charged and neutral B meson decay. Here, one B
in the event must be reconstructed in order to tag the charge of the other; this will
require the large tagged event sample available to BABAR.
Determination of Vcb from BSL via technique (1) at CLEO and LEP is already dominated by
the theoretical error, while method (2) still o ers room for improvement on the systematic
and statistical errors. In addition, it is important to make better measurements of BSL for
the charged and neutral B mesons separately. These are used to determine the ratio of B
lifetimes, which in turn are used in interpreting exclusive semileptonic and hadronic decay
rates. In these measurements, one B in the event is reconstructed in order to tag the charge
of the other; this will bene t from the large tagged event sample available to BABAR.
Technical Design Report for the BABAR Detector
84
Physics with BABAR
Exclusive semileptonic decays o er a reliable model-independent determination of Vcb within
the framework of heavy quark e ective theory (HQET). The heavy quark symmetry allows
the normalization of the q2-dependent hadronic form factors with good precision at zero
recoil for the charm hadron system. This technique is best suited for the process B ! D``
as the leading corrections to the HQET result arise only at higher order, 1=m2Q. Due to the
presence of the neutrino, reconstruction of exclusive semileptonic channels is more dicult
than for hadronic decays; at present, the B ! D`` mode is the most precisely measured.
One technique at the (4S ) is to determine the missing momentum distribution recoiling
against the D` system. The zero recoil con guration for the charm hadrons results in a
reduction of phase space for the decay (as the B and D mesons have approximately equal
velocities), yielding a statistically limited data sample, and requires good detection eciency
for low momentum pions arising from the D decay. Measurements of other exclusive modes,
such as D`` and D``, will also be precisely studied at BABAR.
3.9.2
Vub
An accurate determination of Vub is important to test closure of the unitarity triangle. Data
near the end-point region of the lepton momentum spectrum in inclusive semileptonic B
decays have established that Vub is nonzero. However, converting the measured rate into a
value of Vub introduces substantial errors. This conversion is highly model dependent due to
the small phase space available at the end-point, the large uncertainties in the calculation
of the rates for the resonant modes, and the relative sizes of the contributions of resonant
and nonresonant modes in this region. The subtraction of background from the small data
sample is an additional large source of error. The present experimental error on the ratio
jVubj=jVcbj is comparable to the theoretical uncertainty, and thus new, less model dependent
techniques for extracting this CKM element are necessary.
Measurement of exclusive semileptonic decays (B ! Xu``, where Xu = ; , or !) requires
a large data sample and should provide an improved determination of Vub. Such decays
have yet to be observed, but model predictions indicate that the rate should be distributed
over many exclusive channels, with no dominant modes. The theoretical uncertainties are
expected to be lower than in the inclusive case. Further reductions in the theoretical errors
can be obtained via measurements of the form factor q2 distributions in c ! d transitions such
as D ! Xu``. Given adequate statistics and understanding of experimental systematics
from vertex separation and =K identi cation, such form factor measurements can be made
at BABAR.
An alternative method [Bar90] of extracting Vub from semileptonic B decays is to measure
the invariant mass spectrum of the nal state hadrons below the charm hadron threshold,
Technical Design Report for the BABAR Detector
3.10 Rare
B Decays
85
mX < mD . A study of this method at BABAR has yet to be made, but the advantages of
separate vertexing of the two B s will probably be important for this technique to be feasible.
i.e.,
Vtd
3.9.3
It has recently been shown [Atw94, Gol94, Des94] that the proposed method of obtaining the
ratio jVtd=Vtsj from the ratio of branching fractions B (B ! )=B (B ! K ) is not valid
due to the potentially large long-distance contributions to these exclusive B decay modes.
The sizes of these long-distance contributions could be determined with a measurement of
exclusive radiative charm decays, such as D ! , which are expected to have no shortdistance contamination. Separate measurements of B , ! , and B ! as well as
Bs ! ; K may also provide an estimate of the long-distance rates.
A measurement of Bs -Bs mixing could also yield a value for this ratio of CKM elements as
2
0
0
0
0
0
Bs
BfBs jVtsj
xs = mBs QCD
1
=
s
Bd
xd mBd QCD BfBd jVtdj
[(1 , ) + ] :
2
2
2
2
(3.18)
2
2
2
2
The factor s, which multiplies the ratio of CKM elements, measures the amount of SU (3)
breaking e ects and is estimated to be 1.3. Determination of fB s from the decay B s !
will greatly reduce the theoretical uncertainty in this mixing ratio. The allowed region of
the plane, determined above in the Standard Model, yields xs = 10{57. A precise
measurement of Bs mixing requires a substantial run on the (5S ). It is needed in order to
obtain a clean determination of jVtd j=jVtsj. If Vts is relatively large, a more accurate value of
jVtd=Vtsj could be obtained from a measurement of ,=, for the Bs meson [Bro95].
( )
( )
2
3.10
Rare
B Decays
Rare B decays are an important testbed of the Standard Model. They o er a complementary
strategy in the search for new physics by probing the indirect e ects of new interactions in
higher order processes. In particular, the probing of loop-induced couplings can provide
tests of the detailed structure of the Standard Model at the level of radiative corrections,
where the Glashow-Iliopoulos-Maiani cancellations are important. The recent observation
of radiative B decays by CLEO has already provided bounds on the CKM ratio jVts=Vcbj as
well as powerful constraints on new physics. The high luminosity and separated B vertices
available at PEP-II and improved particle identi cation capabilities of BABAR will be needed
to allow accurate measurements of many rare modes. No full detector study has yet been
Technical Design Report for the BABAR Detector
86
Physics with BABAR
made for any of these channels. However, BABAR is designed to give as complete information
on all tracks as can reasonably be achieved, and hence, is well-suited to extend our knowledge
of these modes.
3.10.1
B
!
The decay B ! is of interest because the branching ratio will yield a value of the product
fB2 jVubj2 . The general expression for the decay width for the decay B ! ` is
2
m2l 2;
,(B + ! l+ ) = G8F fB2 mB m2l jVubj2 1 , m
2
b
where GF is the Fermi constant, fB is the B -meson coupling constant, mB is the mass of
the B meson, ml is the lepton mass, and Vub is the CKM-matrix element. For the decay
B ! , the branching ratio is about 8 10,5.
Leptonic decays of the B meson are expected to be best identi ed by fully reconstructing
one charged B meson (the tagging B ) in events in which the only observed particle coming
from the decay of the second B meson is a charged lepton. The separation of the decay
vertices of the two B mesons, along with the excellent mass resolution of the detector, allows
an unambiguous identi cation of the tagging B meson. The problem then is to reject decays
of the nontagging B mesons which are not purely leptonic but in which only a single lepton
is identi ed in the detector. A classic example of this type of decay is the semileptonic decay
B ! ` KL0 , in which neither the nor the KL0 are detected. The BABAR detector needs to
have excellent muon and KL0 detection and rejection capabilities in order to veto these types
of decays with high eciency.
For the decay B ! , the branching ratio is about 4 10,7, a factor of 200 smaller
than that for B ! . However, B ! decays should be much easier to detect and
discriminate from backgrounds because they are two-body decays, and hence the muons are
close to monoenergetic. Further Monte Carlo simulation work needs to be done before a
prediction can be made of the level of sensitivity of the BABAR detector to B ! ` decays
and to determine whether or searches give better sensitivity. However, it is already
clear that the general design of the detector and the separate vertexing of the two B s make
BABAR competitive with any other experiment for this search.
Technical Design Report for the BABAR Detector
3.11 CP Asymmetries in Charged B Decays
3.10.2
B
87
! X `+`,
s
This decay receives short-distance contributions from electromagnetic and Z penguin diagrams as well as W box diagrams, and long-distance contributions from the process B !
K followed by ! ` `, and from cc continuum intermediate states. The short distance contributions lead to the inclusive branching fractions B (B ! Xs` `,) = (15; 7; 2)
10, for ` = e; ; , and will likely be observed before PEP-II is operational. However,
the best method of separating the long and short distance contributions, and observing any
small deviations from the Standard Model, is to measure the various kinematic distributions
associated with the nal-state lepton pair, such as the lepton pair invariant mass distribution, the lepton pair forward-backward asymmetry, and the tau polarization asymmetry.
Measurement of these distributions requires the high statistics samples available at PEP-II.
Here, the long-distance contributions dominate only in the m(` `,) regions near the J= and
0 resonances, and observations away from these peaks cleanly separate the short distance
physics. The values of the forward-backward and tau polarization asymmetries are large for
currently favored values of the top quark mass around 175 GeV=c [Abe94]. Measurement
of all three kinematic distributions would provide a unique determination of the sign and
magnitude of each contributing short distance operator, and hence, would provide a stringent
test of the Standard Model.
0
( )
(0 )
(0 )
+
+
6
+
2
3.10.3
B
! X
s
The decay B ! Xs proceeds through Z -penguin and W -box diagrams yielding an
inclusive branching fraction of 510, . Measurement of this process would help to determine
the size of Z -penguin e ects in rare B hadronic decays, such as B ! . A search for such
decays will focus on events with large missing ET opposite a reconstructed B . High integrated
luminosity, good detector hermeticity, and ecient KL identi cation for background rejection
are important for this mode. The search will require at least 10 reconstructed B mesons.
The reduction in continuum background problems o ered by the separate vertexing of the
two B s in an event will be a signi cant advantage for this channel.
0
5
0
5
3.11 CP Asymmetries in Charged B Decays
CP asymmetries in charged B decays are another important and as-yet-unexplored region of
physics. While these asymmetries o er no direct measurement of CKM parameters, unlike
the neutral B case, their observation is nonetheless essential to our full understanding of the
physics of B decays. Any observed asymmetry in charged B decays is a direct CP -violation
Technical Design Report for the BABAR Detector
88
Physics with BABAR
e ect and would, for example, rule out the superweak alternative theory of CP violation.
These asymmetries require interference between two amplitude contributions that have both
di erent weak phases and di erent strong phases. In the Standard Model they are expected
to be largest in a channel such as K in which tree amplitudes are CKM suppressed relative
to penguin contributions, and hence the relative strengths of the two types of terms are
comparable. Unknown strong phases make even this statement a crude estimate rather than
a rigorous prediction. However, if factorization proves to be a good approximation, then,
within the Standard Model, this expectation for K becomes quite reliable. Of course, the
total rate for such a channel is also suppressed, so it remains to be seen whether it will be
easier to measure a small asymmetry in a relatively high rate channel or a larger asymmetry
in a rare decay mode.
There are aspects of detector design of particular importance. Just as with the neutral
channels, issues such as particle identi cation for distinguishing K 0 from 0 decays,
are crucial to any such search. Clearly, one also needs to be able to detect 0 s and
other calorimeter signatures to perform such studies. Likewise, good detector resolution
is needed to allow cuts that eliminate light-quark-produced backgrounds without excessive
deterioration in signal. Thus, a detector designed to perform well in the neutral B case is
also well-adapted for a study of charged B decays and a search for possible CP -violating
e ects in these channels.
3.12 Charm Physics
BABAR's expected high D meson production rates, good time resolution between decays,
particle identi cation systems, and near hermetic eciency should make high sensitivity
searches into DD mixing, rare decay physics, and CP violation in the charm sector feasible.
3.12.1
D0 D0
Mixing
Within the Standard Model, D0D0 mixing is expected to be very small. The amount of
mixing is governed by xmix = m=, and ymix = ,=,, where m is the mass di erence
between D0 , D0 states, and , is the decay rate. In the Standard Model, estimates for
xmix and ymix are in the range 10,3{10,5 [Liu94]. If mixing is enhanced, this would be
a sign of new physics. Unless the DD mixing rate is exceptionally large, it will not be
signi cantly tested in the near future. BABAR can play an important role in placing a limit
on the experimental sensitivity of DD mixing at about xmix 10,2 , 10,3, a range of
interest for some nonstandard model predictions [Hal92].
Technical Design Report for the BABAR Detector
3.12 Charm Physics
89
The experimental limitations on mixing sensitivity come primarily from two sources, doublyCabibbo-suppressed decays (DCSD), which mimic mixing in pure hadronic channels, and
experimental backgrounds. The mixing term behaves as sin(mt). Thus, with small xmix ,
one looks for mixing with optimum sensitivity at or beyond two decay lifetimes. Even after
a few lifetimes, one is sensitive to uctuations in the tails of the DCSD rate. In addition,
experimental backgrounds can result from combinatorial counting, poor vertex reconstruction, particle misidenti cation, and confusion with other long-lived decay products. Such
backgrounds can be reduced by vertex separation cuts, particle identi cation requirements,
and rejection of mass re ections, etc.
Either resonant or continuum production of Ds can be used for mixing studies in BABAR.
Commonly used decay chains such as D+ ! D0 +, then D0 ! K , +, K ,,+,, K ,l ,
and even K , +0, require measurement of the sign of the from the D decay to de ne the
initial D0 avor. In the boosted system of BABAR, this usually has enough momentum to
be reconstructed with high eciency. The good particle identi cation properties of BABAR
are important in analyzing such channels. The BABAR vertex detector will have adequate
proper time resolution to resolve the decay time development of the DCSD from that of
decay after mixing.
With suciently high statistics samples, searches via semileptonic decays may ultimately
provide better mixing limits. Semileptonic decay channels have no DCSD contributions, but
experimental backgrounds due to incomplete reconstruction of the fully leptonic nal state
can dominate the measurement error.
3.12.2 CP Violation in Charm Decays
CP -violating e ects are also predicted to be very small in the charm sector, although some
have argued that they may be signi cantly enhanced without requiring physics beyond the
Standard Model [Gol89, Buc93]. Again, with high reconstruction eciencies for charm decays
and ecient tagging mechanisms, BABAR will be sensitive to observation of CP -violating
e ects in the charm sector.
In order for CP violation to occur, there must exist two transition amplitudes to a nalstate f which interfere with nonzero relative phase. In the Standard Model, neutral D decay
CP -violating asymmetries from mixing are suppressed by the low mixing rate and are not
expected to be seen. However, direct CP -violating e ects in Cabibbo-suppressed modes,
due to interference between the suppressed tree amplitude and an unsuppressed penguin
amplitude, may be as high as a few times 10,3 [Buc93, Fry93]. Strong candidates for
searches include D+ ! K 0 K +, Ds+ ! K +0 decays, and D0 ! K + K ,. Limits set by
current experiments [Fra94] have not tested the ACP 10,3 range, and BABAR should be
able to make major contributions.
Technical Design Report for the BABAR Detector
90
Physics with BABAR
3.13
Tau Physics
Tau physics is now precision physics. There are no longer any obvious discrepancies between
experimental tau physics and Standard Model predictions: lepton universality has been
con rmed to better than 1%; world average exclusive branching fractions sum to within 1%
of 100%. Further progress in tau physics will come from high precision, detailed studies of
both leptonic and semihadronic tau decays, studies of tau spin structure, and searches for
rare or forbidden decays. Taus are still a rich eld of physics, and the potential still exists
for uncovering physics beyond the Standard Model.
The cross section for tau pairs at Ecm = 10:58 GeV is 0.91 nb, so an integrated luminosity
sample of 30 fb,1 will contain 27 106 tau pairs. (The current total world sample is on the
order of 6 106 tau pairs, dominated by CLEO.) With such a large sample, signi cant
contributions can be made to all of the topics discussed below. Many studies will be
dominated by systematic errors, so it is important to critically evaluate the strengths and
weaknesses of BABAR for addressing these issues.
Tau physics is more dicult at BABAR, than at LEP because: at BABAR there will be
signi cant backgrounds from qq events and two-photon physics, which require relatively
hard, mode-dependent cuts to isolate a signal; 0 s are softer and therefore more dicult to
detect reliably; and electrons and muons are softer and are thus more dicult to distinguish
from pions. However, the extremely large data sample will make it possible to trade statistics
for systematics, for example, in the use of clean tags such as ! e , so that improved
systematic errors can be achieved. More importantly, the scope of interesting physics goals
will be more focused; BABAR will concentrate on some important measurements that can be
done very well with its large data sample. Some of the topics that will be of interest in
the twenty- rst century, to which BABAR can make a signi cant contribution, are discussed
below.
Second-class currents (which violate isospin) have not yet been observed in tau decay.
The decay , ! , is predicted to proceed at the 10,5 level. The decay , !
! ,
occurs predominantly through the vector current; the decay via the axial-vector
current, which violates isospin, can be observed by analyzing the angular distribution.
Detailed studies of tau decays to three pseudoscalar mesons ( , K , and
KK ) will yield information on the Weiss-Zumino anomaly, isospin violation, SU (3)f
breaking, K1 mixing, and other aspects of these hadronic currents.
Detailed studies of the spectral functions in decays to 3 , 4 , 5 , and 6 , in
comparison with data from e+e, annihilation, permit precise tests of CVC and PCAC.
Technical Design Report for the BABAR Detector
3.14 Two-Photon Physics
91
Studies of the spectral functions, carefully separated into vector and axial vector
components, and strange and nonstrange components, permit tests of Weinberg, DMO,
and other sum rules.
Precision measurements of the Michel parameters , , , and in leptonic tau decays
and of the parity-violating neutrino helicity in semihadronic decays test the V{A
Lorentz structure of the , W , , coupling, and are thus sensitive to the presence of
scalar (charged Higgs) or V+A (WR ) currents. The spin-dependent Michel parameters
and can be measured by exploiting the well-understood (in QCD) correlations
between tau spins in e+ e, annihilation through the analysis of momentum and angle
correlations between the decay products of the two taus in the event.
One can search for CP violation in the leptonic sector in a variety of ways. A nonzero
electric dipole moment will increase the total e+e, ! + , cross section and angular
distribution and a ect the energy distribution of the , in the decay , ! , . It
will also produce nonzero expectation values for CP -odd correlation tensors between
the decay products of the two taus in the event.
In tau decays, nal states containing no neutrinos violate lepton family number. Limits
on branching fractions for these processes are at the few 10,6 level. Some models, with
massive Majorana neutrinos or mass-dependent couplings, predict branching ratios
near this level. It is thus of interest to push these limits down by another order of
magnitude.
The current best direct bound on the tau neutrino mass is around 25 MeV=c2 . Constraints from Big Bang nucleosynthesis and SN-1987A exclude long-lived Dirac tau
neutrinos below 15 keV. However, if the tau neutrino has a mass between 1 and
25 MeV=c2 and a lifetime in the range 103{106 seconds, it can evade nucleosynthesis
bounds, contribute to dark matter, and help explain how type II supernovae explode.
It is thus of great interest to search for a tau neutrino mass in this range.
3.14
Two-Photon Physics
Monte Carlo studies [Bau92] have shown that a high-luminosity B Factory will extend the
study of exclusive two-photon physics from the present 2 GeV=c2 up to at least 5 GeV=c2
in mass. This will likely reveal many of the exotic bound-states now being sought, such
as glueballs, qqg hybrids, and four-quark resonances. Measurements of the two-photon
couplings of light quark and charmonium resonances will continue to be very important
for understanding quark dynamics. Furthermore, a real challenge to perturbative QCD
predictions of exclusive hadron production from two-photon interactions can be made by
Technical Design Report for the BABAR Detector
92
Physics with BABAR
Untagged Reactions
Published Data Expected B Factory (100 fb,1)
2000
300,000
0
600
10
30000
10
800
Single-Tagged Reactions
Published Data Expected B Factory (100 fb,1)
300
6000
30
1100
1
3000
|
100
! 0 ! +,
! (1:42 GeV) ! KK
! c ! all modes
! + , (W > 3 GeV)
! 0 ! +,
++
! 1 (1:42 GeV=c2 ) ! KK
! c ! all modes
! + , (W > 3 GeV=c2 )
Comparison of current published event samples for two-photon physics with
the number of events expected at BABAR with 100 fb,1 of luminosity.
Table 3-20.
obtaining statistics in the > 3 GeV=c2 mass range. Finally, there is a strong incentive to study
single-tagged two-photon reactions (where one of the scattered es is detected) since, when
one of the photons is far from the mass shell, spin-1 resonances can be isolated (especially
J PC = 1,+ qqg hybrid states) and unique QCD predictions can be tested.
Table 3-20 summarizes Monte Carlo event rate predictions for 100 fb,1 of data at PEP-II,
along with present world average data samples. Clearly, at least a factor of 30 improvement
in statistics can be made in all of these exclusive processes. Furthermore, new channels
will be available, with a high-quality detector, which have not been explored in the past.
Most of this can be accomplished with little impact on the optimal detector for studying
CP violation. The main requirement is a very exible trigger which can allow detection
of low-mass and low-multiplicity nal states with balanced transverse momenta. However,
study of the single-tagged reactions require detection of low-angle scattered positrons in the
backward direction down to polar angles of 300 mr. Those topics will need to wait for a
detector upgrade which adds a backward calorimeter.
3.15 Summary
The variety of physics accessible at PEP-II demands a general-purpose detector, as is typical
in colliding beam experiments. The detector discussed here has been optimized for CP violation studies. However, since so many modes contribute to such studies, this process
automatically leads to a detector which can address a much wider range of physics issues.
The set of topics presented in this chapter is in no way intended to be an exhaustive list
Technical Design Report for the BABAR Detector
3.15 Summary
93
of all the physics studies possible with this detector, but rather an overview showing that
a rich program is expected. The BABAR detector will undertake all possible aspects of this
program.
Technical Design Report for the BABAR Detector
94
REFERENCES
References
[Abe94] F. Abe et al.,
73, 2667 (1994).
[Alb94b] H. Albrecht et al. (ARGUS Collaboration), Z. Phys. C62, 371 (1994).
[Ale91] R. Aleksan, I. Dunietz, B. Kayser, and F. LeDiberder,
B361, 161
(1991).
[Ale92] R. Aleksan, A. Gaidot, and G. Vasseur, \Probing the Penguin Contribution in
B ! at an Asymmetric B Meson Factory," DAPNIA/SPP 92{19 (1992).
[Ale93] R. Aleksan, L. Oliver, O. Pene, and J.C. Raynal,
B317, 123 (1993).
[Atw94] D. Atwood, B. Blok, and A. Soni, SLAC{PUB{6635 (1994).
[Bar90] V. Barger, C.S. Kim, and R.J.N. Phillips,
B251, 629 (1990).
[Bat93a] M. Battle et al.,
71, 3922 (1993).
[Bau92] D.A. Bauer, SLAC{400, 608 (1992); also SLAC{373, 173 (1992).
[Bro93] T.E. Browder et al., \A Review of Hadronic and Rare B Decays," CLNS 93/1261
(1993).
[Bro95] T.E. Browder and S. Pakvasa, University of Hawaii Report UH{511{814{95
(1995).
[Buc93] F. Buccella et al.,
B302, 319 (1993).
[CER93] GEANT Detector Description and Simulation Tool, Version 3.15, CERN Program
Library W5103, CERN (1993).
[Cha95] J. Chauveau and P. David, \Reconstruction of the Vertex in Lepton-tagged B
Mesons," A AR
# 224 (1995).
[Cou95] D. Coupal, \Study of B 0 ! D+D, and B 0 ! D+D, in A AR,"
A AR
# 211 (1995).
[Dea93] A. Deandrea, N. Di Bartolomeo, R. Gatto, and G. Nardulli,
B318,
549 (1993).
[Des94] N.G. Deshpande, X.G. He, and J. Trampetic, University of Oregon Report OITS{
564 (1994).
[Des95] N.G. Deshpande and X.G. He,
74, 26 (1995).
Phys. Rev. Lett.
Nucl. Phys.
Phys. Lett.
Phys. Lett.
Phys. Rev. Lett.
Phys. Lett.
B B
Note
B B
B B
Note
Phys. Lett.
Phys. Rev. Lett.
Technical Design Report for the
A AR
B B
Detector
REFERENCES
95
P.L. Frabetti et al.,
D50, 2953 (1994).
J.R. Fry and T. Ruf, \Prospects of Studying CP Violation and Mixing in D
Decays at a -charm Factory," Third Workshop on the -charm Factory, Marbella,
Spain (1993).
[Gol89] M. Golden and B. Grinstein,
B222, 501 (1989).
[Gol94] E. Golowich and S. Pakvasa, University of Massachusetts Report UMHEP-411
(1994).
[Gro90] M. Gronau and D. London,
65, 3381 (1990).
[Hal92] L. Hall and S. Weinberg,
D48, 979 (1992).
[Har95] P. Harrison, \Study of B ! J= KS0 , J= K and ," A AR
# 218 (1995).
[Inn93] W. Innes, \TRACKERR, A Program for Calculating Tracking Errors,"
A AR
# 121 (1993).
[Jad90] S. Jadach and Z. Was, \KORALB 2.1: An Upgrade with Tauola Library of Tau
Decays," Comput. Phys. Comm. 64, 267 (1991).
[Jaf94] D.E. Ja e, F. Le Diberder, M.H. Schune, and G. Wormser, \Treatment of
Weighted Events in a Likelihood Analysis of Bs Oscillations or CP Violation,"
A AR
# 132 (1994).
[Liu94] T. Liu, \The Future of High-Sensitivity Charm Experiments," in Proceedings
of the CHARM2000 Workshop, edited by D. Kaplan and S. Kwan (Fermilab,
Batavia, Illinois), pp 7{9 (1994).
[Nir93] For a Review of Standard Model Predictions on CP Violation in B Decays see
Y. Nir and H.R. Quinn, Ann. Rev. Nucl. Sci. 42, 211 (1992), and references
contained therein.
[PDG94] L. Montanet et al. (Particle Data Group),
D50 (1994).
[Pia94] M.G. Pia, \B Tagging with Muons," A AR
# 192 (1994).
[Pla95] S. Plaszczynski, M.-H. Schune, and G. Wormser, \An Improved Tagging for the
A AR Experiment," A AR
# 209 (1995).
[Qui93] H.R. Quinn and A. Snyder,
D48, 2139 (1993).
[Ric95] J.D. Richman and P.R. Burchat, \Lepton and Semileptonic Decays of Charm and
Bottom Hadrons," submitted to
[Fra94]
[Fry93]
Phys. Rev.
Phys. Lett.
Phys. Rev. Lett.
Phys. Rev.
B B
B B
Note
B B
Note
Note
Phys. Rev.
B B
B B
B B
Note
Note
Phys. Rev.
Rev. of Mod. Phys.
Technical Design Report for the
A AR
B B
Detector
96
[Sjo93]
[SLA89]
[Sjo93]
[Sny90]
REFERENCES
T. Sjostrand, \PYTHIA 5.7 and JETSET 7.4 Physics and Manual," CERN{TH
7112/93 (1993).
For earlier studies see SLAC{353 (1989), SLAC{400 (1992), and SLAC{419
(1993).
T. Sjostrand, \JETSET 7.4," CERN-TH 7111/93, (1993).
A. Snyder and S. Wagner, \An Alternate Method of B -Tagging for CP Violation
Studies," A AR
# 25 (1990).
A. Snyder, \CP Extraction Tool," A AR
# 188 (1994).
A. Snyder, \CP Reach Calculation of the TDR," A AR
# 219 (1995).
R. Waldi, \Flavor Tagging of B 0 Decays: Using All Available Information,"
A AR
# 190 (1994).
R. Waldi, \Flavor Tagging Studies for the A AR TDR," A AR
# 204
(1995).
M. Weaver, \Tag Mode Vertex Resolution Studies," A AR
# 227 (1995).
L. Wolfenstein,
D3, 2381 (1985).
D. Wright, \BEGET: The B Factory Event Generator," A AR
# 149
(1994).
D. Wright, \KL0 identi cation in B 0 ! J= KL0 Decays," A AR
# 201
(1994).
B B
[Sny94]
[Sny95]
[Wal94a]
B B
[Wea95]
[Wol85]
[Wri94a]
[Wri94b]
Note
B B
B B
[Wal95]
Note
Note
Note
B B
B B
B B
Note
Note
Phys. Rev.
Technical Design Report for the
A AR
B B
Detector
B B
Note
B B
Note
4
Vertex Detector
4.1 Vertex Detector Requirements
T
he major motivation for building BABAR is to nd and precisely measure CP violation
in the decays of neutral B mesons. To this end, one needs to determine for each event
the time interval between the two B -meson decays. This is accomplished by reconstructing
the two primary decay vertices, which is the main task of the vertex detector. For particles
with low transverse momenta (p < 100 MeV=c), which may not be reconstructed in the
drift chamber, the vertex detector should also provide full tracking information. Due to this
capability to perform stand-alone tracking, the BABAR vertex detector is known as the silicon
vertex tracker (SVT).
Below, we will discuss the requirements for the SVT and give a concise overview, including
performance studies. A detailed discussion of all design aspects completes this chapter.
t
4.1.1
Resolution
Without a measurement of the B decay vertex, no useful CP asymmetries can be extracted
at the (4S ). Therefore, the most important function of the SVT is the determination
of the B decay positions, especially along the beam direction. The measurement of CP
asymmetries does not place a very stringent requirement on the intrinsic position resolution;
there is less than a 10% loss in precision in the asymmetry measurement due to imperfect
vertex resolution if the separation between the B vertices z is measured with a resolution
of approximately one half the mean separation, which is 250 m at PEP-II [Sny94]. This
translates into a single vertex resolution of better than about 80 m both for CP eigenstates
and for tagging nal states. Vertex resolution in this range is readily achieved with silicon
strip detectors.
We can bene t from additional precision over that required for the measurement of z.
Pattern recognition, vertex reconstruction, and, in particular, background rejection will all
be improved by better intrinsic position resolution, which in turn will improve the eciency
98
Vertex Detector
and purity of the samples used for the CP asymmetry measurements. Therefore, our aim is to
achieve the best precision practical. Constraints which limit the achievable resolution include
the total number of readout electronics channels, mechanical constraints, and considerations
of cost and schedule. The multiple scattering in the beam pipe and in the silicon itself sets
a lower limit on the useful intrinsic resolution, corresponding to a point resolution of about
10{15 m for measurements made close to the IP and 30{40 m for the outer layers [For94].
The vertex detector also dominates in the determination of the track angles. The angle
measurements should be good enough that they do not signi cantly contribute to the
experimental uncertainty in the measurements of track impact parameters or the invariant
mass of combinations of tracks. This requirement does not impose a more stringent limit on
point resolution than that determined through consideration of the issues discussed above.
The tracking resolution needed for matching tracks to the particle identi cation system and
the calorimeter is mostly provided by the drift chamber.
4.1.2 Acceptance
The coverage of the SVT must be as complete as technically feasible, given the location of
the B1 magnets below 17.2 (300 mr) with respect to the beam line in both the forward and
backward directions. To maximize the space available for the SVT and its readout electronics
in the boost direction, machine components such as cooling manifolds and vacuum anges
are located in the backward region. Therefore, it is only possible to extend the backward
coverage to within 30 of the beam line. In the (4S ) center-of-mass, this corresponds to
,0:95 < cos cm < 0:87.
There should be as little material as possible within the active tracking volume. Special
attention must be paid to minimizing the amount of material between the IP and the
rst measurement in order to reduce multiple scattering. The beam pipe itself contributes
0.006X0 at normal incidence. Material located beyond the inner layers does not signi cantly
degrade the measurement of track impact parameters, but does a ect the performance of
the overall tracking system and leads to increased photon conversions in the active region.
4.1.3 Eciency
Our goal is to achieve close-to-perfect track reconstruction eciency within the active volume
of the tracking detectors when information from both the drift chamber and the SVT is used.
The pattern recognition capabilities of the combined tracking system must be robust enough
to tolerate background levels up to 10 times nominal, where nominal is de ned for a 1 nTorr
vacuum pressure in the beam pipe within 30 m of the IP. Low momentum particles that do
Technical Design Report for the BABAR Detector
4.1 Vertex Detector Requirements
99
not traverse many drift chamber planes, such as many of the charged pions from D decays,
must be reconstructed in the SVT alone. For this category of tracks, with pt less than
100 MeV=c, we want to achieve reconstruction eciencies of at least 80{90%. The SVT must
also be ecient for particles such as KS0s that decay within the active volume. Together,
these determine the number of measurements along a track and the necessary single-hit
eciency.
The impact parameter resolutions are determined by the precision of the measurement
closest to the IP. The performance of the inner vertex layers must therefore be optimized
for both high eciency and good point resolution. It follows that redundancy for the rst
measurement is an important design requirement.
4.1.4
Radiation Tolerance
The expected machine-related backgrounds set the requirements for the data transmission
bandwidth and radiation resistance of all components located close to the interaction region [Lev94]. As described in Chapter 12, the expected dose in the innermost layer of the
SVT averages about 33 krad/yr. The radiation is highly nonuniform in azimuth, peaking in
the bend plane of the accelerator with a local maximum of up to 240 krad/yr over a small
region covering approximately 6 in azimuth. At the location of the front-end electronics for
Layer 1 the maximum dose is 110 krad/yr, with an average value of 47 krad/yr. Detector
and front-end electronics are speci ed to be able to withstand at least 10 times the annual
radiation dose. The readout electronics must be fabricated with radiation-hard technologies,
and special attention must be paid to the sensitivity of the detector performance to radiation.
4.1.5
Reliability
The SVT is mounted inside a support tube of radius 20 cm, which also supports and
aligns the machine elements closest to the IP. Access to the SVT is not possible without
a major shutdown involving removal of the support tube from the detector. The reliability
requirements for the SVT are therefore more stringent than usual for such a device, with
implications for engineering design at all levels. The detector layout must provide redundant
measurements wherever possible; the electronic readout must be very robust and have
redundancy built in for critical data and control lines; and the functionality of all components
must not be compromised by exposure to the expected radiation levels. The detector
monitoring and interlock system must serve as a safeguard against catastrophic failure in
the event of a component malfunction or a simple human error.
Technical Design Report for the BABAR Detector
100
4.2
4.2.1
Vertex Detector
Vertex Detector Overview
Choice of Technology
The SVT design is based on double-sided silicon microstrip detectors. The characteristics of
this technology that make it attractive for the BABAR detector are: high precision for measuring the location of charged particles, tolerance to high background levels, and reduction in
mass made possible through double-sided readout. The process for the fabrication of doublesided silicon detectors is now mature enough to be employed in a large-scale application and
to meet the performance standards outlined above.
4.2.2
Detector Layout
The SVT will provide ve measurements, in each of two orthogonal directions, of the positions
of all charged particles with polar angles in the region 17:2 < < 150. A three-dimensional
cut-away view of the SVT is shown in Figure 4-1. Each of the three inner layers has six
detector modules, arrayed azimuthally around the beam pipe, while the outer two layers
consist of 16 and 18 detector modules, respectively. A side view of the detector is shown in
Figure 4-2, and an end view is shown in Figure 4-3.
The inner detector modules are traditional barrel-style structures, while the outer detector
modules employ a novel arch structure in which the detectors are electrically connected
across an angle. The bends in the arch modules minimize the area of silicon required to
cover the solid angle and also avoid very large track incident angles.
In order to satisfy the requirement of minimizing material in the detector acceptance region,
one of the main features of the SVT design is the mounting of the readout electronics entirely
outside the active detector volume. There is a 1 cm space between the 300 mr stay-clear and
the B1 magnet in the forward direction; all of the forward electronics are mounted here. In
the backward direction, there is space below about 500 mr. In both directions, space is very
tight, and the electronic and mechanical designs are closely coupled in the narrow region
available.
The layout speci cations for this ve-layer design are given in Table 4-1. The strips on the
two sides of the rectangular detectors in the barrel regions are oriented parallel ( strips)
or perpendicular (z strips) to the beam line. In the forward and backward regions of the
two outer layers, the angle between the strips on the two sides of the trapezoidal detectors
is approximately 90, and the strips are tapered. Floating strips are used to improve
the position resolution for near-perpendicular angles of incidence; the capacitive coupling
Technical Design Report for the BABAR Detector
4.2 Vertex Detector Overview
Figure 4-1.
Figure 4-2.
101
Three-dimensional cutaway view of the SVT.
Cross-sectional view of the SVT in a plane containing the beam axis.
Technical Design Report for the BABAR Detector
102
Vertex Detector
Figure 4-3. Cross-sectional view of the SVT in a plane perpendicular to the beam axis.
The lines perpendicular to the detectors represent structural beams.
between the oating strip and the neighboring strips results in increased charge sharing
and better interpolation. For larger incident angles, the wider readout pitch minimizes the
degradation in resolution that occurs because of the limited track path length associated
with each strip. These issues are discussed in more detail in Section 4.3.
The design has a total of 340 silicon detectors of seven di erent types. The total silicon area
in the SVT is 0.94 m2 , and the number of readout channels is 150,000.
4.2.3
Electronic Readout
As emphasized above, all readout electronics are located outside the active volume, below
300 mr in the forward direction and below about 500 mr in the backward region. To accomplish this, strips on the forward or backward half of a detector module are electrically
connected with wire bonds. This results in total strip lengths associated with a single readout
channel of up to 14 cm in the inner three layers and up to 24 cm in the outer two layers.
Technical Design Report for the BABAR Detector
4.2 Vertex Detector Overview
Quantity
Radius (mm)
Wafers/Module
Modules/Layer
Silicon Area (cm2)
Overlap in (%)
Readout pitch (m):
103
Layer Layer Layer
1
2
3
32
40
54
4
4
6
6
6
6
457 683 1072
2.4
1.8
1.8
Layer
4a
120
7
8
1506
4.0
Layer
4b
127
7
8
1582
4.0
Layer
5a
140
8
9
2039
2.0
Layer
5b
144
8
9
2082
2.0
z
50
100
50
100
50
100
65{100
200
65{100
200
z
|
1
|
1
|
1
1
1
1
1
Floating Strips:
Intrinsic
Resolution(m):
10
10
10
10{12
10{12
12
12
12
25
25
R.O. Section/Module 4
4
4
4
4
ICs/R.O. Section
6
8
10
4
4
Readout Channels
18432 24576 30720
32768
36864
Strip Length (mm):
95
115 136 177/223 186/232 232/241 241
z
39
48
64
35{52 35{52 35{52 35{52
z Ganging:
Forward 2
34% 18% 7%
82%
88%
71%
60%
Forward 3
29%
40%
Backward 2
34% 18% 7%
82%
71%
60%
60%
Backward 3
18%
29%
40%
40%
z
Parameters of the SVT layout. See text for more detail on the meanings of the
di erent quantities. The intrinsic resolution is calculated at 90 track incidence assuming
S=N = 20 : 1. The z -ganging numbers represent the percentage of detector strips connected
to one other strip (2) or two other strips (3).
Table 4-1.
Technical Design Report for the BABAR Detector
104
Vertex Detector
The signals from the z strips are brought to the readout electronics using fanout circuits
consisting of conductive traces on a thin exible insulator (for example, copper traces on
Kapton). The traces are wire-bonded to the ends of the z strips. To read out all the z strips
with the same number of electronics channels as the strips, some z strips will be electrically
connected, or ganged, together. The length of the z strips is much shorter, typically 5 cm
in the inner layers and either 10 or 15 cm in the outer layers where there is either 2 or 3
ganging.
Front-end signal processing is performed by ICs mounted on hybrid circuits that distribute
power and signals, and thermally interface the ICs to the cooling system. The signals from the
readout strips, after ampli cation and shaping, are compared to a preset threshold. The time
interval during which they exceed the threshold (time over threshold, or TOT) is an analog
variable related to the charge induced on the strip. This time interval is digitally recorded
prior to readout. Unlike the ordinary peak-amplitude measurement at the shaper output,
the TOT technique has a nonlinear input-to-output relationship which is approximately
logarithmic. This is an advantage since it compresses the dynamic range and allows one to
achieve good position resolution and large dynamic range with a minimum number of bits.
TOT readout is less complicated than a linear analog readout system, resulting in less
development time, lower cost, and smaller space and power requirements. At the same
time, it is a signi cant improvement over a purely digital (hit/no-hit) readout, and provides
sucient analog resolution for position interpolation, time-walk correction, and background
rejection. The readout IC is expected to be about 8{9 mm long and to dissipate no more
than 2.0 mW per channel. The total power that will be generated by the SVT readout is
300 W.
For each channel with a signal above threshold, the strip number within the readout module,
the time of arrival of the signal, and the digital value of the TOT will be read out. There
are four readout sections per detector module, where the module is divided in half along z,
and the and z strips are grouped together separately. The data from one-half of a detector
module will be transmitted from the hybrid on a exible cable to a transition card located
approximately 40 cm away, where the signals are converted to transmission via conventional
cables.
4.2.4
Mechanical Support
The silicon detectors and the associated readout electronics are assembled into mechanical
units called detector modules. Each module contains several silicon detectors glued to lowmass beams constructed of carbon and Kevlar ber-epoxy laminates. The beams are attached
to the hybrid electronic circuits at each end. A ceramic or aluminum substrate for the hybrid
provides precise mechanical mounting surfaces and is the heat sink for the electronics.
Technical Design Report for the BABAR Detector
4.3 Detector Performance Studies
105
The inner layer is supported from the second layer; the detector modules from Layers 1 and
2 are glued together with rigid beams, forming sextants which are then mounted from the
support cones in the forward and rear directions. Each detector module of the third, fourth,
and fth layer is mounted on the support cones independently of the other modules. In the
fourth and fth layers, there are two di erent types of modules in each layer, an inner one,
labeled a, and an outer one, labeled b, occupying slightly di erent radial positions. Thus
there are seven di erent types of detector modules.
The support cones are double-layered carbon- ber structures which are mounted from the
B1 magnets. Cooling water ows between the two carbon- ber layers around aluminum
mounts which protrude through the outer surface of the cone. Mounting pins in the hybrid
structure provide the alignment between the modules and the aluminum mounts in the cone,
and thermal contact is made to provide cooling for the front-end electronics located on the
hybrid. The support cones are divided to allow the vertex detector to be assembled in two
halves and then mounted on the B1 magnets by clam-shelling the pieces together.
The sti ness of the overall structure is provided by a very low mass space frame, constructed
of carbon- ber tubes, connecting the forward and backward support cones. It consists of
rings at each end held rigid by 12 struts spanning the length of the detector. The rings are
connected to the support cones by an additional series of 12 struts at each end. All material
is carbon- ber laminate. Preliminary nite element calculations show that this structure
meets the tolerances for rigidity. The motivation for this space frame stems mainly from
the possible relative motion of the two B1 magnets during the assembly procedure. Cooling
water, power, and signal lines are routed along the B1 magnets to points outside the active
region where manifolds for the cooling water and drivers for the electronics are located.
4.3 Detector Performance Studies
Cost, engineering complexity, space, and cooling requirements each constrain performance.
In this section, we discuss how the SVT performance has been optimized within these
constraints. Design and performance studies have considered the following:
the impact of the number and locations of detector layers on track parameter resolution
and pattern recognition;
the e ect of di erent readout schemes for recording charge information on intrinsic
resolution, background rejection, and ultimately, on pattern recognition; and
optimization of the readout pitch, taking into account the required resolution, the
number of readout channels, and, in the case of the outer layers, the signal loss that
occurs when the readout pitch becomes very large.
Technical Design Report for the BABAR Detector
106
4.3.1
Vertex Detector
Resolution
Intrinsic Resolution
We have studied the e ects on intrinsic position resolution of the strip pitch, readout
pitch, and threshold levels, as well as various schemes for recording pulse information. The
deposition of charge in a silicon strip detector was simulated by a Monte Carlo program,
taking into account the e ects of Landau uctuations in energy loss, di usion, channelto-channel gain variations (5%), and noise. The simulation is in good agreement with
experimental data on the energy loss in silicon, and with experimental measurements of
position resolution in strip detectors.
The intrinsic position resolution was studied as a function of the track incident angle in
the plane normal to the strips. Charged particles most often cross the detectors at close to
normal incidence in the plane perpendicular to the strips; however, the angle of incidence
of tracks in the plane perpendicular to the z strips extends to 300 mr from the beam line,
or = =2 , 0:3 1:3 rad, where is the dip angle measured relative to normal incidence.
The resolution degrades signi cantly at large dip angles, especially for small readout pitch.
This is due to ineciency in the readout for strips with small signals deposited by tracks at
grazing angles.
Figure 4-4(a) shows the intrinsic resolution for 300 m-thick silicon as a function of the
incident angle for four di erent strip and readout pitches. Strip and readout pitch coincide
in the absence of oating strips; oating strips, when present, increase capacitive chargesharing and thereby improve the point resolution. The simulation assumed a noise level of
1200 e,, corresponding to a signal-to-noise ratio of about 20 for perpendicular tracks, and
the threshold was set to 4 times the noise. The algorithm employed to determine the position
uses the digital centroid for the central strips and applies a correction based on the relative
amount of charge in the two edge strips of a cluster:
X = X0 + 0:5 p
Q1 , QN
;
Q1 + QN
where X0 is the digital centroid (i.e., geometrical center) of strips 2 through N , 1; Q1 (QN )
is the charge deposited on the rst(last) hit strip, and p is the readout pitch.
As seen in Figure 4-4(a), the resolution degrades signi cantly as the angle of incidence
increases for 50 m strip and readout pitch. With 100 m strip and readout pitch, the
resolution does not degrade as much with larger incident angles but, as expected, the
resolution at normal incidence is about a factor of 2 worse than that for 50 m pitch. On
the other hand, 100 m readout pitch with one oating strip gives a resolution equivalent
to 50 m readout pitch without oating strips for angles smaller than 45( < 0:8). Above
45, it is equivalent to 100 m strip pitch. Therefore, oating strips allow us to retain
Technical Design Report for the BABAR Detector
4.3 Detector Performance Studies
107
(a) Intrinsic resolution for a 300 m-thick silicon detector as a function of the
dip angle for various strip con gurations, assuming perfect analog readout. (b) Same as as
(a), assuming 100 m readout pitch and one oating strip, for various readout techniques.
Figure 4-4.
Technical Design Report for the BABAR Detector
108
Vertex Detector
good resolution for small incident angles with fewer readout channels and less degradation
in resolution for large incident angles. The upper curve in Figure 4-4(a) shows that for
a 200 m readout pitch with one oating strip, the resolution is about 25 m at normal
incidence. This con guration is used for the z strips in layers 4 and 5.
We have also studied the e ect of the nonlinearity and quantization error of a TOT scheme
compared both to an ideal analog readout and to a digital readout in which no information
is recorded about the signal size [Roe94b]. The position is determined using the algorithm
described above, but using the TOT, rather than the charge, to weight the edge strips.
The resolution is shown as a function of track incident angles in Figure 4-4(b). The signal
shaping was assumed to have a peaking time 2.5 times longer than the period of the clock
used in the digitization of the pulse length for this simulation. This value is consistent with
that expected in the inner vertex layers, where fast shaping is necessary to cope with larger
backgrounds. The resolution for a TOT readout scheme is practically the same as ideal
analog readout. Digital readout, which provides hit/no-hit information only, shows much
poorer resolution in comparison.
In addition to the advantage of improved resolution, analog readout may be useful for
background rejection of non-MIP signals, for pattern recognition techniques which correlate
pulse height with angle, and for correction of the time-walk which will a ect the time-stamp
information generated by the readout chip.
Track Parameter Resolution
The track parameter resolutions described in Chapter 2 were calculated using an analytical
technique based on the method of Billoir [Bil84] which can be applied to arbitrary detector
geometries in the program TRACKERR [Inn93]. Figure 4-5 shows the resolutions for the
track impact parameter along the beam (z ) and in the plane perpendicular to the beam
(xy ) for various momenta. The resolution is dominated by multiple scattering and not by
the intrinsic position resolution of the silicon detectors, except for the highest momentum
tracks from B decays.
A Monte Carlo analysis was performed to determine how the intrinsic resolution and the
number of detector layers a ect the track parameter resolution. The study included the
e ect of a catastrophic failure in Layer 1, the most important for vertex resolution. The
intrinsic resolution was varied on the rst two layers in the range 5{25 m and on the outer
two layers in the range 10{50 m. The point resolution for Layer 3 was kept xed at 10 m.
The quoted resolutions are for normally incident tracks; these were degraded as the dip angle
increased, so that = 0 (1 + 1:5).
Technical Design Report for the BABAR Detector
4.3 Detector Performance Studies
109
(a) Impact parameter resolution along the beam (z ) and (b) in the plane
perpendicular to the beam (xy ) vs. dip angle for three di erent momenta: 100 MeV=c (solid
line), 1:0 GeV=c (dotted line), and 3:0 GeV=c (dashed line).
Figure 4-5.
Technical Design Report for the BABAR Detector
110
Vertex Detector
In addition to the baseline geometry with ve layers, we considered a four-layer layout
which was obtained by removing the third layer. The basic conclusions of these studies are
summarized as follows (see Reference [For94] for a full discussion):
The angle and momentum resolution are essentially independent of the intrinsic resolution.
The z impact parameter, on the other hand, does depend on the intrinsic resolution as
shown in Figure 4-6(a) (the xy impact parameter has a similar behavior). In changing
the intrinsic resolution on Layers 1 and 2 from 10 to 20 m, respectively, z degrades
by about 15{20% for 1 GeV=c and by almost 40% for 3 GeV=c tracks. For soft tracks
(< 100 MeV=c), no dependence on the intrinsic resolution is observed because multiple
scattering dominates. Most tracks produced in BABAR will be soft (< 500 MeV=c);
B vertices, however, will be determined by the hardest tracks in the event.
The requirement that the track parameter resolution be dominated by multiple scattering for most tracks therefore leads us to choose the best intrinsic resolution technically feasible in Layers 1 and 2. Within the constraints imposed by the mechanical,
electronic, and silicon detector performance, the best intrinsic resolution achievable
in the inner layers is 10 m in (employing 50 m readout pitch) and 12 m in
z (employing 100 m pitch with one oating strip). A larger pitch in z reduces the
amount of ganging required and provides good resolution for tracks at small crossing
angles.
If for some reason Layer 1 is not usable, the tracking resolution is degraded, as shown
in Figure 4-6(a). The degradation in the z-impact parameter resolution is 19% at
3 GeV=c and 22% at 1 GeV=c; thus, Layer 2 is an e ective backup in case of a failure in
Layer 1.
The resolution on Layers 4 and 5 does not signi cantly a ect any of the track parameters (Fig 4-6(b)), and the choice of the readout pitch in the outer layers is determined
by other considerations (i.e., noise, occupancy, and detector design considerations).
Given the relatively soft momentum spectrum expected in BABAR, we were concerned
that the material in the third layer (whose main motivation is pattern recognition)
could degrade the tracking resolution. TRACKERR studies have shown [For94] that
the e ect of removing Layer 3 is in fact small (at most 5%) for high momentum tracks
and negligible for soft tracks.
Technical Design Report for the BABAR Detector
4.3 Detector Performance Studies
111
parameter resolution as a function of the point resolution (a) in
Layers 1 and 2 for the baseline design and for the case of a Layer 1 failure, and (b) in
Layers 4 and 5. All values are at 90 deg for 1 GeV=c and 3 GeV=c tracks.
Figure 4-6.
z -impact
Vertex Resolution
The z vertex resolution has been calculated using a parameterized Monte Carlo simulation
for three B decay modes: B 0 ! +,, B 0 ! J= KS0, and B 0 ! D+D,. For each decay, the
distribution of the di erence between the reconstructed and generated B 0 vertex positions
in the z direction is t with a sum of two Gaussian functions. The results are shown in
Table 4-2. The results for B ! +, have been checked with a full GEANT-based Monte
Carlo simulation, which gave reasonable agreement for both n and w . In all cases, the
expected rms resolution (obtained by taking the weighted sum of n and w ) is better than
the 80 m required for CP violation studies.
The z vertex resolution for the tagging B has also been studied with both parameterized
and GEANT-based Monte Carlo simulations. The resolution depends upon which type of
tag one employs. For direct leptons, we nd n 60 m, while for kaon tags, n 75 m,
with about 70{80% of the tags contained in the narrow Gaussian.
Good vertex resolution in the plane perpendicular to the beam pipe is also important. This
information will help in associating tracks from B and D decays to the correct secondary
or tertiary vertices, thus reducing combinatorial backgrounds. Figure 4-7(a) shows the
distribution in the xy plane of the distance between two D vertices in the decay B 0 ! D+D,.
Technical Design Report for the BABAR Detector
112
Vertex Detector
Mode
+,
Ks
+
D D,
n( m)
27:5 0:4
42:1 0:7
46:5 0:6
w ( m) fn (%)
rms
85:8 3 84:0 0:4 36:8 0:6
168 7 88:0 0:2 57 1:0
159 10 94:0 0:1 53:3 0:8
A two-Gaussian t to the z vertex resolution for various CP eigenstates; n
and w are the rms of the narrow and wide Gaussians, respectively, and fn is the fraction
of the area contained in the narrow Gaussian. The rms is the weighted sum of n and w .
Table 4-2.
500
300
250
400
200
300
150
200
100
100
50
0
0
0.05
0.1
0.15
(a)
0.2
0.25
-2
x 10
0
-0.5
-0.4
-0.3
m
-0.2
-0.1
0
(b)
0.1
0.2
0.3
0.4
0.5
-3
x 10
m
(a) Separation between D vertices in the xy plane. (b) Resolution on the
separation between the D vertices in the x-y plane. Note the di erent scales on the abscissa.
Figure 4-7.
The distribution is t to a function of the form A xe,x= and yields a value of = 276 m.
For comparison, the expected resolution on the separation between the two D decay points
in the x-y plane is plotted in Figure 4-7(b) for the decay D ! K. The rms resolution is
about 125 m.
4.3.2
Pattern Recognition
The pattern recognition capabilities of the BABAR vertex detector have been studied using a
GEANT simulation with a detailed model of the detector geometry. The simulation includes
Technical Design Report for the BABAR Detector
113
1
ε
ε
4.3 Detector Performance Studies
1
(a)
(b)
0.9
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.1
0
0.2
5 Layers, 95% Hit Efficiency
5 Layers, 90% Hit Efficiency
4 Layers, 95% Hit Efficiency
4 Layers, 90% Hit Efficiency
5
6
7
8
5 Layers, 95% Hit Efficiency
5 Layers, 90% Hit Efficiency
4 Layers, 95% Hit Efficiency
4 Layers, 90% Hit Efficiency
0.1
9
10
11
Number of Hits Found in Silicon
0
5
6
7
8
9
10
11
Number of Hits Found in Silicon
Figure 4-8. Track nding eciency for tracks from generic B decays: (a) all tracks,
(b) tracks measured in the SVT only (pt < 90 MeV=c).
detector overlaps, inactive regions near the edges of the wafers, strip ineciency, background
hits, and the e ects of ganging the z strips. A method based on Kalman ltering techniques
was used to locate tracks in the SVT. For tracks that went a sucient distance into the drift
chamber (corresponding to pt > 90 MeV=c), the results of a t to the drift chamber track
segment were used as a starting point for the silicon track nding. Hits in the SVT that
were best associated with each of these track segments were added to the track. Once this
was nished, these hits were removed from the SVT and a vertex-only pattern recognition
algorithm was used on the remaining hits to nd low-pt tracks. This vertex-only algorithm
was based on the same Kalman ltering technique but used combinations of hits in the outer
layers of the SVT to generate the initial track parameters.
All tracks were required to cross ve layers of silicon. The eciency for nding tracks was
computed by requiring a minimum number of silicon hits per track and by requiring each
track t to pass a fairly loose 2 cut (2 =Ndof < 10). A single hit in the silicon was de ned
as either a or a z hit, so that a track going through ve layers of silicon would generate ten
hits. Six silicon hits was considered to be the minimum required in order for the number of
degrees of freedom for a t to the silicon-only tracks to be 1. The track nding eciency
for all tracks from generic B decays is plotted in Figure 4-8(a) as a function of the number
of hits required. For this simulation, two geometries were studied: the baseline ve-layer
SVT, and a four-layer version in which Layer 3 was removed. The single hit eciency was
set to either 90% or 95%, and additional hits from beam-induced backgrounds at a level
corresponding to 10 times nominal were overlayed on the hits from the B 0 B 0 event. The
Technical Design Report for the BABAR Detector
114
Vertex Detector
eciency for the ve-layer geometry remains high for up to eight required hits per track,
while for the four-layer device, the eciency is high only for the six-hit requirement.
In Figure 4-8(b), the eciency is plotted for the same two geometries, but only for tracks
which do not traverse enough layers in the drift chamber to be reliably found there. The
overall eciency is somewhat lower, and the same general behavior is observed, namely, a
plateau for six, seven, or eight required hits in the ve-layer detector and a sharply falling
eciency for more than six required hits in the four-layer version. The algorithm employed
for nding tracks using the vertex detector only is quite sensitive to the detector eciency, as
seen in the gure. For comparison, in a perfect detector with 100% eciency, the silicon-only
track- nding eciency would be 95% for the ve-layer detector when requiring seven hits.
Both the silicon-only and the combined silicon plus drift chamber track- nding eciencies
are quite insensitive to the background level.
The frequency at which fake tracks are found is also important to quantify. In this study, a
fake track was de ned as a track which had less than half of its hits from a single real track.
A loose impact parameter cut was imposed, rejecting any tracks originating more than 5 cm
from the nominal beamspot in z or more than 2 cm in xy. The average number of fake tracks
per event is shown in Figure 4-9 as a function of the number of hits found in the silicon. This
measurement was also performed assuming a background level that was 10 times nominal. In
order to achieve reasonable eciency for good tracks in the four-layer geometry, one cannot
ask for more than six hits to be found per track. Therefore, the relevant rate in the four-layer
geometry is 0.13 fake tracks per event. The ve-layer geometry provides a relative reduction
in the fake rate of a factor of 3 if one requires seven or eight hits on a track, without any
signi cant reduction in track- nding eciency. Approximately half of the fake tracks have
p (as measured in the silicon) large enough to generate a corresponding track segment in
the drift chamber. Most of these fakes should therefore disappear when attempts are made
to match them to drift chamber hits.
t
4.3.3 Solid Angle Coverage
As discussed above, the SVT must have high eciency over the region 17:2 < < 150.
The stay-clear at 17:2 (300 mr) in the forward region does not allow greater coverage, and
beam-line components in the backward region prevent extension below 500 mr.
The geometry of the SVT in the inner layers is hexagonal, so the solid angle coverage varies
as a function of azimuth [Roe94a]. To meet the solid angle coverage requirements, the rst
two layers are extended along +z so that a track emitted at 300 mr from z = 0, at any
azimuthal angle, is in the SVT acceptance. There is then some coverage down to 245 mr for
tracks in certain azimuthal regions.
Technical Design Report for the BABAR Detector
115
Avg. Fakes per Event
4.4 Silicon Detectors
0.2
5 Layers, 95% Hit Efficiency
4 Layers, 95% Hit Efficiency
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
Figure 4-9.
5
6
7
8
9
10
11
Number of Hits Found in Silicon
Fake track rate in the SVT at 10 times nominal background rate.
In the outer layers it is not possible to extend the detectors beyond the 300 mr stay clear.
However there are many more staves per layer, and coverage extends to between 300 and
310 mr everywhere.
In the backward direction, there is a similar variation of coverage with azimuthal angles. At
the minimum radii of the inner layers there is nominal coverage to within 30 of the beam
line. In the center-of-mass, the SVT covers ,0:95 < cos cm < 0:87, so the acceptance in the
backward direction is actually greater than that in the forward direction in this frame.
4.4
Silicon Detectors
The SVT will be constructed from double-sided, AC-coupled silicon strip detectors. These
solid state devices are a technically mature solution to the requirements the SVT must meet
to provide precise, highly segmented, robust tracking near the interaction point. The detailed
requirements which the detectors must meet are discussed below.
4.4.1 Requirements
Readout Strip Eciency.
The silicon detectors must maintain high single-point eciency in order to achieve the requirements given in Section 4.1 for high overall track
reconstruction eciency and good tracking resolution. Loss of eciency can occur from
Technical Design Report for the BABAR Detector
116
Vertex Detector
intrinsic strip ineciencies, from bad interconnections, or from faulty electronics channels.
Intrinsic strip ineciencies can occur due to production defects which result in strips with
unacceptably large leakage currents, from accidents during assembly causing the strip to be
physically damaged, or from a breakdown in the AC-coupling capacitor. The latter problem
is referred to as a pinhole and is due to a small hole in the oxide separating the implant
from the metal readout strip above it. Pinholes can occur during fabrication, or they can be
generated later on. Understanding and controlling the level of pinholes is one of the primary
concerns in our program of silicon detector R&D.
Our goal is to achieve an overall single detector strip failure rate of less than 1%. Data from
a large production of double-sided DC-coupled detectors (ALEPH) show that 60{70% can
be achieved with a maximum ineciency of 1%. On this basis, we expect that a 50% yield
can be achieved for double-sided AC-coupled detectors while maintaining similar standards.
Point Resolution. As described in Section 4.3, we have determined from Monte Carlo
simulations [For94] that the intrinsic point resolution should be 15 m or better in both z and
for the inner layers. These are the point resolutions for tracks at near-normal incidence.
As the angle between the track and the plane normal to the strip increases, the resolution
degrades. We require the resolution to degrade by no more than a factor of approximately
3 for angles up to 75 ( 1:3) from normal.
Radiation Hardness. A further requirement is that the quoted resolution values hold up
to an integrated dose of 2 Mrad of ionizing radiation (electromagnetic in origin). This requirement leads to the use of AC-coupled detectors in order to avoid the problems associated
with direct coupling of the large leakage currents which can occur at such large doses. It
also has implications in the choice of the biasing scheme.
Minimum Mass. To achieve good vertex resolution, it is especially important to minimize
the material up to and including the rst measurement. This requirement, and the need to
provide precise vertexing in both z and , leads to the choice of double-sided detectors. We
plan to use 300 m-thick silicon wafers, which are a standard choice and present acceptable
handling properties. While it may be possible to go to 200 m-thick silicon, the gain is only
by the square-root of the total material before the rst measurement (including the 0.006X0
in the beam pipe), while the fabrication and handling will be much more problematic.
Technical Design Report for the BABAR Detector
4.4 Silicon Detectors
4.4.2
117
Silicon Detector Design
From the above requirements and from the discussion in Sections 4.1{4.3, we have arrived
at the detector speci cations and design parameters which are described in this section. A
more complete discussion can be found in Reference [Bat94b].
Substrate and implant type. The wafers will be n-type, with a resistivity in the range
4{8 k cm, corresponding to a depletion voltage of 40 to 80 V. These values seem to be a
reasonable compromise between the need to have a low depletion voltage and the need to
avoid type inversion in the presence of radiation damage.
We will employ p+ strips on the junction side and n+ strips on the ohmic side, with p+blocking implants in between; see Figure 4-10 for a cross-sectional view. This choice has
proven to be a reliable technology [Bat94a] which is directly available without extensive
R&D.
Coupling to preampli er. The strips are connected to the preampli ers through a
decoupling capacitor. AC coupling prevents the ampli er from integrating the leakage
current with the signal; handling high leakage currents due to radiation damage imposes
an additional burden on the preampli er design and has other undesirable operational
implications.
The value of the decoupling capacitance must be much larger than the total strip capacitance,
which is as large as 35 pF. It is possible to use DC-coupled detectors with external capacitors
on a separate chip [Bat92]; however, with an electronics pitch of 50 m, we would need a
12 mm-long chip to get a CAC = 180 pF decoupling capacitor.
These dimensions are
prohibitively large given the limited space available in the hybrid region. There would also
be a signi cant increase in the number of wirebonds. Capacitors which are integrated on the
detectors are the most compact solution, minimizing the number of wirebonds and yielding
a value for the decoupling capacitance of 30{80 pF/cm [Ton94], depending on the implant
width.
Bias resistors. The bias resistors must be between 4 and 20 M . The lower limit is
determined by two factors. The noise has a 1= RB dependence, and if several strips are
ganged together, the e ective resistance is correspondingly decreased. Another factor is the
requirement that, for oating strips, the product RB CTOT must be much larger than the
ampli er peaking time (100{400 ns) to allow for capacitive charge partition. The upper limit
(20 M ) is dictated by the allowable potential drop due to the strip leakage current, which
p
Technical Design Report for the BABAR Detector
118
Vertex Detector
Resistive
materials
Metal (Al)
Polysilicon
Dopings in
Silicon
Insulating
materials
n- bulk
Silicon dioxide
p+
Silicon nitride
implants
n+ implants
p
W
P - SIDE
p
W
p stops
N - SIDE
Figure 4-10.
Artist's conception of a silicon detector cross section.
Technical Design Report for the BABAR Detector
4.4 Silicon Detectors
119
is taken to be 100 nA at maximum. A good target value is 8 M . A nal requirement is that
the bias resistor be quite stable for the expected radiation doses.
To meet these requirements, we plan to use polysilicon bias resistors. In prototype CMS
detectors [Ton94], values for the sheet resistance of polysilicon of 40 k = were achieved,
and 50 k = is feasible. Thus, it is possible to fabricate an 8 M resistor with a 6 m-wide,
960 m-long polysilicon resistor. With a suitable shaping of the polysilicon line, the space
required by the resistor will be 480 m for a 25 m implant pitch.
For the junction side, punch-through biasing [Hol89] could also be chosen if it is con rmed
to be suciently radiation hard. Punch-through biasing requires less space, and it can also
reduce the parallel noise on the junction side, since the dynamic resistance is much higher
at low currents (70 M at 1 nA, decreasing to a few M at 100 nA).
Considering the space needed to accommodate the biasing resistors and to gracefully degrade
the electric eld close to the edge with a guard ring structure, we specify the active region
of the detectors to be 1.4 mm smaller than the physical dimensions (700 m on each edge).
Optimization of z and readout strips. A major issue is which side of the detector
(junction or ohmic) should read which coordinate (z or ). The capacitance, and consequently, the noise is smaller on the junction side than on the ohmic side, and the strip pitch
on the junction side can be 25 m, while on the ohmic side, it is limited to about 50 m
because of the p-stop implant. For these reasons and because the z vertex measurement is
more important from the point of view of physics, we use the junction side for the z strips
on the inner layers. The better performance of the junction side also helps compensate for
the additional resistance and capacitance imposed by the longer z fanout circuit.
In order to maintain acceptable signal-to-noise ratios for tracks at large dip angles, we employ
a 100 m readout pitch for these z strips with one oating strip interleaved between every
two readout strips. We have considered using a wider readout pitch, for example, 200 m
for the very forward and backward regions in order to increase the signal at large dip angles.
However, this would involve yet another detector design, and based on our present estimates
of achievable electronic noise, it does not appear to be necessary.
Acceptable resolution can be obtained for the strips on the inner layers using the ohmic
side. Two solutions are possible; either a 50 m readout pitch without oating strips, since
there is no room for them on the ohmic side, or a 100 m readout pitch with one oating
strip. Either solution is feasible, and they should give roughly equivalent position resolution
for single tracks. Double-track resolution is better for the rst solution, and the noise
contribution due to detector leakage currents is doubled in the latter solution. Therefore,
preference goes to a 50 m readout pitch without oating strips. Although this choice has
twice as many readout channels, the cost implications are not very important because the
Technical Design Report for the BABAR Detector
120
Vertex Detector
Detector Type
I
II
III
IV
V
VI
VII
z -readout Side
Cint
CAC
Rseries
-readout Side
Cint
CAC
Rseries
(pF/cm) (pF/cm) ( /cm) (pF/cm) (pF/cm) ( /cm)
1.3
40
7
2.8
40
7
1.3
40
7
2.8
40
7
1.3
40
7
2.8
40
7
1.5
80
3.5
1.3
40
7
1.5
80
3.5
1.3
40
7
1.5
80
3.5
1.3
30
9.2
1.5
80
3.5
1.3
37
7.5
Table 4-3.
Electrical parameters for the di erent detector types.
electronics cost is dominated by the development e ort and consequently the per channel
incremental cost is not signi cant.
On the outer layers, the strips are quite long (up to 24 cm), and their capacitance can
become very high on the ohmic side. The position resolution is not as important in these
layers, and the maximum track crossing angle is about 45 for the z strips. Therefore, we
choose instead to put the strips on the junction side in order to optimize the signal-tonoise ratio for these very long strips. A 100 m readout strip pitch with one oating strip is
foreseen in the rectangular sections, decreasing to about 65 m at the ends of the trapezoidal
detectors.
For the design of strips on the trapezoidal detectors, experimental investigations are needed
to decide whether to keep the width of the implants constant, the size of the gaps constant,
or to maintain a constant ratio of the gap to the implant width. In theory, a constant
ratio between width and pitch is favored because most electrical parameters (i.e., depletion
voltage and interstrip capacitance) approximately scale with this quantity and therefore
would remain constant along the length of the wedge.
For the z strips in the outer layers, the ohmic side is chosen with a readout pitch of 200 m
and one oating strip. The larger readout pitch is employed in order to minimize the ganging
of z strips.
The geometrical layout of the seven di erent types of wafers, including strip pitches and
number of oating strips, is summarized in Table 4-1 (Section 4.2). In Table 4-3, the interstrip
capacitance, coupling capacitance, and series resistance per unit length are summarized for
each detector type.
Technical Design Report for the BABAR Detector
4.4 Silicon Detectors
121
Det. Type
I
II
III
IV
V
VI
VII
z Length (mm) 47.7 57.6 45.5 46.0 55.0 68.1
63.4
Width (mm) 39.8
49.4 65.4 52.5 52.5 35.0{44.0 44.0{52.5
Junction on
z
z
z
Layer 1
24
|
|
|
|
|
|
Layer 2
|
24
|
|
|
|
|
Layer 3
|
|
36
|
|
|
|
Layer 4a
|
|
|
24
0
16
16
Layer 4b
|
|
|
8
16
16
16
Layer 5a
|
|
|
9
27
18
18
Layer 5b
|
|
|
0
36
18
18
Total
24
24
36
41
79
68
68
Total w/Spares 28
28
42
46
88
76
76
Batches
3
3
4
4
8
7
7
Shape, multiplicity, and number of batches for the di erent detector types,
with one spare module per module type.
Table 4-4.
Table 4-4 summarizes the shapes and number of detectors
for the baseline design. The quoted multiplicities refer to the installed modules plus one
spare per module type. The number of fabrication batches quoted assumes a yield of 50%,
i.e., 12 good detectors per batch of 24 wafers.
Using the numbers in Table 4-4, we see that the current design employs seven di erent
types of detectors (i.e., seven sets of masks) and needs 36 fabrication batches, for a total
of 340 installed detectors. Having so many types of detectors complicates both the design
and production phases, especially for prototype and spare production. A reduction in the
number of detector types would be most welcome; however, this represents the minimum
which we have been able to achieve in our present baseline design.
Wafer sizes and quantities.
4.4.3
Fanout Circuit Design
The SVT front-end electronics will be located outside of the active tracking volume to minimize the amount of material crossed by particles within the acceptance. As a consequence,
the signals from both the z and microstrips must be carried to the front-end chips by
exible fanout circuits. While the fanout circuits are just one-to-one connections a few
centimeters long, the z fanouts are more complicated. They must extend the full length of
the detector modules in order to connect all the z strips and to provide interconnections, or
ganging, in cases where the number of available readout channels exceeds the number of z
Technical Design Report for the BABAR Detector
122
Vertex Detector
Figure 4-11.
Schematic drawing of fanout connection to z strips showing ganging.
strips. Figure 4-11 depicts the concept of ganging. Two or three z strips (dashed lines) are
connected in series by the fanout traces (solid lines), bringing the signals to the end of the
detector module where they can be wire-bonded to the front-end electronics IC.
Table 4-5 gives a detailed list of the required fanout circuits with their geometrical features.
There are four possible types of fanout circuits per layer (z/ and forward/backward), though
in many cases, the forward (F) and backward (B) circuits are identical. The number of input
readout strips is shown together with the number of front-end electronic channels available in
the corresponding readout sections. The number of z strips always exceeds the corresponding
number of readout channels: therefore ganging is required. The number of strips is often
smaller than the corresponding number of available channels. In such cases, some of the
readout channels are not used. Typical pitches at the input and output bonding pads are
given; the line pitches can be smaller in critical places. The total number of fanout circuits
is 208; the number of di erent circuit layouts is 28.
Similar circuits have already been realized by industry as very thin, high precision Kapton
ex cables [Lev] for the L3 Silicon Microvertex Detector [DiB95] and the Aleph VDET200
detector [Bag94]. Alternative solutions exist but are not suited to our design. Double-metal
fanouts integrated on the detectors exhibit much larger parasitic capacitances. A thin glass
substrate has interesting properties, but in our case exible circuits are needed to route the
signals to the readout electronics.
The base material for the z fanout of Reference [DiB95] is 50 m-thick type-E kapton,
plated with a 0.25 m nickel adhesion layer followed by 2.5 m of electroplated copper. High
precision masks were realized by electron beam photolithography with a typical geometrical
resolution of 1 m. The required conducting lines were chemically etched to a width of about
11 m, starting from line widths of about 20 m on the masks. The pitch of the lines could
be as small as 33 m. Bonding pads were electroplated with 2 m nickel followed by 1 m of
high-purity amorphous gold.
This technology is satisfactory both for the mechanical and the electrical properties. The
average radiation thickness of the fanout circuit is only 2:3 10,4X0 , and the degradation
of the signal-to-noise ratio due to the fanout contribution to the capacitance and resistance
Technical Design Report for the BABAR Detector
4.4 Silicon Detectors
123
Layer Fanout Length Number of Readout
Typical Pitch at
Number
Type ( cm) Strips Channels Input( m) Output ( m) of Circuits
1
z , F+B 12.5
950
768
100
50
12
, F+B
3.0
768
768
50
50
12
2
z , F+B
, F+B
14.5
3.0
1150
960
1024
1024
100
50
50
50
12
12
3
z , F+B
, F+B
15.6
2.0
1360
1280
1280
1280
100
50
50
50
12
12
4a
z, F
z, B
19.7
24.3
2.0
885
1115
512
512
512
512
200
200
65
50
50
50
8
8
16
z ,F
z ,B
20.6
24.2
2.0
930
1160
512
512
512
512
200
200
65
50
50
50
8
8
16
z, F
z, B
, F+B
25.2
25.1
2.0
1160
1205
512
512
512
512
200
200
65
50
50
50
9
9
18
z , F+B
, F+B
26.1
2.0
1205
512
512
512
200
65
50
50
18
18
, F+B
4b
, F+B
5a
5b
Table 4-5.
Summary of fanout circuit characteristics.
is acceptable. For example, the ALEPH Kapton fanout circuits contribute 0.6 pF/cm to
the interstrip capacitance and 0.1 pF/cm to the strip-to-back capacitance. An acceptable
production yield (about 50%) could be achieved [DiB95] by allowing a maximum of 1.5% of
optically detected defects (shorts or breaks in the lines), most of which could be repaired.
One potential drawback of this approach is that the additional wirebonds required to connect
the fanout can result in a loss of channels due to faulty connections if great care is not taken
in assembly. These aspects must be further investigated in 1995. It is also necessary to
perform more accurate measurements and calculations of interstrip and back capacitance for
the fanout circuits. In addition, mechanical and electrical tests will assess possible limits in
the sharpness of bending. Other substrates, such as Upilex, may o er superior characteristics
and better production yield; these are currently being investigated.
Technical Design Report for the BABAR Detector
124
Vertex Detector
4.4.4 R&D on Detectors and Fanouts
Although the technology is rather stable and reliable, some R&D and prototyping must be
done before freezing the design by the end of 1995. We plan to submit a prototype batch
of AC-coupled, double-sided detectors early in 1995. The mask design will include several
structures, including a detector which is almost full size (15 cm2 ) for use in a test beam in
the summer of 1995, as well as several test structures to measure capacitances and leakage
currents for various strip pitches and implant widths. In addition, a small prototype wedge
detector and various test devices for process characterization and radiation studies will be
included. Prototype fanouts will also be employed in this test beam.
Following the rst prototype submission, there will be a second submission consisting of two
batches, with one full-size detector per batch. These detectors will be put into a later test
beam together with prototype readout electronics.
4.5
Electronic Readout
4.5.1 Introduction
The SVT readout electronics must:
Process the signals that appear on the readout strips with the aim of digitizing the
charge;
Temporally correlate the detector signals with an externally generated Level-1 trigger;
Format the vertex data, which consists of the channel ID and the digital values of the
charge and time-stamp, and transmit them to the DAQ system upon receipt of a read
command; and
Remain functional up to 10 times nominal background with graceful degradation at
higher background levels.
The charge information is retained for several reasons. One important reason is to correct
for comparator time-walk; the pulse arrival time must be measured oine as accurately as
possible in order to minimize the number of background hits associated with an event. A
second reason is to improve, by interpolation, the position measurement of hits for which
there is charge-sharing between adjacent readout strips. Finally, charge information may
also be used to reject background hits and to correlate - and z-side hits.
Technical Design Report for the BABAR Detector
4.5 Electronic Readout
125
The readout electronics are subdivided into readout sections. A readout section services
the or z strips for one-half of a detector module (either the forward or backward end).
The vertex readout electronics consists of a monolithic readout chip; a hybrid circuit, which
provides the mechanical support for the chips and the substrate for connections to the chip;
a transition card, which interfaces the front-end readout to the DAQ system; and ancillary
units, such as power supplies and detector bias supplies. The readout chip, the hybrid, the
transition card, and the power supplies will be described in the remainder of this section.
For further information on the choices that underlie the design of the readout electronics
and the relevant architectural details, the reader is referred to the relevant BABAR notes
[Lev94, Roe94b, Joh95a, Joh95b].
4.5.2
Readout Chip
Functional Overview
The readout chip must amplify and digitize the input signals for all channels in parallel.
In addition, the signals must be bu ered for the duration of the Level-1 trigger latency.
For those hits selected by the trigger, the digitized data must be bu ered until a read
command is received. Upon receipt of the read command, only signals which exceed a prede ned threshold should be read out. Since the time between beam crossings will be very
small (4.2 ns), none of the operations of the chip can be restricted to occur between beam
crossings. Furthermore, the maximum expected Level-1 trigger rate is too high to allow data
acquisition to be suspended during readout operation. Therefore, during normal operation
of the chip the analog front-end is actively acquiring data while digitization, bu ering and
readout occur.
Requirements
The requirements for the readout chip are summarized below. For more information, see
Reference [Joh95a].
1. Mechanical requirements:
Number of channels per chip: 128;
Chip size: total width < 6:2 mm, and total length < 9 mm; and
Channel-to-channel pitch: 50 m.
2. Operational requirements:
Technical Design Report for the BABAR Detector
126
Vertex Detector
Operating temperature: maximum 40C;
Radiation tolerance: up to 2 Mrad (ten years at 200 krad/yr); and
Power dissipation: not to exceed 2 mW/channel.
Dynamic range: Minimum input charge is 0.75 fC (0.20 MIP), and maximum is 19 fC
(5 MIP); circuit must accept either positive or negative signals. It is desirable to
increase the dynamic range to 38 fC (10 MIP) if feasible.
Analog resolution: 5 bits on a linear scale are required to achieve the required resolution
of 0.25 MIP for the above dynamic range, while a minimum of 3 bits are necessary for
a nearly optimal position measurement with range compression in a nonlinear system.
Time-stamp resolution: 100 ns time-stamp resolution is required for the inner layers,
increasing to a maximum of 400 ns in the outer layers to match the pulse shaping times.
Noise performance: The noise must not exceed 800 e, rms at a risetime of 100 ns with
an external capacitance of 10 pF, an external resistance of 100 , and a leakage current
of 100 nA. It must not exceed 1200 e, rms at a risetime of 400 ns with an external
capacitance of 50 pF, an external resistance of 200 , and a leakage current of 200 nA.
(These speci cations refer to the inner and outer layers of the SVT, respectively.)
Maximum data rate: Simulations show that the lost-particle backgrounds dominate
the overall rates; at nominal background levels, the maximum hit rate per strip is
about 300 kHz.
Deadtime limits: The maximum total deadtime of the system must not exceed 10% at
a 10 kHz trigger rate and at 10 times the nominal expected background rate.
Double-pulse resolution: The maximum ineciency from pulse overlap must be less
than 3%.
Trigger speci cations: The trigger has a nominal latency of 9.5 s, a maximum jitter
of 0.5 s, and the minimum time between triggers is 1.5 s. The maximum Level 1
trigger rate is nominally 2 kHz, but a conservative upper limit of 10 kHz is assumed.
Output data format:
only hits above threshold are transmitted (sparse readout); and
the data packet should include a synchronization code, chip ID, time stamp, data,
status word, and a trailer.
Calibration requirements: Provision must be made for injecting a known amount of
charge at the inputs of an arbitrary, remotely selectable set of channels.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
Technical Design Report for the BABAR Detector
4.5 Electronic Readout
Figure 4-12.
127
Block diagram of the silicon vertex detector readout chip.
13. Failure recovery: It is desirable to reduce the risk of failure by employing conservative
design methods and building in redundancy where possible.
Readout Chip Implementation
The chip described at a block diagram level in Figure 4-12 is intended for implementation
in a 0.8 m CMOS radiation hard process and has been designed to comply with the
requirements discussed above [Joh95b]. Each chip consists of 128 channels together with
additional circuitry that is common to all channels. The chip operation is synchronized
by an external clock, which is frequency-divided on the chip to obtain synchronization of
di erent operations at di erent frequencies.
Each channel consists of an analog section, a comparator with a setable threshold (Vth),
a trigger latency bu er, a counter, and a back-end bu ering section. The analog section
consists of a charge-sensitive preampli er followed by a CR(RC)2 shaping ampli er, which
implements a second-order semi-Gaussian impulse response. The shaping has a twofold
Technical Design Report for the BABAR Detector
128
Vertex Detector
Layers 1{3 z
Layer 3
Layer 5 z, x3
Layer 5
tpeak ( ns) Length ( cm) p/n Cap. (pf) ENC(e, ) Signal Loss
100
200
400
400
Table 4-6.
5
15
15
25
p
n
n
p
7.7
35.5
29.3
24.7
400
730
630
590
10%
29%
10%
33%
Equivalent noise charge for various strip con gurations.
purpose: to improve the signal-to-noise ratio (S/N) over that obtained at the preampli er
output, and to provide a signal shape suitable for the TOT processing in the next section
of the analog circuitry. The equivalent noise charge (ENC) after the shaper depends on the
front-end ampli er/shaper characteristics and on the characteristics of the detector to which
it is connected. Table 4-6 illustrates the ENC, which is calculated using a detailed SPICE
model of both the analog front-end and the detector, for a variety of strip lengths found
in the
p
SVT. The simulation assumed an equivalent noise at the input transistor of 1.9 nV/ Hz and
a post-irradiation leakage current of 1 A/cm2, corresponding to approximately 1.5 Mrad of
ionizing radiation. The signal loss is the fraction of the signal which is lost to the back of the
detector and therefore will not be collected at the preampli er. There is additional charge
lost to the neighboring strips; however, this is typically smaller and will be seen, provided it
is above threshold.
The function of a channel can be understood by following the signal path through it. The
fast detector signal with charge Q appears at the output of the shaping ampli er with a time
dependence given by:
V (x) = k Q x2 e,2x ;
(4.1)
where k (in Volts/Coulomb) is the charge sensitivity of the preampli er-shaping ampli er
combination, and x = t=tp is the time normalized to the peaking time tp. The value of
tp is selectable by an external chip control in order to comply with noise, deadtime, and
double-pulse-resolution requirements.
The shaped pulse is compared with the preset threshold Vth, and if it exceeds Vth , it produces
a TOT at the comparator output. The TOT is an analog variable which carries information
about Q. The relationship between Q and TOT is nonlinear and is not far from a logarithmic
dependence. This is desirable because it provides good resolution for small Q and reduced
resolution for large Q, e ectively compressing the range into fewer bits.
A simulation was performed to evaluate the ratio of the pulse width to the sampling period
which was required to achieve good position resolution using TOT information [Roe94b].
The position resolution as a function of the crossing angle between the track and the silicon
strip detector was evaluated for several di erent assumptions: perfect linear analog readout,
digital readout (hit/no-hit information only), and TOT with variable tpeak =tclock . For the
Technical Design Report for the BABAR Detector
4.5 Electronic Readout
129
TOT simulation, a CR-RC2 shaper was assumed. The results are shown in Figure 4-4; it is
observed that TOT with tpeak =tclock = 2:5 is essentially as good as a perfect linear analog
readout.
It should be pointed out that an ENC-based noise speci cation makes sense so long as it
is restricted to the linear section consisting of the preampli er and shaping ampli er and
refers, for example, to the threshold at which the resulting pulse is discriminated. For the
accuracy of the TOT-based charge measurement, a di erent S/N must be de ned:
(S=N )TOT = T OT = < >;
(4.2)
where < > is the rms, noise-induced dispersion on the TOT. As a rst approximation,
(S=N )TOT = (Q , Vth =k)=ENC , where k is the gain of the preampli er/shaper combination
in volts/Coulomb.
The TOT information lends itself to a straightforward digitization, for instance, by comparison with the period of a reference clock. TOT digitization and trigger delay compensation are
performed by the revolving bu er of Figure 4-12, in which the status of the TOT comparator
is stored under the supervision of a write pointer operating at the frequency of the reference
clock. The number of locations in the revolving bu er is given by the maximum expected
trigger delay, taking into account the trigger latency, trigger jitter, and the front-end chip
jitter, divided by the period of the writing clock.
The readout operation is initiated when a trigger is received, beginning with the generation
of the event time stamp, obtained by latching the contents of a 16-bit counter (one per chip)
which continuously counts the write clock (the time stamp counter in Figure 4-12). A timed
sequence ensues, controlled by the trigger-jitter counter and the TOT full-scale counter, both
of which are common to all channels in the chip. The read pointer for the revolving bu er
is set at the position corresponding to the earliest possible data in the bu er, taking into
account the xed trigger latency, the trigger jitter, and the position of the write pointer at
the instant of the trigger arrival.
The read pointer is controlled by a clock at a higher frequency than the write pointer in
order to minimize deadtime, because no new triggers can be received while the read pointer
is active. The data associated with the trigger are retrieved from the revolving bu ers by
detecting a zero-to-one transition in the revolving bu er. When a zero-to-one transition is
found, three actions take place:
A hit ag is generated;
The TOT counter is enabled to count the number of ones in the bu er; and
A hit time stamp is latched.
Technical Design Report for the BABAR Detector
130
Vertex Detector
The number of bits counted corresponds to the digital value of the TOT and hence, of the
charge Q. Based on simulation of the TOT technique [Roe94b], it has been determined
that an adequate resolution for position measurement is obtained with between 3 and 5 bits,
so 4 bits are envisaged to represent the digital value of the TOT. The hit information in
the individual channels consists of 10 bits; ve of them represent the hit time stamp, four
represent the TOT value, and one is for the hit ag.
The back-end section of each channel includes additional bu ers into which the hit data can
be stored while awaiting readout. If there is no backlog of data to be read out, the hit data are
completed by adding the information about the channel number, which requires 7 additional
bits, and transferred into a special bu er where sparsi cation takes place. Removing the hit
ag, there are 16 bits per hit channel which are then to be transmitted upon receipt of a
data transmit command.
4.5.3
Hybrid Design
Functional Overview
The hybrid is composed of several components which are hybridized into one electrical
unit [Col94]. These include the high-density interconnect (HDI), the front-end chips and
other components that are mounted on the HDI, the thermal interface between the chips,
and the water-cooling system, and the cable or tail which connects the HDI to the transition
card.
The hybrid is connected to the silicon detectors by exible fanout circuits which bring the
detector signals to the front-end chips. The hybrid is mechanically mounted onto the ber
composite beams which provide structural rigidity for the detector modules. The mounting of
the detector modules onto the cooling/support cones is accomplished with pins and screws
which are inserted from inside the cone through the aluminum mounts of the cone. The
screws pass through the thermally conductive substrate of the hybrid and into the rigid
mechanical structure of the module. This mechanical support thus provides the thermal
contact that allows heat from the ICs to ow into the cooling water circulating around
the aluminum mounts. From this brief description of the functionality of the hybrid, it is
apparent that the hybrid design is complicated by the fact that the requirements imposed
by the mechanical support, the electronic readout, and the cooling must all be met in a
self-consistent solution. The physical space available for the hybrid is very limited, imposing
yet another constraint on the design.
Technical Design Report for the BABAR Detector
4.5 Electronic Readout
131
High-Density Interconnect and Thermal Interface
The HDI is a simple fanout circuit on which the readout chips are mounted. It provides
connections for each chip to the data and control lines and to the low voltage supplies and
grounds. It may also include some passive ltering and any external components required by
the front-end chip. It is important that the layout of the HDI minimize crosstalk and noise,
especially on the analog power and ground planes. The HDI must also provide connections to
the detector for the detector biasing voltages and provide a local, low-impedance connection
from the -side to z-side preampli ers for the detector current return path. There must also
be a low-impedance thermal path from the front-end chips to the aluminum mounts which
protrude from the cooling/support cone; this thermal path will either be through the HDI
itself or through a substrate to which it is laminated.
There are ve di erent HDI layouts required, one for each of the ve layers. The same HDI
layout is used for both - and z-side readout within a layer. There are several options for the
implementation of the HDI. It may be a exible kapton circuit laminated to an aluminum
heat sink, or a thick- lm ceramic circuit fabricated from a thermally conductive material
such as AlN or BeO. Thin- lm circuitry on a ceramic substrate is also an option. At present,
these design options are being studied, and an optimum choice in terms of performance, ease
of fabrication, and cost is being determined.
The thermal path from the front-end chips to the heat sink has been evaluated with Finite
Element Analysis (FEA) simulations, assuming a exible kapton circuit laminated to an
aluminum substrate. A simple model was built and tested experimentally, and the FEA
calculation was found to agree well with the measured temperatures. Having validated
the simulation, a more detailed FEA calculation was performed, taking into account the
precise dimensions of the hybrid and all of the layers in the multilayer kapton circuit. The
calculations indicate a temperature drop of about 15C from the IC to the cooling water.
Given the maximum chip operating temperature of about 40C, this is more than adequate.
The design is being further re ned, and more experimental tests are planned.
Hybrid Tail
The tail is the connection between the HDI and the transition card (described below). It
is envisioned as a exible multilayer circuit which will have power and ground planes and
one or more signal planes. It will be fabricated separately from the HDI itself and will be
connected to two HDIs by means of bump-bonds, solder, or other permanent connection. The
connection to the transition card will utilize a mateable connector. The biggest constraint
on the design of the tail is the limited space available for cables from the inner SVT layers
to pass between the cooling/support cone and the outer detector modules. This restricts the
width of the cable to less than about 0.75 cm, and the thickness to less than about 0.3 mm.
Technical Design Report for the BABAR Detector
132
Vertex Detector
The trace impedance, which is about 30 , must be matched to the drivers and receivers at
either end.
4.5.4 Data Transmission
Data Transmission Requirements
The main requirements for the data transmission system are given below.
1. Data rate: between 60 and 80 Mbits/s/readout section;
2. Power consumption (inside the detector volume): below 5 W/readout section;
3. Radiation resistance (inside the detector volume): the system must be functional within
the speci ed performance after being exposed to a 100 krad integrated dose of ionizing
radiation; and
4. Error rate: less than 10,11/bit. This would imply roughly 1 bit error in every 106
events transmitted.
Functional Overview
Two di erent designs have been studied for transmitting data from the SVT front-end
readout IC to the DAQ system and also to send clock, trigger and control signals from
the DAQ system to the readout IC. The two designs di er primarily in where the transition
between electrical and optical transmission occurs.
In the rst design which was considered, the transition from electrical to optical transmission
was made on transition cards which sit inside the support tube at the outboard end of the
B1 magnets. A exible circuit carries electrical signals between the HDI and the transition
card, and optical bers connect the transition card to the DAQ system. This option has
the advantage of avoiding noise pick-up in the 5{6 m cable run from the transition card to
the outside of the detector, and reducing the volume of cable in the space-constrained area
inside of the support tube. However, it requires customized packaging and use of radiationhardened opto-electrical components on the transition card which are very costly.
In the second design, electrical transmission is used from the HDI all the way outside of the
detector to a point where crates can be mounted to support a commercial optical transmission
system. The transition cards inside of the support tube are still needed; however they would
have little or no active circuitry. Because this option can make use of commercial ber optic
Technical Design Report for the BABAR Detector
4.5 Electronic Readout
133
systems, the cost is reduced and it is possible to use optical transmission systems common
to other sub-detectors. The opto-electric system is also accessible for repair and does not
need to be radiation hard.
The transition cards are required in both designs in order to provide a place where the
hybrid tail can terminate. In the second option, if the impedance of the hybrid tail and the
transmission cables extending beyond the transition card can be matched suciently well,
all active components can be removed from the transition card thus improving the reliability
of the system inside the support tube where access is limited.
After studying the space budget for cabling beyond the transition card and the feasibility
of the readout ICs transmitting signals electrically out of the support tube area, the second
option has been chosen as the baseline design.
4.5.5
Baseline Design
The baseline design for transmitting data out of the detector starts with the readout ICs
transmitting di erential signals on pairs of traces up the hybrid tail to the transition card.
The signals are there transferred to conventional paired cables for transmission to an interface
card outside of the detector. Likewise, clock and control signals coming from the interface
cards will be transmitted as di erential signals on conventional cable to the transition card
inside the support tube and then transferred onto the hybrid tail to the HDI and the readout
ICs.
The hybrid tail is still under development, and its precise electrical characteristics are still
to be determined. The design of the hybrid tail is constrained by mechanical requirements
and limited by the available technology for producing exible cables. If the impedance of
the hybrid tail can be closely matched to that of available cabling, the signal transmission
through the transition card will be only passive. If the impedance mismatches are too
great, commercially available bipolar components have been identi ed to receive and retransmit signals on the transition card. Both the design of the transmitting and receiving
circuits on the readout ICs and the choice of components for the transition card have been
made to maintain the consistent di erential, low power, low noise transmission signals. The
identi ed bipolar components for the transition card will meet the speed, power and radiation
requirements of the system.
The transition cards will not only provide the physical connection between hybrid tail and
conventional cable and any necessary impedance matching for signal transmission, it will also
provide necessary ltering for all the DC voltages bussed to the detectors and the readout
ICs. There will be one transition card for each hybrid tail. They will be located inside the
support tube, approximately 20 cm from the outside layer of the SVT. They are mounted
Technical Design Report for the BABAR Detector
134
Vertex Detector
Figure 4-13. Side view of the connections from the readout electronics to the transition
card, which is a triangularly shaped circuit mounted from the B1 magnet.
from a cooling ring which is attached to the far end of the B1 magnet. See Figure 4-13
for a side view showing the connections from the readout electronics to the triangularly
shaped transition card. A multi-conductor cable bundle for each transition card will carry
all necessary power, sense and data lines from the transition card through the support tube
to a point outside the detector.
The interface cards will provide transition between electrical signals and optical signals. The
exact location for these cards has not been determined but they will be located approximately
5{6 m from the interaction point, outside the detector. Eight data lines will be multiplexed
onto one gigabit ber optic link for transmission to the DAQ system. Likewise, clock and
control signal coming from the DAQ on a ber will be fanned out and transferred to electrical
cables going into the detector. The exact level of multiplexing between electrical signals and
optical is made to match bandwidths and partitioning of DAQ functions.
Technical Design Report for the BABAR Detector
4.5 Electronic Readout
135
4.5.6 Power Supplies
The power supply system for the SVT must provide the voltage sources with adequate current
levels for two purposes: biasing the silicon detectors, and supplying power to the front-end
chip, transition card, and remote opto-electronic circuits. The low-noise electronic readout
for the silicon vertex detector imposes stringent requirements on the power supplies, and
they must be carefully designed to prevent common-mode noise from entering the system.
Both supply subsystems must be built up from modular, fully oating units without galvanic
connection between the utility side and the load side, with the two sides electrostatically
shielded. Preference is generically given to continuous-type voltage regulators in view of
their lower noise levels. However, tests are foreseen to determine whether smaller switching
regulators with better power yield can also provide satisfactory performance.
The power supplies must also feature remotely controlled, digitally preset voltages. Optical
connection of the voltage controls is recommended. The power supplies must be capable of
sweeping the output voltage throughout the nominal range as speci ed below.
The speci cations for the two di erent supply subsystems are as follows. For the detector
bias supply, the nominal full scale voltage is 80 V per module, with a setting resolution of
7 bits and maximum current of 0.1 A. Ripple must be less than 10 mV peak to peak, and
noise must be less than 2 mV rms. For the electronics power supply, the nominal full scale
voltage is 6 V per module, with a setting resolution of 8 bits and maximum current of 100 A.
Ripple must be less than 2 mV peak to peak, and noise must be less than 500 V rms.
4.5.7 Electronics R&D
The R&D to support the development of the electronic readout system described above has
already begun. Characterization of radiation-hard CMOS processes is well underway. Test
structures have been submitted to Honeywell and to UTMC to characterize their radiationhard processes. This work was begun as part of an R&D program for SDC. The UTMC
circuits have already been evaluated, and the work is continuing with irradiation and testing
of the Honeywell chips. Honeywell has an 0.8 m, triple-metal process which is attractive
both for its small minimum feature size and for the third metal layer which is useful for
minimizing the circuit size and may also prove useful for shielding. Initial results with
this process are promising. Figure 4-14 gives comparison of the noise performance of the
Honeywell process before and after irradiation, plotted versus frequency. For the 100 ns
shaping time which is foreseen for the inner detector layers, the corresponding frequency is
about 3 MHz. At this frequency, the noise increases by about 15% after 500 krad and by
Technical Design Report for the BABAR Detector
136
Vertex Detector
Figure 4-14.
Noise levels for Honeywell transistor before and after irradiation.
about 30% after 1.5 Mrad of ionizing radiation. Further tests up to a total integrated dose
of 5 Mrad are planned.
A rst round of test chips on the TOT concept has been submitted and is being tested in
early 1995. These test chips will be used to evaluate the TOT technique and to measure
the extent to which digital activity on the chip during data acquisition can contribute to
front-end noise. Work on a rst full-scale prototype of the TOT front-end chip is underway,
and the rst submission is expected in the spring of 1995. This chip will be operated on a
test bench in the fall of 1995 with prototype BABAR detectors and will then go into a test
beam.
For the hybrid, a mechanical prototype is foreseen early in 1995 as a means of qualifying
a vendor. This will be followed by a working electronic prototype in mid-1995, which is
designed to work together with the rst prototype TOT chip.
4.6 Mechanical Support and Assembly
An overview of the SVT mechanical support is provided in Section 4.2. In this section
we provide a more detailed account of the constraints on the mechanical design due to
Technical Design Report for the BABAR Detector
4.6 Mechanical Support and Assembly
137
the accelerator components near the IP and describe the details of the detector assembly,
installation, survey, and monitoring.
4.6.1 IR Constraints
The support structure design and con guration of the SVT is dictated by the con guration
and assembly procedure of the machine components near the interaction point, as well as
by the SVT geometry. The close spacing of the bunches in PEP-II (4.2 ns) and the desire
to avoid parasitic crossings dictate the need for a pair of permanent dipole magnets located
about 20 cm from the IP on either side. These B1 magnets occupy most of the region below
17.2 (300 mr). In order to minimize the mass inside the active tracking volume, it is desirable
to mount all of the electronics below the 300 mr line. In the forward direction, this requires
that electronics, cooling, cabling, and support be con ned in a volume one centimeter thick
around the B1 magnet. The use of this small space below the 300 mr line must be carefully
coordinated with the needs of the accelerator. In the backward direction, coverage to within
30 of the beam pipe leaves room for some machine components which are moved out of the
forward region. The solid angle coverage of the SVT is therefore restricted to the region
17:2 < < 150.
A further constraint on the SVT is imposed by the PEP-II support tube, which is a tube of
radius 20 cm that aligns the nal magnets closest to the IP. The central part of the support
tube is a thin carbon- ber structure, while the outer sections are stainless steel. The SVT
must be supported inside the carbon- ber support tube and therefore cannot be mounted
from the drift chamber endplates. In the baseline design, the SVT is supported from the
B1 magnets. Since the SVT must be installed with both B1 magnets in place, it must be
assembled in two halves and then clam-shelled around the beam pipe. The assembly and
alignment of the beam pipe, B1 magnets, Q1 magnets, and SVT inside the support tube will
take place in a staging area away from the interaction hall. The entire assembly will then
be transported and installed in the interaction hall. This procedure is reversed in order to
gain access to the SVT for repair.
4.6.2 Module Assembly
The SVT is constructed from detector modules, each of which is mechanically and electrically
independent of the other modules. Each module consists of silicon wafers bonded to ber
composite beams, with a high density interconnect (HDI) electronic hybrid at each end.
The HDIs are electrically connected to the silicon strips by means of exible circuits and are
mechanically supported by the ber composite beams. The entire module assembly is a rigid
Technical Design Report for the BABAR Detector
138
Vertex Detector
Figure 4-15. Detector module from Layer 3, consisting of six silicon detectors wirebonded
together and read out at each end.
structure that can be tested and transported with relative ease. A drawing of a detector
module from Layer 3 is shown in Figure 4-15.
Assembly of the detector modules begins with the preparation of the necessary parts. The
silicon detectors must be fully tested, including a long-term test under full bias voltage.
The fanout circuits will be optically inspected and single strip tested for shorts/opens.
The readout hybrids must be assembled and tested, starting with the HDI circuit, the
front-end chips, any additional passive components, and the hybrid support. Finally, the
completed beams, which provide mechanical sti ness, must be inspected to ensure they meet
speci cations. These individual parts will be fabricated at di erent institutions from which
they can easily be shipped to the institutions at which the module assembly is carried out.
The hybrids will be retested after shipment.
The assembly of the inner barrel-shaped modules and the outer arch-shaped modules is
necessarily di erent. However, there are common steps. Generally, the procedure is as
follows:
1. The z and fanouts are glued to the detectors and wire-bonded to the strips. The
ganging bonds between strips are performed.
2. The silicon detectors and readout hybrids are held on a suitable xture and aligned
relative to each other. The fanouts are glued to the hybrids and wire-bonded to the
input channels of the readout ICs. Electrical tests, including an infrared laser strip
scan, are performed and the detector-fanout assemblies (DFAs) are visually inspected.
3. The nal assembly stage is di erent for di erent layers. For modules of Layers 1 and
2, the DFA is bonded to the ber composite beams with appropriate xtures ensuring
alignment between the mounting surfaces on the HDI and the detectors. The module
is again tested. A module from Layer 1 and Layer 2 are then joined together by gluing
Technical Design Report for the BABAR Detector
4.6 Mechanical Support and Assembly
139
the beams on the top of Layer 1 to the bottom of Layer 2. The combined structure is
called a sextant module. For Layer 3, the DFA is bonded to the ber composite beams
as with Layers 1 and 2.
For the modules of Layers 4 and 5, the DFA is held in a suitable xture and bent at the
corners of the arch and at the connection to the HDI. The module is tested. The ber
composite beams are bonded to the module with xtures assuring alignment between
the silicon detectors and the mounting surfaces on the HDI. This procedure has been
performed with real detectors that have been successfully operated in a test beam.
Once completed, these detector modules are extremely rigid devices that can be stored and
subjected to long-term testing. They are then shipped to the location for nal assembly and
installation of the detector.
4.6.3 Detector Assembly and Installation
Half-Detector Assembly
The detector is assembled in halves in order to allow the device to be clam-shelled around
the beam pipe. The detector modules are supported at each end by cooling/support cones
constructed from double-walled carbon- ber laminates. Cooling water circulates between the
two carbon- ber shells and around aluminum mounting pieces which protrude through the
outer shell. The cones are split along a vertical plane and have alignment pins and latches
that allow them to be connected together around the B1 magnets. See Figure 4-16 for a
drawing of the cooling cone. The two carbon- ber support cones are mechanically connected
by a low-mass carbon- ber space frame.
During the half-detector assembly, the two half-cones will be held in a xture which holds
them in precise relative alignment. The detector modules are then mounted to the half-cones
at each end. A manipulating xture holds the detector modules during this operation and
allows for well-controlled positioning of the module relative to the half-cones. Pins located
in the aluminum mounting pieces provide precise registration of the modules, which are then
screwed down. Accurate alignment of the mounting with respect to the silicon wafers is
achieved by a pair of mating xtures. One is a dummy module and the other simulates the
mating surfaces on the cone. These xtures are constructed together and mate perfectly.
One is used to verify the machining of the aluminum mounting pieces on the cones. The
other is used to position the mounting points on the HDI during the assembly of the modules
as mentioned above.
The connection between the module and the cone (called the foot) provides for accurate and
reproducible alignment of the module and conduction of heat from the HDI heat sink to the
Technical Design Report for the BABAR Detector
140
Vertex Detector
Schematic view of the cooling/support cone. The cone is constructed in
two halves and clam-shelled together. Water circulating between the two carbon- ber shells
cools the readout electronics, which are mounted on aluminum pieces protruding through
the outer shell.
Figure 4-16.
cooling water circulating between the cone layers. A detail of the foot region, which contains
the readout electronics and the mounting pieces, is shown in Figure 4-17.
After veri cation of the alignment, the connection between the HDI and the support beams is
permanently glued. The glue joint allows for the correction of small errors in the construction
of either the cones or the modules. After the beam is glued, the module may be removed and
remounted on the cone as necessary. The design of the foot allows this glue joint to be cleaved
and remade should major repair of the module be required. After each detector module is
mounted, it is electrically tested using a laser scan to verify its functionality. As each layer
is completed, it is optically surveyed and the data are entered into a database. Finally,
the two half-cones are connected together with the space-frame, resulting in a completed
half-detector assembly.
Mount to B1 Magnets and Transport to IR
When the two half-detector assemblies are complete, they are brought to the staging area
where both B1 magnets and one of the Q1 magnets have been assembled onto the beam
Technical Design Report for the BABAR Detector
4.6 Mechanical Support and Assembly
Figure 4-17.
module.
141
Exploded view of foot showing mounting pins and connection to the detector
pipe. Fixtures are employed to hold the cones as they are brought together and clam-shelled
around the beam pipe. The two half-detector assemblies are mated, and the latches between
them are closed. The cables from the HDIs are routed to the transition cards, which are
mounted in cooling manifolds at the ends of the B1 magnets. The entire detector is then
thoroughly tested, and an optical survey is performed. After the survey is complete, the
support tube is slid over the entire assembly and mated to the Q1 magnet. The nal Q1
magnet is then installed. At this point, the assembly is relatively rigid and can be transported
to the interaction hall and installed in the accelerator.
The detector assembly as described above forms a rigid structure as long as the cones and
space frame are connected together. This structure is supported on the B1 magnets. During
transport of the support tube assembly to the interaction hall, it is possible for the magnets
to have as much as a 1 mm relative motion [Bow95]. This motion is reversible, and they will
return to their original alignment when installed in the accelerator. In addition, di erential
thermal expansion may a ect the relative alignment of the magnets during periods in which
the temperature is not controlled, and relative motion of the magnets and the beam pipe
may occur should there be seismic activity.
The support of the detector from the magnets must allow for this motion without placing
stress on the silicon wafers. In addition, the position relative to the IR must be reproducible
when installed in the accelerator. These constraints are met by mounting the support cones
Technical Design Report for the BABAR Detector
142
Vertex Detector
on a pair of gimbal rings. One gimbal ring connects the forward cone to the B1 magnet so
as to constrain its center in x, y, and z, while allowing rotation about the x and y axes. A
second set of gimbal rings supports the cone in the backward direction in a similar manner,
with an additional sleeve that allows both for motion along z and rotation about the z axis,
relative to the B1 magnets.
Installation of Complete Assembly into Detector
The clearances between the vertex detector and the beam tube and B1 magnets are on the
order of 1 to 3 mm. During transport of the support tube assembly, the critical clearances
must be monitored in real time to ensure that no accidental damage to the detector occurs.
In its nal position, the support tube assembly will be supported from the ends. Thus it is
necessary that it is always supported only from the ends during installation. One possible
installation scenario employs beams attached to the ends of the Q1 magnets which support
the entire assembly from the ends. The beams are threaded through the drift chamber,
resting on wheeled guides or tracks, which have been previously aligned with respect to
the BABAR detector. Once the support tube is through and supported at both ends, the
temporary assembly beams are removed.
4.6.4 Detector Placement and Survey
Placement Accuracy
The SVT must provide spatial resolutions on the order of 10 m. Final locations of each of
the wafers relative to each other and to the IR will be determined by track survey. This
requires a certain degree of overlap of the modules within a layer. There must be overlap
in z as well as , so as to accurately locate the z positions of the wafers in a single module
with respect to each other. These requirements are discussed in more detail below.
Mechanical tolerances and measurements must be such that the process of track survey
converges in a reasonable time. Placement of the wafers within a module should be within
25 m and optically surveyed to a few m. Placement of a module relative to other modules
should be on the order of 100 m and be optically surveyed to within 25 m.
Stability of the positions should be such that resurvey with tracks is rarely required. This
leads to the requirement that the relative positions of the various wafers be stable to the
5 m level over long periods of time (months or more). The position of the entire detector
structure with respect to the IR can be followed more easily, so that variations on the order
of a day can be tracked. However, longer time constants are desirable.
Technical Design Report for the BABAR Detector
4.6 Mechanical Support and Assembly
143
Stability of the detector components at the 5 m level requires a stable operating temperature. Preliminary calculations for the thermal expansion of the entire structure predict on
the order of 0.5 m/C over the length of the active region of the detector. If the temperature
inside the support tube is maintained at 1 C, thermal expansion will not be a problem.
Survey with Tracks
Previous experiments have shown that a silicon vertex detector is best aligned using track
data together with self-consistency constraints, because the point resolution in silicon is much
more precise than that from a drift chamber. The most powerful alignment constraint comes
from the overlap between wafers adjacent in . Another important constraint comes from
muon-pair events, where knowledge of the mass, boost, and common origin of the two-track
system provides a means of linking hits on opposite sides of the vertex detector. A more
detailed discussion of this subject can be found in Reference [Bro95].
The SVT alignment is parameterized as a pair of nested transformations for global and
local coordinates. The global alignment of the SVT describes the position and orientation
of each detector layer, considered as a rigid body, relative to the drift chamber. Previous
experiments have shown that a good global alignment helps speed the convergence of the
local alignment and allows a simple, low-statistics check of time-dependent e ects such as
detector motion. The procedure depends completely on the drift chamber tracking and
simply requires consistency between SVT hits and the extrapolated drift chamber track
positions. Systematic e ects limit the accuracy of the global alignment to 50 m. Assuming
an accuracy of 2 mm for the drift chamber track extrapolation in the z direction, this implies
that a global alignment can be performed with 2000 tracks. Further assuming two useful
high momentum tracks per event, we nd that 1000 hadronic events will suce to perform
a global alignment.
The local alignment describes the position and orientation of each silicon wafer relative
to its nominal position in the layer. The local alignment of the SVT derives most of its
statistical precision from the active area overlap between -adjacent wafers. The overlap
directly constrains three of the six geometric degrees of freedom of each wafer, namely those
that de ne its position in the nominal wafer plane. A further strong constraint on the wafer
translation in the out-of-plane direction comes from combining the overlap information for
all wafers in a given layer. This is essentially a circumference constraint, using the fact that
the size of the wafers is precisely known.
The number of events needed to perform the local SVT alignment can be roughly estimated.
In the following, we assume that each wafer is aligned separately for six geometric degrees
of freedom, and we further assume four useable medium-momentum tracks per event. We
require that the alignment should not contribute signi cantly (less than 5%) to the SVT
Technical Design Report for the BABAR Detector
144
Vertex Detector
Layer
# of Wafers # of Tracks/Wafer # of Hadronic Events
1
2:4%
24
80
20K
2
1:8%
24
80
27K
3
1:8%
36
80
40K
4
4:0% (4.8%)
112
40
28K (24K)
5
2:0% (1.9%)
144
40
74K (77K)
Fractional overlaps ( ) of the SVT layers and the resulting number of hadronic
events required for an accurate alignment. The values in parentheses for Layers 4 and 5
correspond to the wedge modules.
Table 4-7.
point resolution, implying that each wafer position must be known to better than 30% of
its average point resolution. The overlap solid angle and number of wafers are di erent for
each layer and are given in Table 4-7 together with the number of events needed. Layer 5 is
seen to set the limit, requiring 75K hadronic events to be precisely aligned. This is mainly
because of the large number of wafers in this layer.
Overlaps do not constrain the relative positions of the di erent layers, nor the relative
positions of the di erent annuli which comprise the layers. An annulus is here de ned
as the set of all wafers in a layer having the same z position. The relative positions of
these sets of wafers are best obtained using e+ e, ! +, events; we require an alignment
precision of 30% of the intrinsic resolution for each annulus of a given layer. To estimate the
approximate number of muon pairs required, we assume that the three parameters de ning
the muon trajectory have been measured with essentially no error on one side of the SVT,
and this trajectory is compared to hits on the opposite side. The tracks are presumed to have
2.5 GeV=c momentum, and the material of each layer is presumed to be 0.5% of a radiation
length on average. This gives the resolution of the track extrapolation due to multiple
scattering as presented in Table 4-8, which is seen to dominate the total (extrapolation
intrinsic) resolution. Layer 5 is again seen to require the most events to be accurately aligned,
namely 7000 muon pairs or the equivalent of 27,000 hadronic events. Thus the overlaps set
the overall minimum number of events needed for alignment.
4.6.5
Detector Monitoring
Position Monitoring Systems
Although the nal placement of the silicon wafers will be measured and monitored with
charged particles which traverse the silicon detector and drift chamber, two displacement
monitoring systems will be designed to measure relative changes in the position of the silicon
Technical Design Report for the BABAR Detector
4.6 Mechanical Support and Assembly
Layer Track
1
25 m
2
30 m
3
50 m
4
80 m
5
100 m
145
Hit # of Annuli # of Muon Pair Events
10 m
4
225
10 m
4
325
10 m
6
1300
30 m
7
4000
30 m
8
7000
Properties of the di erent SVT layers, and the number of muon-pair events
these imply for an accurate alignment.
Table 4-8.
detector with respect to the machine elements and the support tube. One displacement
monitoring system will be used to monitor relative positions during transportation of the
support tube with the silicon detector inside it and during data taking. This system consists
of either capacitive displacement monitors or LED-photodiode re ection monitors which are
sensitive to relative displacements between the silicon detectors and the machine components
such as the beam pipe, magnets, and support tube.
In addition, a laser system will monitor displacements of the outer layer of detectors with
respect to the drift chamber during data taking. Given that the SVT layers are not mounted
on the same support as the drift chamber, it is possible that motion between the two will
occur. To monitor this motion, short infrared laser pulses are brought in with ber optics
(e.g., 50 m core diameter) which are attached to the drift chamber. The laser light shines
through small holes in the support tube and reaches the outer layers of the silicon detector.
The resulting signals are read out with the normal silicon detector readout system.
Displacement monitoring systems based on capacitive sensors and laser pulses have been used
successfully by many experiments in the past [Acc94, Bin93, Cac92, Bre91]. The two systems
are complementary in technology and in their sensitivity to di erent kinds of displacements.
For example, the L3 experiment [Acc94] attains a resolution of a few microns with the laser
monitoring system for transverse displacements. With their capacitive monitoring device,
they attain a resolution of 1{2 m in the radial direction and 5{10 m in the transverse
direction.
Radiation Monitoring
To protect the silicon detector system against potentially damaging beam losses, and to
monitor the total radiation dose that the detectors and electronics receive, silicon diodes will
be installed close to the beam pipe in the vicinity of the SVT. If the radiation dose exceeds
a certain threshold, a beam-dump signal will be sent to the PEP-II control room. This sort
of radiation protection system is already used in all of the LEP experiments. The silicon
Technical Design Report for the BABAR Detector
146
Vertex Detector
diodes will be read out with various gains so that the dynamic range of the full set of sensors
covers quiet running and potentially damaging conditions.
4.6.6 R&D Program
The following R&D projects are planned before the design of the SVT mechanical con guration is nalized.
Cable. Prototypes of the cable from the hybrid to the transition card will be constructed.
This will allow proof of the details of cable routing and mechanical robustness. It will also
allow the electrical properties to be measured to verify simulations.
Hybrid. Realistic mechanical modules of the high-density interconnect (HDI) are required.
The HDI is a critical element both in the cooling of the electronics and the mounting of the
detector modules. Models will be tested for heat transfer capability and for module mounting
schemes.
Inner Layer Sextant. A full-scale mock up of the inner layer sextant will be constructed.
It will be used to verify thermal stability calculations and to investigate the e ects of
nonuniform beam pipe cooling. It will also be used to test and practice assembly techniques.
Arch Modules. Full-scale mock ups of the arch detector modules will also be constructed
and used with the prototype cones to verify cooling and mounting techniques.
Cones and Space Frame. A set of prototype cones and a space frame will be built to
provide realistic tests of cooling, mechanical rigidity, and thermal stability. In addition, they
will be used to design assembly xtures and test assembly techniques for mounting modules
onto cones.
Full-Scale Model of the IR. A model of the B1 magnets and the beam pipe near the IP
will be constructed. This will aid in identifying interference problems and verifying mounting
schemes. It will also provide a test bed for the design of various installation xtures.
Technical Design Report for the BABAR Detector
4.7 Services, Utilities, and ES&H Issues
147
4.7 Services, Utilities, and ES&H Issues
4.7.1 Services and Utilities
The vertex detector requires the following services, which must be brought inside the support
tube to a location near the outboard end of the B1 magnet.
Data and Control Lines. Approximately 624 wire pairs are required on each side of the
detector to service the 52 transition cards. These consist of 2 clock, 2 control and 2 data
links per readout section providing redundancy for each signal. The interface cards will
multiplex these into 26 optical links for data and 26 optical links for control on each side of
the detector.
Power. The readout ICs will require three low-voltage power supplies (two analog, one
digital), and the transition card will require at least one power supply. This amounts to
four power and return cables per readout section, or 416 supplies and returns for each half
of the detector. Each line carries only a few watts, except for the transition card supply
which could be a few tens of watts. In addition to these low voltage supplies, we require
52 detector bias supplies per detector half. The bias voltage ranges from 40{80 V, and the
detector draws very little current. The power supplies will all be specially procured to the
vertex detector speci cations in order to control electronic noise.
Cooling Water. The readout electronics and transition cards will be water cooled. Two
sets of water connections will be required for each cone (since each cone is constructed from
two halves), and one water connection for each transition card support. The cooling water
will be supplied by a special low volume chiller system dedicated to the vertex detector.
Dry Air or Nitrogen. The vertex detector requires a dry, stable environment, and dry
air or nitrogen from each side is planned.
4.7.2 ES&H Issues
There are very few ES&H issues which impact the construction, assembly, and operation of
the SVT. The detector bias voltages can exceed 50 V and therefore qualify as high voltage,
though they are extremely low current. There is a potential for water leaks from the cooling
Technical Design Report for the BABAR Detector
148
Vertex Detector
systems. This water would be con ned to the volume inside the support tube and would
pose a danger primarily to the vertex detector itself. The laser monitoring system would
employ Class 1 lasers which pose essentially no danger.
Technical Design Report for the BABAR Detector
REFERENCES
149
References
[Acc94]
[Bag94]
[Bat92]
M. Acciarri et al., CERN{PPE/94{122, to be published in
A.
G. Bagliesi et al., \Test Beam Results from Prototypes of the Upgraded ALEPH
Vertex Detector," ALEPH Note 94{069 (1994).
P. Coyle et al., \ALEPH Vertex Detector," presented at the Sixth Pisa Meeting
on Advanced Detectors, Elba, Italy (1994).
G. Batignani et al., \Experience with the ALEPH Silicon Vertex Detector,"
A315, 121{124 (1992).
G. Batignani et al., \Double-Sided Silicon Strip Detectors in Pisa," presented at
the Sixth Pisa Meeting on Advanced Detectors, Elba, Italy (1994).
G. Batignani et al., \Speci cations of the Silicon Detectors for the Vertex
Detector," A AR
# 196 (1994).
P. Billoir,
225, 352 (1984).
N. Bingefors et al.,
A328, 447 (1993).
G. Bowden, \IP Beam Pipe/Vertex Detector Clearance" (1995) (unpublished).
A. Breakstone et al.,
A305, 39 (1991).
D. Brown, \ A ARVertex Detector Alignment Requirements," A AR
# 148
(1995).
M. Caccia et al.,
A315, 143 (1992).
T. Collins, \ A AR Hybrid Design Considerations," A AR
# 222 (1994).
D. DiBitonto et al., \Ultra-thin, High Precision Flex Cable for the L3 Silicon
Microvertex Detector," to be submitted to
A.
F. Forti, \TRACKERR Studies for Optimization of Vertex Detector Resolution,"
A AR
# 195 (1994).
P. Holl et al., IEEE Trans. Nucl. Sci. 36, N. 1, 251 (1989).
W. Innes, \TRACKERR, A Program for Calculating Tracking Errors,"
A AR
# 121 (1993).
Front-End Working Group, \Requirements Speci cations for Silicon Vertex
Detector Readout Chip," edited by R. Johnson, A AR
# 213 (1995).
Nucl. Instr. Methods
Nucl.
Instr. Methods
[Bat94a]
[Bat94b]
B B
[Bil84]
[Bin93]
[Bow95]
[Bre91]
[Bro95]
[Cac92]
[Col94]
[DiB95]
Note
Nucl. Instr. Methods
Nucl. Instr. Methods
Nucl. Instr. Methods
B B
B B
Note
Nucl. Instr. Methods
B B
B B
Note
Nucl. Instr. Methods
[For94]
[Hol89]
[Inn93]
[Joh95a]
B B
Note
B B
Note
B B
Note
Technical Design Report for the
A AR
B B
Detector
150
REFERENCES
[Joh95b] Front-End Working Group, \Target Design Speci cations for Silicon Vertex
Detector Readout Chip," edited by R. Johnson, BABAR Note # 214 (1995).
[Lev]
Max Levy Autograph, Inc., Philadelphia, PA (USA).
[Lev94] M. Levi, \The Impact of Backgrounds on the Silicon Architecture and Detector
Trigger," BABAR Note # 136 (1994).
[Roe94a] N. Roe and L. Kerth, \Vertex Detector Solid Angle Coverage," BABAR Note # 160
(1994).
[Roe94b] N. Roe, \Silicon Strip Resolution for Time-Over-Threshold Readout,"
BABAR Note # 161 (1994).
[Sny94] A. Snyder, \The E ect of Vertex Cuts on CP Reach," BABAR Note # 177 (1994).
[Ton94] G. Tonelli et al., \Development of Double-Sided strip Detectors for High
Radiation Environment," presented at the Sixth Pisa Meeting on Advanced
Detectors, Elba, Italy (1994).
Technical Design Report for the BABAR Detector
5
Drift Chamber
5.1 Physics Requirements and Performance Goals
T
he goals of the BABAR experiment require that exclusive nal states from B 0 decays be
reconstructed eciently and with high resolution. This places stringent demands on the
performance of the main tracking chamber. It must provide maximal solid angle coverage,
good momentum resolution at all momenta, and ecient reconstruction of tracks as low as
100 MeV=c. In addition, the tracking chamber provides one of the two major triggers for
the experiment. These requirements are met through the use of a small-cell, low-mass drift
chamber which, in addition to providing excellent momentum resolution for low-momentum
tracks, minimally degrades the performance of the calorimeter and particle identi cation
device.
Figure 5-1 shows the integrated eciency for a variety of all-charged B decay modes as a
function of the momentum and polar angle acceptance. The channels B 0 ! +,, B 0 !
J= KS0 with KS0 ! + , , and B 0 ! D + D , with D + ! K , + , have been used in the
upper two plots in Figure 5-1 to demonstrate the importance of the forward angle acceptance
limit. For example, a factor of 6.5 would be lost in the D+D, eciency if good momentum
and particle identi cation information were not available beyond cos = 0:8. Thus the
drift chamber must cover the polar angle range down to the beam-line components in the
forward direction, namely 300 mr or cos < 0:955. The backward angle requirement is
considerably less stringent. The high- and low-momentum requirements are illustrated in the
lower two plots of Figure 5-1; the channel B 0 ! D+D,, with its subsequent cascade decay
D + ! + D 0 producing a soft pion, places the greatest demand on the reconstruction of low
momentum tracks. The ability to nd and reconstruct charged particles down to 100 MeV=c
is clearly required for good eciency in this mode and in other channels containing one or
more D+ mesons.
In order to reconstruct exclusive nal states with minimal background, the chamber must
provide excellent momentum resolution. For tracks with p above 1 GeV=c, such as those
arising from the decay B 0 ! +,, a resolution of pt =pt ' 0:3% pt is anticipated. For
the momentum range 0.1 to 1.0 GeV=c, relevant for tracks from the decay B 0 ! D+D,,
pt =pt ' 0.3{0.4% should also be achievable. The expected resolution on the reconstructed
152
Drift Chamber
100
Efficiency (%)
80
60
(b)
(a)
π+π–
ψKs
D+D– or D*+ D*–
40
20
0
0
0.2
0.4
0.6
cosθmin
1.0 –1
0.8
0
cosθmax
1
100
(d)
(c)
Efficiency (%)
80
60
40
20
0
0
6-93
1
2
3
p min (GeV/c)
4
5
0
1
2
3
p max (GeV/c)
4
5
7468A1
for charged particle detection from B 0 ! + ,
(solid), B 0 !
!
! 0+, (dash-dot),
and B 0 ! D+D, , D+ !
,
+
+
+
,
+
K for the angular distributions and B ! D D , D
! + D0 , D0 ! K ,+
for the momentum distributions (dashed). Each set of curves is shown as a function of:
(a) minimum detectable ; (b) maximum detectable ; (c) minimum detectable momentum;
and (d) maximum detectable momentum.
Figure 5-1.
B reconstruction eciency
J= KS0 , J=
`+ `, , KS0
B0
mass for these decays and the dependence on various drift chamber detector parameters
is given in Section 5.3.
At PEP-II, the average charged particle momentum is less than 1 GeV=c. For most particles,
the major limitation on track parameter resolution in the drift chamber will be multiple
scattering. In order to minimize this e ect, the chamber will be made of low-mass materials.
A thin inner wall facilitates matching drift chamber and vertex detector tracks, improves
the contribution of the high-precision measurement in the outer layer of the SVT to the pt
resolution, and minimizes backgrounds due to photon conversions in the chamber wall. The
material in the outer walls and endplates of the chamber must also be kept to a minimum
Technical Design Report for the BABAR Detector
5.2 Tracking Chamber Overview
153
in order to reduce the impact of the drift chamber on calorimeter and particle identi cation
performance.
The drift chamber is expected to participate in the detector trigger by providing one of two
complementary trigger streams. The drift chamber must furnish a charged track trigger with
a latency of less than 9.5 s and a maximum jitter of about 0.5 s. The latency constraint
results from the bu ering in the silicon vertex detector, which therefore requires a strobe
at less than 10 kHz. The drift chamber will trigger on one and one-half tracks, i.e., one
track traversing the full chamber radius together with one track traversing at least half the
chamber radius.
The main tracking chamber serves several other functions as well. For low momentum
particles, the chamber will provide particle identi cation through ionization loss (dE/dx)
measurements. For the helium-based gas mixtures under consideration, a resolution of just
under 7% should be attainable for the dE/dx measurement, allowing =K separation up
to 700 MeV=c. This capability is complementary to the DIRC in the barrel region, but is
essential in the extreme backward direction where no dedicated particle identi cation device
is foreseen.
The drift chamber also allows the reconstruction of secondary vertices, such as decays of
KS0 s outside the silicon detector volume. For this purpose, the chamber should be able to
measure not only the transverse coordinate, but also the longitudinal position of tracks with
good (1 mm) resolution. Good z resolution also aids in matching drift chamber and silicon
tracks, and in projecting tracks to the DIRC and calorimeter.
Finally, the chamber must be operational with a margin of safety at the expected backgrounds, which are predicted to be 5 kHz/cell for the innermost layers. For a ve-year
period of operation, the integrated charge is expected to be 0.025 C/cm for sense wires in
the innermost drift chamber layers, where variations in z and have been included. Details
are described in Chapter 12.
5.2 Tracking Chamber Overview
In order to meet the requirements outlined above, a small-cell drift chamber has been chosen
for the BABAR detector. Low-Z materials, including a helium-based gas and aluminum wires,
will be used in the design. The chamber, illustrated in Figure 5-2, is 280 cm in length and
occupies the radial space between 22.5 cm and 80 cm. It is bounded by the support tube at
its inner radius and the particle identi cation device at its outer radius. Forty layers of wires
will ll this volume, providing up to 40 spatial and ionization loss measurements for charged
particles with transverse momentum greater than 180 MeV=c. Two alternatives have been
considered for the wire and layer arrangement: a baseline axial/stereo chamber containing
Technical Design Report for the BABAR Detector
154
Drift Chamber
Tracking chamber geometry showing the conical endplates. The chamber
is o set in z from the interaction point and extends 1.66 m in the forward direction and
1.11 m in the backward direction. A ten-superlayer structure is indicated. The drift chamber
envelope includes a 16 cm space at the backward endcap for readout cards, cables, and an
rf-shield. The forward endcap includes a 6 cm space for feedthroughs and an rf-shield.
Figure 5-2.
four superlayers of axial wires and six superlayers of stereo wires, and an alternative allstereo chamber containing only layers with positive and negative stereo-angle wires. The
wire arrangements are discussed in detail in Section 5.4. For an outer radius of 80 cm, there
are about 7000 sense wires of 20 m-diameter gold-plated tungsten and about 45,000 eld
wires of 55 m-diameter gold-plated aluminum. The exact ratio of the number of eld to
sense wires is still under study.
A chamber length of 166 cm in the forward direction has been chosen. This ensures sucient
coverage for forward-going tracks to avoid compromising the invariant mass resolution, while
at the same time does not threaten the electrical stability of the chamber or lead to a
prohibitively expensive calorimeter. In the backward direction, a length of 111 cm means
that particles with polar angles down to 435 mr penetrate as far as the mid-plane of the
chamber, which yields good tracking and dE/dx information.
The endplates, which carry an axial load of 2400 kg, are designed as truncated cones, with the
outer part inclined at 22:7 to the vertical. The conic shape provides a substantial increase in
mechanical strength and allows particles to enter the forward particle identi cation device
at close to normal incidence. In the forward direction, the tip of the endplate cone lies
on the 300 mr line. The inner sections of the endplates will be conically shaped as well.
This decreases the wire lengths, permitting larger stereo angles. A detailed comparison of
endplate shapes is given in a later section.
Technical Design Report for the BABAR Detector
5.3 Projected Performance
155
Momentum resolution expected for pions at = 90 for the ve-layer silicon
vertex detector and the 40-layer drift chamber in a 1.5 T magnetic eld. For the drift
chamber, an averaged radiation length of 340 m and a single-cell spatial resolution of 140 m
have been assumed. Both the axial/stereo and all-stereo chamber designs give the resolution
shown in the gure.
Figure 5-3.
The inner cylindrical wall of the drift chamber, which sits 15 mm outside the 0.005X0 support
tube, acts as a gas seal and is not foreseen to provide structural support. The goal is therefore
to minimize the inner wall material and the resultant contribution to multiple scattering.
To that end, the inner wall will be made out of carbon ber.
The outer cylindrical wall bears the wire load from the endplates. It is desirable to make this
wall as thin as possible, so that the performance of the calorimeter is not degraded. Carbon
ber will therefore be used for the outer wall as well as the endplates.
The chamber will be lled with a helium-based gas at or near atmospheric pressure. In order
to minimize the amount of material before the forward endcap, all readout electronics and
HV distribution will be placed on the backward endplate.
5.3 Projected Performance
The expected transverse momentum resolution for the ve-layer silicon vertex detector and
the 40-layer drift chamber in a 1.5 T magnetic eld is shown in Figures 5-3 and 5-4. The
Technical Design Report for the BABAR Detector
156
Drift Chamber
Figure 5-4. Transverse momentum resolution vs. center-of-mass angle for pions produced
with three di erent center-of-mass momenta. The dotted curves show the total momentum
resolution p =p assuming a 100 m vertex constraint.
large, low-mass tracking volume provides excellent resolution1. For simplicity, a radiation
length of 340 m has been used in the calculations: the helium-based gas represents an 800 m
radiation length, while the wires, when averaged over space, present a slightly larger amount
of material compared to the gas. The inclusion of the e ects of discrete wires would result
in better chamber performance than that shown. However, systematic e ects which would
be encountered when reconstructing tracks from actual data have not been included. Such
calculations therefore represent the optimal performance of the tracking system, and have
been used in a comparative way when exploring various design choices.
Figure 5-4 shows that nearly symmetric forward-backward coverage has been achieved in the
center-of-mass frame for momenta above 200 MeV=c. For a He:C4 H10 80:20 gas mixture, with
the material from the wires and gas averaged together and with a single-cell spatial resolution
of 140 m, the expected resolution varies from 0.2% for low momentum tracks to 0.5% for high
momentum tracks in the central region. The resolution degrades considerably for low-angle
tracks, which exit the chamber through the endplates. At low momenta, the total momentum
resolution p=p is dominated by the polar angle uncertainty caused by multiple scattering
in the silicon. This uncertainty can be signi cantly reduced by using the primary vertex to
impose a constraint on the track. Conservatively assuming the vertex resolution to be 100 m
yields the dotted p=p curves shown in Figure 5-4. At momenta above 1 GeV=c, however,
The calculations were performed using the TRACKERR [Inn93] package, which employs the Kalmanlter technique to analytically propagate the error matrix. The simulation assumes an average 140 m single
cell resolution and multiple scattering contributions from the drift chamber gas and wires.
1
Technical Design Report for the BABAR Detector
5.3 Projected Performance
[mrad]
157
: cos = 0 gAxial/Stereo
: cos = 0:8
: cos = 0 gAll Stereo
: cos = 0:8
3
2
1250
z
[m]
1000
750
500
1
250
0
1
2
p
3
4
5
[GeV/c]
0
1
2
p
3
4
5
[GeV/c]
Expected angular resolution (left) and expected z resolution, both at the
DIRC, for two di erent polar angles and for the two chamber con gurations.
Figure 5-5.
the error in the transverse component dominates the total momentum resolution. The z
position and polar angle resolutions at the DIRC are shown in Figure 5-5 for various momenta
and polar angles. These estimates were obtained using a modi ed version TRACKERR to
simulate the e ect of propagating errors outward in the drift chamber to the DIRC radius.
In optimizing the performance of the tracking system, a number of global design parameters
have been examined. Neglecting material, the transverse momentum resolution is directly
proportional to the single-cell spatial resolution and inversely proportional to the square
root of the number of measurements, to the magnetic- eld strength, and to the square of
the measured arc length in the transverse plane. However, for most momenta encountered
in this experiment, multiple scattering in the material within the tracking volume is also an
important factor, dominating the resolution below 0.5 GeV=c. The range of mass resolutions
which could be provided by the tracking system is illustrated in Table 5-1. It would be
tempting to enlarge the radius of the drift chamber beyond 80 cm. However, this is a very
expensive direction for design change, since it also drives the volume of CsI required for the
calorimeter. Removing the PEP-II support tube is an equally expensive way to improve
tracking resolution, since it leads to a much higher risk of reduced luminosity from the
machine. This has left magnetic eld strength as the most cost-e ective way to obtain
suitable B mass resolution for CP studies.
Technical Design Report for the BABAR Detector
158
Drift Chamber
Outer
Support- Gas- Resolution B Field Mass Resolution (MeV/c2 )
Radius (cm) Tube Wires
(m)
(Tesla) B 0 ! +, B 0 ! D+D,
80
Yes
He-Al
140
1.0
30.7 0.3 5.69 0.09
80
80
95
95
80
80
Yes
No
Yes
No
Yes
Yes
He-Al
He-Al
He-Al
He-Al
Ar-Cu
He-Al
140
140
140
140
140
210
1.5
1.5
1.5
1.5
1.5
1.5
20.8 0.2
17.1 0.1
14.0 0.1
12.5 0.1
25.2 0.2
26.0 0.2
4.70 0.08
4.37 0.08
4.07 0.07
3.73 0.06
5.76 0.09
5.09 0.09
The e ect of chamber parameters on mass resolution for B ! + , and
B ! D+ D,. The entries in bold represent the parameters of the baseline design. The
underlined values denote a variation with respect to the baseline. The cases with no support
tube are unrealistic to the extent that the e ects of a nite thickness for the drift chamber
inner wall are not considered.
Table 5-1.
5.4 Drift System Design
5.4.1
Cell Design
The drift chamber uses small cells arranged in 40 concentric layers about the axis of the
chamber. We have chosen a rectangular cell shape, approximately 13 mm by 19 mm along
the radial and azimuthal directions, respectively. In the baseline arrangement, drift cells are
arranged in superlayers with either an axial or stereo orientation.
Each cell has one sense wire surrounded by a rectangular grid of eld wires, as shown in
Figure 5-6. The sense wires are made of gold-coated tungsten, 20 m in diameter, tensed
with a weight of 50 g. The de ection due to gravity is 120 m at the mid-length of the
longest wires. Wire lengths vary from 2.6 m to 2.8 m, due to the conic shape of the end
plates. Approximately +1:8 kV is applied to the sense wires with the eld wires at ground
potential, giving an avalanche gain of approximately 5 104.
The eld wires are gold-coated aluminum with a diameter of 55 m. This diameter is
suciently large to keep the electric eld on the wire surface below 20 kV/cm, in order
to avoid whisker growth on the wires. A tension of 53 g is applied to the eld wires to match
the gravitational sag of the sense wires. This tension is roughly one-half the yield tensile
strength of the wire.
Technical Design Report for the BABAR Detector
5.4 Drift System Design
159
.0
-90.6
.0
89.2
.0
-87.7
.0
86.2
-52.0
-84.7
-50.1
83.2
-48.3
-81.6
-46.5
80.0
48.7
46.7
-78.4
76.8
44.7
-75.1
42.7
73.3
.0
.0
-71.5
69.7
.0
-67.8
.0
65.9
Layers of rectangular cells grouped into axial-stereo (left plot), and all stereo
(right plot) con gurations. The axial-stereo case has four layers of cells per superlayer and
an extra layer of eld wires at each axial-stereo boundary to enclose the cells. In the allstereo case, the layers alternate between positive and negative stereo angles. The stereo
angles ( mr) of the layers are also shown for each case.
Figure 5-6.
5.4.2 Layer Arrangement
Two di erent con gurations of wire layers are being considered (Figure 5-6). The baseline
con guration has both axial (A) and stereo (U and V) superlayers, while an optional layout
having only stereo layers that alternate with positive and negative angles from layer to layer
has also been studied.
Technical Design Report for the BABAR Detector
160
Drift Chamber
Axial-Stereo Layer Arrangement
The baseline version of the chamber layout contains ten superlayers with four layers per
superlayer. The layers within each superlayer have the same cell count, with cells staggered
by half a cell from one layer to the next. This provides a constant cell pattern at all
azimuthal angles, which allows local segment nding and left-right ambiguity resolution
within a superlayer. The stereo angles of the superlayers alternate between axial (A) and
stereo (UV) pairs, in the order AUVAUVAUVA, as shown in Figure 5-6. The di erence
between the minimum and maximum cylindrical radius of a stereo-layer is known as its
stereo sagitta. For each UV pair of superlayers, a stereo sagitta of 5.5 mm is chosen at the
innermost layer. This sets the stereo angle for the inner layer, while stereo angles in the
other layers of the UV pair are determined by requiring that the wires have the same twist
angle (the change in azimuthal angle over the length of the wire) to preserve the cell shapes
along the wires. The tangent of the stereo angle varies as the square root of the layer radius
within the UV superlayers.
The charged track trajectories in the magnetic eld appear circular in the axial views. Three
of the ve track parameters (the curvature , azimuthal angle , and distance to origin r0)
can be determined from a circle t in the axial view, while the z position and polar angle
are determined using the stereo layers. The hardware trigger performs a fast transverse
momentum determination using only the axial layers for tracks traversing a signi cant
portion of the chamber. The cell widths for layers in the innermost superlayer vary by
7%, requiring separate time-distance relationships for each layer. This variation decreases
at larger radii. Because of the hyperboloidal geometry of stereo layers, the radial space
between layers at an axial-stereo boundary varies along the wire, so an extra enclosing layer
of eld wires is added at the boundaries to provide cell shape uniformity. These additional
eld layers introduce 5 mm of dead space at six axial-stereo boundaries. A bias voltage of
230 V is applied to the corner eld wires in the enclosing layers to sweep out the ions in the
dead regions. Table 5-2 gives the radii and stereo angles of all the layers.
The 5 mm enclosing space and the 5.5 mm stereo sagitta result in 10.5 mm radial regions of
free space at three radial locations on the end plates, which are otherwise fully populated
with feedthroughs. This space may be used to our advantage for gas ports and mechanical
stand-o s for cable supports.
All-Stereo Layer Arrangement
The all-stereo optional layout also has 40 layers, but with alternating positive and negative
stereo angles. The cell width is nearly constant in all layers, with the cell count increasing
by ve from one layer to the next layer outward in radius. This constant cell size throughout
the whole chamber allows (to rst order) a single time-distance relationship to be used for
Technical Design Report for the BABAR Detector
5.4 Drift System Design
161
Layer Number Radius at Radius at Twist Stereo Cell Width
of
z=0
z = 1380 (mrad) (mrad)
(mm)
cells
(mm)
(mm)
1
90
241.8
241.8
.0
.0
16.9
2
90
254.4
254.4
.0
.0
17.8
3
90
266.9
266.9
.0
.0
18.6
4
90
279.5
279.5
.0
.0
19.5
5
108
297.0
302.6
386.0
42.0
17.3
6
108
309.3
315.2
386.0
43.8
18.0
7
108
321.7
327.7
386.0
45.5
18.7
8
108
334.0
340.3
386.0
47.3
19.4
9
126
347.3
352.9
-356.8 -45.3
17.3
10
126
359.7
365.5
-356.8 -47.0
17.9
11
126
372.0
378.0
-356.8 -48.6
18.6
12
126
384.4
390.6
-356.8 -50.2
19.2
13
144
408.2
408.2
.0
.0
17.8
14
144
420.8
420.8
.0
.0
18.4
15
144
433.3
433.3
.0
.0
18.9
16
144
445.9
445.9
.0
.0
19.5
17
162
463.4
469.0
308.7
52.2
18.0
18
162
475.8
481.6
308.7
53.6
18.5
19
162
488.3
494.1
308.7
55.0
18.9
20
162
500.7
506.7
308.7
56.4
19.4
21
180
513.7
519.3
-293.1 -54.9
17.9
22
180
526.2
531.9
-293.1 -56.2
18.4
23
180
538.6
544.4
-293.1 -57.6
18.8
24
180
551.0
557.0
-293.1 -58.9
19.2
25
198
574.6
574.6
.0
.0
18.2
26
198
587.2
587.2
.0
.0
18.6
27
198
599.7
599.7
.0
.0
19.0
28
198
612.3
612.3
.0
.0
19.4
29
216
629.8
635.4
264.7
60.7
18.3
30
216
642.3
648.0
264.7
61.9
18.7
31
216
654.8
660.5
264.7
63.1
19.0
32
216
667.2
673.1
264.7
64.3
19.4
33
234
680.1
685.7
-254.7 -63.0
18.3
34
234
692.6
698.3
-254.7 -64.2
18.6
35
234
705.1
710.8
-254.7 -65.3
18.9
36
234
717.6
723.4
-254.7 -66.5
19.3
37
252
741.0
741.0
.0
.0
18.5
38
252
753.6
753.6
.0
.0
18.8
39
252
766.1
766.1
.0
.0
19.1
40
252
778.7
778.7
.0
.0
19.4
Table 5-2.
Radii and stereo angles of all layers in the axial-stereo chamber.
Technical Design Report for the BABAR Detector
162
Drift Chamber
all the cells. The allowed stereo sagitta of 15 mm determines the stereo angles for each layer,
for xed length wires. The stereo angles range from 66 to 115 mr, from the inner to the outer
layer. The layer arrangement is shown in Figure 5-6. Table 5-3 gives the radii and stereo
angles of all layers. Alternative arrangements with as few as 36 layers could be considered
in this layout. Likewise, consecutive layers could be arranged as superlayers with the same
cell count and stereo angle, by sacri cing a small amount of cell uniformity.
Each stereo view has 20 equally spaced layers in the baseline design. In this geometry,
tracks are found in each of the views using local-continuity algorithms. For tracks with large
curvature, or where overlaps occur, this geometry may be advantageous. Given the projected
resolutions in azimuthal angle and curvature , the probability of confusion in matching
the two views is small due to the relatively low density of tracks in an (4S ) event. Cell
widths in this design vary over the length of the chamber by a maximum of 3% for the
innermost layer, decreasing towards larger radii. The cell height in the radial direction is
constant.
5.4.3
Total Channel Count
The total channel count in either of the internal layout schemes described above depends on
the choice for the total number of layers in the chamber, the channel increment from one
layer to the next (or from one superlayer to the next), and the number of channels assigned
to the rst layer of the chamber. The number of options for these three design parameters
is quite limited, as shown in Table 5-4.
The choice of 40 layers is mandatory for an axial-stereo layout. This allows four axial
superlayers to complete the primary function of providing a p measurement, along with
three pairs of UV stereo superlayers (with just 12 layers per view). Four layers per superlayer
allow three-out-of-four majority logic for triggering and oine segment nding. In principle,
the all-stereo design would allow a reduction to as few as 36 layers; this would then be
comparable to the ARGUS drift chamber, but with only two rather than three views.
The cell aspect ratio, i.e., ratio of width to height, is de ned by the choice of the increment
in the cell count, , from one layer to the next in proceeding radially outward. In the case
of a superlayer design, the cell count is kept xed for four and is then increased by 4.
The simplest designs keep = xed over the entire chamber, leading to a uniform cell
size. Choosing = 6, as has been used for both ARGUS and CLEO II, produces a nearly
square cell in cross section. This is well-suited to the low average momentum in (4S )
decays, which results in large angles of incidence on the drift cells with respect to the radial
direction. The nearly circular isochrones which dominate a large fraction of the drift-cell
response for a square cross section are optimal for handling such arbitrary entrance angles.
t
i
i
Technical Design Report for the BABAR Detector
5.4 Drift System Design
163
Layer Number Radius at Radius at Twist Stereo Cell Width
of
z=0
z = 1380 (mrad) (mrad)
(mm)
cells
(mm)
(mm)
1
80
242.2
258.7
718.2
65.8
19.0
2
85
255.5
272.0
-700.2 -67.5
18.9
3
90
268.8
285.3
683.5
69.2
18.8
4
95
282.1
298.6
-667.9 -70.8
18.7
5
100
295.5
312.0
653.4
72.4
18.6
6
105
308.8
325.3
-639.7 -74.0
18.5
7
110
322.1
338.6
626.9
75.5
18.4
8
115
335.4
351.9
-614.8 -77.0
18.3
9
120
348.8
365.3
603.4
78.5
18.3
10
125
362.1
378.6
-592.6 -79.9
18.2
11
130
375.4
391.9
582.4
81.4
18.1
12
135
388.7
405.2
-572.7 -82.8
18.1
13
140
402.1
418.6
563.4
84.1
18.0
14
145
415.4
431.9
-554.6 -85.5
18.0
15
150
428.7
445.2
546.2
86.8
18.0
16
155
442.0
458.5
-538.2 -88.1
17.9
17
160
455.4
471.9
530.5
89.4
17.9
18
165
468.7
485.2
-523.1 -90.7
17.8
19
170
482.0
498.5
516.0
91.9
17.8
20
175
495.3
511.8
-509.2 -93.1
17.8
21
180
508.7
525.2
502.7
94.4
17.8
22
185
522.0
538.5
-496.4 -95.6
17.7
23
190
535.3
551.8
490.3
96.7
17.7
24
195
548.6
565.1
-484.5 -97.9
17.7
25
200
562.0
578.5
478.8
99.1
17.7
26
205
575.3
591.8
-473.4 -100.2
17.6
27
210
588.6
605.1
468.1 101.4
17.6
28
215
601.9
618.4
-463.0 -102.5
17.6
29
220
615.3
631.8
458.1 103.6
17.6
30
225
628.6
645.1
-453.3 -104.7
17.6
31
230
641.9
658.4
448.7 105.7
17.5
32
235
655.2
671.7
-444.2 -106.8
17.5
33
240
668.6
685.1
439.8 107.9
17.5
34
245
681.9
698.4
-435.6 -108.9
17.5
35
250
695.2
711.7
431.5 110.0
17.5
36
255
708.5
725.0
-427.5 -111.0
17.5
37
260
721.9
738.4
423.6 112.0
17.4
38
265
735.2
751.7
-419.8 -113.0
17.4
39
270
748.5
765.0
416.1 114.0
17.4
40
275
761.8
778.3
-412.5 -115.0
17.4
Table 5-3.
Radii and stereo angles of all layers in the all-stereo chamber.
Technical Design Report for the BABAR Detector
164
Drift Chamber
Nlayers N1
40
40
36
36
51
w
Total
Comments
Ratio
(cm)
Channels
90 (18)/4 1.343{1.550 1.69{1.95
6840
Axial-Stereo Design
80
5
1.308{1.429 1.74{1.90
7100
All-Stereo Design
72
4
1.571
2.15
6000 (5760)
90
5
1.257
1.72
7500 (7200)
108
6
1.047
1.43
9000 (8640) Ideal
65
4
1.571
2.39
4860
82
5
1.257
1.91
6102
All-stereo Option
98
6
1.047
1.59
7308
All-stereo Option
60
6
1.047
1.88
5940
ARGUS
96
6
1.047
1.50
12240
CLEO II
w=h
Table 5-4. Total channel counts for the baseline drift chamber designs, compared with
various alternative solutions with di erent numbers of layers (Nlayers ), cell aspect ratios
(w=h) as derived from possible layer-to-layer cell increments (), or rst-layer channel
counts (N1 ). The total channel counts in parentheses are for an axial-stereo design, with a
four-layer constant-channel superlayer structure.
The present designs for the BABAR chamber compromise on the choice of layer increment
in two ways: (1) by choosing to vary slightly from one layer to the next, and (2) by
deviating signi cantly from the ideal square-cell design. Drift cell simulations have shown
that increasing the w=h ratio beyond 1.65 begins to result in performance degradation.
The drift cells in either of our design layouts are distinctly rectangular, with the width
being 30{55% larger than the height. These are already close to the maximum feasible for
good performance in the (4S ) environment, as shown in Figure 5-7, leaving little room for
additional reduction in channel count.
The last parameter to be considered is the number of cells in the rst layer of the chamber.
Here, the robustness of the inner layers of the chamber with respect to occupancy is of
primary concern. The design choice for BABAR lies midway between ARGUS and CLEO II,
despite the fact that circulating beam currents in PEP-II will be about 50 times larger
than at DORIS II. The ARGUS value was chosen to accommodate the charged multiplicity
in (4S ) decays. At BABAR, considerable e ort has been made to understand and control
beam-related backgrounds. However, the potential for higher rates has led to a conservative
choice of N1 = 80 or 90 in our design.
Technical Design Report for the BABAR Detector
5.4 Drift System Design
165
Layer eciencies (upper) and distortion free regions (lower) as a function of
the cell width. Layer eciencies account for cases where 2 adjacent cells in a layer contain
information. Distortion free is de ned to be cases with less than 100 m corrections for
angle of incidence.
Figure 5-7.
Technical Design Report for the BABAR Detector
166
Drift Chamber
Isochrones for rectangular cells in superlayers. An endplate view is shown of
the stereo layers on either side of a four-layer axial superlayer, with enclosing eld wires. A
230 V potential on the cell's corner wires in the enclosing layers clears out the ions between
these layers. The plots show equal drift time contours at 100 ns intervals.
Figure 5-8.
5.4.4
Cell Studies
A simulation package (DCSIM), which models drift chamber responses to charge deposition
for given wire con gurations, has been used to study various cell properties, including drift
time isochrones, time-distance relationships, distortions, and gain variations.
Isochrones for the rectangular cells in axial-stereo superlayers are shown in Figure 5-8. A
similar plot for the all-stereo design is shown in Figure 5-9. The cells have circular contours
near the sense wire but become distorted near the eld wires. In the superlayer case, a
Technical Design Report for the BABAR Detector
5.4 Drift System Design
167
Figure 5-9. Isochrones for all-stereo cells. The plots show equal drift time contours at
100 ns intervals.
potential of +230 V at the cell corners in the enclosing layers clears out the electrons in the
dead space between these enclosing layers.
Although isochrones show drift times of electrons anywhere in the cell, they do not easily
describe the collection of electrons originating along tracks. The time-distance relationship
of tracks in small cells has been simulated and compared with measurements made with the
Prototype I chamber. The simulation program generates tracks and computes the average
drift time of electrons over a short track segment centered about the minimum arrival point.
The length of a short segment used in the simulation was made equal to the mean free
path between primary electrons. Figure 5-10 shows the agreement between the simulated
time-distance curves and measurements in Prototype I with an 80:20 mixture of He:C4 H10
gas.
These simulated time-distance relationships have then been used to study other cell properties. The time-distance dependence on the track entrance angle has been calculated, along
with the distortions which would arise if a single time-distance function were used over a
range of track angles. For each drift time in a cell the track distance from the wire at various
Technical Design Report for the BABAR Detector
168
Drift Chamber
12
Distance (mm)
10
105
75
90
8
6
4
2
0
0.0
0.2
0.4
0.6
0.8
1.0
Time (µs)
Figure 5-10. Simulated and measured time-distance functions. The solid curves are the
simulated time-distance curves for track angles of 75, 90, and 105 , where 90 is along the
radial direction. Measurements in a prototype chamber with no magnetic eld are shown
by the dashed curve.
track angles is calculated, and the average and rms spreads in distance over a range of track
angles are determined. The rms spread is a measure of the distortion that a single timedistance function would have due to the e ects of track angles in the cell. Figure 5-11 shows
the rms spread in distance over track angles within 30 of the radial direction for the four
layers within a superlayer. The left plot has all eld wires at ground potential, while the
right plot has compensated voltages of +230 V at cell corners on enclosing layers and ,120 V
on the pairs of wires between the sense wires. The distortions are smaller in the latter case.
One can compare distortions for various cell geometries by recording the distance of closest
approach (DCA) at which the rms variation is 100 m. For the rectangular cell shown, this
is at 7.4 mm for vertical tracks, which is 78% of the 9.5 mm geometric half-width of the cell.
Table 5-5 shows the distance of closest approach values at the distortion limit of 100 m for
various cell and layer geometries.
The spatial resolution has been determined for those parts of the cell in which a sucient
number of electrons is collected within the leading edge of the signal, de ned to be the case
in which two electrons arrive within 10 ns of the theoretical minimal drift time. The DCA
Technical Design Report for the BABAR Detector
5.4 Drift System Design
169
0.8
0.8
0.7
0.7
0.6
1
4
3
1
0.5
2
0.4
RMS (mm)
0.4
RMS (mm)
2
0.6
4
0.5
3
0.3
0.2
0.1
0.3
0.2
0.1
0.0
0.0
0
1
2
3
4
Track DCA (mm)
5
6
7
8
9 10 11 12
0
1
2
3
4
5
6
7
8
9 10 11 12
Track DCA (mm)
Distortions in a rectangular cell vs. the track's distance of closest approach.
The distortions are calculated over a range of track angles with a single time-distance
relationship. The four curves show distortions in the four layers in a superlayer in the axialstereo design, for tracks which are within 30 of the radial direction (i.e., pt > 350 MeV=c
in an 80 cm drift chamber). The left plots are for the case with all eld wires at ground
voltage, while the right plot has +230 V on the cell corners along the enclosing layers and
,120 V on each pair of eld wires between the sense wires.
Figure 5-11.
Cell Size Distortion Performance
Cell Geometry
dr dz
% of cell
% of cell
(mm mm) (see text) (see text)
Rectangular 6:1 15.0 20.2
78.
73.
Square 5:1
15.0 15.4
86.
71.
Square 4:1
15.0 15.4
76.
69.
Hexagonal 5:1 15.0 20.2
80.
74.
Table 5-5. Time-distance distortions and resolution performance of cells. The distortion
column gives the percentage of the cell width that has less than 100 m of rms variation
in the time-distance relationship. The performance column gives the percentage of the cell
width that produces signals with at least two primary electrons collected within the rst
10 ns of the pulse shape. These values are for tracks within 30 of the radial direction in
an 80:20 He:C4 H10 gas mixture and a 1.5 T magnetic eld.
Technical Design Report for the BABAR Detector
170
Drift Chamber
Spatial resolution performance in a rectangular cell, as described in the
text. The jagged contour depicts track angles and impact parameters at which an average
of at least two primary clusters will contribute to the rst 10 ns of the pulse shape.
Figure 5-12.
has been calculated at each track angle for regions which satisfy this criterion. Tracks with
larger DCA values generally have shorter collection regions within the 10 ns window, and
therefore poorer resolutions, while tracks with smaller DCA values have better resolutions.
A contour plot of these DCA performance limits as a function of the track angle is shown in
Figure 5-12 for rectangular cells. The ratio of the area within the contour and the geometric
area of the cell is a gure of merit for the cell. Table 5-5 shows these values for various cell
geometries with the number of eld wires per sense wire shown (6:1). The 5:1 square cell has
the least distortion. The rectangular and hexagonal cells are somewhat worse. The square
cell (4:1 eld wire to sense wire ratio), having only one eld wire between sense wires, has
the most distortion.
5.4.5
Gain Variations
Gain variation as a function of z position has been studied using DCSIM, which computes
charge densities on sense wires. The gain changes within cells are based on the approximate
relation that a 1% change in the charge on the wire results in a 20% change in avalanche
gain. In the axial-stereo design, the charge on the wire varies due to the varying cell widths
within superlayers and the varying radial size between enclosing layers at boundary cells.
The largest gain variation of boundary cells is found to be 3%, while the largest gain
Technical Design Report for the BABAR Detector
5.4 Drift System Design
171
variation between an inner layer and an outer layer in a superlayer is 4%. In the all-stereo
case, the 3% increase in cell width at the ends of the wire length leads to a gain variation
of less than 2% along the wire length. Gain changes due to eld wires' phase locations
relative to the sense wires from cell to cell in a layer are miniscule in the all-stereo case.
5.4.6
Electrostatic Forces and Stability
Electrostatic forces and stability limits have been simulated using DCSIM. The accuracy of
the program has been veri ed with observed wire de ections in the 100-wire test chamber,
Prototype I. For this purpose, the 2.5 m-long sense and eld wires were tensioned with only 9 g
so that wire instability could be observed at nominal high voltage. For the BABAR chamber,
at the nominal stringing tension, the predicted de ection due to electrostatic forces is 160 m
for the worst case wire, and the instability point is far from the operating point. Therefore,
for the present cell con guration, it appears that there is no problem with electrostatic
de ections or stability.
5.4.7
Pattern Recognition Studies
Axial-Stereo Case
In similar chambers in other detectors, pattern recognition is performed by nding track
segments within each superlayer, and then linking segments to form tracks. The left-right
ambiguity resolution is done during the segment nding. The axial segments are linked
to form circles in the x-y plane, while the stereo segments provide the z and coordinates.
After segments have been linked, a full three-dimensional t is made to all the wire hits. Part
of this pattern nding and tting has been done for the prototype chamber. Some losses
may arise at lower momentum due to the reduced eciency for segment nding within a
superlayer or diculties in segment linking for tracks with signi cant curvature, particularly
for those particles which do not originate from the interaction point.
All-Stereo Case
Several studies concerning pattern recognition and track reconstruction in an alternatinglayer all-stereo chamber have been made in order to investigate potential limitations of having
only two views. In particular, two independent approaches to the pattern recognition problem have achieved eciencies between 95% and 99% (depending on method and backgrounds)
for nding tracks despite the noncircular projections in the stereo views. The absence
Technical Design Report for the BABAR Detector
172
Drift Chamber
of superlayers also raises a concern about unresolvable left-right ambiguities in a single
view. This has been shown to occur for less than 0.5% of tracks from generic (4S ) events,
and appears to be further reduced by a factor of 2 by the careful choice of the azimuthal
o sets of each layer. Most of the ambiguities that do occur are expected to be resolved by
information from the other view. Finally, the possible mismatching of track projections found
in the two views has been investigated using a parameterized approach [Bri93] and with full
pattern recognition and tting. Both studies indicate that genuine mispairing of projections
from opposite views happens in about 0.5% of all cases. With extreme backgrounds, and
depending on the matching criteria, mismatches for up to 2.8% of tracks have been found.
Most of these, however, arise from the inclusion of small segments of an additional track, or
fake track, in one view; the actual contribution to track- nding ineciency is estimated to
be less than 1%.
5.5
Gas Choice and Properties
The choice of gas for the drift chamber is driven primarily by the needs to reduce the
total amount of material, minimize multiple scattering for low momentum tracks, and to
operate eciently in a 1.5 T magnetic eld. These requirements are quite well satis ed
by mixtures of helium and hydrocarbons. In particular, mixtures with 10{30% of various
hydrocarbons a ord a small Lorentz angle, good resolution, and low multiple scattering. In
Table 5-6, the properties of the baseline gas choices are shown, along with other possibilities
which have been considered for use in the BABAR Drift Chamber. The drift velocities and
Lorentz angles are determined with the Boltzmann integration code MAGBOLTZ [Bia89].
The dE/dx calculations are performed with a modi ed version of a program from Va'vra et
al. [Vav82]. The table shows that the helium mixtures under consideration have a radiation
length more than ve times larger than that of HRS gas, Ar:CO2 :CH4 89:10:1, a commonly
used argon-based mixture. Figure 5-13 shows the calculated drift velocity vs. electric eld
for four of the gases in the table. The helium-based gases are not saturated, but lead to
better performance than typical argon mixtures, since the smaller Lorentz angle results in a
more uniform distance-time relationship.
The two gases which have received the most attention are an 83:10:7 mixture of He:CO2 :C4 H10
and an 80:20 mixture of He:C4 H10. These gases and the benchmark argon mixture, HRS
gas, have been studied in a prototype chamber, which is described in Section 5.11. The
results for the spatial resolution obtained in these studies are summarized in Figure 5-14.
Both helium-based gases provide good spatial resolution. The 80:20 He:C4 H10 mixture is our
preference based on measured spatial resolution and simulated dE/dx resolution, although
there are concerns regarding the safety implications if this proves to be a ammable gas.
We have also studied whether these gases exhibit gain changes or high-voltage breakdown
Technical Design Report for the BABAR Detector
5.5 Gas Choice and Properties
Gas Mixture
Ar:CO2 :CH4
He:C2 H6
He:DME
He:C3 H8
He:C4 H10
He:CO2 :C4 H10
173
Separation
Ratio X0 Primary
vd
L
Resol. p for 3 # of at
(m) Ions/cm (m=ns) (deg) (%) (MeV=c) 2.6 GeV=c
89:10:1 124 23.6
49
52
7.3
665
2.4
50:50 686 23.1
31
45
6.6
720
2.1
70:30 723 22.4
6
8
6.7
720
2.1
70:30 733 24.1
24
36
6.5
730
2.2
80:20 807 21.2
22
32
6.9
710
2.1
83:10:7 963 13.8
19
26
8.5
660
1.7
dE/dx
K=
Properties of various gas mixtures at atmospheric pressure and 20 C. The
drift velocity (vd ) and Lorentz angle (L ) are given for an electric eld of 600 V/cm with
no magnetic eld and with 1.5 T, respectively. The dE/dx resolution is calculated for a
minimum-ionizing particle. Also listed are the momenta below which there is at least 3
K= separation, and the K= separation at 2.6 GeV=c.
Table 5-6.
Calculated and measured drift velocities as a function of electric eld for
zero magnetic eld. The calculations use the code of Biagi (see text). The solid curve and
crosses [Pla92] are for the 80:20 He:C4 H10 mixture; dot-dashed curve and squares [Boy92] are
for 78:15:7 He:CO2 :C4 H10 ; the dotted curve and diamonds [Cin91] are for 70:30 He:DME;
and the dashed curve and circles [Boy92] are for (89:10:1) Ar:CO2 :CH4 (HRS gas).
Figure 5-13.
Technical Design Report for the BABAR Detector
174
Drift Chamber
HRS Ar:CO2:CH4 [89:10:1]
1775 V
LoZ1 He:CO2:i-C4H10 [83:10:7]
1775 V
LoZ2 He:i-C4H10 [80:20]
1800 V
Uno93 He:C2H6 [50:50]
2300 V
Pla92 He:DME [70:30]
1650 V
Cin91 He:DME [70:30]
300
Spatial Resolution (microns)
250
200
150
100
50
0
0
2
4
6
Drift Distance (millimeters)
8
10
Figure 5-14. The prototype drift chamber results for HRS gas, He:CO2 :C4 H10 , and 80:20
He:C4 H10 . Points represent data from the Prototype I chamber; curves are the results of
studies performed elsewhere.
after a period of irradiation. The aging studies were performed with a small proportional
counter and an 55 Fe source. All of the helium-based gases in Table 5-6 were tested except
for the DME mixture. The isobutane and propane mixtures showed negligible aging. The
50:50 He:C2 H6 and 78:15:7 He:CO2 :C4 H10 mixtures both had some gain loss, equivalent to
30% for an accumulated charge of 1 C/cm. These studies indicate that all of these gases
are appropriate for our purposes since the expected charge accumulation over the lifetime of
BABAR should be under 0.1 C/cm. Further long-term aging studies are planned.
Technical Design Report for the BABAR Detector
5.6 Mechanical Design
175
Angle (Degrees) Thickness De ection Maximum Stress
Outer/Inner
(mm)
(mm)
(MPa)
0/20
6.5
36.2
154
5/20
6.5
18.3
75
10/20
6.5
6.6
50
20/20
6.5
2.2
32
30/20
6.5
1.2
24
20/20
4.7
3.4
51
Finite-element calculations of de ections and stress for various endplate
shapes. A 2400 kg load was evenly applied between 22.5 and 80 cm and supported at
the outer radius. The endplate is supported at the outer radius only, and the maximum
de ection occurs at the inner radius.
Table 5-7.
5.6
5.6.1
Mechanical Design
Endplates
The chamber endplates must support the wire load of 2400 kg yet be of negligible thickness
in radiation lengths to minimize photon conversion before the calorimeter. The o set of the
chamber and the beam energy asymmetry at PEP-II lead to boosted events such that photons
with cos cm > 0:67 exit through the forward endplate, while only those with cos cm < ,0:91
exit through the rear endplate. The two endplates will be identical, but the readout and
high-voltage connections will be installed on the backward endplate only.
The choice of carbon ber as the structural material and the application of the double cone
geometry shown in Figure 5-2 allow for extremely thin endplates. The material proposed is
an intermediate modulus graphite ber IM7 [IM7] coupled with a toughened epoxy (977-2)
or a cyanate ester resin system (954-2). The mechanical advantage of the double-cone shape
is illustrated in Table 5-7 which shows material thicknesses, de ections under wire load,
and maximum stress for various endplate con gurations. The entries in the angle column
represent the angles of the outer and inner cones with respect to a at endplate|thus 0/0
would be a at end plate, and 20/20 is closest to the 22.7 symmetric cone illustrated in
Figure 5-2 as the baseline design. The scale for acceptable de ections is set by the 9 mm
elongation of the aluminum eld wires under the 53 g stringing tension. The factored limit
stress in carbon ber is estimated at 61 MPa, which includes a de-rating factor of 3 for the
holes, and a safety factor of 3.
Technical Design Report for the BABAR Detector
176
Drift Chamber
The double-conical endplate provides substantial mechanical advantages without seriously
degrading the forward tracking length, or introducing too large an angle for axial drilling of
the feedthrough holes. The radius of curvature at the cone tip (the transition between the
inner and outer cone) has been limited to 50 mm in order to reduce the stress associated
with sharper angles. The curvature also allows better laminate properties in the transition
area by reducing ber breakage due to expansion and contraction during the cure and allows
the hole pattern to be drilled without any discontinuity.
The endplate thickness of 6.5 mm corresponds to 2.8% of a radiation length; the feedthroughs
(1.2%), rf-shield (0.6%), and cable support spider (0.4%) roughly double the material thickness of the forward endplate to about 0.065X0. Cables, preampli ers, and HV boards add
an additional 0.1X0 to the backward endplate. The endplates will be laid-up on a mandrel,
covered with rubber cure pads, then bagged and cured in an autoclave. Both of the resin
systems under consideration are cured at 350F under a pressure of 100 psi. A 3 mm sacri cial
layer of nonstructural material will be applied to the outsides of the endplates. A considerable
fraction of this layer is then removed as at annuli are machined for each wire layer as shown
in Figure 5-15. These surfaces allow machining tools to make a perpendicular approach to
the surface when drilling the feedthrough holes and provide a registration surface for the
feedthroughs.
The holes for the eld wire feedthroughs will be approximately 2.5 mm in diameter, while the
sense wire holes will be 5{7 mm in diameter, depending on the choice of restringing options
currently under consideration. The holes will be drilled axially, which simpli es the drilling
process, the hole survey, and connections to the electro-mechanical boards. The wires exit
from the feedthroughs at the appropriate stereo angle, and the wire position accuracy is
therefore weakly coupled to the insertion depth of the feedthrough. The depth of insertion
is controlled by the machined annuli, which act as registration surfaces.
5.6.2
Inner Wall
For good pt resolution, it is essential to minimize the material between the silicon vertex
detector and the drift chamber. The inner wall of the chamber lies immediately outside the
0.005X0 support tube and represents a small amount of material by comparison. The inner
wall is not required to carry any of the wire load but must support any di erential pressure
between the inside and outside of the chamber.
An engineering case study has been performed [Smi94], which assumed that the inner
wall must withstand a test pressure of 2 kPa (20 mbar), and the results are summarized
in Table 5-8. While beryllium would provide a minimal amount of material, this option
would be very expensive and unlikely to be feasible with the required thickness of less than
0.5 mm. Therefore, the inner wall will be constructed from 0.4 mm of the same carbon ber
Technical Design Report for the BABAR Detector
5.6 Mechanical Design
177
Figure 5-15. Cross section through the endplate cone tip showing the feedthrough
seating. Narrow annuli are milled to seat feedthroughs in the cathode layers and alternate
with wider annuli containing the sense wire feedthroughs and, in the case of a 6:1 cell design,
pairs of eld shaping wires.
as the rest of the chamber. An rf-shield will be required which will contribute an additional
0.1% of a radiation length, and concerns about helium permeability may lead to a Mylar
layer as a gas seal.
5.6.3
Outer Wall
The outer wall bears the axial wire load between the endplates. The large circumference
allows this load to be supported by as little as a 1.6 mm-thick (0.006X0) carbon- ber tube. In
order to be robust against local impacts, a 3.2 mm-thick (0.012X0) option is being pursued
as more realistic. This is still a small amount of material in comparison to the 0.18X0
Technical Design Report for the BABAR Detector
178
Drift Chamber
Material
Thickness % of Radiation
(mm)
Length
Radel [RAD]
2.45
0.82
Vitrex PEEK 450G [VIT]
2.19
0.73
Carbon Fiber IM7/977-2 [IM7]
0.40
0.15
Beryllium
0.32
0.09
Inner wall thickness required to withstand a test di erential pressure of
20 mbar, determined from Roark's Formulae [You89] assuming no sti eners and ends held
circular.
Table 5-8.
contributed by the DIRC, which is located between the outer wall of the tracking chamber
and the calorimeter.
During operation, the chamber will be maintained at a constant temperature. However,
during transport, downtime, and other anomalous situations such as magnet quenches, relatively large temperature uctuations might be encountered; in such cases, thermal expansion
can lead to signi cant changes in wire tension. An aluminum outer wall would provide a
perfect thermal match to aluminum eld wires but lead to an increase in the tungsten sense
wires' tension of about 0.5% per C. The thermal expansion coecient of carbon ber is
well matched to that of the sense wires but leads to a decrease in the aluminum eld wire
tension of 0.8% per C. The sign of this e ect and the fact that the eld wires rather than
the sense wires are at risk make carbon ber preferable to an aluminum shell.
The outer tube will be constructed of the same material as the endplates. An aluminum foil
will be adhesively bonded to the outer surfaces to provide rf-shielding. The inner surface
may be treated with a carbon spray to ensure adequate electrical conduction to drain charge.
The low moisture absorption characteristic of the carbon ber is expected to eliminate any
concerns about humidity control.
5.6.4
Joints
The inner and outer walls will be installed after the chamber wires have been strung. The
outer joint is within the physics acceptance, so every e ort must be made to minimize the
amount of material. Both joints must occupy minimal radial space to preserve the maximum
possible tracking length. The outer wall will be attached using a double-labyrinth gluing
technique that has been used on several previous chambers. A conceptual sketch of the
outer joint is shown in Figure 5-16. For gluing, the chamber is turned vertically and epoxy
is injected via many syringes into the lower labyrinth and eventually ows to the upper.
Technical Design Report for the BABAR Detector
5.6 Mechanical Design
179
Glue Labyrinth
Carbon Fiber Cylinder
Gas Port / Stud
Cathode Wire
Sense Wire
Endcone
Figure 5-16.
Conceptual design for the outer-wall/endplate joint.
The glue injection ceases once the upper recess is lled and glue starts to appear at small
air-escape holes.
The inner joint is not within the acceptance region of the detector, so the amount of material
is not critical. A removable joint, using an O-ring gas seal, will be used. If at some future
date the support tube were removed, this would allow the option of replacing the inner wall
with a lower-mass alternative.
The inner and outer joint regions must also accommodate gas ttings and attachment points,
the latter being used both for supporting the chamber and for pretensioning the endplates
prior to stringing. A ring of studs will be embedded in the carbon ber endplates at the
outer and inner radii during the fabrication process. The studs alternate in an appropriate
ratio with gas ttings. The advantage of embedding during fabrication is that none of the
structural bers are broken.
5.6.5 R&D Program on Structural Components
The proposal to use carbon- ber laminates as the main structural component of the chamber
requires a thorough research and development program. Such materials are widely used for
applications such as satellite structures [Bra94], in which extreme physical conditions are
coupled with zero access for maintenance. For high-energy physics applications, the long
radiation length of carbon ber (about 25 cm) is an additional attraction. The R&D program
will examine basic material properties and fabrication techniques using small samples, or
coupons, followed by the evaluation of increasingly substantial engineering models which
address speci c areas of concern, leading eventually to the construction of a full-sized
prototype endplate. In parallel with this process, theoretical models will be developed to
allow the calculation of properties using nite element analysis and classical lamination
theory. The interaction between these parallel developments should lead to a complete
Technical Design Report for the BABAR Detector
180
Drift Chamber
mechanical speci cation and an integrated nite element analysis model of the entire chamber
prior to the construction of the full-sized prototype.
At the coupon level, investigations will include evaluation of basic material properties with
standard tests (American Standard Test Methods) such as tensile strength (ASTM D3039),
humidity absorption (ASTM D560), and adhesive tensile lap strength (ASTM D1002).
Design-speci c tests such as the evaluation of outgassing e ects on chamber aging and the
helium permeability of the proposed laminates will be performed. Extensive drilling tests can
also be carried out at this stage in order to investigate hole accuracy, ovality, breakout, toolbit wear, lubricant absorption, procedure elapsed time, and quality assurance techniques.
Engineering models will be developed to address the joint design with respect to strength,
gas-seal, and assembly procedure. Model tubes will be constructed to verify compression
and buckling calculations. The insertion, seating, and gas-seal of the feedthroughs will be
investigated, and a full segment of the endplate will be constructed to con rm the laminate
lay-up and cure procedures.
After the engineering model tests and the laminate design have been completed, and after
a manufacturing technique, a machining strategy, and quality assurance methods have been
developed, a full-sized prototype endplate will be constructed and partially drilled to verify
the entire fabrication procedure. Such a prototype may subsequently be further drilled to
investigate de ection properties and could be shipped to the stringing site for evaluation of
the proposed prestressing and load transfer technique.
5.6.6
Wires
Transverse momentum resolution is dominated by multiple scattering up to the highest
momenta of interest. This has led to the design of a chamber with low-mass eld wires
and a helium-based gas. Aluminum has a relatively long radiation length and is a good
eld wire candidate. Bare aluminum is not suitable, however, because oxidation forms an
insulating layer on the wire surface leading eventually to electrical discharges. Previous
chambers have used gold-coated aluminum, but thin uniform layers of gold have been hard
to achieve. We plan to use 55 m-diameter aluminum wire with a 0.50 m gold coating.
The contribution of the gold to the total radiation length is roughly equal to that of the
aluminum. Unfortunately, gold coatings this thin tend to ake o and provide nonuniform
coverage. Other alternatives, such as silicon carbide wires and nickel-coated aluminum wires,
have been investigated but proved unacceptable.
Aluminum wires also su er creep, causing wires to lose tension over time. With careful
choice of the wire type, this can be limited to an acceptable level of less than 0.1% of the
wire length per year after an initial rapid increase in length over the rst few hundred hours.
Technical Design Report for the BABAR Detector
5.6 Mechanical Design
181
The relative softness of aluminum means that considerable care must be taken to design a
reliable crimping procedure.
For the sense wires, 20 m tungsten is proposed, which would contribute approximately onehalf the material contained in gold-plated aluminum eld wires. Although tungsten-rhenium
has recently become popular as a wire material, its use has been ruled out for the BABAR drift
chamber due to the 50% increase in resistance, which would degrade the signal-to-noise
performance of the chamber. Some consideration is being given to a larger 30 m wire to
improve signal preservation of signal amplitude from the forward end of the chamber. This
would increase the sense wire material by a factor of 2.25, making it comparable to the eld
wires.
Wire aging and cross-talk tests will be performed, and a full understanding of the mechanical and electrical wire/feedthrough/endplate interface will be sought using a full-length
prototype chamber, Prototype II.
5.6.7
Feedthroughs
Feedthroughs insulate eld and sense wires electrically from the endplates and provide for
accurate wire placing. They must be inexpensive to manufacture, must provide a gas seal,
and must not react chemically with the chamber gas.
A preliminary eld-wire feedthrough design is shown in Figure 5-17. The body of the
feedthrough is a single piece of injection-molded plastic with a 2.5 mm precision outer
diameter and a length of about 15 mm. Wire positioning is provided by a separate brass
insert containing a 100 m-diameter precision hole. This has a 1 mm bend radius on the
feedthrough exit end to avoid kinking stereo wires. A metal insert (brass) was chosen to
provide low electric eld density at the interface with the plastic body, avoiding discharge
problems in the insulator. Delrin or celenex plastic insulation is presently being considered.
The wire is secured in the feedthrough using a gold-plated aluminum crimp pin of a design
similar to that used successfully in several TRIUMF-built drift chambers. The TRIUMF pins
were actually made of brass, which has the advantage of providing a more rigid contact for
connecting the electro-mechanical boards, but provides for a narrower tolerance for crimping
wires than a softer material such as aluminum. Good results in terms of wire breakage have
been obtained at SLD where aluminum wires were crimped in softer aluminum crimp pins.
The small diameter region of the crimp pin is the crimp locator, while the large diameter
region is used for electrical connections. This separate-function design is expected to reduce a
crimp-pin breakage problem observed at SLD, where the mechanical connections were made
in the same region as the crimp.
Technical Design Report for the BABAR Detector
182
Drift Chamber
Field-wire feedthrough design, showing crimp pin, Delrin sleeve and
precision brass insert for wire location.
Figure 5-17.
Wire positioning is provided by accurate centering of the 100 m locator hole in the brass
insert of the feedthrough. Measurements of 1000 feedthroughs manufactured using a similar
design for the Beijing Energy Spectrometer (BES) showed that the concentricity of the
precision hole with respect to the feedthrough outer diameter was 11 m. The feedthrough
itself is required to have an outer-diameter tolerance of 10 m, as was achieved with the BES
feedthroughs, and must be tted tightly in the endplate hole.
The gas seal is provided by a slight (1:48) taper at the back of the feedthrough, which may
be adjusted to provide the desired degree of press t into the endplate. The crimp pin is
press t in the plastic body and may be sealed with epoxy if necessary.
Technical Design Report for the BABAR Detector
5.6 Mechanical Design
183
The sense-wire feedthrough design will have a larger (5{7 mm) outer diameter and a longer
body (25 mm). This provides both thicker insulating walls and a longer projection from the
endplates to better shield the HV from the grounded endplate. The precision inserts will be
identical to those used for the eld wires, and a brass/copper crimp will be used.
Feedthrough prototypes are expected to be available in early 1995. Planned tests include dimensional tolerance checks both at the time of molding and over long term to study stability,
as well as checks of the electrical properties of the feedthroughs. The Prototype II chamber
will use the new feedthroughs to test their performance in a real chamber environment,
including electrical isolation and gas leakage tests.
5.6.8
Stringing
The chamber will likely be strung in a large clean-room at TRIUMF in Vancouver. The
7 82 11 45 m, Class 10,000 clean-room has a ceiling structure containing lighting xtures
and 18 distributed HEPA- lter/air intake blowers. The stringing stand proposed for the
chamber would be about 2 0 4 7 m and would require a ceiling height of about 6.0 m when
rotated into the vertical position. This would necessitate raising the ceiling of the TRIUMF
clean room by about 3.0 m. Temperature control (2 C) will also be installed.
The chamber will be strung horizontally without the outer or inner cylindrical shells in place.
Each endplate will be attached from its outer edge to a central axle of the stringing stand by
means of external support spiders. The endplates will be preloaded at their inner radii, and
possibly at intermediate positions as well. All loads are transferred to the central axle by
the spiders. Stringers will be able to work between and under this support structure, while
the interior region of the chamber is completely unobstructed.
Two teams of three people (or two people and a robot) each may work simultaneously as the
chamber is strung from the inner radius out. The wire is inserted through one endplate by a
stringer, transported to the other endplate either by the robot or by the third stringer, and
is then inserted though the appropriate hole for the second stringer. Both stringers thread
the wire onto a feedthrough/crimp-pin assembly and insert the assembly into the endplates.
When the rst stringer has crimped one end of the wire, the second stringer may tension and
crimp the other. A robot would o er a smooth method of transporting the wire and would
ensure that the correct hole was used. Depth perception is notoriously bad when looking at
a eld of strung wires, and a robot would help to avoid stringing errors. The chamber can
also be kept cleaner by enclosing the space between the endplates, including the robots, in
a very clean tent within the outer clean room.
Wires which fail during the stringing process will be restrung with the chamber rotated into
a vertical position, aligned at the appropriate stereo angle. A magnetized needle will be
:
:
:
:
Technical Design Report for the BABAR Detector
184
Drift Chamber
The stringing and pretensioning stand for chamber assembly, showing the
normal horizontal position for wire insertion with two stringing teams and a vertical position
for wire replacement.
Figure 5-18.
Technical Design Report for the BABAR Detector
5.6 Mechanical Design
185
lowered through the top hole and captured by a similar needle inserted into the lower hole.
The stringing stand illustrated in Figure 5-18 has a dual support at one end, allowing the
chamber to be cantilevered for the installation of the inner and outer shells. The inner wall
supports no axial load and is installed after stringing to avoid accumulation of dust.
5.6.9
Endplate Connections
Electro-Mechanical Boards
The electro-mechanical boards form the interface between the feedthroughs and the electronics. They perform the following functions:
Provide a ground plane for the cathode wires at both ends of the chamber. Insulating
feedthroughs are being used for the eld wires to allow tension measurements to be
performed and broken wires to be located. In the axial-stereo layout, some of the
cathode wires will be biased at a few hundred volts, which also requires insulated
feedthroughs.
Provide high voltage to the sense wires. This will be done on the backward endplate
only.
Provide collection of signals from sense wires and distribution of calibration pulses to a
standard preampli er card. All signals will be extracted from the backward endplate.
No cables will be present on the forward endplate.
Thin cathode boards are mounted on both ends of the chamber and separate sense wire
boards are mounted on the backward endplate only. The cathode boards are coppercoated, printed circuit boards approximately 0.25 mm thick, providing a low-resistance, lowinductance connection. A good quality ground is essential to avoid cross-talk between sense
wires arising from induced transient signals on neighboring eld wires (Figure 5-19). The
boards are mounted parallel to the endplate and connect to the cathode wire feedthrough
pins. Each cathode board is connected to neighboring boards with removable jumpers for
the ground and bias voltages.
The 0.25 mm-thick sense-wire circuit board is mounted on the backward endplate parallel
to the cathode boards but separated by approximately 6 mm to avoid high voltage (HV)
breakdown. The connection to the sense-wire crimp pins is made through the large holes
in the cathode boards as illustrated in in Figure 5-19. At the forward endplate, the sensewire crimp pins will be capped to prevent corona discharges into the air. The sense wire
boards will provide a standard connection for the preampli er cards, for ease and safety of
Technical Design Report for the BABAR Detector
186
Drift Chamber
Figure 5-19. One concept for endplate cathode boards for grounding eld wires and
HV/signal distribution boards. In this example, the cathode boards connect together
12 cathode wires and are jumpered together to form the ground plane. The sense wire
boards mount over the ground plane and provide a standardized connection between the
preampli ers and ve sense wires.
installation and replacement. The cathode boards and sense wire boards will rarely, if ever,
be removed.
The sense wire boards will also include a HV bus, as well as the isolating resistors and
capacitors that couple to the sense wires. Jumpers will connect adjacent boards in the direction, allowing distribution of high voltage. This scheme allows the HV to be applied
without the preampli er cards attached, which o ers greater exibility and safety during
testing or troubleshooting.
Preampli er Cooling
At a constant voltage, the gas gain depends strongly on the gas density, which is proportional
to pressure and inversely proportional to temperature. Therefore, for good dE/dx resolution,
both the pressure and temperature must be kept constant or at least be well-monitored.
While pressure is always uniform across the chamber, temperature di erences at the chamber
surface may generate di erences within the gas volume. The total heat generated by the
preampli ers is on the order of 35 W and may be removed by owing nitrogen gas.
Technical Design Report for the BABAR Detector
5.7 Front-End Electronics
187
Cables
The coaxial HV cables and ribbon-style signal cables will be attached to the backward
endplate only. A typical HV cable as considered here has a 3 mm outer diameter and
corresponds to about 6% of a radiation length. Approximately 220 cables will be routed
radially over the endplates to the various HV distribution points. The volume of HV cables
is suciently small that only a single layer will be formed at the outer radius of the chamber.
The ribbon cables supply the preampli ers with power, carry test lines, and the preampli er
outputs. A typical unshielded PVC coated ribbon cable is 33 mm wide, 2 mm thick, and
corresponds to 0.75% of a radiation length. The cables are stacked six deep, corresponding
to 4.5% of a radiation length at the outer radius.
A low-mass structure is required to support the cables over the endplate. The anchor points
will be the studs at the inner and outer wall joints. Cables are supported on a spiderlike structure which will be constructed of three rings, one each at the inner radius, the
cone tip and the outer radius, connected by spokes. In the axial-stereo case, intermediate
connections to the endplate are possible at the three hole-free superlayer transition regions.
Approximately 6 cm of clearance is required between the endplate and the support to allow
for card replacement.
5.7
Front-End Electronics
The front-end electronics provide ampli cation and shaping of the signals for optimal timing
and pulse height (dE/dx) resolution. We have identi ed parameters of the drift chamber
which will impact the design of the front-end electronics. For example, the combination of
wire resistance and terminating resistor is being optimized for best signal/noise behavior.
The shaping time and required dynamic range depend on the drift gas selected and the mode
of drift time measurement (TDC, FADC) and are the subject of simulations. In order to
optimize power and noise performance, the on-chamber, front-end electronics will consist of
a bipolar ASIC chip serving several channels. The front-end electronics are mounted on PC
boards, which connect directly to the feedthrough pins on the backward endplate.
The ampli ed signals are transmitted about 30 m to the readout cards using shielded twisted
pairs. The signals are di erential to minimize the emission of electromagnetic interference
and to allow for common mode rejection at the ends of the transmission lines. They are
sent to a readout card containing discriminators and pulse-height digitizers based on a
deadtimeless FADC system. The discriminator outputs are used by the trigger segment
nder logic, and the FADC samples are stored continuously in a bu er memory. When a
trigger is received, the FADC samples are read from the dual port bu er memory, and the
Technical Design Report for the BABAR Detector
188
Drift Chamber
drift time and charge information is extracted from the FADC raw data and transmitted to
the event builder. The sampling clock will probably run at 79 MHz, a submultiple of the
238 MHz machine frequency. This will allow the suppression of recurrent machine-related
noise and yield a timing resolution on the order of 2 ns, which is sucient to match the
intrinsic position resolution of the chamber (about 125 m). Further details of this readout
system can be found in Chapter 10.
5.8 High Voltage System
The high voltage (HV) system will provide 2 kV to the 7000 sense wires while minimizing
the impact of HV problems, such as sparking in the chamber, broken wires, and failed HV
capacitors. Forty HV supplies will be used to drive approximately 220 HV cables. This
number ensures that a very low current is drawn even with high background levels. A
sensitive trip level will protect against wire breakage due to sparking. Forty supplies will
also provide enough current to ramp the HV in less than ten seconds, thereby minimizing
detector downtime during frequent injections. A possible implementation would involve
using the LeCroy 1454 system, consisting of a mainframe interfaced to VME or CAMAC
which houses four HV cards, each having 12 channels capable of providing up to 3 kV and
2.5 mA.
In the example illustrated in Figure 5-19, a set of ve sense wires receives voltage from a
single HV-distribution card located on the backward endplate. The HV card contains a
HV bus with jumpers to adjacent cards, and an isolation resistor and one or two decoupling
capacitors for each wire. The resistor values for the sense wires will be approximately 0.5M
per wire, ensuring that the voltage drop across the resistor is 1 V. The induced current
in the rst sense-wire layer has been estimated to be approximately 50 nA for the nominal
background levels. Two HV capacitors mounted in series would mean that a single failure
(short) would not a ect the performance of the chamber; however, space limitations on the
endplate may not allow this solution. The capacitance value will be selected on the basis of
signal propagation measurements and calculations.
Each HV cable supplies four such HV distribution cards; the jumpers are used to distribute
the voltage between cards. If there is a HV problem, 32 sense wires (0.3% of the total)
will be a ected until the next access to the drift chamber endplate.
Technical Design Report for the BABAR Detector
5.9 Calibration and Monitoring
5.9
5.9.1
189
Calibration and Monitoring
Calibration
The timing characteristics and the gain of the electronics chain will be calibrated with a
pulser system. The input of every preampli er channel is connected via a coupling capacitor
and isolating resistor to a calibration bus. When a voltage step is applied to the bus, a
de ned charge is injected into the front-end of the preampli er, and the whole electronics
chain can be calibrated. The matching of these calibration capacitors on bipolar ASICs is
excellent. It has been shown to be advantageous to connect neighboring channels to di erent
calibration buses. It is planned that every chip will be serviced by four buses, and that these
buses will be distributed to a limited area of the drift chamber, similar to the case of the
high voltage and preampli er power supplies.
5.9.2
Slow Controls
The slow control system will monitor the operation of the drift chamber, the high voltages,
power supplies, and test pulse distribution. The front-end hardware can be monitored with
the help of suitable online histograms. However, it is essential that the proper hardware
diagnostic tools exist in order to detect the failure of a particular component as soon as
possible. The histogram observation method frequently implies rather substantial delays in
discovering a component failure. A hardware sensor can set o an alarm instantly when, for
example, a power supply fails.
The slow control system is designed to respond to expert commands from a local console,
generate programmed sequences in response to a command issued from the run control,
and take corrective actions in the event of an abnormal situation (voltage trips, etc.). This
requires that the system have its own local processor, a set of readout and control cards to
communicate with the appropriate sensors and control devices, and a link to a local area
network (LAN) to receive the run control commands.
The slow controls also generate alarm messages to the central error reporting system as
well as status reports. Most of these messages can be handled through the LAN, but some
alarms|the hazardous gas monitors, for example|will act on speci ed hardware interlocks.
A link is also required to a database containing the mapping for the physical channels, the
calibration constants, and, if required, some voltage or current limits for a channel.
Technical Design Report for the BABAR Detector
190
Drift Chamber
High Voltages. The sense wires, nominally at 2 kV, will be monitored for the current
they draw. The slow control system provides fast ramp-up and ramp-down of voltages. The
eld wires, maintained at ground potential, do not require monitoring except in the case of
small bias voltages for the axial-stereo design.
Low Voltages and Currents. The slow control system will also be used to monitor the
voltages and currents in the preampli ers and the local CAMAC, VME, and VXI crates. This
will be achieved with slow 64-channel ADCs having individually programmable gains and
di erential inputs. About 150 channels are required for the front-end crate power supplies,
and about 450 to monitor each preampli er card, assuming 32 channels and two voltages
per card.
Temperature, Pressure, and Gas Flow. The same slow ADCs will be used to monitor
ambient temperatures and pressures at various locations and the response from a gas gain
monitor. Air ow switches will be handled with parallel input registers.
All of the above controls can be realized using commercial CAMAC modules controlled by a
VME processor via a standard A2 CAMAC controller and a VME CAMAC branch driver.
Slow Control via Data Monitoring. The performance of all cells will be monitored
locally by histogramming of hits in each cell, track segment density, amplitude and time
distributions, and other relevant distributions. These histograms are inspected either visually
or, in some cases, automatically if the shape can be compared to a template histogram.
5.9.3
Monitoring
The slow controls described above provide one means of monitoring gas properties such as
drift velocity as a function of electric eld, the gas gain at the operating point of the chamber,
the composition and purity of the gas, and the level of contaminants (such as O2 and H2O),
which could in uence the operation of the chamber. In addition, the quality of incoming
premixed gas bottles will be monitored by small test chambers before being introduced to
the gas system.
The gas ow and the relative pressures and temperatures must be monitored at various
locations for proper operation of the chamber, since the drift time and gas gain depend on
the gas density. The chamber pressure is expected to follow the ambient pressure, which
uctuates by 4% for this geographical region. These pressure di erences will a ect the
chamber gain and drift velocity. The gain can be corrected by applying a scale factor to
Technical Design Report for the BABAR Detector
5.10 Integration
191
the pulse height measurements. The drift velocity poses a more dicult problem, since
the gas is nonsaturated. Corrections to the track position measurements depend on the
distance from the sense wire, and range by at least a factor of 5 over the cell. The SLD
experiment measured a 2% change in drift velocities due to ambient pressure uctuations,
which were corrected to 0.2%. The e ects of the pressure on the drift velocity are being
investigated using the Prototype I chamber. Simulations will be used to determine the best
parameterization for the correction.
Monitoring some gas properties inside the chamber would be also advantageous. A laser
calibration system is being considered for measurement of the drift time in the chamber
volume by ejecting electrons into drift cells through photoemission from eld wires using a
high-intensity UV laser. The laser beam is transmitted to the chamber via optical bers.
A beam port into the chamber volume will be situated near the midpoint of the outside
wall. This arrangement is compact, with few optical elements, such as a small mirror and
lens, required to direct the beam transverse to the eld wires. The ber would lie along
the outside surface of the chamber with the beam directed into the chamber at 90 by a 45
mirror. The feasibility of this approach will be determined as part of the prototype program.
5.10
Integration
5.10.1 Overall Geometry and Mechanical Support
The mechanical envelope for the drift chamber is shown in Figure 5-2. The envelope includes
the space needed for endplates and front-end electronics in addition to the active detector
volume. It does not include the volume needed for mechanical supports or cable routing.
The method for supporting the drift chamber is still under study. In one scenario, the
chamber is held from the accelerator pedestal in the backward direction and the support
tube column in the forward direction. This arrangement allows the support tube to be
repositioned (by remote-control cams at each end) or removed independently of the drift
chamber. The supports are fastened to the lower edge of the inner radius of each endplate.
In another scenario, the chamber is held at its outer radius by the barrel IFR.
5.10.2 Cable Plant and Utilities Routing
The cables and utilities that must be brought into the drift chamber from outside the detector
include signal, high voltage, and preampli er power cables; calibration and monitoring
Technical Design Report for the BABAR Detector
192
Drift Chamber
signals; drift gas; and preampli er cooling gas. All connections are made to the backward
endplate, except for gas lines and monitor signals which will be at both ends.
The cross-sectional area and mass are dominated by the signal and HV cables. The cable
routing scheme has been laid out assuming 9100 cables of 2.3 mm diameter (signal, monitoring, calibration, and preampli er power) and 220 cables of 5 mm diameter (HV). The 2.3 mm
is approximately the size of RG-174; the actual cables used are likely to be smaller and fewer.
Including an additional safety factor of 2, the total cross sectional area reserved for cables is
866 cm2 . The cables and gas lines will be brought into the detector along 2.5 cm-deep cable
trays that cover approximately 70% of the inner radius of the DIRC support tube (i.e., at
the outer radius of the drift chamber).
The total cable run to the electronics room is approximately 35 m. If thicker cable is needed
to preserve signal quality, a transition box will be located close to the detector.
5.10.3 Access
Access to the drift chamber will be needed to replace failed HV capacitors or preampli ers,
to x gas leaks, and to replace broken wires. None of this is anticipated to be a frequent
occurrence. For example, similar front-end electronics have been used on OPAL chambers
for seven years without access. In that time, only one of 400 cards has failed. This reliability
is achieved both by design|including diode transient suppressors, for example|and by
a thorough program of pre-installation testing. A small, randomly distributed number of
failed preampli er cards will have negligible impact on the detector performance. Two HV
capacitors will be used in series on each sense wire to improve the reliability of the HV
distribution.
A broken wire is a more signi cant problem than a failed preampli er. The proposed
R&D program will ensure that deformation of the feedthrough and wire due to crimping
is understood and will establish the allowed range of crimping force. The crimping tools
will be repeatedly calibrated throughout the chamber stringing to ensure operation within
this range. With careful attention to such details and a testing period of several months
following stringing, experiments such as SLD and BNL-787 have built drift chambers with
large numbers of aluminum wires without incurring a single wire break during operation.
Tests performed at SLAC indicate that broken wires tend to curl up at the feedthrough
rather than fall into the rest of the chamber.
Although regular access will not be required, access procedures have been developed for both
ends of the drift chamber. In the forward direction, the endcap-instrumented ux return
is vertically split and can be withdrawn along rails at an angle relative to the beam line.
Technical Design Report for the BABAR Detector
5.10 Integration
193
Temporary sca olding may then be installed, giving very good access to the drift chamber
endplate.
In the backward direction, the DIRC support cylinder and water tank create a tunnel
containing the Q1, Q2, and Q4 magnets, pumps and masks, ux return iron, and the drift
chamber cable trays. For approximately 70% of the azimuthal angle, the return iron can be
withdrawn on rails mounted on the support cylinder. After clearing the DIRC water tank,
they are then removed by crane.
The time needed to achieve access will be one shift for either endplate. The reassembly time
will be longer by approximately one shift for the forward endplate because of the extra time
needed to realign the forward calorimeter.
5.10.4 Integration Aspects of the Gas System
The drift chamber has two gas systems, the drift gas (He:C4 H10, for example) and the dry
nitrogen used to remove the 35 W produced by the preampli ers. Work is currently underway
to establish the ammability of each of the drift gases under study. If necessary, the rf-shield
at each endplate can be used to make a second gas seal. This volume would then be ushed by
nitrogen to ensure that a ammable concentration of gas cannot accumulate due to a leaking
feedthrough. (In the backward direction, nitrogen is also the cooling gas.) In particular, the
nitrogen gas distribution will be designed to ensure that a ammable concentration cannot
accumulate between the feedthroughs and the printed circuit boards mounted on them. The
exhaust gas and a gas sample pumped from the endplate region will both be monitored for
oxygen and isobutane concentration. Fire-retardant cables are required in any event as a
re safety measure for the detector.
A large leak of helium could damage the DIRC or aerogel photomultiplier tubes. In addition
to the precautions listed above, the photomultiplier tubes may need to be located in a gastight volume.
5.10.5 Installation and Alignment
The drift chamber is installed after the magnet, IFR, calorimeter, and DIRC. The machine
components from Q2 to Q4 are absent, but the accelerator pedestal is installed in the
backward direction. The drift chamber is inserted from the forward end along a beam
passing through the chamber to the accelerator pedestal. The forward support tube column
is then installed and the drift chamber connected to its permanent supports.
Technical Design Report for the BABAR Detector
194
Drift Chamber
There is an assembly tolerance of 15 mm between the chamber inner radius and the accelerator support tube outer radius. This clearance also allows the remote movement of the
tube, which may be as much as 5 mm.
An optical alignment will be performed after the accelerator support tube has been installed
in order to place the drift chamber to within 1 mm, and to align the axis of the chamber to
within a few mrad. The nal, precision alignment will be extracted from data.
5.11 R&D Using Prototype Drift Chambers
Simulation and prototyping e orts are underway in the areas of cell geometry, orientation,
layer packing, pattern recognition, and track reconstruction. Several stand-alone data analysis packages are being used, and there have been recent advances on the incorporation of the
drift chamber into the BBSIM GEANT package. Prototype e orts are continuing at SLAC,
with the current prototype, Prototype I, and further prototyping e orts are being planned.
5.11.1 Prototype I Chamber
A small-cell prototype chamber has been built and is being used to study electrostatic stability, feedthrough design, front-end electronics, cell crosstalk, and wire and gas properties. The
prototype is 2.5 m in length, with at aluminum endplates, 72 anode wires of 20 m-diameter
gold-coated tungsten, and 512 cathode wires of 55 m-diameter unplated aluminum 5056.
The wires are arranged into three superlayers of hexagonal cells with the center superlayer
positioned at a 50 mr stereo angle.
Studies of suitable helium-based drift gases are continuing with Prototype I, using cosmic ray
particles. To date, the three component gas mixture He:CO2 :C4 H10 (83:10:7) and the two
component gas mixture He:C4 H10 (80:20) have been investigated. Preliminary results are
included in Figure 5-14, along with measurements from previous gas studies [Cin91, Pla92,
Uno93]. Plans are underway to study other gas mixtures, including He:C3 H8 and He:C4 H10
with lower partial pressures of isobutane. The possibility of including CF4 as an additive to
counter aging is being considered.
Prototype I has also been the test bed for a study of the achievable position resolution using
a FADC sampling system and a low-density helium-based gas with pronounced discrete
clustering. Data from four anodes, sampled with 8-bit resolution at 250 MHz (' faccel),
have been arti cially compressed to 125 MHz (' faccel=2) and 83.3 MHz (' faccel =3), and
several algorithms for extracting the timing information have been applied to the signals.
The results are steadily improving under re nement of these algorithms and currently show
Technical Design Report for the BABAR Detector
5.11 R&D Using Prototype Drift Chambers
195
single wire position resolutions of 127 m (at 125 MHz) and 134 m (at 83.3 MHz) for the
gas mixture He:C4 H10 (80:20). These should be compared to the single wire resolution of
109 m obtained with a discriminator and TDC (500 ps bin size). The timing algorithms
have yet to be fully optimized, and so further improvements in these values are possible. In
early 1995, the preampli ers of Prototype I will be replaced by ones of a better design, and
all 72 channels will be instrumented with 100 MHz FADCs. The Prototype I chamber will
then provide full tracking with FADCs.
5.11.2 Prototype II
A second full-length prototype, Prototype II, will be constructed in 1995 and will incorporate
all design decisions made to date. It will have sloped carbon ber endplates, the new crimpstyle feedthroughs, and gold-coated aluminum cathode wires.
The preampli er electronics for this chamber will be furnished by UCSC, in the form of
an ASIC with adjustable gain, rise time, and shaping time. This preampli er will allow
precise determination of the characteristics of the nal preampli er design in accordance
with the chamber cell response and the timing algorithm used. Cell shape, aspect ratio,
layer arrangement, and the segmentation of high voltage, power, and calibration signals will
re ect the decisions of the group with respect to the internal chamber geometry.
5.11.3 Other R&D E orts
Depending upon the availability of a test beam, a somewhat smaller test chamber could
be used for dE/dx studies. It would also be desirable to use such a chamber for studies
of tracking, dE/dx, and Lorentz angles in helium-based gas mixtures in the presence of a
magnetic eld.
Studies on the aging of the drift gas in a high rate environment are underway at TRIUMF
and at LBL. The most likely gas candidates, as identi ed in the Prototype I studies, will be
subjected to long term aging tests. The usefulness of CF4 as a gas additive to counter aging
will be investigated as well.
Technical Design Report for the BABAR Detector
196
REFERENCES
References
[Bia89]
The MAGBOLTZ Simulation Code was kindly supplied by S.F. Biagi of the
University of Liverpool. Some results from this code are discussed in S.F. Biagi,
A283, 716 (1989).
A. Boyarski, D. Briggs, and P. Burchat,
A323, 267 (1992).
D. Brashford et al., \The Use of High Sti ness Material and Dimensionally Stable
Materials in Spacecraft Applications," in Proceedings of the Int. Workshop on
Advanced Materials for High Precision Detectors, CERN 94{07, p. 9 (1994).
D. Britton and E. Hyatt, \Pattern Recognition in an All Stereo Drift Chamber,"
A AR
# 102 (1993).
V. Cindro et al.,
A309, 411 (1991).
Manufactured by ICI Fiberite, Tempe, AZ.
W. Innes, \TRACKERR, A Program for Calculating Tracking Errors,"
A AR
# 121 (1993).
S. Playfer et al.,
A315, 494 (1992).
Manufactured by Dupont.
C. Smith, McGill Internal Technical Memo, McG{BF{TM{0007, (1994).
S. Uno et al.,
A330, 55 (1993).
J. Va'vra, L. Roberts, D. Freytag, and P. Clancey,
A203,
109 (1982). The code has been modi ed to use the number of primary ions in the
calculations.
Manufactured by ICI Advanced Materials, Exton, PA.
W.C. Young, \Roark's Formulas for Stress and Strain," 6th Edition, McGraw
Hill, (1989).
Nucl. Instr. Methods
[Boy92]
[Bra94]
[Bri93]
Nucl. Instr. Methods
B B
[Cin91]
[IM7]
[Inn93]
Nucl. Instr. Methods
B B
[Pla92]
[RAD]
[Smi94]
[Uno93]
[Vav82]
[VIT]
[You89]
Note
Note
Nucl. Instr. Methods
Nucl. Instr. Methods
Technical Design Report for the
Nucl. Instr. Methods
A AR
B B
Detector
6
Particle Identi cation
6.1 Physics Requirement and Performance Goals
6.1.1
Introduction
E
xcellent particle identi cation for hadrons and leptons over a large range of solid angle and momentum is an essential requirement for meeting the physics objectives of
BABAR. In particular, measurements of CP violation require particle identi cation, both to
reconstruct exclusive nal states and to tag the quark content of the other B in the event.
Information from the drift chamber, calorimeter, and the instrumented ux return can be
used to identify most of the leptons and many of the hadrons. However, these systems are
not sucient to distinguish charged pions from kaons with momenta greater than about
0.7 GeV=c, or protons above 1.3 GeV=c, as is required to obtain ecient tagging and event
reconstruction. To meet these requirements, dedicated particle identi cation will be provided
by a combination of Cherenkov counters: a ring-imaging detector in the barrel region (the
DIRC, for Detection of Internally Re ected Cherenkov light) and a threshold detector in
the forward endcap region (the ATC, for Aerogel Threshold Cherenkov counter) with two
indices of refraction. No dedicated particle identi cation in the backward region is needed.
6.1.2
B Flavor Tagging
The avor of B mesons will primarily be tagged with charged kaons and leptons: a b-quark
decay leads to a K , and/or a direct lepton `,, and a b-quark decay leads to a K + and/or
a direct lepton `+. B - avor tagging thus relies on the identi cation of kaons and leptons in
an environment of inclusive B decays with large multiplicities of low-momentum daughters,
where relative abundances of pions, kaons, and protons are approximately 7:1:0.2. Because
of the boost ( = 0:56), the average (K ) momentum depends on the polar angle, and
198
Particle Identi cation
Inclusive momentum spectrum for kaons as a function of polar angle in the
laboratory frame. The banded region is an enhanced population of kaons from the decay
B ! K.
Figure 6-1.
ranges from about 0.3(0.45) GeV=c in the backward direction to about 0.75(1.1) GeV=c in the
forward direction.
The e ective tagging eciency is de ned as eff = tag (1 , 2w)2, where tag is the
fraction of tagged B events, and w is the fraction of tagged events which are incorrect.
The e ective eciency is limited by physics itself, for example, by wrong-sign contamination
from Cabibbo-suppressed decays or from kaon or lepton pairs. In the case of kaon tagging,
even assuming perfect particle identi cation and no kaon decays in ight, the maximum
tagging rates and minimum wrong-tag fractions which can be achieved within the BABAR
acceptance are 36.3% and 6.6%, respectively. This leads to a maximum e ective eciency
of 27.3%. Kaon decays result in a further 20% reduction of eciency. The loss of eciency
introduced by the PID system must be small in comparison. Given the relative abundances
noted above, a small misidenti cation rate for pions is particularly crucial.
The measurement of ionization in the drift chamber will provide =K separation better than
3 up to 0.7 GeV=c, but will be of limited assistance in the relativistic rise region at higher
momenta. As a consequence, the dE/dx measurement is not itself sucient, since 48% of
all kaons would remain untagged, as demonstrated by the momentum distribution shown in
Figure 6-1.
Technical Design Report for the BABAR Detector
6.1 Physics Requirement and Performance Goals
199
Kaon tagging thus requires dedicated PID systems that achieve high eciency and purity.
Furthermore, because the forward region subtends a large fraction of the center-of-mass solid
angle, tagging eciency is sensitive to the forward PID coverage. In the proposed design, the
DIRC covers 86.5% of the full solid angle, down to = 445 mr, and, combined with dE/dx,
provides an e ective tagging eciency of 18.5%. The very forward part (300 < < 412 mr)
covered by the ATC represents 5% of the full solid angle, and a dedicated PID system in
this region, in combination with dE=dx, increases the e ective tagging eciency by 3.4% to
19.1%. To make this contribution, the forward aerogel needs to be able to tag kaons up to
1.5{2 GeV=c through the use of an aerogel layer with an index of refraction n = 1:055.
In the case of lepton tagging, the dE/dx measurement ensures electron identi cation over
the full momentum spectrum. However, muon identi cation by range measurement in the
ux return is ecient only above about 0.7 GeV=c. At lower momenta, the DIRC, primarily
designed for its =K separation capabilities, will also provide = separation, complementary
to the IFR.
6.1.3
Exclusive
B decays
The search for CP -violating modes, such as B 0 ! +, or B 0 ! , and the measurement
at much higher momentum of Vub through charmless B decays, places greater demands
on the performance of the PID system than those required for tagging. The exclusive
branching ratios of interest are very small (10,5), and CP violation studies are made even
more challenging by the need to tag the other B in (4S ) ! B 0B 0 decays. While much
of the background from the continuum can be suppressed by simple kinematic cuts, the
reduction of correlated backgrounds from B 0 ! +K ,, B 0 ! K , and B 0 ! K (892)
demands that pions be positively identi ed, with a minimum of kaon misidenti cation, in
the high-momentum range up to 3{4.5 GeV=c. Non-B physics will also be demanding in
terms of kaon identi cation. A challenging example is the study of rare strange decays of
the tau lepton in which the momentum spectrum is very hard (up to 8 GeV=c).
An example of a detailed analysis of a two-body mode requiring high-momentum particle
identi cation is given in Chapter 3. A study of the decay B 0 ! +, is reported which relies
on the combination of DIRC and ATC to suppress the correlated K +, background. These
ASLUND studies show that dE/dx and kinematics, while providing signi cant capability
for identi cation of the B 0 ! +, exclusive channel, are not sucient to guarantee
both high eciency and low background. Requiring a reasonable separation between the
competing mass hypotheses, e.g., by demanding that the 2 for the +, assignment exceeds
alternatives by at least four, quickly reduces acceptances from 92% to 80% in the case of
idealized particle tracking and dE/dx performance; accounting for reasonable degradation in
the actual running detector lowers these estimates to unacceptable levels. The addition of
Technical Design Report for the BABAR Detector
200
Particle Identi cation
the DIRC/ATC system provides a clear margin of safety. The acceptance for +, remains
at the 95% level given this and even more stringent background suppression requirements.
6.1.4 Summary of Requirements
A dedicated PID system is required both for B - avor tagging and to identify exclusive
nal states. These two purposes place di erent demands on the system: tagging requires
positive identi cation of kaons and leptons with minimum pion background, mostly in the
low-momentum range; measurement of exclusive decays requires positive tagging of pions
with minimum kaon misidenti cation, mostly in the high-momentum range, resulting in the
requirement of highly ecient =K separation over the full spectrum.
These requirements also place constraints on the detectors and material located inside the
PID system, in order to bene t fully from the intrinsic performance capability. Key issues
include the accuracy of track extrapolation to the DIRC and a low level of backscattering
and -ray backgrounds in the threshold ATC counter.
Likewise, the PID system must not signi cantly degrade the performance of the detectors
located outside; both the physical volume occupied and material budget in radiation lengths
are crucial in this respect. Of primary concern is the importance of calorimetry in measuring
exclusive decays such as B 0 ! 00 and B 0 ! , for which gamma detection and mass
resolution are essential.
6.2 Particle Identi cation Overview
To meet the requirements described above, a system containing two types of Cherenkov
detectors has been incorporated into BABAR. The barrel region (25:5 < < 147) is covered
by a DIRC [Rat92], which provides good performance over the whole momentum range
while occupying only a thin radial region. An ATC covers the forward region (17:1 < <
23:6) augmenting the kaon tagging coverage and providing =K separation up to 4.3 GeV=c.
The boundary between the systems was chosen to maximize the acceptance for particle
identi cation within the constraints of magnet length and calorimeter position and angle.
Both systems maintain low mass in order to minimize their e ect on low-momentum photon
calorimetry. The layout of the main components of the particle identi cation system is shown
in Figure 6-2.
Technical Design Report for the BABAR Detector
6.2 Particle Identi cation Overview
201
Elevation view of the PID system geometry showing the barrel, DIRC, and
forward ATC systems.
Figure 6-2.
6.2.1
The DIRC
A charged particle traversing a DIRC quartz bar with velocity in a medium of refractive
index n produces Cherenkov light if n 1. The DIRC Cherenkov radiators are 4.7 m-long
rectangular quartz bars oriented parallel to the z axis of the detector. Through internal
re ections, the Cherenkov light from the passage of a particle through the DIRC is carried
to the ends of the bar as shown schematically in Figure 6-3. At the readout end, photons are
detected outside the magnetic eld region by conventional photomultiplier tubes (PMTs); a
re ecting mirror at the other end of the bar returns forward-going light to the PMTs. For
each track, the Cherenkov image expands from the end of the source bar across a stando
region lled with water to a toroidal surface of closely packed PMTs.
The DIRC uses as a radiator 156 quartz bars arranged in a 12-sided polygon around the beam
line. This maximizes azimuthal coverage, simpli es construction, and minimizes edge e ects.
For suciently fast charged particles, some part of the Cherenkov radiation cone emitted
by the particle (c(E ) = cos,1[1= n], with n = 1:474) is captured by internal re ection in
the bar and transmitted to the photon detector array located at the backward end of the
detector. (Forward-going light is rst re ected from a mirror located on the end of the bar.)
The high optical quality of the quartz preserves the angle of the emitted Cherenkov light.
The measurement of this angle, in conjunction with knowing the track angle and momentum
from the drift chamber, allows a determination of the particle velocity. An advantage of the
DIRC for an asymmetric collider is that the high momentum tracks are boosted forward,
Technical Design Report for the BABAR Detector
202
Particle Identi cation
tz
n2
Quartz
n3
n1
n3
Pa
rti
c
z
le
Tr
aje
c
y
to
ry
ty
Detector
Surface
θD
z
Side View
n2
Particle Trajectory
n3
tx
n1
n3
x
z
Plan View
5-94
7172A5
Schematic of a single radiator bar of a DIRC counter. The particle trajectory
is shown as a connected line of dots; representative trajectories of Cherenkov photons are
shown by lines with arrows.
Figure 6-3.
which causes a much higher light yield than for particles at normal incidence. This is due to
two e ects: the longer path length in the quartz and a larger fraction of the produced light
being internally re ected in the bar.
Each quartz bar is 1.75 cm thick, 3.5 cm wide, and 470 cm long, and is constructed by gluing
together four shorter bars. The total radial thickness of the DIRC, including quartz thickness,
sagitta from the polygonal shape, mechanical supports, and a 1 cm clearance on each side,
is 10 cm. This material represents 0.18X0 at normal incidence. An e ort has been made
to minimize both the radial thickness and the amount of material, since these increase the
radius and cost of the barrel calorimeter while degrading its performance for soft photons.
The photon detector consists of 13,400 conventional 1.125-inch-diameter phototubes. They
are organized in a close-packed array at a distance of 120 cm from the end of the radiator
bars. The phototubes, together with modular bases, are located in a gas-tight volume as
protection against helium leaks from the drift chamber.
The photo-detection surface approximates a partial cylindrical section in elevation and a
toroid when viewed from the end. The stando region has re ecting surfaces along the inner
Technical Design Report for the BABAR Detector
6.2 Particle Identi cation Overview
203
and outer radii to reduce the number of phototubes required. The opening angle between
these surfaces is still under review but is taken to be 55 at this time. To maintain good
photon transmission for all track dip angles, the stando region is lled with water. The
water seal occurs at a quartz window that is glued to the end of the bar assemblies. Most of
the mass of the stando box structure is high permeability steel, which provides adequate
magnetic shielding for the phototubes.
The DIRC radiators are supported on the central support cylinder, which is cantilevered
from the strong support tube. Alternative designs that do not require a cantilevered support
for the central support cylinder are also under consideration. Both the strong support tube
and the stando box are supported by a yoke directly to the barrel IFR. Since the DIRC bars
penetrate the backward endcap return iron, the DIRC mechanical design interacts strongly
with other structures at the backward end of the detector (Section 6.4.1).
A single-bar DIRC using 47 phototubes with an air stando has demonstrated this detector
concept (Section 6.4.5). A larger prototype with a realistic water stando region will shortly
be tested in a particle beam. It will be capable of imaging almost all Cherenkov patterns
on a track-by-track basis. This prototype and other laboratory measurements will provide
a detailed test of the optical quality of the quartz, glue joint, and mirrored surfaces, and
establish the performance (single photoelectron response, quantum eciency, and timing
resolution) using approximately 500 PMTs. The prototype will also be used to test frontend electronics.
6.2.2 The ATC
The forward PID detector is a silica aerogel Cherenkov counter covering the polar angle
range 17:1 < < 23:6. Both kaon tagging and pion identi cation are achieved with two
refractive indices, n ' 1:0065 and n ' 1:055.
There are two key components in the ATC system: the aerogel radiator and the readout
device. Candidates for both are commercially available from well-established suppliers.
Several suppliers of aerogel with a polymer gel structure have been identi ed; material
with the appropriate density range, transmittance, purity, and anticipated stability has been
obtained in small quantities.
Aerogel with good optical quality is produced by Airglass, Lockheed, Jet Propulsion Laboratory (JPL), Lawrence Livermore National Laboratory (LLNL), Aerojet, and the Budker
Institute of Nuclear Physics (BINP) at Novosibirsk. Caltech's Space Radiation Laboratory
regularly uses large aerogel blocks from Airglass in balloon-based Cherenkov detectors.
Lockheed's production facility can, in principle, produce enough high quality aerogel for
our needs in 10 months [Men94]. The aerogel produced by JPL is hydrophobic, has good
Technical Design Report for the BABAR Detector
204
Particle Identi cation
optical transmittance [Tso94], and has been extensively tested in our beam measurements. A
two-step process for aerogel production has been developed by LLNL and is being transferred
to the commercial company Aerojet [Hru94].
In order to collect Cherenkov photons eciently inside the solenoid, the photon detector
must work in magnetic elds up to 1.5 T, have a large UV-sensitive area with high quantum
eciency, and have low noise. The baseline design uses bialkali ne mesh photomultiplier
tubes (FM PMT) from Hamamatsu with a newly developed high quantum eciency photocathode. Hamamatsu's FM PMTs with conventional bialkali photocathodes have been
widely used in high-energy experiments [Gil88] for many years.
The average thickness in front of the endcap calorimeter is 9.5% of a radiation length.
However, the present design places phototubes in the active region in front of the DIRC.
The implications of this for DIRC and calorimeter performance need further study.
6.3 Projected Performance
6.3.1 Simulation Based on Prototype Results
The barrel DIRC and forward aerogel detectors are being studied using the simulation
programs ASLUND and BBSIM. The former, based on fast parameterizations of the detector
responses, allows studies requiring high statistics, and the GEANT-based BBSIM program
provides a tool for simulating whole events more realistically. In the latter, other subdetectors
in BABAR are de ned with materials and geometries consistent with their present designs.
This provides a fairly realistic source for secondary processes representing event-related
backgrounds. Primaries are tracked as well as the secondary products they produce, such as
-rays, pair production, hadron scattering, backsplash, and decays. PID detector geometries
are de ned to a high level of detail, and simulation of response is tuned to the prototype
test results (Section 6.4.5).
Barrel DIRC Detector
ASLUND Simulation. The behavior of the DIRC has been simulated in the ASLUND
program; much of it is calculated analytically, and part is done approximately. The parameters for some of the approximations are adjusted to agree with the results of a stand-alone
simulation of the DIRC. This program includes the propagation of photons to phototubes
placed on a toroidal detector surface and the calculation of errors in the measured angles;
it accounts for absorption in the quartz, in the water, and at surfaces; it calculates the /K
Technical Design Report for the BABAR Detector
6.3 Projected Performance
205
separation taking into account all ambiguities. Some aspects, such as the 12-sided geometry
and the wavelength dependence of the absorption, are done in an approximate way. This
simulation ignores background e ects from -rays and other tracks, so its estimates are
somewhat optimistic. In the future, GEANT simulations that include background e ects
will be used to determine the parameters used in the ASLUND simulation.
BBSIM Simulation. The barrel DIRC has been implemented in BBSIM with a realistic
geometry. The detector description is parameterized in an easily modi able form to create a
description of the detector close to the engineering drawings. The active length of 178.6 cm
in the forward direction provides an angular coverage to cos = 0:91, and 2 cm gaps between
the modules result in a 4% loss in azimuth. The radiator represents 0.18X0 at normal
incidence. All physics parameters, such as absorption lengths, refractive indices, and overall
photoelectron yield, are tuned to the prototype test results.
The generation and propagation of Cherenkov photons are fully simulated. After tracking
is complete, all particles crossing the DIRC with momenta above Cherenkov threshold
are processed through a customized simulation. Photons are generated with a Poisson
distribution along the curving track trajectory through the radiator. Dependence of the
refractive index n(E ) of the radiator on photon energy is used to generate photons emitted
at the Cherenkov angle c(E ). Photons produced in the DIRC bars are propagated until they
reach the phototube detector surface. Light transport includes total internal re ection inside
the quartz bars, di raction at the quartz/water interface, and possible re ections from the
mirrors at the forward end of the bar or within the stando tank. Losses due to absorption
in the bars or the water are included as functions of the photon energy. Photon absorption
at the mirrors, scattering by surface imperfections, re ection of light at the quartz-water
boundary (including the e ects of photon polarization and angle of incidence), and the
quantum eciency of the PMT photocathode are also taken into account. Photons which
strike the sensitive area of a PMT are recorded with digitized spatial and time information.
A typical event with one B decaying to an arbitrary channel and the second B to +, has
been projected onto the phototube detector plane in Figure 6-4(a). The time window for
background in the simulation is 15 ns. Four charged tracks reach the DIRC: (4.1), (2.1),
K (0.61), and (0.25); the momenta, in GeV=c, are indicated in parentheses.
Measurements made in cosmic ray Prototype-I tests (Section 6.4.5) have been used to tune
the absorption coecients and the re ection eciencies at the bar surfaces and mirrors. The
resulting comparison between predicted and measured numbers of photoelectrons at various
positions and angles are in excellent agreement, as illustrated in Figure 6-5.
Technical Design Report for the BABAR Detector
206
Particle Identi cation
(b)
(a)
16
(c-1)
14
30
10
(c-2)
25
8
12
7
20
10
(c-3)
9
6
15
8
5
4
6
10
3
4
2
5
2
1
0
0
400
500
600
700
0
600
700
800
900
600
700
Cherenkov angle
800
900
(mrad)
(a) PMT hits projected onto the x-y plane. The particles producing the hits
have distinctive markers: Cherenkov images (conic sections) from the B ! + , decays
are shown as solid circles, hits from the two other charged tracks as open circles, and from
the secondary tracks as `+'. (b) PMT hits projected in Cherenkov angle space with the four
tracks superimposed at the center. The hits from the two pions from the B decay overlap
at the largest radius, c 820 mr. The `+' marker in this case corresponds to assignment
ambiguities. (c) Cherenkov angle projections for three of the tracks and the associated
ambiguities.
Figure 6-4.
Technical Design Report for the BABAR Detector
6.3 Projected Performance
207
90
θD = 30°
Photo–electrons
80
70
60
50
0
11–94
7641A13
50
100
150
200
Z (track position from the bar end) (cm)
Simulated photon yield (solid curve) compared to track position dependence
of the photoelectron yield at a dip angle of D = 30 as measured (circles) in Prototype-I
using cosmic rays.
Figure 6-5.
Forward Aerogel Detector
The forward aerogel PID simulation in ASLUND is based on test beam results and the
measured single photoelectron response of a FM PMT. The photoelectron yield of the
phototube is described by a Poisson distribution with an excess noise factor of 2. The mean
of the distribution is determined by the length of the trajectory through the aerogel block,
where the full length at normal incidence corresponds to 10 photoelectrons. The output
of the simulation is the pulse height in photoelectrons re ecting the single photoelectron
pulse-height distribution of the FM PMT.
A simulation of the forward PID system has been developed in BBSIM. The geometrical
description will correspond to the present design of the ATC with two layers and include
a realistic simulation of the mechanical structure fabricated in carbon or berglass, as well
as two rings of FM PMTs located on the inside and outside of the two aerogel layers. The
geometry is fully parameterized for easy testing of options.
6.3.2
Pattern Recognition in the DIRC
The complex event of Figure 6-4(a) illustrates that the Cartesian coordinates of the hits are
not the easiest representation with which to do pattern recognition, i.e., to assign hits to
tracks. The most appropriate change of variable appears to be from Cartesian coordinates
Technical Design Report for the BABAR Detector
208
Particle Identi cation
to Cherenkov angle c. In the representation in which the track segment de nes origin and
dip angle, the matching hits cluster at the radius corresponding to the particle species. This
not only leads directly to a measurement of Cherenkov angle, but also obviates the need for
sophisticated pattern recognition (see Figure 6-4(c)).
Cherenkov Angle Measurement and Mass Hypothesis Testing. Photons are pre-
sumed to originate from the centers of curved track segments. They are seen by a PMT as
an image on the far side of the plane interface between the stando tank and the quartz
bar. A ray can be traced back from the PMT to the center of the quartz bar with a single
refraction at the quartz-water interface. The angle between this ray and the track segment
is the measured value of c. The azimuthal angle c is also determined in this manner. The
optical distance (corrected for refractive index) from track to PMT is used to estimate the
arrival time of the PMT signal.
Up to a 16-fold ambiguity exists in the determination of c because the number of bounces
o the quartz walls and the mirrors is unknown. Only one solution is correct in each case,
provided the photon is truly associated with the given track. Ambiguities can be eliminated
when the corresponding photon path yields unphysical values for re ection, refraction, or
Cherenkov emission angle. Timing information will also be used to reject solutions when
the computed and measured time of arrival for a photon at the PMT are inconsistent. On
average, only two solutions remain with a timing cut of 7:5 ns. Tighter timing cuts may
be available, depending on the choice of phototube and readout system.
For each track, the distribution of the Cherenkov angles of all potentially associated photons
shows a clear peak, corresponding to the parent particle type, above a background of wrong
track-hit associations (Figure 6-4(c)). The peak position and the number of photoelectrons
are then compared to that expected for each particle type and a mass assignment made on
the basis of an hypothesis test. Consistency between angles and timing information will also
be checked using the expected relationship between c and c for a hit and the propagation
time for the photon to reach the PMT (Figure 6-6).
Intrinsic Resolution. The full DIRC pattern recognition algorithm has been applied to
single pion samples to estimate the intrinsic resolution of the Cherenkov angle c. The
single photoelectron c resolutions are approximately 9.5 mr, largely independent of track
momentum and dip angle. The number of associated photoelectrons per track is given in
Table 6-1. In this study, the availability of precise timing information was not taken into
account. The intrinsic resolution of the DIRC detector is obtained assuming that the NPE
individual measurements of c are uncorrelated.
Technical Design Report for the BABAR Detector
6.3 Projected Performance
209
80
80
70
70
60
60
50
50
40
40
30
30
20
20
0
0.5
1
1.5
2
80
80
70
70
60
60
50
50
40
40
30
30
20
0
0.5
1
1.5
2
0
0.5
1
1.5
2
20
0
0.5
1
1.5
2
Figure 6-6. Measured time at the PMT face as a function of the photon Cherenkov
azimuthal angle c for several track polar angles. The track momentum is 4 GeV=c, and the
value 'c = is de ned for a photon whose initial direction projected onto the bar plane is
towards the mirror.
Momentum Sine of Particle Dip Angle
( GeV=c) {0.5 0 0.5
0.85
0.500
39 22 25
|
0.800
34 31 28
61
1.000
36 35 29
58
2.000
39 38 31
61
4.000
41 41 32
60
Mean value of the number of detected associated photoelectrons as a function
of pion momentum and sine of the dip angle. The e ective packing fraction, i.e., the ratio
of active to total surface, is assumed to be 53% for this estimate.
Table 6-1.
Technical Design Report for the BABAR Detector
210
Particle Identi cation
Momentum Sine of Particle Dip Angle
( GeV=c) ,0:5 0 0.5 0.85
0.500
7.1 2.9 7.0
|
0.800
4.7 2.6 5.1
4.6
1.000
4.0 2.5 4.2
4.4
2.000
2.8 2.2 2.9
2.7
4.000
2.2 2.1 2.4
2.0
Resolution on the c (mr) measurement per pion from BBSIM. Charged tracks
are extrapolated from the vertex to the DIRC bars.
Table 6-2.
E ective Resolution. The DIRC performance is not dependent on intrinsic qualities
alone. At low momentum, the resolution of the Cherenkov angle measurement is mainly
determined by resolution of the track extrapolation, in addition to multiple scattering in the
DIRC itself.
Using the reconstruction algorithm which is currently available in BBSIM, track parameters
are obtained at the point of closest approach to the IP by a progressive tting procedure using
drift chamber and silicon vertex detector hits. Therefore, the procedure for swimming the
track from the primary vertex to the DIRC includes angular errors coming from Coulomb
scattering in the support tube and the inner wall of the drift chamber. As a result, the
angular errors at the DIRC are signi cantly overestimated by BBSIM, especially in azimuth.
Table 6-2 gives the resulting e ective c resolution per track. A procedure with better
accuracy would be to use a Kalman lter t to extrapolate track parameters from the
outer drift chamber layers to the DIRC. Track extrapolation errors would then include only
the e ect of the material between the last drift chamber hit and the quartz bar, which is
equivalent to 0.02X0 at normal incidence. From preliminary results, the quoted errors on the
extrapolated angle should be reduced by 10 to 40% using such a technique. A constrained
t to the Cherenkov image may further improve upon the errors.
6.3.3 Particle Identi cation in the DIRC
The ultimate performance of the DIRC is limited by the Cherenkov angle di erence between
particle species. In a quartz radiator (with n = 1:474), the di erence in Cherenkov angle
between a pion and a kaon is fairly large at low momenta but is as small as 6.5 mr at 4 GeV=c.
The same di erence occurs between a muon and a pion at 770 MeV=c.
Figure 6-7 shows the K= separation as a function of track momentum and cos , obtained
with the ASLUND simulation of DIRC performance. The separation is very clear in the
Technical Design Report for the BABAR Detector
6.3 Projected Performance
211
100
#σ
50
0.5 GeV/c
1.0 GeV/c
1.5 GeV/c
10
2.5 GeV/c
5
3.5 GeV/c
1
-1
-0.5
0
0.5
cos θ
1
Predicted K / separation performance of the DIRC, quoted in terms of the
number of standard deviations, vs. cos , for di erent momenta.
Figure 6-7.
momentum region that is most interesting for tagging kaons in the barrel, 0.5 to 1.5 GeV=c,
which is why the kaon tagging eciency is so close to the ideal case. The separation is always
> 4 for particles within the kinematic limits for B decays, as can be seen in Figure 2-13.
Table 6-3 shows similar information on =K separation, but for both ASLUND and BBSIM
simulations of single track samples. This comparison is meant to illustrate the range of
performance estimates obtained from current simulations. The ASLUND results in column
(a) are known to be somewhat optimistic due to a lack of background, but on the other
hand do not incorporate potential improvements brought about by a constrained t to the
Cherenkov image. The BBSIM results in column (b) include the e ects from a realistic
simulation of background hits but overestimate the contribution due to the track angle
uncertainty because of the extrapolation of errors from the vertex and also do not incorporate
a constrained t. The actual =K separation is probably somewhat better than the ASLUND
result, but the comparison between the two simulations is an instructive measure of the
reliability of predictions from our present understanding. It is clear that good =K separation
is achievable in most of the phase space of the asymmetric collider.
BBSIM as well as ASLUND studies show that a = separation with greater than 2 can
be achieved by the DIRC at momenta below about 500 MeV=c. The separation power is
somewhat spoiled by the track direction uncertainty due to multiple scattering. However, a
separation of more than two standard deviations up to 500 MeV=c may be of great interest for
tagging purposes and is likely to improve with more accurate track ts and extrapolations.
Technical Design Report for the BABAR Detector
212
Particle Identi cation
Sine of Particle Dip Angle
Momentum
,0:5
0
0:5
GeV=c
a
b
a
b
a
b
0.500
38 68 60 167 38 69
0.800
35 35 40 63 33 32
1.000
27 26 30 42 25 25
2.000
11.0 9.3 12.2 11.7 9.7 8.9
4.000
2.9 3.0 3.1 3.0 2.3 2.7
0:85
a b
| |
24 36
23 24
9.7 9.6
3.9 3.3
=K separation in standard deviations as a function of track momentum and
the sine of the dip angle from a simulation using (a) ASLUND and (b) BBSIM. The
entries are outside the kinematic region populated by B decays.
Table 6-3.
6.3.4 Particle Identi cation in the ATC
Projected particle identi cation capabilities of the aerogel system are based on beam tests.
The n = 1:0085 aerogel counter's response to 3.5 GeV=c positive pions and protons (below
Cherenkov threshold) was measured (Figure 6-8). If a cut is made at 55 ADC counts,
corresponding to 1.5 photoelectrons, one obtains a 97% eciency for pion detection with a
3% misidenti cation of below threshold particles. Similar separation between pion and kaons
was also obtained at 3.5 GeV=c after subtraction of the background due to electronic noise,
wrongly tagged pions, and kaon decays.
For the current design, the Monte Carlo simulation tuned with our test beam results shows
that a =1 charged particle passing through the aerogel produces a signal of more than
ten photoelectrons in both the low and high refraction index aerogel blocks. With a ten
photoelectron signal, the aerogel PID system will give /K separation up to ' 4:3 GeV=c
[Oya94], as shown in Figure 6-8.
6.3.5 Performance Requirements for Track Reconstruction
In a simple model, the measurement error of the Cherenkov angle per track in the DIRC can
be written
!2
P E
2
2
2
c = p
+ trk
+ ext
;
NP E
p
2
is the track angular error at the last drift
where P E = NP E is the intrinsic resolution, trk
2
chamber layer, and ext
is error on extrapolation to the DIRC due to multiple scattering
in the outer wall of the drift chamber and in the DIRC supports. The study of errors due
Technical Design Report for the BABAR Detector
6.3 Projected Performance
213
1000
900
800
100
π
600
500
400
Efficiency (%)
3.5 GeV/c protons
700
300
K
60
40
π
3.5 GeV/c pions
1.055
K
20
200
1.0065
100
0
80
0
0
0
100
200
300
400
500
600
700
800
900
1000
1000
Responses of a prototype
1.0085 aerogel counter to 3.5 GeV=c pions
(solid line) and to the below threshold
protons (dashed line) in a 1.3 T magnetic
eld.
3000
4000
5000
Momentum (MeV/c)
ADC counts
Figure 6-8.
2000
Threshold curve for pions and kaons as a function of momentum for a two-index aerogel system.
Figure 6-9.
to multiple scattering in the quartz bar itself, which contribute to PE but are not fully
uncorrelated, requires further work. In principle, these e ects can be reduced by tting the
Cherenkov image.
Clearly, it would be desirable to maintain the track error contributions, due to the resolutions
on the projected angles and , below the intrinsic resolution (i.e., below 1.0 mr) at least
for large momenta at which the =K Cherenkov angle di erence is small. In this context, it
is important to minimize the amount of material between the drift chamber and the DIRC
radiator. For example, increasing the amount of material to 8% of a radiation length at a
60 dip angle (i.e., twice the present design) would lead to more than 1 mr uncertainty for a
4 GeV=c track due to multiple scattering. The present design of the drift chamber and DIRC
support achieves this goal, as shown in Figure 5-5.
Technical Design Report for the BABAR Detector
214
Particle Identi cation
Background Source
Primary Track-Equivalent/Event
Event-Related Background
47
Beam-Gas EM (10nom.bkg)
0 60
Radiative Bhabha
0 11
PMT Noise
0 020
Cosmic Rays
0 00016
Others
Negligible
:
:
:
:
:
Table 6-4. Backgrounds in primary track-equivalent units (33 photoelectrons per track)
for a readout time of 50 ns.
6.3.6 E ects of Backgrounds on PID Detector Performance
E ects of Backgrounds on DIRC Response
Several background sources can a ect the performance of the DIRC, such as event-related
backgrounds, synchrotron radiation photons, beam-gas interactions, cosmic rays, phototube
dark current BBSIM has been used to simulate the detector response. Each random background source has been integrated over an assumed 50 ns readout time, and the probability
of a double hit in the same PMT in one event has been studied for readout times of 50 ns
and 100 ns.
Event Related Backgrounds. Based on a BBSIM simulation, a typical (4S ) ! B 0 B 0
event produces an average of 345 photoelectrons detected in the DIRC. Of these, 190
are attributed to an average of 5.7 primary tracks which lie within the DIRC acceptance
and have momenta above Cherenkov threshold. Thus, each primary particle contributes
33 photoelectrons. The other hits are due to secondary tracks produced in the detector,
mostly in the form of backsplash from the electromagnetic calorimeter, that have much
softer momentum spectra than the primary tracks. Their random angular distribution gives
an event-related background in the DIRC, which manifests itself in the angular coordinate
system used for pattern recognition as a at contribution under the Cherenkov peaks for
primary particles. This background can be subtracted easily.
Other Backgrounds. The machine is a source of backgrounds from lost particles, radiative Bhabhas, and other backgrounds as discussed in Chapter 12. The largest source of
machine-related background is the beam-gas electromagnetic term as shown in Table 6-4.
Other backgrounds such as synchrotron radiation, cosmic rays, and PMT dark current have
very small or negligible impact on the DIRC response.
Technical Design Report for the BABAR Detector
6.3 Projected Performance
215
Multihit Probability. On average, 175 PMTs are hit by the 190 photons associated with
measurable tracks in a BABAR event, giving a double-hit probability of 8%. Including hits
due to event-related background, the number of PMTs with a double hit increases to 16
and the probability becomes 9%. If beam-gas interactions are included at 10 times nominal
background, the double-hit probability is 9.2% for a read-out time of 50 ns and 9.5% for a
readout time of 100 ns.
E ects of Backgrounds on ATC Response
Since each cell is read out by a PMT placed outside the Cherenkov radiator, charged particles
passing through a PMT pose little problem for particle identi cation in the ATC. One
background source is shower leakage from the calorimeter. For incident photons, this is
generally not a problem, as no track will point to that cell; one can actually use the photons
to determine the size of the e ect. Backsplash from electromagnetic showers consists mostly
of photons and electrons below Cherenkov threshold. The photons cause a problem only to
the extent that they convert in the material of the PID system.
Another background is -ray production, since a below-threshold kaon can be misidenti ed
as a pion by knocking out an electron with momentum above Cherenkov threshold. Both
an analytic calculation [Oya94, Gro92] and a detailed GEANT simulation [Shi94] have been
used to study the e ect. The results are in good agreement with measurements in beam tests.
In the forward direction, the study shows that 1% of 4 GeV=c kaons may be misidenti ed as
pions due to -ray production in the drift chamber wall and backsplash from the calorimeter.
With 97% detection eciency for pions, the system still provides 4 =K separation
at 4 GeV=c. Though -ray production and backsplash cause about a 2{3% probability of
misidentifying a low momentum (p < 1:2 GeV=c) kaon as a pion, they do not a ect B !
and B ! K separation since both daughters from these decays have p > 1:5 GeV=c. In the
worst case, such K ! misidenti cation results in a B tagging ineciency of less than 3%.
Synchrotron radiation photons are not expected to cause problems. Their energies lie
between 4 and 100 keV, and cannot produce any above threshold (1 MeV for n = 1:06)
electrons. Their wavelengths are also not in the photocathode-sensitive region.
Beam-gas background should not be a problem either with a 100 ns integration time (the
PMT's pulse width is a few tens of ns). The average number of hits per microsecond in the
whole CsI calorimeter is about 0.8. Therefore, fewer than 0.08 hits per 100 ns are expected
over the 144 aerogel cells, corresponding to 0:0005 hits per cell during one readout cycle.
Technical Design Report for the BABAR Detector
216
Particle Identi cation
PMT Module
,,
,,,
,,
,, ,,
,
,,
,,
,, ,
,,
,
,
,
,,
,
,
,
Quartz Bar Sector
Hinged Cover (12)
,,,
,,,,
Plane Mirror (12)
Standoff Cone
~5
m
~2 m
Figure 6-10.
6.4
6.4.1
Mechanical elements of the DIRC.
The DIRC Detector
DIRC Mechanical Design
Principles of Operation
The major mechanical elements of the DIRC, shown schematically in Figures 6-10 and 6-11,
include the following.
Quartz Radiator Bars and Sectors. The bars must have small attenuation loss for
300 nm. The surfaces must be accurately rectilinear and smooth to preserve light angle
and intensity after many re ections. The choice of bar thickness and width involves tradeo s among light intensity, angular resolution, material ahead of the calorimeter, and cost.
The bars are assembled into 12 separately covered and mounted sectors of 13 bars each. The
bars of a sector are united at the water boundary by a glued quartz window which is sealed
by an O-ring to the assembly ange.
Technical Design Report for the BABAR Detector
Figure 6-11.
Schematic of the DIRC support system.
Support Tubes and Assembly Flange. The central support tube is an aluminum
honeycomb between cylindrical skins which holds the sectors within the sensitive volume
of the detector. Its material thickness in radiation lengths will be small compared to the
quartz bars. The central support tube will be cantilevered from the strong support tube.
This steel tube, axially aligned with the central support tube, also holds the quartz sectors.
It is mounted inside the backward IFR and is the central structural element of the DIRC,
strong enough to support the beam line elements and the backward poletip. The steel tube is
supported independently by the barrel IFR. The assembly ange is a short aluminum section
rigidly connecting the strong support tube to the stando region. Its functions are to reduce
the magnetic ux coupled to the PMTs, to seal the sector windows, and to support and seal
the inner and outer radius boundaries of the stando volume.
Stando Region and Mirrors. A water- lled volume between the windows at the ends
of the quartz sectors and the PMTs permits the Cherenkov image to expand. The volume
is de ned by a cylindrical inner boundary and a conical outer boundary that nearly meet
at the sector windows, plus a back toroidal surface covered by the PMTs. The walls are of
high-quality magnetic steel to eliminate the e ects of stray magnetic elds on the PMTs.
Pure water is inexpensive, gives plenty of light transmission, and matches quartz fairly well
in index of refraction and dispersion. There are 24 at mirrors azimuthally aligned with the
sectors. Half are mounted on the inner radius cylinder, the others on the cone at a 55 angle
to the beam line. The choice of this angle is a compromise between the number of PMTs
needed and ambiguities in the Cherenkov images and is still under review. To minimize
mirror distortion from stress, a neutrally buoyant honeycomb structure is proposed.
Technical Design Report for the BABAR Detector
218
Particle Identi cation
Phototubes. The PMTs need good single photoelectron response and good quantum
eciency for 300 nm. The PMT faces are immersed directly in water, covering the
entire back toroidal surface in a close-packed geometry. The module geometry for the PMTs
has not been nalized. If individual tubes (or very small modules) are used, the toroidal
steel PMT support plate (Figure 6-12), located at a radius somewhat larger than that of the
PMTs, can accommodate the necessary access holes without sacri cing either mechanical
structure or magnetic shielding at the PMT face. Steel cover plates that would limit access
will be avoided.
Quartz
The quartz bars will be made from bulk natural fused quartz (e.g., Vitreosil 055). This
material has an index of refraction of 1.474, good light transmission for 300 nm, is
reasonably radiation hard, and is relatively free of bubbles and inclusions. This bulk material
will be used to produce bar pieces of 117:5 3:5 1:75 cm3 , which will be glued to form
156 full-length bars. These dimensions were chosen with cost, photoelectron yield, angular
resolution, and minimization of material in front of the calorimeter in mind. The bars need
to be extremely rectilinear to preserve the Cherenkov image, and well polished to preserve
the light after many re ections. We have set up an optical laboratory for quality control
and characterization of all the optical properties of the quartz bars. Cosmic ray tests, as
described in Section 6.4.5, indicate that commercially produced bars meet the requirements.
The decision to glue short bars together to form a full-length radiator is driven by the
excessive cost of tooling needed to cut and polish longer bars at standard optical industry
shops. An epoxy with excellent optical properties and sucient mechanical strength to join
the short bars together has been selected and tested. Alignment xtures and gluing jigs
will be used in the process. Radiation hardness of the quartz, the epoxy, and its long term
properties will be veri ed. Since the bars are read out in one direction, mirrors will be needed
on the other end. We will use glass mirrors coated on the front surface with aluminum and
silicon dioxide. For maximum re ection, the mirror will not be coupled optically to the bar
end since many photons are internally re ected.
Sectors and Covers
There are several reasons that the quartz bars are arranged in azimuthal subassemblies or
sectors. First, for structural reasons, it is necessary to allow azimuthal gaps for ribs to couple
the inside and outside radii of the DIRC support tubes. Second, a water-tight seal is needed
at the junction with the stando region. Third, the optics of the stando region work best
to maintain the resolution and minimize reconstruction ambiguities if the mirrors are at
and aligned with the sectors. These conditions tend to drive the sector multiplicity down.
Technical Design Report for the BABAR Detector
6.4 The DIRC Detector
219
Sector Cover
Assembly Flange
Quartz Bar
Quartz Window
Window Frame
Figure 6-12.
Schematic view of bars assembled into a mechanical and optical sector.
On the other hand, they increase the radial width of the DIRC, not only because the sagitta
is proportional to the square of the sector width, but also because the sti ness of the sector
covers is a sensitive function of the width. For the parameters considered here, the radial
budget is: sagitta, 3 cm; quartz and gaps, 2 cm; sector covers, 2 cm; and support tubes, 1 cm;
for a total of 8 cm. The minimal azimuthal gap, between the inside corners of quartz bars in
neighboring sectors, is 1 cm, which is sucient to accommodate side-by-side O-rings. The
average coverage loss between sectors is then approximately 2 cm, less than 5% of the total
circumference.
The bars are assembled into a mechanical and optical unit or sector on a long optical table.
At the end near the stando region, each bar is glued to a common 3 mm-thick quartz
window. The window area is somewhat larger than that of the bars so that light can emerge
unimpeded at large angles from all parts of the sector. Because the bars are close together
with sides parallel, care must be taken to avoid optical crosstalk. For this reason, the bars
are separated at the support points by thin, small-area plastic wafers aluminized on both
sides. To seal the sector at the stando region, an aluminum window frame is sealed by
O-rings; rst to the back of the window and then to the assembly ange as the nal part of
sector installation. Some details are shown schematically in Figure 6-12.
In our present design, each bar is supported at 0.6 m intervals, 0.3 m from the glue joints.
In this geometry, each joint is normally at a point of maximal (but very small) stress, but it
is subjected to zero additional stress due to a misalignment of one of the nearest supports.
Technical Design Report for the BABAR Detector
220
Particle Identi cation
For the part of the sector inside the sensitive region of the detector, 3 m in length, the
periodic bar supports are mounted to top and bottom plates of low density honeycomb, 1 cm
thick by 42 cm wide, which are coupled together by thin strips on the sides. The periodic
supports are strengthened by a transverse sti ener, which is a few centimeters wide axially.
The bars do not touch the honeycomb or its sti eners directly but are spaced below by an
aluminized 1 mm-diameter ber and clamped on top by a threaded screw, aluminized on the
tip. Transverse support is supplied by stressing the spaced coupling strips on the sides, so
that the cover supports the sector independent of orientation. The part of the sector outside
the sensitive region of the detector, 2 m in length, has top and bottom cover plates of solid
aluminum.
To protect the glue joints during extreme (e.g., seismic) axial motion, each bar is compressed
at its own weight ( 100 kg/sector) by spring loading the front surface mirror at the nonreadout end of each sector. The force is returned to the window by the tension in the thin
skin (0.1 mm stainless steel) outer layer of the sector cover. Because the covered sectors can
easily be made gas tight, the internal atmosphere is readily controlled by a nitrogen purge.
<
Support Tubes and Assembly Flange
Figures 6-10 and 6-11 illustrate the concepts discussed in this section. Very di erent structures are needed to support the sectors in the sensitive (central support tube) and insensitive
(strong support tube) parts of the detector.
The central support tube must be as thin as feasible. The problem is made somewhat
dicult by the long distance from the major structural elements of the DIRC. To avoid
coupling this relatively fragile assembly simultaneously to structurally independent ends of
the detector, one of the proposed designs uses a cantilever. The major structural elements
are aluminum: 1 mm-thick skins at inner and outer radii; 3 mm-thick circumferential ribs
slotted for the sectors and spaced every 0.6 m; and 2 mm-thick axial ribs every 30 linking
the circumferential ribs. The average material thickness of the support tube is 0 04 0.
Carbon- ber support structures will be studied in order to further reduce the radiationlength thickness. Studies using a nite element analysis show that this design will carry
the required load of 17 kg per meter per sector. The structure is predicted to de ect a few
millimeters when loaded, with buckling the nearest failure mode. The load limit can be
increased by distributing the forces at the corners of the ribs. Inside the central region, but
beyond the maximum angle of the barrel calorimeter, the central support tube is a heavy
aluminum structure like the steel strong support tube to which it is joined. This design
reduces the required length of the low density cantilever.
The strong support tube, located inside the IFR, requires steel for strength and for returning
magnetic ux from the poletip. To accommodate whatever solution is adopted for beam line
:
Technical Design Report for the BABAR Detector
X
6.4 The DIRC Detector
221
support, the strong support tube is designed to support, in addition to the DIRC, the
backward beam line elements and the poletip, which feels an axial force of 300 tons at 1.5 T.
Given septa of 1 cm width every 30 to couple the inner and outer skins, careful analysis
shows that the tube is strong enough for this worst case. Unlike the central support tube,
the strong support tube is not limited to a radial width of 8 cm. At the outside radius, the
strong support tube is attached by a ange to the yoke supported by the barrel IFR. The
inner radius matches that of the assembly ange and the cylinder of the stando region,
which are discussed below. The preferred fabrication method for the strong support tube is
to cut slots for the sectors into annular plates of convenient thickness. These are pinned for
alignment and held together by as many as 24 axial bolts (12 each on the inside and outside
radii).
The 15 cm-deep aluminum assembly ange is rigidly supported from the back of the strong
support tube and has two major functions. First, it breaks the magnetic circuit between the
stando region and the poletip, IFR, and strong support tube; and second, it provides the
seal and supports for the sectors and stando regions. Based on the magnetic simulation
done to date, it is known that the eld at the PMTs is acceptable with the gap; whether it
is essential has still to be determined. If it is not, this assembly ange could be eliminated,
saving 15 cm in the length of the quartz bars. The strong support tube could handle the
major mechanical functions, which would save a signi cant amount of radial space.
Stando Region
The stando volume is lled with approximately six tonne of water, contained by an inner
radius cylinder, an outer radius cone at 55, and a toroidal back PMT support plate. They
are of high quality magnetic steel; simulations using 5 cm-thick steel (permeability 104)
gave B 0:05 T at the PMTs with a central eld of 1.5 T. Tests on conventional 29 mmdiameter PMTs showed no e ect at this eld, independent of orientation. It is expected that
further optimization through simulation will lead to reduced steel thickness in some regions.
In any case, it is essential that the joints between the PMT plate, the cone, and the cylinder
provide good ux transfer. With 5 cm-thick walls, the total weight is 15 tonne. The water
reservoir is mounted below the stando region, so the forces do not change signi cantly
when water is transferred. We plan to use de-ionized reverse-osmosis water, which has an
attenuation length of 10 m at = 300 nm. The water will recirculate slowly through a lter
and under a UV lamp, and the attenuation will be monitored oine. In addition, there will
be an emergency pressure relief valve, an expansion reservoir for temperature changes, and
a dam to channel water in case of leaks.
Technical Design Report for the BABAR Detector
Figure 6-13.
Schematic of a possible PMT module assembly design.
Mirrors
The mirrors in the stando region are needed to minimize the required number of PMTs;
they have the same azimuthal symmetry as the sectors. The 0 dip angle mirror is essential
because half the light from each sector emerges from the quartz bar in a direction which is
pointed towards a smaller radius. The mirrors could be aligned from within the stando
volume, although it would be convenient to use sealed adjustments for outside control. The
mirrors on the conical surface are presently assumed to be placed at a 55 angle to reduce
the number of PMTs.
The mirror design is complicated by the need for immersion in water for a long period of
time. If a front surface coating is used, it must be covered for protection. A likely possibility
is to use aluminum with a coating of either silicon dioxide or a series of dielectric layers.
An alternative is to coat the back of a sheet of thin UV-transmitting quartz, which is then
sealed against water. Mirror \coupons" could be immersed in the water circuit and removed
periodically to measure the degradation of re ectivity with time. The mirrors will be made
neutrally buoyant by using a hexcell support structure to minimize the distortions.
PMTs and Modules
Approximately 13,400 PMTs of 29 mm diameter will be needed to cover the toroidal surface
in a close-packed array. The packing fraction depends upon design details but will be 87%.
Typically, the PMTs are only sensitive inside a diameter of 25 mm. To recover this lost
region, each PMT will carry a re ecting cone with a depth of 4 mm and radial width 2 mm
Technical Design Report for the BABAR Detector
6.4 The DIRC Detector
223
to collect most of the light hitting the cone at dip angles less than or equal to 30. The
individual PMTs will be mounted into modules that can be joined together to provide a
leak-tight interface between the water of the stando tank and the phototube array. Several
di erent schemes are now under evaluation. One of these is shown in Figure 6-13.
6.4.2
Photodetectors and Readouts
DIRC Readout
In principle, the Cherenkov photon arrival time contains information about the Cherenkov
angle. This information becomes competitive with the position information obtained from
the array when the time resolution approaches 100 ps for a single photoelectron hit. If
cost were not an issue, modern instrumentation and very good PMTs could provide time
resolution at this level. However, it is substantially less expensive to determine the angle information from the position on the PMT array plane and not attempt such precise
timing. Timing information, however, is still a powerful tool for distinguishing signal from
background. For each track, the dominant background comes from the event itself, which
occurs over a time window of about 50 ns. Thus, to gain a signi cant factor in signal to
noise on these backgrounds requires a timing resolution of a few nanoseconds. The readout
electronics will provide a timing resolution on the order of 1 ns. Although the pulse height
carries no additional information for a single hit, it is useful for calibration purposes. We
propose implementing a system which o ers the option of measuring the pulse height during
dedicated calibration runs.
Figure 6-14 is a block diagram of the readout chain for a single PMT. During normal physics
running, the time is recorded in a FIFO whenever the signal at the output of the PMT
crosses a preset threshold. The FIFO is large enough to store the maximum data for the
duration of the Level 1 trigger latency, 9.5 s. Upon receiving a Level 1 trigger signal, the
data in the FIFO falling within a 0:5 s window around the trigger are sent to the DAQ
readout module where they are bu ered before being shipped over the FDDI network to the
Level 3 farm for further processing. The baseline design calls for the front-end electronics to
be located near the PMTs, with 64 channels per card and one ber-optic link to the DAQ
readout modules for each group of 960 channels. The VME-based DAQ readout modules,
located in the electronics house, will receive four optical bers each. The baseline design of
this system and a description of continuing research are presented in Chapter 10.
The criteria of greatest importance for the DIRC photosensors are high quantum eciency, maximal coverage of the photon image plane, adequate
gain, good timing resolution (1 ns), low noise, and low cost. The performance of candidate
Photodetectors and Bases.
Technical Design Report for the BABAR Detector
224
Particle Identi cation
L1 Trigger
Discriminator
TDC
Control
FIFO
PMT
MUX
Peak
sensing
ADC
DAQ
Readout
Block diagram of the readout chain for a single PMT. The FIFO is capable
of storing data for the duration of the Level 1 trigger latency.
Figure 6-14.
PMTs from three leading manufacturers has been tested: Hamamatsu R268, Philips XP2982,
and EMI 9124A. These PMTs are all of similar design: 29 mm in diameter with bialkali
photocathodes. As a result of these investigations, the Hamamatsu R268 has been chosen
for the forthcoming prototype with 500 tubes. These tubes are, therefore, included in the
DIRC baseline design. The market will be researched again before the nal choice of PMTs
is made.
Both timing and pulse height information were obtained for the three tube types using a
prompt Cherenkov light signal. This was produced by a particle emitted from a 106 Ru
source traversing a thin quartz plate. The source was suciently weak that nearly all
detected signals were the result of single photoelectrons. The absolute gain was derived
from the measured pulse height distribution and from the known ADC charge sensitivity.
The time distributions were obtained with the same system. A very fast TDC start signal was
generated by stopping the particle in a scintillator. Gaussian ts to these time distributions
were used to extract their rms time spread. The noise (dark current) and the photocathode
response, were measured. (The latter is de ned to be the mean PMT pulse height divided by
the gain when the photocathode is uniformly illuminated.) These phototube characteristics,
which measure photon detection ability, have been compared for the three tubes using the
same light source operated at the same intensity.
The results of the measurements are collected in Table 6-5. The voltages given in the table
are the highest voltages at which the PMTs may be safely operated according to criteria
which are believed conservative. The dark current can vary from tube to tube and is shown
only to indicate that it is small enough to be of no concern. Except for the time resolution,
the performance results favor the R268 tube. The improvement in time resolution that would
come from using the Philips tube does not compensate for the loss in response.
Technical Design Report for the BABAR Detector
6.4 The DIRC Detector
225
Tube
HV Gain Noise Rate Pulse Height Resol. Time Resol. Relative
(kV) (108)
(kHz)
=PH(%)
(ns)
Response
R268 1.6 1.1
2.6
42
1.2
1.6
XP2982 2.0 1.2
3.4
38
0.55
1.0
9124A 2.0 5.7
2.2
20
1.3
1.3
Table 6-5.
Comparison of the three tested PMTs.
The details of both the electrical and mechanical design of the PMT base depend on the
particular device chosen and the mounting scheme. However, neither of these choices
poses particularly dicult problems. A base composed of two printed circuit boards has
been developed for the current choice of PMT (Hamamatsu R268) in the 500 PMT DIRC
Prototype-II. An alternate scheme [Hub92] using a voltage multiplier is also being studied.
The HV feed to 24 individual bases will be provided by a single PC board, serviced by one
or two HV supply channels. The same board may also be used to route the signal lines onto
a single 24-pair ribbon cable (as with the prototype) or to signal processing electronics on
the same board. This board can be mounted directly behind the PMT support.
The PMTs are located in the stando box. Their backs and bases will be easily accessible
for repair and/or replacement.
6.4.3 Laser Flasher Monitoring and Calibration System
The DIRC laser asher system will be used to measure the gain and relative quantum
eciencies of each tube and to monitor the optical performance of the quartz bars, glue
joints, mirrors, and windows. The conceptual design for this system uses one or more pulsed
nitrogen lasers as the light source. The wavelength, 337 nm, is characteristic of Cherenkov
photons. The pulse duration is shorter than the timing resolution of the DIRC. The number
of lasers will be determined by their repetition rate and the desired trigger rate.
The laser light will be distributed by optical bers. Two methods are under study:
Locate the ber between photomultiplier tubes (see Figure 6-15). The light will be
directed towards a scattering surface located above the bar window.
Locate the bers at the junction of the quartz bar exit and the stando box with one
ber directed along the bar towards its far side and the other directed into the water
through a scattering surface. (The use of LEDs or Xenon ashers in place of lasers is
also under study.)
Technical Design Report for the BABAR Detector
226
Particle Identi cation
optical fiber
fan
out
laser
water tank
scattering
surface
pmt
quartz bar
mirror
mirror
Figure 6-15. Conceptual design of the DIRC laser asher system. The 337 nm light is
distributed by four bers per sector and re ected from a di use scattering surface, providing
a reasonably uniform distribution of single photons to all phototubes.
Each of the bers of a sector will be a di erent length to provide a timing di erence that
can be used to distinguish pulses from di erent bers.
The scattering surfaces will give a reasonably uniform illumination of all phototubes. The
light level will be adjusted so that the probability per pulse of any particular PMT being
struck by a photon is small. This ensures that essentially all hits will be due to single
photoelectrons.
The gain of each PMT will be derived from the single photoelectron pulse height spectrum;
the number of hits is a measure of the stability of the relative quantum eciency.
Light will travel the length of the bar and re ect o the far mirror before striking the PMT.
The ratio of the number of these hits to direct hits is proportional to the transmission of the
bar, glue joints, window, and the re ectivity of the mirror. The short pulse duration of the
light source is needed so that these hits can be distinguished on the basis of timing.
It is anticipated that a asher calibration will take approximately 10 minutes and be performed once per week. The laser trigger will be used also as a level 1 detector trigger, with
a partition for DIRC data only.
6.4.4
Integration Issues
Oine DIRC Assembly and Test
A set of twelve aluminum boxes will be built to simulate the sector boxes containing the
quartz bars. These will be dimensionally correct and loaded to the weight of a real sector.
These boxes will be preassembled into the central support tube where clearance checks and
modi cations can, if needed, be made. These boxes will also serve to align the light section
Technical Design Report for the BABAR Detector
6.4 The DIRC Detector
227
of the central support tube with the heavy aluminum section. At this point, the central
support tube is rigidly held o the oor by a spider and framework. The central support
tube and the strong support tube are mated into an accurate cylinder 5 m long.
The sector boxes will all be installed at the top of the support tube; the support tube will
be rotated for each new sector box. We begin by removing a dummy aluminum box and
bringing up a sector on an adjustable assembly strongback. The box is precisely aligned
with the slot and moves in on rollers pulled by a line. Compressed air ow around the sides
reduces friction. After installation, the sector will be sealed to the assembly ange and the
next insertion will take place.
The stando region, which weighs approximately 15 tonne, could be preassembled and moved
by crane. A major advantage of this approach is that it would allow an inspection of the
PMTs and electronics before moving onto the beam line. At this stage, the weight of the
device is supported by the oor. The 24 mirrors in the stando are installed and aligned,
after which the PMT support plate is bolted to the cone. The PMTs and electronics are
installed and a dummy assembly ange mounted. The water reservoir and other devices can
be attached for calibration, monitoring, and possibly cosmic ray tests.
Online DIRC Installation and Test
The support tube and sector assembly is inserted into the detector, supported by a yoke
directly from the barrel IFR. After surveying, shims are used to align the central support
tube with the drift chamber. The stando structure is bolted directly to the assembly ange
and sealed. The procedures described above are repeated for inspection and testing and
upon successful completion, the beam line elements, poletip, and utilities are placed inside
the DIRC tunnel.
Cabling and Access
The signals from the 13,400 DIRC phototubes are carried by ribbon wax or twisted pair
cables to the front-end electronics, located in nearby crates. They are multiplexed onto a
small number of optical bers for transmission to the electronics hut. The cable plant is
dominated by the 900 HV cables. There are, in addition, 144 optical bers for calibration
and additional cables for monitoring and charge-injection calibration. The cable routing is
straightforward and presents no interference to other systems.
All active components of the DIRC are located in the stando box and can be accessed by
removing the electronics covers|no other detector subsystems must be moved. Access is
expected to be required infrequently, since the failure of a small number of isolated channels
will have no impact on the physics performance. No access is required to the radiators.
Technical Design Report for the BABAR Detector
228
Particle Identi cation
6.4.5 Research and Development Program
Prototype-I: Proof of Principle
The goal of the conceptual prototype (Prototype I) was to demonstrate the basic proof of
principle of the DIRC [Ast94]. Because no detector like the DIRC had been built before, this
prototype needed to address many fundamental issues, and be easily modi ed to respond
to new issues that arose during the studies. The detector was placed in a cosmic ray beam
hardened to 1 GeV by an iron stack. A series of scintillation counters served as the trigger,
and a set of straw tube chambers was used for tracking in the angular resolution studies. The
quartz bar used initially was 120 4 7 1 7 cm3. It was supported at a few contact points
inside a light-tight box. Later, two bars were glued together to make a bar 240 cm in length.
A mirror was attached to the non-readout end in order to make studies at an e ective
distance of 4 m from the readout end in the reverse con guration. For the photoelectron
yield and attenuation length studies, a Burle 8850 51 mm PMT was glued to the end of the
bar. For the angular resolution measurements, an array of 47 EMI 9124A 29 mm PMTs was
placed at varying distances from the end of the bar with air lling the stando region.
The photoelectron yield was studied as a function of dip angle and position along the bar.
The results of one of these measurements are shown in Figure 6-16, for a position in the
middle of a bar. These results are consistent with the Monte Carlo prediction, shown as a
solid line. The behavior at large dip angles demonstrates a major advantage of the DIRC
at an asymmetric collider in that the highest momentum tracks, which are forward-going
at large dip angles, produce large numbers of photoelectrons. The measured yield with the
PMT directly glued to the bar does not represent the number expected in the real DIRC at
the same distance and angle, since there are additional losses from the packing of the array
tubes, water in the stando region, and transition e ects.
There are also length-dependent losses from attenuation in the quartz, both from light
absorption in the bulk material, and from losses at the surface. To study these e ects,
a series of measurements was made with both the 120 and 240 cm bars, in both forward and
mirror re ected con gurations at a constant dip angle. As expected, when characterized by
a simple exponential, the measured attenuation lengths become somewhat longer as the path
lengths increase. However, with this readout system, they are approximately 10% per meter,
consistent with a bulk transmission length of greater than 10 m, and an internal re ection
coecient of more than 0.9995. This performance, which is completely adequate for the
DIRC needs of BABAR, is shown in Figure 6-5.
The single photoelectron angular resolution was studied for a variety of dip angles, positions
of the beam along the bar, and stando distances in air. Major contributors to the resolution
are source and detector size, and chromatic dispersion. Lesser contributions come from the
momentum spread in the beam, multiple scattering, tracking, and reconstruction errors
=c
:
:
Technical Design Report for the BABAR Detector
6.4 The DIRC Detector
229
50
40
Photo-electrons
150
σ = 8.5 ± 0.6 mr
30
100
20
50
10
0
–25
11-94
0
25
50
θD (degree)
75
7641A3
Observed photoelectron
yield from a 1.7 cm-thick DIRC bar at a
track position 60 cm from the photodetector end as a function of the track dip
angle D . The solid line is a Monte
Carlo simulation; the statistical errors are
smaller than the data points. There is
a scale error of 3% due to calibration
uncertainty.
0
0.8
0.9
1
Figure 6-16.
Cherenkov Angle (radians)
Single Cherenkov photon
angle distribution at D = 30 , a stando
distance of 90 cm, and a track position
205 cm from the end of a 240 cm-long bar.
The line shows a t to the data obtained
with a Gaussian plus polynomial form.
Figure 6-17.
in the telescope. A typical result is shown in Figure 6-17. The measurement resolution
of 8:5 0:6 mr is in good agreement with Monte Carlo simulations for the Prototype-I
con guration, but di ers from the resolution which will be obtained with the actual DIRC.
In addition to the cosmic ray tests, a test in a proton beam was performed using a quartz
bar identical to the one described above. A PMT was attached directly at each end of the
bar with a silicone optical joint, and both amplitude and timing information were recorded.
Proton beams of = 0.8 and 0.95 traversed the quartz bar at various positions and angles
along the quartz length. The number of photoelectrons detected by the PMT, towards which
the emitted Cherenkov light propagates by internal re ection, was measured to be 65 for
= 0.95 at a 30 angle of incidence. The observed absorption was only 3% per meter. The
fact that the number of photoelectrons was slightly smaller and the absorption length longer
than in the cosmic test is thought to be due to the di erent absorption properties of the
optical couplers between the bar and PMT.
Technical Design Report for the BABAR Detector
230
Particle Identi cation
A signi cant amount of light was also detected by the opposite PMT. The direction and
speed of propagation of this light has been measured, and two components have been clearly
identi ed. The rst corresponds to re ection from the other PMT of the normal Cherenkov
light. In this setup, its fraction amounted to 8% of the emitted light.
The second component cannot be produced by Cherenkov emission of the incident particle
since it travels in the wrong direction, with a low longitudinal propagation speed that does
not depend on the incident track angle, and is seen even below the Cherenkov propagation threshold. This second component, produced isotropically in the bar, was present in
approximately 30% of the events and corresponded to about four photoelectrons (with an
exponentially falling amplitude spectrum). In BABAR, this component will lead to an extra
background of 2 to 3 photoelectrons per track, if a simple extrapolation holds, and therefore
will not degrade the DIRC performance.
The conceptual prototype test program, which is now completed, was essential as the proofof-principle test, demonstrating that the DIRC can produce enough photoelectrons with
sucient angular resolution for all track angles and positions to provide good =K separation
for BABAR.
Prototype-II: Large-Scale Engineering Prototype
After the proof of principle, the next important step is a proof of viability for a large,
complete detector system. To that end construction is nearing completion on a large scale
prototype (Prototype-II), which approximates a full size section of the eventual con guration.
It provides a realistic model to test many of the principal engineering and fundamental
system performance issues. In particular, it will allow the observation of a nearly complete
Cherenkov image on a track-by-track basis and the direct measurement of the resolution,
in mass or , also on a track-by-track basis. Moreover, because of its size, the prototype
provides a realistic assessment of the impact of the mechanical structure on the performance
of the nal device. Examples of components which may a ect performance are glue joints
between bars, quartz windows, water- lled stando box, mirrors, PMTs, supports, and
interface.
A cut-away isometric view of Prototype-II is shown in Figure 6-18. The prototype consists of
two 1:67 4:6 120 cm quartz bars glued together to form a composite bar 240 cm long. This
bar is coupled to a water- lled stando region with one mirror inside and 500 PMTs (29 mm
in diameter) on the back plane of the stando region for photodetection. The phototubes
are read out into standard modular electronics.
The dimensions of the stando box are 2.4 m wide by 1.2 m on each edge, with a 60 angle
between each of the large surfaces. There are two possible ports at which the quartz bar can
be connected to the stando box, one centered and one o center. The latter port allows
Technical Design Report for the BABAR Detector
6.4 The DIRC Detector
Figure 6-18.
231
Isometric view of the Prototype-II detector.
the study of the extreme end regions of the Cherenkov conic images outside the acceptance
of the back plane when the quartz bar is in the center port. The quartz bar is connected to
the port by an interface consisting of a quartz window sealed in a watertight frame bolted
to the stando box. The port and interface are the same size as in the nal con guration,
which will accommodate a complete sector of quartz bars.
For simplicity, the back surface is a plane, instead of a curved toroidal shape, in the nal
con guration. Because of the limited number of PMTs for the prototype, they must be
specially con gured on the back plane for each beam dip angle to subtend the speci c band
of light corresponding to the Cherenkov image produced at that angle. To provide maximum
exibility for this, the PMTs are separated from the water volume by a 2 cm-thick quartz
window barrier, made up of nine window panes as shown.
Construction of Prototype-II at LBL is scheduled for completion at the end of February,
1995. It will then be installed in a hardened cosmic ray telescope at SLAC for tests from
March through early May. It will then be shipped to CERN for extended tests starting in
early June of 1995, with beams of pions, kaons, and protons below about 4 GeV/c.
Future upgrades of this prototype will include lengthening the quartz bar to 480 cm, attaching
bar mirrors, adding more bars side by side (up to 12) to make up a quartz bar sector as in
Technical Design Report for the BABAR Detector
232
Particle Identi cation
Aerogel 1.055
Aerogel 1.005
Figure 6-19.
R- view of the ATC detector
the nal con guration, adding more PMTs, testing various calibration schemes and di erent
electronics. The versatility of this prototype construction also permits future tests of di erent
PMT mounting and interface schemes with the water volume, such as having the PMT front
faces immersed in the water.
6.5 The ATC Detector
6.5.1 ATC Mechanical Design
The forward PID system consists of 144 blocks of aerogel arranged in two rings that are two
layers deep, as shown in Figures 6-19 and 6-20. The front layer, 146 mm long, has a refractive
index of 1 006, and the back layer, 60 mm long, has an index of 1 055. The detector will
be built in two C-shaped pieces, as shown in Figure 6-21, as is the endcap electromagnetic
calorimeter, to which it is attached.
Each half endcap is composed of:
:
Technical Design Report for the BABAR Detector
:
6.5 The ATC Detector
233
Figure 6-20.
R- view of one ATC module.
Six modules, subtending 30 in , each housing six aerogel blocks of two indices. The
block dimensions are shown in Figure 6-20. The modules, shown in Figure 6-21, are
made of a light composite material covered by a water-resistant PTFE Te on with
a re ectance of about 95% over the region 300{700 nm. PTFE te on has been used
extensively in our beam tests. Any di use re ector used for the aerogel wrapping must
be of low density and high re ectivity. It must remain stable over a long period. Other
candidates are Millipore paper and Kodak paint. All are hydrophobic and commercially
available. Re ectivities of these materials are under study.
Each aerogel block is read out by a Hamamatsu ne mesh photomultiplier through a
di using box located on the top and bottom of the module, as shown in Figure 6-19.
Phototubes of 52 and 35 mm diameter are used for the 1.0065 and 1.055 blocks,
respectively. The e ect of the material in the phototube on the performance of the
DIRC and calorimeter is being studied.
The modules are glued on a half cone skin made of 2 mm-thick composite material.
This front skin supports kapton high voltage and signal cables.
The entire front section, equipped with readout and aerogel blocks, is mounted inside
a conical backing structure arranged as a half-shell in azimuth. This support provides
rigidity for each half endcap, supports cables and services, prevents light leaks between
cells, and provides the mechanical connection to the supporting calorimeter endcap.
Technical Design Report for the BABAR Detector
234
Particle Identi cation
Generated heat is removed by blowing nitrogen at the top and bottom of the detector around
all tubes. The nitrogen ow also protects the aerogel against humidity.
The size of the aerogel blocks is a compromise between minimizing the number of readout
channels and maximizing the number of photoelectrons. A low-density block of transverse
size 10 10 cm has been demonstrated in a test beam to have acceptable performance. A
block of this size at 170 cm from the interaction point corresponds to a 3 2 angle. Only
2.1% of kaons overlap with another charged particle within this angular range. This fraction
drops to 0.4% for a 1 6 cone, corresponding to the case in which two particles fall in the
same block, whatever the impact point. Larger block sizes also lead to poor and nonuniform
photon collection. Simulations show that going from a 10 10 cm2 to a 10 20 cm2 block
reduces the average number of photoelectrons by a factor of 3. For these reasons, a 10 10 cm
transverse block dimension has been adopted for the baseline design.
:
:
6.5.2
Photodetectors and Readouts
In order to collect Cherenkov photons eciently inside the solenoid, the photon detector must
work in magnetic elds of up to 1.5 T, have a large UV sensitive area with high quantum
eciency, and have low noise. The baseline design uses 19-stage ne mesh photomultiplier
tubes (FM PMT) from Hamamatsu with a newly developed high quantum eciency bialkali
photocathode. Hamamatsu's FM PMTs with a conventional bialkali photocathode have
been widely used in high-energy experiments [Gil88]. The newly developed FM PMT gives
much higher quantum eciency (see Figure 6-22). Using the new FM PMT, 1.7 times as
many Cherenkov photons were observed from a prototype aerogel counter in our beam test
at the KEK PS facility than were observed with an older version.
Our recent studies using a spectrometer magnet at SLAC [Gea94] con rm that a conventional
16-stage FM PMT has a gain of 4 105 at 1.5 T, if tilted at 45 with respect to the eld
[Kic93, Suz86, Ham94]. At 5 the relative gain drops by factor of 200 for 1.35 T, as
shown in Figure 6-23. Without the magnetic eld, the FM PMT has a gain on the order
of 107, an intrinsic excess noise factor (ENF) of 2 due to the ineciency of the mesh
dynode structure, and negligible electronic noise. When operated in a magnetic eld, the
FM PMT's gain drops substantially, its e ective ENF increases due to the in uence of the
eld on electron trajectories in the tubes, and the overall electronic noise increases because
an ampli er is needed to increase the signal. As a consequence, even though the FM PMT
gives the highest gain at a 45 angle to the magnetic eld, an optimal operation angle must be
determined for single photon response. A series of tests have been performed which indicate
that a 200 FM PMT tilted 5 with respect to the eld gives a reasonable photoelectron yield
along with an optimal pedestal-to-signal separation for the aerogel counter readout.
Technical Design Report for the BABAR Detector
6.5 The ATC Detector
Figure 6-21.
235
Three-dimensional view of one sector of the forward ATC.
Technical Design Report for the BABAR Detector
236
Particle Identi cation
40.00
QE (new, glass)
QE (conventional, glass)
1
Relative Gain
Quantum Efficiency (%)
50.00
30.00
20.00
0.1
0.01
R5065 n=1.0085
10.00
KEK single pe
R5065 LED
0.001
0.00
100 200 300 400 500 600 700 800
0
Wavelength (nm)
6-22. Comparison of the
quantum eciency of the newly developed ne mesh PMT (solid line) vs. a
conventional PMT (dashed line). Both
PMTs have glass windows.
Figure
5
10
15
B (KGauss)
Measured relative gain
of a ne mesh PMT operated in magnetic elds up to 1.35 T.
Figure 6-23.
More compact readout devices, such as hybrid photodiodes (HPDs) and microchannel plates
(MCP PMTs) are also being considered. The characteristics of HPDs and MCP PMTs,
and their performances in magnetic elds, have been extensively studied by the aerogel PID
group and by others at KEK, DESY [Kic93, Suz86], and CERN [Des94a]. An outline of their
characteristics is provided here. More detailed descriptions can be found in Section 6.5.5.
The HPD has the advantage of being a ENF1 photon detector, while its overall e ective
electronic noise is on the order of a few photoelectrons. Its low gain, 3 103, requires the
use of a charge-sensitive ampli er. However, the overall e ective electronic noise has been
reduced to less than or equal to 1 photoelectron in a recent test. The improvement in the
signal-to-noise ratio of the HPD comes from using a photodiode with a thicker depletion
layer, applying more voltage in the vacuum gap, and reducing the thickness of the inactive
layer in front of the photodiode [Des94b]. Improving the HPD's performance has been one
of our major R&D e orts.
Newly developed GaAsP micro-channel plate photomultiplier tubes (MCP PMT) from Intevac have quantum eciencies of 10% at 350 nm and 40% at 500{600 nm [Edg92],
providing a better average collection eciency for Cherenkov light than FM PMTs. The
Technical Design Report for the BABAR Detector
6.5 The ATC Detector
237
Electronics Hut
beam expander
Laser
beam splitter
Detector
optical fibers
20 meters
aerogel
blocks
FM
PMT
photodiode
ADC
trigger
inputs
gate
FM
PMT
Figure 6-24.
Conceptual design of the ATC laser asher system.
MCP PMT has a gain on the order of 106, an ENF measured to be 1.4, and is immune
to magnetic elds [Kic93]. The Novosibirsk Institute of Semiconductor has successfully
produced an InGaAsP MCP PMT with a quantum eciency of 50% at 600 nm [Alp94].
6.5.3 Monitoring and Calibration System
A laser-driven optical ber will be attached to each aerogel block to monitor the stability of
the aerogel and photodetector on a short term basis. The system will use a pulsed nitrogen
laser, which has a 337 nm wavelength. The laser beam is optically split into two beams: one
is monitored by a photodiode to determine the stability of the laser pulse, and the second
is expanded and coupled into an optical ber which is attached to the back of each aerogel
block (Figure 6-24). The long-term calibration of the system will be performed using Bhabha
events, dimuons, and pions from KS0 decays.
6.5.4 Integration Issues
ATC Assembly, Installation and Test
After positioning and gluing the six modules of each endcap segment, the cells are lled with
the aerogel blocks, and photodetectors are mounted inside the di using boxes. Bases and
pre-ampli ers are also mounted during this operation. The detector is closed by attaching
the back cover of the aerogel structure. Each half endcap is assembled and transported
independently. Special tools have to be developed for positioning and gluing the modules.
Technical Design Report for the BABAR Detector
238
Particle Identi cation
Each ATC endcap is mounted on the front face of the endcap calorimeter, and aligned with
respect to it, before installing the two detectors.
Cabling and Access
The front-end electronics for the ATC are situated on the detector, and for any access to
this part, the ATC and the electromagnetic endcap calorimeter have to be removed. After
dismounting the back cover of the aerogel structure, access to all electronics and aerogel
blocks is possible.
Cables from the detector to the DAQ, which is located outside the BABAR detector, run
between the beam pipe and the inner radius of the ATC.
6.5.5 Research and Development Program
Prototype Beam Test Results
Proof-of-principle requirements have been met through several beam tests of prototype
aerogel counters read out with ne mesh photomultipliers in a strong magnetic eld at
CERN and KEK. More than 10 photoelectrons have been observed with both n = 1:0085
and n = 1:063 aerogel counters. Threshold curves have been mapped out, and 4 separation
between below threshold particles and above threshold particles has been obtained. In
magnetic elds up to 1.0 T, 13 photoelectrons were measured when a 5 GeV=c charged pion
traversed a masked 3 ne mesh PMT with a UV window and a sensitive diameter of 44 mm.
Both n = 1:0085 and n = 1:063 aerogel counters with sizes of 10 10 10 cm and 10 10
4 cm, respectively, were wrapped with eight layers of 1.5 mil PTFE covered with a thin layer
of aluminized mylar and placed into the beam. Minimum ionizing particles were selected by
four scintillation trigger counters and identi ed by two CO2 gas Cherenkov counters operated
at about 3 atm. A lead brick with a thickness of 2 cm, corresponding to 3.6X0, was placed
behind the aerogel counter, between the fourth trigger counter and a veto counter, to veto
electrons as well as to simulate backsplash from the calorimeter.
Light yields of 9.2 photoelectrons were obtained by reading out the n = 1:0085 counter with
a 2 RCA Quantacon PMT (C31000M); 10.2 photoelectrons were obtained from the n =
1:063 aerogel counter by a masked RCA Quantacon PMT with e ective sensitive diameter
of 27 mm. The momentum-dependence of the light yield demonstrates that both aerogel
counters behaved as threshold Cherenkov devices, as shown in Figure 6-25. The measured
refractive indices 1:0087 0:004 and 1:061 0:005 were in good agreement with the refractive
indices provided by the aerogel manufacturer, 1.0085 and 1.063, respectively.
00
00
Technical Design Report for the BABAR Detector
6.5 The ATC Detector
239
20
Number of Photoelectrons
120
Relative Photon Yield (%)
100
80
60
40
π
p
n=1.0087
20
15
10
5
1.0085
1.063
0
0
0
1
2
3
4
5
6
0
2
Measured threshold
curves of the aerogel Cherenkov counters.
6
8
10
12
B (KGauss)
Scaled Momentum [m(π)*βγ)] (GeV/c)
Figure 6-25.
4
Photoelectron yield of
the aerogel counters read out by ne mesh
PMT's in the magnetic eld.
Figure
6-26.
The aerogel detectors were also read out by ne mesh PMTs operated in magnetic elds up
to 1.0 T. The setup is very similar to the beam tests described above, except that the aerogel
counters were sandwiched between two 10 10 10 mm trigger counters placed inside the
magnet to have a better determination of the path of charged particles in the eld. The
1.06 aerogel counter used in this beam test is the same as the one used in the previous test,
while the dimensions of 1.0085 aerogel counter were changed to 12 cm long and 10 8:5 cm
in transverse dimensions. The photoelectron yields of n = 1:0085 and n = 1:063 aerogel
counters at di erent magnetic elds are summarized in Figure 6-26.
Another beam test of the prototype 1.0085 aerogel counter was carried out at the KEK
PS test beam facility. Sixteen photoelectrons were measured for 3.5 GeV=c charged pions
using a newly developed ne mesh PMT with a glass window and an e ective sensitive
diameter of 44 mm in a 1.3 T eld. This corresponds to a light yield of 18 photoelectrons for
= 1 particles. A clear separation of the above and below threshold particles was obtained,
Figure 6-8. The uniformity of the light collection eciency was also studied in the n = 1:0085
aerogel counter by running the beam at various transverse positions across the counter. The
average light yield is plotted as a function of the distance from the beam to the PMT in
Figure 6-27. The measured response is determined to change by less than 12% over a 4 cm
distance.
Technical Design Report for the BABAR Detector
240
Particle Identi cation
Number of Photoelectrons
20
15
10
5
0
0
1
2
3
4
5
6
7
8
Distance to PMT (cm)
Figure 6-27.
Uniformity of the light collection eciency.
The HPD option
The proof of principle that an aerogel-counter-based PID system can work eciently in the
BABAR environment, namely that more than ten photoelectrons are read out in a magnetic
eld of 1.5 T has been obtained using Fine Mesh PMTs. There are, however, good reasons
to continue the investigation of other photon detector candidates. Among them, the most
promising seems to be the Hybrid Photodiode (HPD) because of its substantial immunity
to magnetic eld e ects, its wide angular acceptance and its extreme compactness (25 mm
total length).
The HPD tube consists of a normal photocathode coupled to an anode made out of a
silicon photodiode separated by a thin vacuum gap. An incoming photon is converted into a
free electron at the photocathode, accelerated toward the silicon diode, and stopped in the
depletion volume. A gain of about 3000 is achieved for a 15 kV electric eld. The charge is
collected on the anode and read out on a charge preampli er.
Results of measurements in a magnetic eld. Several 1 diameter HPD tubes have
00
been tested in a magnetic eld. The tubes were operated at 8 kV. A simple setup allowed
variation of the angle between the HPD axis and the eld direction with a precision of about
1{2 degrees. A green LED attached to the HPD was used as the source of light.
Technical Design Report for the BABAR Detector
241
Relative Gain
Relative Gain
6.5 The ATC Detector
1
0.8
0.8
0.6
0.6
(a)
0.4
0.2
0
(b)
1
0.4
0.2
0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
B Field (Tesla)
0
0
10
20
30
40
50
60
70
80
Angle (degrees)
(a) HPD response versus magnetic eld for a tube with a 5.6 mm gap; (b)
HPD response versus angle with the eld for two tubes with di erent gaps.
Figure 6-28.
A scan in angle was performed, at values between 0 and 71 and at di erent values of the
eld between 0 and 2 T [Cal95]. The results of these measurements are given in Figure 6-28.
From these measurements it is clear that HPDs show no loss of signal in magnetic eld at 0
angle, and a small decrease at higher angles (30% at 45{50, depending on the individual
device).
To nd the optimal parameters for an HPD tube to be used as an aerogel photon detector
we intend to perform a series of laboratory tests in close contact with the manufacturer. A
potential of about 15 kV will be used to operate the tubes in order to increase the gain and
thereby reduce the importance of the electronic noise. However, it has yet to be demonstrated
that 15 kV is possible without extending the gap between photocathode and anode beyond 5{
8 mm and without any induced noise on the diode and the preampli ers. Other parameters,
such as diode thickness (relevant for noise) and window thickness, must also be optimized.
We expect to receive the nal tube design by the spring of 1995; in the meantime, several
prototypes that are already available will be tested in the laboratory, using LEDs and aerogel
cosmic rays tests, to evaluate progress.
Tests at CERN
In 1995 we plan a test of all elements needed for the nal choice of the detector con guration
using a beam at CERN. In particular, aerogel blocks from di erent suppliers and a variety
of photon detectors and preampli ers will be studied. Based on these results, we plan to
Technical Design Report for the BABAR Detector
242
Particle Identi cation
build a full (30) section of the ATC module at the beginning of 1996 and test it in a beam
of pions, kaons, and protons to study all aspects of the detector performance in a realistic
environment.
Technical Design Report for the BABAR Detector
REFERENCES
243
References
[Alp94]
[Ast94]
[Ben91]
[Cal95]
V. Alperovich et al.,
A340, 429, (1994).
D. Aston et al., \Test of a Conceptual Prototype of a Total Internal Re ection
Cherenkov Imaging Detector (DIRC) with Cosmic Muons," SLAC{PUB{6731,
(1994).
B. Bencheikh, R. DeSalvo, W. Hao, C. Xu, and K. You, \A Simple Light Detector Gain Measurement Technique," CERN{LAA{HC/91{007, (1991). CERN{
PPE/91{143, (1991).
M. Calvi et al., \Hybrid Photodiode Performance in High Magnetic Field,"
A AR
# 208 (1995)
R. DeSalvo et al., CERN{LAA/HC/94{04 (1994); CERN{PPE/93{101 (1993),
DEP Scienti c Technical Information (1993); DIRC R&D Plans, Internal Note,
Paris (1994).
R. DeSalvo, private communication.
J.P. Edgecumbe et al., Proceedings of SPIE (1992).
Magnet was provided and operated by R. Gearhart, SLAC.
P. Bailly and J.-F. Genat, \A 100 Picosecond Resolution, 6 Microsecond Full Scale
Multihit Time Encoder in CMOS Technology," Proc. of the Third Int. Conf. on
Electronics for Future Colliders, Chestnut Ridge, 57 (1993).
See also: J.-F. Genat, \High Resolution Digital Time to Digital Converters,"
Proc. of the First Int. Conf. on Electronics for Future Colliders, Chestnut Ridge,
(1991).
T.T. Giles et al. (CLEO-II),
A252, 41 (1988); S. Ahmad et
al. (STAR),
A330, 416 (1993).
J.E. Grove et al.,
A314, 495 (1992).
Hamamatsu, private communication.
L. Hrubesh, LLNL, private communication.
L. Hubbeling, \Large Photomultiplier Systems{A New Approach," CERNPreprint ECP 92{10, (1992).
H. Kichimi et al.,
A325, 451 (1993).
Nucl. Instr. Methods
B B
[Des94a]
[Des94b]
[Edg92]
[Gea94]
[Gen92]
[Gil88]
Note
Nucl. Instr. Methods
Nucl. Instr. Methods
[Gro92]
[Ham94]
[Hru94]
[Hub92]
[Kic93]
Nucl. Instr. Methods
Nucl. Instr. Methods
Technical Design Report for the
A AR
B B
Detector
244
REFERENCES
[Men94] D. Mendez, Lockheed, private communication.
[Oya94] J. Oyang, A AR
# 137 (1994).
[Pur93] Purschke et al., \A New High Voltage Supply for Large Photomultiplier Systems,"
International Conf. on Electronics for Future Colliders, CERN Preprint, (1993)
[Rat92] B. Ratcli , A AR
# 92; P. Coyle et al., \The DIRC Counter: A New Type
of Particle Device Identi cation for the B Factory,"
A343,
292 (1994).
[Shi94] X. Shi, A AR
# 142 (1994). \Status Report on the Design of a Detector
for the Study of CP Violation at PEP-II at SLAC," SLAC{419 (1993).
[Suz86] S. Suzuki et al., IEEE Trans. Nucl. Sci. NS-33, 377 (1986); H. Saito et al.,
A270 (1988) 319; J. Janoth et al., DESY 93{119 (1993).
[Tso94] Dr. Tsou, JPL, private communication.
B B
Note
B B
Note
Nucl. Instr. Methods
B B
Note
Nucl.
Instr. Methods
Technical Design Report for the
A AR
B B
Detector
7
Electromagnetic Calorimeter
7.1 Physics Requirements and Performance Goals
T
he introduction of high resolution, highly ecient CsI calorimeters has already led to
major advances in the areas of conventional B meson, charmed meson, charmed baryon,
and physics at the (4S ) [Sch94]. CsI calorimetry is also ideally suited to the primary
physics goals of the BABAR detector, which require the tagging and reconstruction of B events
containing CP eigenstate decays. The need to reconstruct CP eigenstates containing one or
more 0 decays makes excellent electromagnetic calorimetry essential to the success of the
experiment. High eciency for low-energy photons is required to o set the small branching
fractions typical of all CP eigenstates, and to make it possible to reconstruct nal states
containing several 0 s. Good energy and angular resolution improve the 0 and B mass
resolutions, in turn improving the signal-to-background ratios of these rare decays. By
facilitating lepton identi cation (e= and e= separation), the calorimeter also provides one
of the tags needed for every CP analysis.
7.1.1 Physics Processes In uencing Performance Goals
Energy and Angular Resolution
The understanding of CP violation in the B meson system requires the reconstruction of
nal states such as B 0 ! J= KS0 , B 0 ! J= K 0 , B 0 ! , B 0 ! D+D,, B 0 !
D+ D, , B 0 ! 0 , and B 0 ! KS0 KS0 . These channels either contain a direct 0 or can be
reconstructed through a decay chain containing one or more daughter 0s. For example, the
decay K 0 ! KS0 0 occurs about one-sixth of the time (which is one-half of its decays to CP
eigenstates), and many D and D decays with large branching fractions contain 0s.
Generic B decays contain an average of 5.5 charged tracks and 5.5 photons, with about 50% of
the photon energies below 200 MeV (Figure 7-1(a)). Because the B meson is moving relatively
slowly in the lab frame, the photon energy resolution, rather than the angular resolution,
determines the B 0 mass resolution [Ale94b]. This is true even for the kinematically most
246
Electromagnetic Calorimeter
Figure 7-1. Photon energy spectrum
in (a) generic B decays and (b) B 0 !
0 0 events.
0
(dashed line) and B
meson reconstruction eciency (solid
line) vs. photon energy threshold for
generic B decays. (Taken from reference [HEL92]. Eciencies exclude geometric acceptance.)
Figure 7-2.
extreme channel of interest, B 0 ! 00 (used to untangle penguin contributions to B 0 !
+ , ), in which both 0 s have momenta above 1 GeV=c, and photon energies extend up to
4.5 GeV (Figure 7-1(b)).
Photon Eciency and Solid Angle Coverage
Sensitivity to low-energy photons is critical for ecient 0 detection. The 0 and B meson
reconstruction eciency for generic B decays, as a function of the minimum detectable
photon energy, is shown in Figure 7-2. Reconstruction eciency is also sensitive to solid angle
coverage. The reconstruction eciency, as a function of both the minimum and maximum
measured photon energy and the minimum and maximum detectable polar angle, is shown
in Figure 7-3 for (1) B 0 ! 00 , producing four relatively high-energy photons, (2) B 0 !
J= KS0 , KS0 ! 0 0 , producing four relatively soft photons, and (3) B 0 ! , producing
only two photons. Losses in the forward direction (Figure 7-3(a)) are not as severe as for lowmultiplicity all-charged modes, though 0 0 is somewhat worse than +,. In the backward
direction (Figure 7-3(b)) the demands on solid angle coverage are more modest.
Lepton tagging requires the calorimeter to reject pions misidenti ed as electrons to a level
of 10,3.
Technical Design Report for the BABAR Detector
7.1 Physics Requirements and Performance Goals
247
100
Efficiency (%)
80
60
(b)
(a)
π oπ o
ψK s
ρ+π–
40
20
0
0
0.2
0.4
0.6
cosθmin
100
0.8
1.0 –1
1
0
cosθmax
(c)
(d)
Efficiency (%)
80
60
40
20
0
0
6-93
1
Emin (GeV)
2
0
1
2
3
Emax (GeV)
4
5
7468A2
Dependence of B reconstruction eciency on (a) minimum detectable ;
(b) maximum detectable ; (c) minimum detectable energy; and (d) maximum detectable
energy for decay photons from B 0 ! + , (solid), B 0 ! J= KS0 , J= ! `+ `, , KS0 ! 0 0
(dash-dot), and B 0 ! , ! 0 (dash).
Figure 7-3.
The requirements placed on the calorimeter by non-CP physics are generally similar to those
from CP physics. One exception is the need for ecient tagging of outgoing leptons in twophoton events. A backward endcap, which would contribute little to overall B reconstruction
eciency, would be important for this purpose. Another exception occurs in low-multiplicity
nal states that do not arise from B decays (e.g., Bhabha scattering, and charm and tau
decays), for which the energy range can extend up to the higher beam energy. For these
processes, leakage, and hence detector depth, is the primary consideration.
Technical Design Report for the BABAR Detector
248
Electromagnetic Calorimeter
7.1.2 Summary of Performance Targets
These physics requirements lead to a calorimeter design based on quasi projective CsI(Tl)
crystals over the entire available solid angle.
The target energy resolution for photons at a polar angle of 90 is:
1% 1:2%:
E
q
(7.1)
=
4
E
E(GeV)
The constant term arises from front and rear leakage ( 1:0%), inter-calibration errors
(0.25%) and light collection nonuniformity ( 0:5%). This expression does not include
electronic noise, which is estimated to be 150 keV per crystal in the proposed readout
system, because this contribution is negligible for typical clusters of 16{25 crystals, even
at low energies. The resolution degrades at small polar angles, as discussed in subsequent
sections, as a consequence of the staggered arrangement of crystals.
For comparison, this performance represents a considerable improvement over that of the
CLEO-II calorimeter. A number of factors contribute to the improvement: the use of
projective geometry in the forward endcap, a reduction of back leakage and losses in inactive
material in front of and between the crystals, and reduction of electronic noise (allowing more
crystals to be summed). In addition, the usable solid-angle coverage has been increased.
The angular resolution is determined by the transverse crystal size and average distance to
the interaction point. The target angular resolution is:
3 mr 2 mr:
= q
(7.2)
2
E(GeV)
As in CLEO-II, the minimum detectable energy for photons is about 10{20 MeV. It is expected to be largely determined by beam- and event-related backgrounds in the calorimeter,
since the intrinsic CsI(Tl) eciency should be close to 100%, even at energies as low as a
few MeV.
The solid angle coverage of the calorimeter extends from a polar angle of 250 mr in the
forward direction (cos = 0:97) to , 653 mr in the backward direction (cos = ,0:80). A
backward endcap has been designed to cover the angular range down to , 560 mr, primarily
for two-photon tagging. It is not included in the baseline detector, but is a candidate for a
future upgrade.
The performance of the calorimeter is summarized in Table 7-1.
Technical Design Report for the BABAR Detector
7.2 Calorimeter Overview
249
BABAR Design
Performance
E
(Stochastic Term) at 90
1%
E
E
(Constant Term) at 90
1.2%
E
at 1 GeV at 90
5 mr
Eciency at 20 MeV at 90
85%
Eciency at 100 MeV at 90
95%
=e Rejection at 500 MeV=c
few10,3
Minimum Detectable Energy 10{20 MeV
Electronic Noise/Crystal
150 keV
Parameter
Table 7-1.
Target performance for the CsI(Tl) calorimeter.
7.2 Calorimeter Overview
7.2.1
Technology Choice
The considerations leading to the choice of CsI(Tl) for the BABAR calorimeter are discussed in References [Ale94a] and [Hea94]. The BABAR physics goals require an electromagnetic calorimeter with excellent photon energy resolution and eciency at low energies
( 100 MeV). The calorimeter sits inside the solenoid and thus requires a readout system
that operates in a magnetic eld. These requirements, and considerations of cost and
engineering impact upon other detector subsystems and the accelerator, lead to the choice
of a CsI(Tl) crystal calorimeter with photodiode readout. The relevant properties of CsI(Tl)
are summarized in Table 7-2.
The radiation environment at PEP-II is more severe than that at existing e+e, colliders
because of much higher beam currents. Given the radiation hardness of CsI crystals, the
calculated background rates during luminosity running, discussed in Chapter 12, imply a
wide safety margin. However, in actual storage ring operations, experience shows that most
of the radiation exposure comes during beam injection and machine studies. Scaling typical
CESR operation to the beam currents expected at PEP-II gives a radiation dose of about
1.5 krad/yr at a radius of 45 cm and 0.5 krad/yr at 100 cm [Blu86]. The PEP-II masking
system has been designed with graded apertures to prevent the deposition of large amounts
of radiation in the interaction region due to accidental beam loss. Nonetheless, it is desirable
Technical Design Report for the BABAR Detector
250
Electromagnetic Calorimeter
Properties
Radiation Length (cm)
Absorption Length (cm)
Light Yield (Photons/ MeV103 )
Light Yield Temperature Coef. (%/C)
Moliere Radius (cm)
Peak Emission (nm)
Lower Wavelength Cuto (nm)
Refractive Index at Emission Maximum
Decay Time (ns)
Density (g/ cm3)
Hygroscopic
Table 7-2.
CsI CsI(Tl) CsI(Na)
1.86 1.86
1.86
34.2 34.2
34.2
2{10 50{60 38{44
0.1
0.1
0.1
3.8
3.8
3.8
320
565
420
260
320
300
1.95 1.79
1.84
10
940
630
4.51 4.51
4.51
slight slight
weak
Properties of pure and doped CsI.
to use detector components that are intrinsically radiation hard and can survive doses of tens
of krad over the lifetime of the experiment.
7.2.2 Description of the Calorimeter
The baseline design of the CsI(Tl) calorimeter consists of a cylindrical barrel section and
a forward conic endcap, as shown in Figure 7-4. Radially, the barrel is located outside
the particle ID system and within the magnet cryostat. It weighs 24.2 metric tonne and
is supported o each end of the coil at two points by eccentric pins. The barrel has an
inner radius of 90 cm and an outer radius of 135.6 cm. It is located asymmetrically about
the interaction point, extending, at its inner radius, 117.8 cm in the backward direction
and 180.1 cm in the forward direction. The barrel covers a solid angle corresponding to
,0:80 cos 0:89 in the laboratory frame (Figure 7-4).
The barrel is built out of 250 m-thick carbon ber composite (CFC) compartments that
house individual crystals (Figure 7-5). At 90 to the beam, a compartment extends 43 cm
radially. Aluminum support elements bonded to the back of each compartment allow them
to be grouped into modules three crystals wide and seven crystals long, and to carry loads
to a cylindrical strongback structure, which is xed to the coil. By supporting the crystal
at its rear, minimal material is placed in front of the crystals. The front material consists of
Technical Design Report for the BABAR Detector
7.2 Calorimeter Overview
Figure 7-4.
endcap.
251
Side view showing dimensions (in mm) of the calorimeter barrel and forward
two 1 mm-thick cylinders of aluminum, separated by foam, which provide a gas seal and rf
shielding; cooling, cables, and services are located at the back of each module.
The forward endcap is a conic section, with front and back surfaces tilted at 22.7 with respect
to the vertical, which conforms to the drift chamber endplates, minimizing the distance to
them. The endcap weighs about 4 metric tonne. It is supported o the coil rather than the
ux return end doors, and is precisely aligned to the barrel to minimize the gap between
the barrel and forward endcap. To further reduce the gap's e ect on photon eciency, it
is nonprojective. The forward endcap starts at an inner radius of 50.0 cm from the beam
line, making an angle of 250 mr with respect to the beam. It covers the solid angle from
cos = 0:97 to cos = 0:89 in the lab.
The forward endcap is composed of two identical monolithic sections consisting of loadbearing outer containers lled with a honeycomb of 250 m-thick CFC compartments, each
containing one crystal package. The forward endcap is segmented vertically into two pieces,
each of which can be retracted separately using a set of removable rails installed after the
doors are opened. This allows fast access to the barrel end region.
Material in front of the calorimeter is kept to a minimum in order to achieve the required
performance. In the barrel, the major source of such material is the DIRC (0.18X0).
Including the beam pipe, the vertex detector, and the drift chamber, the total amount of
material represents 0.23X0 at normal incidence. The endplate of the drift chamber and
the forward PID system contribute most of the material in front of the forward endcap. The
endplate constitutes 0.065X0. The aerogel represents an average of 0.095X0. However,
Technical Design Report for the BABAR Detector
252
Electromagnetic Calorimeter
Figure 7-5.
Side view of typical barrel module and strongback.
the ATC material is lumped: the aerogel itself represents 0.035X0, with the balance
concentrated in the phototubes.
In both the barrel and endcap, each crystal is wrapped with a di use, re ecting material
on its sides, with a re ector on its front face. Diode readouts and the preampli er packages
are located at the rear of each crystal box. Each crystal is held loosely in its box by a rear
sti ening ring that also acts to locate and support the diode/preampli er package. Cables
for diode bias, preampli er power and cooling, and calibration signals enter the back of
the compartment. The crystal and electronics are shielded by a foil liner and a metal cap,
respectively.
There are 5880 crystals in the barrel, arranged in 49 rows of distinct sizes, each having 120
identical crystals in azimuth. The forward endcap has a total of 900 crystals, made up from
nine distinct radial rows (3 120; 3 100; 3 80) arranged to give approximately the same
crystal dimensions everywhere.
The crystals are tapered along their length with trapezoidal cross sections. The average area
of the front faces of the crystals is 4:8 4:7 cm, while the back face area is 6:1 6:0 cm.
They vary in length in 0.5X0 steps from 17.5X0 in the forward part of the barrel to 16.0X0
in the backward part, with 17.5X0 in the forward endcap. The barrel and endcap have total
crystal volumes of 5.2 m3 and 0.7 m3, respectively. Table 7-3 summarizes the crystal counts.
To minimize the loss of tracks that traverse inactive material between crystals, the crystals
are arranged to be slightly nonprojective in with respect to the interaction point.
Technical Design Report for the BABAR Detector
7.2 Calorimeter Overview
253
Forward
Barrel
Endcap
Number of Crystals (17.5X0)
840
900
Number of Crystals (17.0X0)
840
0
Number of Crystals (16.5X0)
840
0
Number of Crystals (16.0X0)
3360
0
Total Number of Crystals
5880
900
3
Total Volume (m )
5.2
0.7
Number of Crystals/Modules () 120/40 120, 100, 80
Number of Crystals/Modules () 49/7
9
Module Organization
3 7 Two C-sections
Table 7-3.
7.2.3
Summary of crystal counts and sizes in the barrel and endcap.
Readout Chain and Trigger
The electronic readout chain is shown schematically in Figure 10-13 and described in detail
in Chapter 10. Two independent photodiodes view the scintillation light from each crystal.
Each diode has an independent dual-range preampli er with calibration input. This twofold redundancy is based on reliability issues presented in Section 7.4.4. Shaped, ampli ed
signals are carried di erentially to the end of the detector where they are input to ADCs
which sample the data at a rate between 3 MHz and 5 MHz. The signals are digitized to an
18-bit e ective dynamic range.
Digitized signals are passed out of the detector on ber optic cables to the DAQ receiver
modules, where calibration and waveform processing is performed. The data are then stored
in video RAM. Trigger sums of 25 crystals are formed digitally on the DAQ cards. These
data are passed along to the trigger for the full trigger decision.
The barrel digitization electronics is housed in 20 bays at the ends of the barrel. Forward
endcap electronics is housed behind the endcap in six bays. Fiber optic cables are routed
through channels between the barrel and endcap ux returns. Table 7-4 summarizes the
channel organization and required space for the on-detector readout electronics.
Technical Design Report for the BABAR Detector
254
Electromagnetic Calorimeter
Barrel Forward
Endcap
Crystals
5880
900
Photodiodes
11760
1800
Preampli er Boards
5880
900
Cables to ADC
11760
1800
Channels/ADC Board
25
25
ADC Boards/Bay
12
6
ADC Bays
20
6
2
Number of Fibers to DAQ (cm ) 240 (4) 36 (1)
Table 7-4.
7.2.4
Summary of front-end electronics components.
Review of Options
The baseline calorimeter has been described above. A number of variations have been
evaluated, with some remaining under consideration. They are described brie y below.
Monolithic versus Modular Support System
Monolithic and modular support structures have been considered. Monolithic support
structures, such as the CFC system used in the L3 BGO system, are fully integrated systems
that combine strength and low mass. The major drawbacks to such assemblies are the
increased complexity of fabrication and the high cost, arising in part from the need to
incorporate proper compensation for the deformation of the structure when it is loaded.
Modular systems require more material to achieve the required strength and resistance to
deformation. They are less expensive because each unit is small and hence easier to fabricate,
handle, and assemble. Fabrication and assembly can occur at many locations simultaneously.
The BABAR calorimeter structure is a hybrid, with a modular system in the barrel and a
monolithic system in the endcap. The relatively small size of the endcap should alleviate the
problems of fabrication anticipated for a larger monolithic system, while reducing installation
and access problems.
Technical Design Report for the BABAR Detector
7.2 Calorimeter Overview
255
Longitudinal Segmentation of Crystals
Segmentation of crystals along their length (at about 3X0 from the front) to provide information on longitudinal shower development has been considered. Some improvement in e=
separation is indicated by Monte Carlo studies, but this is judged not to have a signi cant
impact on the physics performance. Since the cost of implementation is prohibitive, this
option is no longer being considered.
Crystal Cross Section: Trapezoidal Versus Hexagonal
While trapezoidal cross sections are conventional for crystal calorimeters, the use of hexagonal cross sections may o er certain advantages. In the endcap, an arrangement of hexagonal
crystals on a spherical bounding surface can be designed that allows a single crystal size to be
used, eliminating the problems associated with progressively smaller crystals of trapezoidal
cross section and with abrupt transitions across superlayer boundaries.
Laboratory measurements indicate that for hexagonal crystals, about 25% more light is
collected than for trapezoidal crystals. The response along the crystal is also more uniform [Jes94]. This might allow the acceptance criteria for crystal light yield and uniformity
to be relaxed, leading to decreased costs. The crystal cost depends on the growth technique
adopted; this is lower in some cases for hexagonal crystal shapes. However, the decreased
growing cost is largely balanced by the increased cost of cutting and polishing six side surfaces
instead of four.
Engineering design considerations make the cost and complexity of the support structures
for hexagonal crystals greater than those for trapezoidal crystals. Complications also arise
with hexagonal crystals at the interface of the barrel with the forward endcap, where it is
vital to minimize gaps and inactive material.
Overall, we have not been convinced of any cost advantage in using hexagonal crystals.
Therefore, trapezoidal crystals have been adopted in the baseline design for both barrel and
endcap. The study of hexagonal crystals for use in the endcap will continue, subject to
physics, engineering, and cost considerations, since the endcap is not on the same critical
path as the barrel.
Direct Photodiode Versus Wavelength Shifter/Photodiode Readout
Two options for light collection and readout are presently under study. They are large area
photodiodes directly attached to the crystal back face, and smaller-area photodiodes axed
to the edges of a wavelength-shifting plastic plate covering the crystal back face. Both
Technical Design Report for the BABAR Detector
256
Electromagnetic Calorimeter
techniques have produced results consistent with the desired noise performance. The cost,
reliability, and reproducibility of each method are currently being evaluated (Section 7.4.1).
7.3 Projected Calorimeter Performance
7.3.1 Contributions to Photon Resolution and Eciency
A number of e ects beyond shower statistics contribute to the calorimeter resolution, with
di erent e ects dominating in di erent energy regions. Some of these contributions are
linked, in that reducing the e ect of one may facilitate improving another. The e ects
include:
1. Fluctuations in energy loss caused by leakage out of the front and the rear, by losses
out the side due to the staggered geometry, and by losses in the inter-crystal material.
2. Fluctuations in transverse energy spread, with cluster sizes optimized to include e ects
of noise and background.
3. Scintillation light collection eciency.
4. Light collection nonuniformity, coupled to uctuations in shower position.
5. Incoherent electronic noise in the readout device.
6. Coherent noise and pickup.
7. Digitizer resolution.
8. Calibration of the energy scale, including crystal inter-calibration and the e ects of
time and temperature.
9. Material en route to the calorimeter. This a ects eciency as well as resolution.
10. Beam-related backgrounds.
In the high-energy regime, rear leakage (1) dominates the energy resolution. Fluctuations
in energy loss (1), along with contributions from nonuniform light collection and calibration
uncertainties (4) and (8), are expected to yield the constant term shown in Equation (7.1).
Transverse shower spread and choice of reconstruction algorithms (2) also contribute at high
energy, but become especially important to the energy resolution in the intermediate-energy
range.
Technical Design Report for the BABAR Detector
7.3 Projected Calorimeter Performance
257
The low-energy regime is characterized by the importance of noise (5) and (6), and of beamrelated backgrounds (10). Such backgrounds, especially if well above nominal levels, can
also indirectly a ect resolution through the need to reduce the number of crystals summed
into a cluster, or the need to reduce the shaping time for the readout from the optimal
value for maximizing the signal-to-electronic-noise ratio. Electronic noise is determined by
the properties of the photodiode and preampli er/shaper. Its value in units of energy is
inversely proportional to the amount of light collected per unit of energy deposited (3).
Photon eciency is primarily a ected by the amount and distribution of material in front of
the calorimeter (9), i.e., in other detector systems, and also by losses (1) due to geometric
edge e ects.
Photon angular resolution is given by position resolution divided by the distance to the
interaction point. Even for the best possible algorithm, uctuations in shower shape (transverse energy spread) will limit the position resolution. The transverse crystal size at the rear
should be less than two Moliere radii, but returns diminish when that size is comparable to
a Moliere radius.
Contribution to Energy Resolution from Crystal Nonuniformity
The e ect of nonuniform light output along a CsI crystal has been studied using a standalone GEANT simulation. An untapered, 35 cm-long crystal with large lateral dimensions
(40 40 cm) is divided along its length into 100 slices. A nonuniform response is imposed
as a weight on the energy deposited in the slices for each event.
A study of linearly decreasing and increasing eciencies for light collection along the crystal
for photon energies between 200 MeV and 9 GeV has been done. The contribution to the
energy resolution E =E due to crystal nonuniformity is shown in Figure 7-6(a) for three
di erent linearly decreasing weight functions (10%, 5%, and 2.5% over the full length of
35 cm). Nonuniformities with decreases up to 5% over the full crystal length lead to an
energy resolution contribution which meets the requirement stated in Section 7.1.2.
Linearly increasing weight functions lead to changes in the energy resolution that cannot be
presented in the same way; for photon energies above 1 GeV the energy resolution improves.
Figure 7-6(b) shows the resolution E =E as a function of the increase over 35 cm. The result
for higher energies can be understood as a compensation for leakage of the electromagnetic
shower at the rear end of the crystal. A similar behavior has been observed in a Monte Carlo
study for the CLEO-II calorimeter [Blu86].
Technical Design Report for the BABAR Detector
Electromagnetic Calorimeter
(FWHM/2.36)/Energy
Nonuniformity Contr. to σE/E
258
0.014
0.012
0.01
0.008
0.006
0.004
0.012
0.011
0.01
0.009
0.008
0.007
0.006
0.005
0.002
0.004
0
0.003
1
10
0
Incident γ-Energy (GeV)
5
10
15
20
25
Linear increase in %/35 cm
Energy resolution for nonuniform light output along a CsI crystal of 35 cm
length. (a) Contribution to E =E from light output decreasing towards the crystal end
(2.5%, 5%, 10% over the crystal length). (b) E =E for light output increasing towards the
crystal end. Note that the crystal dimensions used for this study are much larger than
baseline lengths and transverse leakage is ignored.
Figure 7-6.
Contribution to Energy Resolution from Beam Backgrounds
Simulations indicate that beam-related background is almost entirely caused by beam-gas
interactions or by o -axis beam particles striking machine elements near the interaction
region. Synchrotron radiation photons are a negligible source of background [Zis91]. The
energy distribution of the shower remnant photons has a median of 500 keV. About
311 depositions per microsecond with a total energy of 0.33 GeV are expected in the barrel,
while in the forward endcap, about 20 depositions per microsecond with a total energy of
0.07 GeV are expected. There are on average 0.7 (0.4) depositions per microsecond above
20 MeV in the barrel (forward endcap) that can produce fake photons in reconstructed
events. The e ect of beam backgrounds on photon energy resolution has been studied.
Lost particle backgrounds corresponding to 1 times nominal and 10 times nominal levels
in 1 s windows were generated together with 100 MeV photons in the barrel and forward
endcap. Comparison with the zero background case shows contributions, in quadrature, to
the resolution = FWHM=2:36 of 0.55 MeV from 1 times nominal background and 1.05 MeV
from 10 times nominal background (i.e., 0.55% and 1.05%, respectively, at 100 MeV).
Technical Design Report for the BABAR Detector
7.3 Projected Calorimeter Performance
7.3.2
259
Modeling
The expected calorimeter performance has been studied using a GEANT Monte Carlo
simulation of the full detector. The GEANT model of the calorimeter contains a description
of the barrel with crystal lengths varying in 0.5X0 steps from 17.5X0 in the forward part to
16X0 in the backward part, with 17.5X0 crystals in the forward endcap. The model includes
500 m of carbon ber composite as structural material between the crystals and a 5 mm
gap at the barrel/forward endcap interface.
The full detector model includes a DIRC particle ID device in the barrel and an aerogel device
in the forward endcap. They are modeled according to the description given in Chapter 2.
The DIRC represents about 0.18X0 at normal incidence. Together with the beam pipe, the
silicon vertex detector, and the drift chamber, this totals 0.23X0 in the barrel region at
normal incidence. In the forward region, the materials incorporated in the detector model,
as represented in Figure 2-16, is di erent from that described in the balance of the Technical
Design Report. Thus, in the forward region, material in front of the endcap calorimeter
totals about 0.35X0.
Photons are reconstructed by forming clusters of contiguous crystals with energies above
0.5 MeV. In studies of the performance for single photons, their energies are determined
by summing up the energy deposited in an area corresponding to 5 5 crystals at 90.
In studies involving 0s, the cluster energy is determined as the sum of the energies of all
contiguous crystals making up the cluster. To deal with overlapping showers, the cluster is
searched for additional local energy maxima, which are then treated as seeds for subclusters.
The subcluster energies are determined in an iterative way by dividing up the energy in
shared crystals according to the seed energy as it develops and adding it to the seed. In
both methods, an incoherent electronic noise contribution of 150 keV per crystal is included.
Since the distribution of deposited energy typically shows a non-Gaussian tail on the low
side, the resolution is de ned as = FWHM=2:36. In order to allow for e ects not included
in the GEANT simulation, such as nonuniform light collection and calibration uncertainties,
0.5% is added in quadrature to E =E .
Eciency for single photons is de ned as the fraction of events with a ratio of measured over
generated photon energy, R = E=E0, above an E0-dependent cut value:
RCUT (E0 ) = 1 , 0:05
E =E (E0 )
;
E =E (E0max )
(7.3)
where E =E (E0 ) is the target energy resolution (Equation 7.1) at energy E0 , and E0max =
5 GeV; therefore, RCUT (5 GeV) = 0:95, and RCUT (100 MeV) = 0:92. (Note that in
calculating R, the measured photon energy distribution has been calibrated so that its peak
value equals E0 .)
Technical Design Report for the BABAR Detector
260
Electromagnetic Calorimeter
7.3.3 Expected Performance for Photons
Initially, the performance of the calorimeter has been studied without any material in front of
it. Figure 7-7 shows energy resolution and eciency, as de ned above, vs. energy, for single
photons with polar angles near 90. Also shown (solid line) is the target energy resolution
(Equation 7.1). The predicted resolution, based on 16.5X0 crystals near 90, is just above the
target resolution. The sawtooth crystal layout at the inner radius of the barrel calorimeter
causes some performance degradation towards small polar angles. Figure 7-8 shows energy
resolution and eciency vs. polar angle for photons with energies of 100 MeV and 1 GeV.
One can see a 40% (20%) worsening of resolution for 100 MeV (1 GeV) photons at cos
values near the end of the barrel. This e ect is due to increased leakage, mostly out of the
uncovered crystal side faces. This degradation is probably acceptable over the limited solid
angle involved. (It could be ameliorated by decreasing the crystals' transverse dimensions
(front-side leakage) and increasing their length (rear-side leakage). Studies have shown that
simply lling in the sawtooth pattern degrades eciency and uniformity of light collection
signi cantly [Spi94]).
A photon that converts in the material between the main tracking volume and the calorimeter
can either be lost entirely or so degraded that it will fail to lead to a satisfactory 0 mass.
While energy loss by the conversion e+ and e, is a contributing factor, even more important
is their de ection by the solenoidal eld, so that one or both of the leptons may not reach
the barrel calorimeter. For the endcap, in contrast, the conversion leptons will spiral around
the eld lines and eventually reach the calorimeter close to the intersection point of the
projected photon. Important considerations besides the amount of material traversed include
the proximity of that material to the calorimeter and whether the particle ID detector can
\ ag" the conversion [Eis93]. The e ect of material in front of the calorimeter has been
simulated using the full GEANT model of all detector systems. The results for energy
resolution and eciency are shown in Figure 7-9. Note that there is only a small e ect
on resolution since most photons pass through the material una ected. However, those
that interact experience a signi cant energy loss, resulting in a reduction in eciency of
2-32% at 100 MeV and 6-12% at 1 GeV, depending on cos . More details can be found in
Reference [BaB94]; the e ect on 0 reconstruction is presented below.
The detailed reconstruction algorithms needed to determine the positions of showers are
still being developed. A crude algorithm being used at the moment is based on a corrected
energy-weighted center-of-gravity method. Due to the nite granularity of the calorimeter,
the resulting shower positions are biased towards the crystal centers. A correction, derived
from Monte Carlo, is applied in an attempt to remove this bias. The resulting resolution in
polar angle is shown in Figure 7-10 as a function of energy, together with a parameterization
of the CLEO-II angular resolution. It is slightly better than the CLEO-II resolution, as a
result of our lower electronic noise, smaller transverse crystal sizes, and longer distances to
Technical Design Report for the BABAR Detector
7.3 Projected Calorimeter Performance
2.0
9
(a)
9
1.8
9
1.6
9
9
1.4
9
9
1.2
FWHM/2.36 (%)
FWHM/2.36 (%)
2.2
261
9
90
80
0.02
9
9
9
(b)
9
9
9
(a)
90
2
2
2
2
3
3
3
3
3
2
3
2
100 MeV
2
3
2
3
2
3
2 222
2 22 2 2222
3
3 3333 33 3
33
3
1 GeV
(b)
Efficiency (%)
Efficiency (%)
1.0
100
2.8
2.6
2.4
2.2
2.0
1.8
1.6
1.4
1.2
1.0
100
3
2
3
2
3
2
3
2
3
2
3
3 333
3 333 3
32222
2 22 232
22
2
80
0.1
0.2
1
2
0
0.1 0.2 0.3
Photon Energy (GeV)
Figure 7-7. (a) Energy resolution and
(b) eciency (as de ned in the text)
vs. energy for photons with polar angles
near 90 and 16.5X0 crystals without
any material in front of the calorimeter.
The solid line shows the target resolution
(Equation 7.1).
0.4 0.5 0.6 0.7
0.8 0.9
1
cos(θ)
(a) Energy resolution and
(b) eciency vs. cos cm for 100 MeV and
1 GeV photons without any material in
front of the calorimeter.
Figure 7-8.
many of the crystals. Figure 7-11 shows resolution vs. polar angle for di erent photon
energies. Note that the angular resolution is comparable for showers in the barrel and in
the endcap since, in contrast to the case at CLEO, the endcap crystals are also arranged in
quasi-projective geometry.
7.3.4 Expected Performance for 0s
Most photons detected in the calorimeter arise from 0 s. To study 0 mass resolution and
eciency and the e ect of material in front of the calorimeter, single, monoenergetic 0s have
been simulated using the GEANT model of the full detector. Figure 7-12 shows 0 mass
resolution and eciency vs. 0 momentum for 0s with cos cm = 0. Mass resolution is again
de ned as = FWHM=2:36; eciency is obtained using a 3 cut around the 0 mass. Also
indicated is the performance without any material in front for 0 momenta of 100 MeV=c and
1 GeV=c. As is the case for single photons, the inactive material has very little e ect on the
resolution, but signi cantly degrades the eciency.
Technical Design Report for the BABAR Detector
262
Electromagnetic Calorimeter
FWHM/2.36 (%)
3.5
2
(a)
3.0
2.5
2.0
1.5
100 MeV
2
3
2
3
1 GeV
2
3
2
2 3 2
3 3 22
33
3
1.0
100
Efficiency (%)
(b)
90
3
3
80
2
2
3
2
70
33 3
3 33
2
222
22
60
0
0.1 0.2 0.3
0.4 0.5 0.6 0.7
0.8 0.9
1
cos(θ)
(a) Energy resolution and (b) eciency for 100 MeV and 1 GeV photons vs.
polar angle with the full detector model.
Figure 7-9.
In addition, several benchmark processes relevant for measuring CP asymmetries have been
used to study 0 mass resolution and eciency: B 0 ! J= K 0 , K 0 ! KS00 ; B 0 ! ,
! 0 ; and B 0 ! 00. The 0s from those processes cover a wide range of momentum,
with B 0 ! J= K 0 yielding the softest 0s and B 0 ! 00 the hardest. Figure 7-13(a)
shows the invariant mass spectrum in the decay B 0 ! J= K 0 , with no material in
front of the calorimeter. (Note that the KS0 is not allowed to decay to 00 in this study,
and the J= decays into lepton pairs only.) The 0 mass resolution and eciency are
m0 = 4:0 MeV=c2 and 74%, respectively. Also shown (Figure 7-13(b)) is the mass spectrum
when the detector material is included in front. Again, the additional material leads to a
relatively small degradation in resolution to m0 = 4:5 MeV=c2 , but causes a very signi cant
low-mass tail which results in a drop in eciency to 59%. The results for all three channels
are summarized in Table 7-5.
In the all-neutral channel B 0 ! 00, we achieve a B 0 mass resolution of mB = 62 MeV=c2
(Figure 7-15) and an overall reconstruction eciency of 62%, using a 0 -mass cut of 110
m 160 MeV=c2 and a B -meson mass cut of 5000 < m0 0 < 5350 MeV=c2 (Figure 7-14).
Technical Design Report for the BABAR Detector
7.3 Projected Calorimeter Performance
263
12
10
σθ (mrad)
8
6
4
2
0
0.02
0.1
0.2
1
2
Photon Energy (GeV)
Resolution in polar angle vs. energy for photons with polar angels near 90 .
The solid line shows a parameterization of the measured CLEO-II resolution in the barrel.
Figure 7-10.
B
Decay Channel
!
!
!
B 0 J= K 0
B 0
B0 00
m0
(MeV=c2 ) Eciency (%)
4.5
59
5.5
65
8.3
72
mass resolution and eciency in various B decay channels, modeled with
the complete detector material in front of the calorimeter.
Table 7-5.
0
Technical Design Report for the BABAR Detector
264
Electromagnetic Calorimeter
12
10
5
2
2
5
σθ (mrad)
8
5
2
5
2
5
2
100 MeV
5
2
55
5
55 5
5
5
2
2
6
4
8
3
3
8
8
3
8
3
0.2
0.3
8
3
8
3
8
3
0.4
0.5
0.6
2
2
2 8
8 88 8882828828
8
3
333333 3
3 333
1 GeV
2
0
0
0.1
0.7
0.8
0.9
1
cos(θ)
Figure 7-11. Resolution in polar angle vs. polar angle for 100 MeV and 1 GeV photons.
The points connected by the dashed lines show the resolution with the distance e ect
removed.
7.3.5
e=
Separation
In the BABAR calorimeter, electrons are totally absorbed, while muons and some pions, having
momenta above a few hundred MeV=c, penetrate to the Instrumented Flux Return (IFR),
depositing only a fraction of their energy.
Since the drift chamber provides a precise momentum measurement, electrons can be identi ed using energy-momentum matching. In addition, electromagnetic showers are usually
narrower than hadronic showers, so the lateral shower pro le can be used to further distinguish electrons from pions. The results of such a study [Zhu95] using a GEANT simulation
of the detector, with electrons of momenta in the range 500 MeV=c to 5 GeV=c produced at
polar angles of 90, indicate that the probability of misidentifying a pion as an electron is
10,3 for an electron eciency of 95% (Table 7-6).
Technical Design Report for the BABAR Detector
7.3 Projected Calorimeter Performance
16
265
(a)
2
FWHM/2.36 (MeV)
14
12
10
8
6
2
1
4
2
Efficiency (%)
100
2
2
2
1
2
2
(b)
80
5
3
3
0.1
0.2
60
5
3
3 3
1
2
3
3
π0 Momentum (GeV/c)
(a) Mass resolution and (b) eciency (as de ned in the text) for single,
mono-energetic 0 s vs. 0 momentum with the full detector model (0 s are generated with
cos cm = 0). The crosses indicate the performance for 100 MeV=c and 1 GeV=c 0 s with no
material in front of the calorimeter.
Figure 7-12.
Momentum
0.5 GeV=c
1.0 GeV=c
5.0 GeV=c
Table 7-6.
Events Electron + ,
Generated 1000 5000 5000
926
3
7
Surviving
968
1
1
972
0
1
Results of a GEANT study of e= separation.
Technical Design Report for the BABAR Detector
266
Electromagnetic Calorimeter
150
(a)
100
Entries/0.001 GeV
50
0
150
(b)
100
50
0
0
0.05
0.1
0.15
0.2
γγ Invariant Mass (GeV)
S!
K0
!
!
invariant mass distribution in the process B 0 J= K 0 , K 0 KS0 0 ,
+
,
for (a) no material in front of the calorimeter and (b) the full detector model.
Figure 7-13.
7.3.6 Performance and Cost Optimization
Taking into account the correlation between maximal photon energy and polar angle caused
by the boost of the center-of-mass system, crystal lengths can be varied as a function of polar
angle without signi cantly a ecting the calorimeter performance in CP -relevant B decay
modes. The process B 0 ! 0 0 has been used as a benchmark. Crystal lengths are chosen
to keep the corresponding energy resolution (Figure 7-16) close to the target resolution at
all polar angles. This mode has also been used to check the e ect of shorter crystals on the
B mass resolution. For a calorimeter in which all crystals are shortened by 1X0 , one obtains
mB = 73 MeV=c2 and an overall reconstruction eciency of 57%. Such a performance is
signi cantly worse than that of the baseline design.
Using the center-of-gravity method described above, angular resolution has been studied as
a function of transverse crystal size (see Figure 7-17). One nds a rather weak dependence,
especially at high energies where there is signi cant energy sharing between the crystals in
Technical Design Report for the BABAR Detector
7.3 Projected Calorimeter Performance
267
Entries / 0.001 GeV
150
100
50
0
0
0.05
0.1
0.15
0.2
γγ Invariant Mass (GeV)
Figure 7-14.
Entries / 0.01 GeV/c 2
detector model.
invariant mass distribution in the process B 0
! , using the full
0 0
80
60
40
20
0
4
4.25
4.5
4.75
5
5.25
5.5
2
π0 π0 Invariant Mass (GeV/c )
invariant mass distribution where 110 < m < 160 MeV=c2 , in the
process B ! , using the full detector model.
0 0
Figure 7-15.
0
0 0
Technical Design Report for the BABAR Detector
268
Electromagnetic Calorimeter
7
3
2
4
18 rad. len.
17 rad. len.
16 rad. len.
15 rad. len.
2
Target Resolution
FWHM/2.36 (%)
4
2
3
7
2
2
3
37
3
7
1
Typical error bars shown
0
1
2
3
4
5
6
Photon Energy (GeV)
Figure 7-16. Energy resolution vs. energy for a number of crystal lengths and for photons
with polar angle near 90 . The solid line shows the target resolution (Equation 7.1).
Material Thickness
0 m
500 m
800 m
(%) Eciency (%)
1.6
82
2.1
79
2.3
78
E =E
Energy resolution and eciency for 100 MeV photons with cos cm
di erent thicknesses of the inter-crystal material.
Table 7-7.
= 0
for
the cluster. The angular resolution for our choice of transverse crystal size is consistent with
the target (Equation 7.2).
The e ect of the CFC structural material and crystal wrappings on energy resolution and
eciency is given in Table 7-7 for photons with energies of 100 MeV and polar angles around
60. Note that the baseline design has an amount of inter-crystal material equivalent to about
800 m of CFC. As can be seen, the performance degrades slowly for material thicknesses
greater than 500 m of CFC equivalent.
Making the inter-crystal cracks nonprojective does not help with the loss of resolution shown
in Table 7-7, but it does allow recovery of much of the lost eciency. This has been studied in
detail for the 800 m case. In both and , most of the loss is recovered with a nonprojective
angle of 15 mr (about 6 times the angle subtended by the crack). The baseline design includes
Technical Design Report for the BABAR Detector
7.4 Crystal Subassemblies and Readout
269
15
100 MeV
σθ (mrad)
10
2
2
2
2
2
2
2
2
3
5
3
3
1 GeV
0
2
3
4
5
6
Crystal Width (cm)
Figure 7-17.
Resolution in polar angle vs. transverse crystal size for 100 MeV and 1 GeV
photons.
15 mr of nonprojectivity in only, and relies on misalignment to produce a smaller amount
of nonprojectivity in . Even without the latter, the photon eciencies are within 2% of the
zero-material values; this loss is small compared to that resulting from material in front of
the calorimeter.
7.4 Crystal Subassemblies and Readout
A crystal subassembly consists of a CsI(Tl) crystal covered with a suitable re ective material
wrapped in a thin aluminum foil and a readout package located at the rear. The readout
package consists of silicon PIN photodiodes closely coupled to low-noise, charge-sensitive
preampli ers, both enclosed in a metal cap. (The exact crystal dimensions are given in
Section 7.6; for a detailed description of the front-end electronics refer to Chapter 10.) All
components are selected to have a uniform and high level of light collection, together with
minimal noise contribution from the readout electronics, in order to achieve optimal energy
resolution.
Technical Design Report for the BABAR Detector
270
7.4.1
Electromagnetic Calorimeter
Photodiode Readout
Signal-to-Noise Considerations
The requirement for the readout system is that at the lowest photon energies of interest
(20 MeV), the electronic noise must not dominate the resolution. In the absence of electronic noise, the resolution at these energies is expected to be on the order of 500 keV, mostly
due to shower leakage uctuations. The electronic noise contribution to the cluster energy
resolution is required to be smaller than 500 keV. This results in an upper limit of about
150 keV per crystal, assuming that less than a dozen crystals are summed into a cluster at
such low photon energies.
Electronic noise has two components, an incoherent one which derives from thermal and
shot noise in the readout, and a coherent one which comes predominantly from pickup. Experiments using arrays of CsI(Tl) similar to the one proposed here have achieved incoherent
noise performances of 250{500 keV per crystal at room temperature [Ake92, Kub92]. Noise
performance at the 150 keV/crystal level should be achievable with improvements in average
crystal light output, surface preparation, light collection, photodiode packaging, low-noise
charge-sensitive preampli ers, and shaper ampli ers. A target at this level has been set for
the readout system. This meets the resolution requirements and allows the use of radioactive
sources for crystal monitoring (with beams o ). Note that lower noise per crystal implies
improved energy resolution because a larger number of crystals per shower can be summed.
Shielding is needed for protection against coherent pickup noise. The entire calorimeter is
enclosed in a metallic Faraday housing. The crystal housings are metallized to reduce cross
coupling, and the diode-to-preampli er connection uses shielded cable of minimum length.
Both the diode and the preampli er are encapsulated in a metal cap for electromagnetic
shielding.
The equivalent electronic noise energy (ENE) for a crystal with a given readout system can
be calculated as
(7.4)
ENE = ENC ;
L(pe= MeV)
where ENC is the equivalent noise charge in number of electrons, and L(pe= MeV) is the light
yield in electrons per MeV of energy deposition. Equation 7.4 indicates that achievement
of low ENE requires not only low intrinsic noise of the electronics (ENC), but also a highly
ecient photodetector coupled with maximal light output of the crystal. The resulting light
yield requirements are detailed in Section 7.4.3.
Each crystal is read out by two independent photodiode-preampli er-shaper channels [Hal95].
The two channels are then summed before digitization. The redundancy is due to reliability
Technical Design Report for the BABAR Detector
7.4 Crystal Subassemblies and Readout
271
requirements (Section 7.4.4), and the noise per crystal is the incoherent sum of the noise of
the two channels.
The principal contributions to the electronic noise are the shot noise from the dark current
of the diode and the thermal noise of the preampli er [Gro84]. The shaping time a ects the
relative contributions of the two components. Two possible readout schemes, one using two
Hamamatsu S3588-03 photodiodes (3.4 30 mm2 active area) mounted on a wavelength
shifter (WLS) and the other using two Hamamatsu S2744-03 photodiodes (10 20 mm2
active area) directly axed to the crystal, are being considered. A preampli er-shaper
hybrid for both options has been developed and tested. The equivalent noise charge per
crystal for the WLS option is 620 e, and for the direct option 680 e,, using a 2 s shaping
time [Jes95a]. Note that a slight cooling of the diodes may further reduce the noise. The
dark current of the photodiodes is halved by a reduction in temperature of as little as 5{7C.
Studies of noise vs. temperature are presently underway.
Wavelength Shifter/Photodiode Readout
Typical silicon PIN photodiodes (e.g., Hamamatsu S 3588-03,S 2744-03) show a broad spectral response peaking at wavelengths around 960 nm. They have a quantum eciency of
about 75%. The photon spectrum of CsI(Tl), on the other hand, has its peak emission
around 560 nm. In addition, there is an order of magnitude mismatch between the lightemitting area of the crystal and the sensitive area of the photodiode. Thus, the purpose of
the WLS is twofold: to shift the wavelength of the CsI(Tl) scintillation light to the more
sensitive range of the photodiode, and to work as a ux concentrator by collecting the light
onto the sensitive area of small, relatively inexpensive photodiodes through total internal
re ection.
Fluorescent dyes are suitable materials for wavelength shifters [Lor86]. They show a large
di erence in absorption and emission wavelengths and re-emit the uorescent light isotropically. The absorption of the dye should match the emission of CsI(Tl), and the re-emission
of uorescent light must take place at wavelengths as long as possible, with good internal
quantum eciency, small luminescent decay time, and low light loss. The overlap of the
absorption and re-emission spectra of the wavelength shifter should be minimal in order to
prevent self-absorption of luminescent light.
Typical wavelength shifters consist of thin plastic tiles doped with dye concentrations of
a few hundred ppm [Lor86, Fis87]. The dyes used so far (BBQ, Y7, BASF Lumogen F
red 339) have Stokes shifts of only 50 nm [Ake92, Kam83] and therefore show a large selfabsorption of about 50% [Kei70]. Figure 7-18 shows the absorption and re-emission spectra
of a wavelength shifter similar to that used by the Crystal Barrel [Ake92]. Transmission as
high as 30% has been achieved with this material.
Technical Design Report for the BABAR Detector
272
Electromagnetic Calorimeter
Figure 7-18. Absorption and uorescence re-emission spectra of the Crystal Barrel
wavelength shifter using BASF Lumogen F red 339 dye. Overlayed are the emission
spectrum of CsI(Tl) and the photosensitivity of the Hamamatsu S 3588 photodiode.
Fluorescent dyes that are used for Nd:YAG pumped dye lasers (e.g., Styryl 8, Styryl 9M)
absorb around 550 nm, re-emit above 750 nm, and have nearly no self-absorption. Acrylic
material, such as PMMA, is used to carry the dye. These dyes must, however, be dissolved
in a polar co-polymer before adding to the PMMA. Whether these dyes have high enough
quantum eciency and the optimal concentration for high absorption and high quantum
eciency are being investigated.
Figure 7-19 shows the proposed WLS setup with two PIN diodes. The WLS covers almost
the entire rear face of the crystal. The re-emitted light is concentrated along the WLS side
faces by total internal re ection and read out using small-area rectangular photodiodes (e.g.,
Hamamatsu S 3588-03, 30 3:4 mm2 ). The diodes are attached to the WLS with Kodak HE80 epoxy or Cargille Meltmount. The thickness of the WLS is primarily chosen to match
the area of the photodiode. For redundancy, and in order to minimize light losses, two
photodiodes are glued to two adjacent side faces of the WLS. To increase the light collection
eciency, a white di use re ector is placed behind the wavelength shifter, with an air gap to
maintain total internal re ection inside the WLS. The other edges are polished and coated
with white di use re ecting paint.
The light yield for a variety of crystals has been measured with the WLS/diode option [Jes95a].
For an 18X0 tapered crystal with a front face 4:54:5 cm2 and rear face 5:85:8 cm2 , wrapped
in three layers of 38 m-thick Te on, a light yield of 4600 pe/ MeV has been measured. The
Technical Design Report for the BABAR Detector
7.4 Crystal Subassemblies and Readout
Side View
AAA
A
A
A
AA
A
AAAAA
A
AAAA
AAAA
A A
AAAA
Teflon AF + Diffuse
Reflector + Al Mylar
Al Mylar
273
Standoff
CsI (Tl)
Air Gaps
PIN Diode
PC Board
with Preamps
Fluorescent
Flux Concentrator
Back View
Fluorescent
Flux Concentrator
Bias, Signal,
Calibration
Preamplifiers
PIN Diodes
Figure 7-19.
CsI(Tl) readout assembly (not to scale).
equivalent noise charge for the two diodes combined is 620 e,, which implies an equivalent
noise energy of 135 keV per crystal. For comparison, the Crystal Barrel Collaboration has
achieved an equivalent noise energy of 220 keV with a similar wavelength shifter geometry
and one diode [Ake92].
Tests of possible radiation damage or aging e ects of the WLS and the epoxy used for the
WLS/photodiode bond must be completed. Radiation damage tests will be carried out using
60
Co sources. There will also be long-term tests at normal and elevated temperatures.
Direct Photodiode Readout
Photodiodes can, alternatively, be placed directly on the back of the CsI(Tl) crystal. To
collect a sucient amount of scintillation light, two large-area photodiodes (e.g., Hamamatsu
S 2744-03, 10 20 mm2 ) must be used. The balance of the back surface is covered with a
di use re ector. This technique is more sensitive than the WLS readout to variations in
light production and collection because the back surface is not fully covered by detectors.
Direct readout with four small-area (Hamamatsu S 1723, 10 10 mm2) photodiodes is used
in the CLEO-II calorimeter, which yields an equivalent noise energy of 500 keV [Kub92].
A variety of crystals have been measured using two Hamamatsu S2744-03 (20 10 mm2)
diodes [Jes95a]. Using the same 18X0 crystal as in the WLS test previously described,
we measure a light yield of 5960 pe/ MeV. The equivalent noise charge of the two diodes
combined is 680 e, to give an equivalent noise energy of 114 keV per crystal. For comparison,
the BELLE Collaboration has achieved 150 keV in similar tests [BEL94].
Technical Design Report for the BABAR Detector
274
Electromagnetic Calorimeter
1200
1.2
1000
1
800
0.8
600
0.6
0.4
400
0.2
200
0
1
2
3
4
5
6
0
1
2
3
4
5
6
(a) Relative light yield vs. shaping time and (b) noise vs. shaping time for
the direct and WLS readout schemes.
Figure 7-20.
Conclusions
In bench tests, the WLS and direct readout options have achieved comparable performance.
Further improvements in performance are anticipated with new WLS dyes and diode cooling. Once noise equivalents of about 150 keV per crystal using WLS readout are routinely
achieved, this solution will be adopted. The overall complexity and reliability of such a
readout system is comparable to that of direct diode coupling but has the advantage of
substantially lower cost.
Note that all tests used a shaping time of 2 s. A lower value may be needed in order to
reduce beam-related backgrounds [Lev94]. Figure 7-20 shows how to scale the light yield
and noise for di erent values of the shaping time.
7.4.2
Light Collection
Surface Preparation, Wrapping, Coating, and Tuning
Each crystal is cut and polished by the manufacturer. The light output of the crystal is
required to be uniform along its length. This is traditionally accomplished by roughening
the crystal surface selectively to reduce locally the re ected light and then wrapping the
crystal in multiple layers of a suitable re ector, such as PTFE Te on. The highest light
yield is obtained by the use of a high-re ectivity white di use re ector with a small airgap
Technical Design Report for the BABAR Detector
7.4 Crystal Subassemblies and Readout
275
Layers of 1.5 mil Teflon
0
5
10
15
20
1.6
Light Output
Relative to 3 Layers of Teflon
1.4
1.2
1
0.8
0.6
PTFE Teflon
0.4
Hard Tyvek
Soft Tyvek
0.2
0
0
5
10
15
20
25
30
Thickness (mils)
Figure 7-21. Light output vs. number of layers of PTFE Te on and DuPont Tyvek. (In
each case, there is one cover layer of aluminum foil.)
between the re ector and the crystal surface. The light yield increases with the number of
Te on layers used to wrap the crystal (Figure 7-21).
One potential wrapping scheme for the crystals uses four layers of PTFE Te on, each layer
being 38 5 m thick. This yields 58% of the available light with a total Te on thickness
of 152 10 m. To provide electronic shielding, each crystal is wrapped in addition with
25 m of aluminum foil.
A more ecient, automated approach employing Tyvek, a spunbonded ole n paper produced
by DuPont, is under investigation. Tyvek has a high di use re ectivity similar to Te on for
the same thickness but is considerably more robust. A further advantage over Te on is that
Tyvek can be printed with a pattern of dots to selectively absorb light, thereby producing
the required tuning pattern in an automated fashion. It is also possible to bond the 25 m
aluminum shielding foil directly to the Tyvek.
Coating CsI crystals with clear Te on AF as a sealant to reduce the e ects of humidity
on the crystal surface is also being evaluated. Te on AF has a low index of refraction
compared to CsI and transmission measurements at the peak wavelength of scintillation
emission (560 nm) indicate that the amount of light leaving the end of the crystal increases
by about 5% compared to an uncoated crystal. A 2.54 cm-diameter by 2.54 cm-height CsI
cylinder with diamond-polished end faces was coated with Te on AF, and its transmission
was measured. The crystal was then placed in a room-temperature chamber with 100%
Technical Design Report for the BABAR Detector
276
Electromagnetic Calorimeter
humidity for three days. No change was observed in the transmission of the crystal after
this exposure. The radiation hardness of Te on coatings has been studied, and no change
in optical properties for exposures of 40 krad of 60 Co -rays has been observed.
Water-based cross-linkable uorocarbon (WXF) o ers the same performance as Te on AF
but with a simpler dipping or brushing application and a lower temperature curing step.
WXF can be applied in thicker coatings as well.
Both Te on AF and WXF can be loaded with white TiO2 powder to provide a highly
re ective, di use coating. We are studying the application of a clear coating (to mimic a
low index of refraction air layer) followed by a white di use coating (to mimic a wrapping
of PTFE lm, for example). This coating system o ers greater control of thickness than is
obtainable with wrappings.
Uniformity Speci cations
The simulation results of Section 7.3.1 allow speci cation of light output uniformity. The
nonuniformity contribution must be well below the target energy resolution (Equation 7.1).
Requiring a contribution to E =E below 0.5% translates into an integral variation of the
light output between ,6% and +6% from the front end to the back end of the crystal, as
shown in Figure 7-22. Variations near shower maximum are the most critical. However,
short distance variations of 2% in the front third of the crystals are tolerable (Figure 7-22).
A procedure for measuring the light output uniformity is described in Section 7.4.3.
7.4.3 Light Yield Measurements
A number of crystals having dimensions similar to our baseline design were obtained from
potential vendors and the light yields measured [Jes95a]. The readout techniques, shaping
times, polishing, wrapping, and preparation are close to that of the baseline design.
The tests were made with a wrapping of three layers of PTFE Te on lm (each 38 m thick)
with one layer of aluminum foil outside. The light yield has been measured in three ways:
1. A 3.4 mm-thick WLS and two 30 3:4 mm2 S3588-03 diodes was epoxied to adjacent
edges. Two standard calorimeter preamps were used, with the WLS positioned 0.5 mm
from the back face of the crystal.
2. Two 10 20 mm2 S2744-03 diodes were directly coupled to the crystal using optical
grease. Two standard calorimeter preamps were used.
Technical Design Report for the BABAR Detector
L /Ln
7.4 Crystal Subassemblies and Readout
277
1.08
1.06
1.04
1.02
1
0.98
0.96
0.94
0.92
0
5
10
15
20
25
30
35
cm
Figure 7-22. Upper limits for the light output nonuniformity in a 35 cm-long CsI(Tl)
crystal, assuming no more than 0.5% contribution to E =E for all photon energies up to
5 GeV.
3. A two-inch photomultiplier tube (Hamamatsu R669) with an extended-red multi-alkali
photocathode (peak sensitivity at 650 nm) was positioned 0.5 mm from the crystal back
face. The light yield is measured relative to that from a small calibration crystal, since
this will be the procedure used for vendor speci cations.
For the photodiode readout, the output signal is shaped using a 2 s shaping time. Figure 7-20 shows how the relative light yield varies with shaping time. The absolute calibration
of electron number per ADC channel for the diode is obtained by using a 59.5 keV x-ray from
a 241 Am source, taking into account that 3.62 eV is required to produce one electron-hole
pair in silicon.
Light Yield Speci cation
The light yield of CsI(Tl) crystals is directly related to the resolution and the energyequivalent electronic noise. To ensure that the calorimeter can achieve the required performance, every crystal must satisfy a light yield speci cation supplied to the crystal manufacturers.
The light yield is sensitive to the thallium iodide concentration in the crystal, the light
attenuation length, the shape and size of the crystal, the surface preparation, the wrapping,
and the position of the source which introduces the scintillation. It is necessary to provide a
practical and economical measurement procedure for the vendor. Therefore, the light yield
Technical Design Report for the BABAR Detector
278
Electromagnetic Calorimeter
Front Rear Length
LY
LY
LY
Dim. Dim.
Rel. to
Direct
WLS
2
2
Vendor
( cm ) ( cm ) (cm) Std. (%) (pe/ MeV) (pe/ MeV)
Kharkov
25
25
34
32
6340
4200
Kharkov
34
20
34
32
5960
4600
Crismatec
28
18
23
27
4820
3650
Crismatec (hex) 28
18
22
31
5300
4250
Horiba
28
18
23
30
8600
5560
Horiba (hex)
28
18
23
50
9800
6800
Polsyscine
34
20
34
35
6500
4550
Light yield (LY) measurements (as described in the text) for full-sized
crystals [Jes95a]. All crystals have a square cross section except where indicated.
Table 7-8.
speci cation is given in terms of light yield relative to a small (e.g., 2.54 cm diameter
2.54 cm height) standard CsI(Tl) crystal, measured using lines from a radioactive source
such as 22Na, and a red-sensitive PMT. Light yield for a full-sized crystal can be de ned
as the average of a set of measurements taken at di erent source positions along the length
of the crystal; such measurements are also used in determining the crystal's light output
uniformity.
Assuming the WLS readout option, with 620 e,s ENC for each crystal, a light yield of
620=0:15 MeV 4100 photoelectrons/ MeV is required to obtain 150 keV incoherent noise.
From Table 7-8, a relative light yield of 30% is therefore required from each wrapped
crystal.
7.4.4 Reliability of Inaccessible Readout Components
The barrel crystals and their front-end readout up to the ADC boards are inaccessible,
barring disassembly of the detector. It is anticipated that this will not take place until ten
years after the commissioning of the detector. Endcap crystals are relatively accessible and
can be repaired on a much shorter time scale. We therefore require that no barrel crystal
should fail in ten years. Achieving this reliability requires screening for defective parts before
assembly and establishing an adequate natural lifetime for the remaining components.
The components most likely to fail are the connectors, bonds, and electronic components.
Mechanical and thermal stresses and chemical aging may all be contributory factors. A
stable thermal and humidity environment can reduce these e ects.
Technical Design Report for the BABAR Detector
7.4 Crystal Subassemblies and Readout
279
A burn-in technique is being developed to screen all components, while random samples
are collected to estimate the natural lifetimes by accelerated aging tests. Screening involves
stressing components to remove those subject to infant mortality.
To establish the natural lifetime of components is dicult. It is empirically observed that, in
the limit of a large number of components, each with a natural lifetime much greater than the
operational lifetime, the failure rate follows an exponential distribution with a time constant
1/MTBF, the mean time between failures. This time is typically 100 years for electronics
components, implying that a joint MTBF in excess of 60,000 years must be established if
one requires less than one failure in 6000 crystals in 10 years. To establish this MTBF with
certainty requires testing 6000 components for 10 years, which is not feasible. There are
three techniques available to estimate the MTBF:
1. Assume aging is proportional to some power of temperature T (the Arhenius rate
reaction formula). Test a number of components at high temperature and scale to the
operating temperature. The power can only be estimated from similar studies.
2. Use published formulae for estimating the failure rates of components based on experience from, e.g., the military [Mil90].
3. Extrapolate failure rates from similar experiments such as CLEO-II, which operates
32,000 photodiodes and preampli ers under similar conditions.
Adopting method (3), Table 7-9 lists cumulative diode and preampli er failures and noisy
diodes over ve years of operation. The initial 48 diode failures are attributed to connector
failures or insucient screening. Considering all 48 plus the 14 noisy diodes as failures, the
preampli er appears far more likely to fail than the diode.
Using Poisson statistics, at the 95% con dence limit, one nds MTBFdiode 1954 years,
MTBFpreamp 832 years and combined, MTBFdiode+preamp > 643 years. The WLS option
uses diodes of similar area and manufacture, implying 93 failures per 6000 crystals after 10
years if a single diode/preamp readout is chosen. Table 7-10 shows the number of failing
crystals as redundancy is increased, assuming MTBFdiode+preamp > 643 years. Twofold
redundancy yields 308 crystals with one dead channel and two crystals with two dead
channels
p after 10 years [Jes95b]. A crystal with
p one dead channel has half the light yield
and 1/ 2 the electronic noise in ENC, hence 2 higher equivalent noise energy.
Technical Design Report for the BABAR Detector
280
Electromagnetic Calorimeter
Time
Diode Noisy Preampli er
(Months) Failures Diodes Failures
0
48
0
17
5
48
2
58
12
48
7
85
24
48
12
100
32
48
12
112
38
48
12
119
50
48
13
128
60
48
14
142
Table 7-9.
to 1994.
Cumulative diode and preamp failures for the CLEO-II calorimeter from 1989
Failed
1 Diode 2 Diodes 3 Diodes 4 Diodes
Channels 1 Preamp 2 Preamps 3 Preamps 4 Preamps
0
5907
5815
5725
5635
1
93
184
271
356
2
1
4
9
3
0
0
4
0
Table 7-10. Number of failed channels on a crystal for various levels of redundancy, after
10 years, assuming the MTBF of (diode+preamp) is greater than 643 years.
7.5
Calibration
7.5.1 Requirements and Ingredients
Several methods will be used in calibrating:
1. Electronics calibration will be done at least weekly with beams o . The 2000 events
needed per channel can be collected in several minutes. Calibration will entail the
injection of charge at the preampli er input, with controllable amplitude, phase (with
respect to digitization), and pattern, using the system described in Chapter 10, in
order to determine the detailed response function for each channel of electronics.
Technical Design Report for the BABAR Detector
7.5 Calibration
281
2. Bhabha events will be used for tracking overall gains. Enough data for this can be
obtained on a timescale of well under a day.
3. Photons and electrons of known energy, as well as minimum ionizing tracks, will be
used to set the energy scale (gain), including inter-channel variation of crystal light
output and optical properties of the photodiode (WLS) readout.
4. Radioactive sources will provide low-energy point(s) for the response of individual
crystals. Energy deposited in the front of a single crystal can be compared to Bhabhas
in order to track optical changes. The source spectrum is a direct monitor of noise in
energy units.
Items (2) and (4) also allow us to track changes over time of the entire optical path, including
possible radiation damage or environmental changes. CLEO-II has achieved a calibration
accuracy of 0:2% at 5 GeV increasing to 1% at 25 MeV, and a time-dependent gain
stability of under 0:25% [Kub92]. BABAR can at least equal this performance and improve
on the accuracy at low energies because lower electronic noise allows the use of radioactive
sources. Although it is not part of the baseline design, a ber optic distribution of light
pulses is still under consideration for monitoring crystals.
The calibration system will be used to set initial gains for all channels to within 10%, to
provide approximate real-time calibration factors for forming trigger sums and to provide
online diagnostics.
7.5.2 Energy Calibration with Beam Events
At design luminosity, the total rate of Bhabha coincidences (above 300 mr) is 92 Hz, with
at least 28 Bhabha e per hour incident on each crystal. Since the mean fraction of energy
in the peak crystal is 0.7, at least half of these events should be usable for calibration
with either an iterative procedure or a t. The goal of 0:25% precision requires about
40 usable events in a crystal, which can be achieved in under three hours. For polar angles
backwards of 133, Bhabha singles must be used. A special trigger can prescale such events
as a function of polar angle to obtain the sample.
The crystal response vs. energy must be mapped using photons of known or constrained
energy. The process e+e, ! provides maximum-energy data at each polar angle. In
e+ e, and
nal states, the energy of the lower energy photon can be computed from
event topology. Radiative Bhabha yields have been calculated for a minimum separation of
15 between the photon and each lepton. At design luminosity, the usable yield of photons
above 1 GeV ranges from 10 to 300 per hour per GeV per 10 of polar angle, varying across
photon phase space. Lower energy photons can be obtained from 0 ! decays; on the
Technical Design Report for the BABAR Detector
282
Electromagnetic Calorimeter
order of one 0 per B B event is expected to be usable. These are the methods by which
CLEO-II presently covers the photon energy range of interest.
High-energy minimum ionizing tracks deposit most of their dE/dx loss, averaging 200 MeV,
in individual crystals, providing a useful high-statistics comparison to Bhabhas. The yield
from the +, nal state alone is at least 0.7 per hour per crystal, but non-interacting
with momentum greater than 2.1 GeV=c, which traverse 2=3 of a crystal in azimuth, will
also be useful.
7.5.3
Source Calibration
Radioactive sources are included in the baseline design for several reasons. First, since the
energy is deposited at the front of a crystal, comparison to Bhabhas will help identify the
origin of a change in crystal gain, e.g., radiation damage or other change in light collection.
Second, a source spectrum provides a direct monitor of noise in energy units. Third, sources
are the easiest way to make initial gain settings. Fourth, sources provide low-energy points
for the responses of individual crystals to actual photons. Fifth, integrated sources do allow
continual system checks after installation and whenever the accelerator is not running. Source
energies of 2 MeV or higher are desirable to allow for cases of bad electronic noise and to
better mimic photons of physics interest.
Source runs will be performed with beams o using a special processing mode which involves
a search for signal waveforms in video RAM memory and the extraction of their amplitudes.
A source activity of 0:01 C for one crystal implies 1000 usable events in ten seconds. An
actual source run will be longer, due to processing limitations, but still well under 20 minutes.
Three options are under study:
1. Mounting 0.01 C 60Co sources (t1=2 5 years) at the front face of each crystal and
using the 2.51 MeV sum line. For adequate eciency, this may entail inserting a source
at the end of a ne needle into a very narrow hole drilled several centimeters into the
front of the crystal. The feasibility of this approach is being evaluated.
2. Pneumatically inserting several 100 C sources at z = 0 just outside the support tube
or within the PID enclosure. A good candidate is 208 Tl, which has a 2.614 MeV line,
driven by an decay chain from 228 Th (t1=2 2 years) or 232 U (t1=2 100 years).
Encapsulation prevents particles or radon from escaping.
3. Circulating about 1000 cm3 of liquid, activated in a shielded neutron source located
outside the detector, through ne tubes running in front of all crystals. Choices include:
(a) A uorine-containing uid, with a metastable state of 16O via 19 F (n, ) 16 N;
Technical Design Report for the BABAR Detector
7.6 Mechanical Support Structure
283
Barrel
Forward Endcap
( cm)
( cm)
Inner Gas Seal
0.5
3.0
Crystal Box Extension
0.1
0.7
Inward Dimension Tol.
0.6
0.6
Crystal Length
29.76{32.55
32.55
Outward Dimension Tol.
0.6
0.6
WLS/Diode Package+Spacers
0.4
0.4
Gap to Preampli er
0.1
0.1
Preampli er/Header
2.6
2.6
Cooling Manifold
1.5
1.2
Module Insertion Fixture/Cables
2.5
Module Strongback
1.6
Alignment Jacks
2.5
0.8
Aluminum Strongback /Frame
3.2
2.0
Total
46.0
44.5
Table 7-11.
Thickness of the calorimeter assembly, barrel, and endcap.
neutrons of 4-6 MeV from a Po-Be source are appropriate. The decay provides a
6.13 MeV line, with t1=2 of only 7 sec. This choice has been assessed as posing no
signi cant radiation safety hazard. (b) Water, with 16 O (n,p) 16N leading to the same
line. Activation could be by 14 MeV neutrons from a DT source. By using saltwater
and a moderated neutron source, one could additionally provide 2.75 MeV and 1.37 MeV
lines (with t1=2 of 15 hours) via 23 Na (n, ) 24Na.
7.6
7.6.1
Mechanical Support Structure
Design Considerations
The mechanical structure of the BABAR calorimeter is designed to minimize inactive material
between individual crystals, while providing hermeticity at the barrel-endcap interface.
Access, testing, repair, and reliability are addressed in the design of the overall structure, as
are the crystal loading procedures, and installation and system integration. RF shielding and
cooling of electronics, cable routing, and humidity control considerations are also discussed
in the following sections. Table 7-11 details the radial dimensions of the barrel and forward
endcap.
Technical Design Report for the BABAR Detector
284
Electromagnetic Calorimeter
Figure 7-23.
Crystal dimensions (Tables 7-12 and 7-13).
Since the detector is located in a seismically active region, safety standards require that the
calorimeter be designed to withstand accelerations of up to 2 g horizontally and 1 g vertically,
without signi cant damage, both during the assembly period and after installation.
Crystal Loading, Sizes, and Tolerances. CsI(Tl) is grown in polycrystalline form and
is a soft, deformable material that cannot support substantial loads. Therefore, the structure
must support individual crystals without transferring the load through neighboring crystals.
The crystal sizes and counts for the barrel and endcap are given in Tables 7-12 and 7-13.
The columns in the tables give the dimensions of the crystal edges as shown in Figure 7-23.
Crystals are required to t loosely into their barrel or endcap compartments. Crystal
dimensional tolerances are summarized in Figure 7-24. The tolerance on crystal transverse
dimensions (225 m) allows the cost of cutting and polishing to be minimized. Consequently, an additional radial space of 0.5 cm is required, as indicated in Table 7-11.
Wall Material and Thickness, and Inter-Crystal Material. Properties of structural
materials under consideration for the compartment walls are given in Table 7-14.
Technical Design Report for the BABAR Detector
7.6 Mechanical Support Structure
θ
285
Row
Number
Needed
Volume
(cc)
A
(cm)
B
(cm)
C
(cm)
D
(cm)
E
(cm)
F
(cm)
Height
(cm)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
882.5
888.4
894.6
901.0
907.6
914.4
921.3
896.7
903.5
910.4
917.3
924.4
931.4
938.3
910.6
916.9
922.9
928.7
934.2
939.3
943.9
911.3
914.6
917.2
919.2
920.6
921.2
882.4
882.4
921.2
920.6
919.2
917.2
914.6
911.3
907.5
903.2
898.4
893.4
888.0
882.3
876.5
870.5
864.4
858.3
852.2
846.1
840.0
834.1
4.956
4.953
4.950
4.947
4.943
4.939
4.935
4.930
4.925
4.919
4.913
4.907
4.900
4.892
4.884
4.876
4.867
4.857
4.847
4.837
4.826
4.814
4.802
4.790
4.778
4.765
4.752
4.740
4.740
4.752
4.765
4.778
4.790
4.802
4.814
4.826
4.837
4.847
4.857
4.867
4.876
4.884
4.892
4.900
4.907
4.913
4.919
4.925
4.930
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.773
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
4.736
5.779
5.812
5.846
5.880
5.916
5.952
5.988
5.994
6.030
6.065
6.101
6.136
6.171
6.204
6.197
6.226
6.254
6.279
6.302
6.322
6.339
6.306
6.315
6.320
6.321
6.318
6.310
6.299
6.299
6.310
6.318
6.321
6.320
6.315
6.306
6.293
6.277
6.258
6.236
6.212
6.185
6.157
6.127
6.096
6.064
6.031
5.998
5.965
5.931
5.523
5.559
5.595
5.633
5.672
5.712
5.753
5.766
5.808
5.850
5.893
5.936
5.980
6.023
6.028
6.069
6.108
6.147
6.184
6.220
6.253
6.237
6.264
6.287
6.307
6.322
6.334
5.896
5.896
6.334
6.322
6.307
6.287
6.264
6.237
6.208
6.176
6.142
6.106
6.068
6.029
5.990
5.949
5.909
5.868
5.827
5.787
5.747
5.708
5.524
5.559
5.594
5.631
5.669
5.708
5.748
5.759
5.800
5.841
5.882
5.924
5.965
6.007
6.009
6.048
6.086
6.123
6.158
6.191
6.222
6.204
6.228
6.248
6.266
6.279
6.289
6.294
6.294
6.289
6.279
6.266
6.248
6.228
6.204
6.177
6.147
6.115
6.081
6.045
6.008
5.971
5.932
5.893
5.854
5.815
5.776
5.737
5.699
32.55
32.55
32.55
32.55
32.55
32.55
32.55
31.62
31.62
31.62
31.62
31.62
31.62
31.62
30.69
30.69
30.69
30.69
30.69
30.69
30.69
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
29.76
Crystal dimensions and counts for the barrel. Dimensions A through F are
de ned in Figure 7-23.
Table 7-12.
θ
Row
1
2
3
4
5
6
7
8
9
Number Volume
Needed
(cc)
80
80
80
100
100
100
120
120
120
741.0
802.4
863.6
739.6
788.3
836.9
737.7
778.4
819.1
A
(cm)
B
(cm)
C
(cm)
D
(cm)
E
(cm)
F
(cm)
Height
(cm)
4.298
4.641
4.982
4.247
4.519
4.790
4.209
4.437
4.665
4.670
4.670
4.670
4.670
4.670
4.670
4.670
4.670
4.670
3.944
4.289
4.632
3.970
4.244
4.517
3.984
4.214
4.445
4.995
5.394
5.792
4.941
5.257
5.573
4.898
5.162
5.427
5.426
5.428
5.430
5.432
5.433
5.433
5.434
5.433
5.432
4.583
4.985
5.386
4.618
4.937
5.256
4.636
4.903
5.171
32.55
32.55
32.55
32.55
32.55
32.55
32.55
32.55
32.55
Table 7-13. Crystal dimensions and counts for the endcap. Dimensions A through F are
de ned in Figure 7-23.
Technical Design Report for the BABAR Detector
286
Electromagnetic Calorimeter
Figure 7-24.
Crystal and wrapping tolerances.
Density
dE/dx
X0
3
2
Wall Materials (g/cm ) (MeV/g/cm ) (g/cm2)
Carbon Fiber
1.68
1.90
45.2
Glass Fiber
1.70
1.87
33.0
Aluminum
2.70
1.62
24.0
Structural-material properties (assumes 40% epoxy and 60% ber loading
for ber materials).
Table 7-14.
CFC with 250 m wall thickness has been chosen for crystal compartments because it has
the best combination of strength and minimum material (X0). There are two such walls
between crystals, resulting in a total amount of structural material between crystals of
2 0:09% X0. Mechanical tolerances of the CFC boxes that form the support structure
can be made small because the boxes are formed on accurate mandrels and assembled in
precision trough xtures. (The fabrication technique in the endcap places a di erent set of
requirements on the dimensions.)
Inactive material between crystals also includes re ective wrappings, shielding foils, and
airgaps due to mechanical tolerances. This adds another 2 0:07% X0 , resulting in a total
of 2 0:16% X0 between crystals.
Technical Design Report for the BABAR Detector
7.6 Mechanical Support Structure
287
Nonprojectivity. To minimize unmeasured energy due to inactive material between crys-
tals, the crystals are arranged to be slightly nonprojective in and projective in with
respect to the IP. The degree of nonprojectivity is set by the requirement that straight-line
tracks cannot be completely contained in the gaps between individual crystals. Simulations
show that degradation of reconstruction eciency begins for photons that traverse more
than one fourth of the gap length before entering a crystal. To prevent this degradation, a
xed o set of 15 mr is added to the projective angle for a given crystal and this o set is
applied to each crystal in the barrel and endcap. The crystals then project to a point that
varies between 2 mm and 5 cm from the IP along the beam line. To minimize the number
of unique crystal shapes, the crystal gap is centered projectively at 90 with respect to the
beam line. The front faces of the crystals are uniform in width and perpendicular to the
crystal axis. In , the crystals are arranged so that the center lines of their compartment
gaps project to the beam axis.
The crystal row of the last barrel and rst endcap at the forward barrel/endcap interface
require special treatment because of the larger gap in this region. They will be shaped so
that the barrel/endcap gap is suciently nonprojective.
RF Shielding of the Electronics. The barrel and endcap are constructed as separate
Faraday cages. The inner rf-shields are provided by a laminate of two layers of 1 mm-thick
aluminum (each about four skin-depths at 100 kHz) separated by a thin foam core. The
interface region between the inner and outer faces of the barrel and forward endcap are tied
together with exible metallic ngers to shield the 3 mm gap. Shielding for inter-crystal
cross-talk and rf noise from the preampli ers and photodiodes of an individual crystal is
provided by aluminum enclosures attached to the 25 m aluminum wrap of the crystal.
Environmental Control for Crystals and Electronics. The calorimeter will be kept at
an appropriate constant temperature in order to keep the leakage current of the photodiodes
below 2 nA, to maintain stable preampli er performance, and to keep a constant CsI(Tl)
scintillation light yield.
The two preampli ers for each crystal generate about 100{250 mW, giving a total heat load
inside the rf-shield of about 2.5 kW. Heat ow calculations on a model crystal assembly with
the conservative assumption of 500 mW total power generated at the electronics indicate
that active cooling is required. Thermal control to better than 1 C is achieved by owing
coolant into the calorimeter along circumferential manifolds into pipes running axially along
the crystal rows. Each crystal's electronic shield is attached by braid to the manifold. The
cooling system output is cycled through an external heat exchanger. Digitizing electronics
located in shielded enclosures at the end of the barrel calorimeter housing and the outer rear
of the endcap are cooled using the same system.
Technical Design Report for the BABAR Detector
288
Electromagnetic Calorimeter
The outer surfaces of the calorimeter form a gas enclosure for humidity control. Dry nitrogen
owing through the volume at a slow rate and slight overpressure is used to maintain a dry
atmosphere to prevent degradation of the crystal surfaces.
Signal and Control Routing. Signal and control cables are passed through the Fara-
day/environmental shield using sealed bulkheads containing signal feedthroughs. Signal
cable lengths between preampli ers and bulkheads are minimized, with the longest run
being approximately 1.5 m. A standard 12-pair shielded cable will be used.
Digitizing electronics cards are attached to the feedthrough connectors in the bulkheads.
Fiber-optic cabling from the digitizer cards of both the barrel and forward endcap is routed
outward between the barrel and endcap ux return iron.
Access and Component Replacement. Individual barrel crystals and their readout
packages are not accessible once the calorimeter has been installed in the solenoid. However,
the modular design facilitates testing and replacement of diode/preamp packages before this
installation. The endcap can be removed for internal maintenance. All calorimeter digitizer
cards can be readily accessed for repair by opening the endcap doors.
7.6.2 Barrel Fabrication, Assembly, and Installation
Barrel System
In order to avoid the transmission of loads through the crystals, the barrel cylinder is divided
into independently supported and aligned modules which have little or no mechanical contact
with each other. Their light weight limits de ections. The structure is composed of CFC
elements between crystals in which minimum material is necessary, and aluminum plates in
which structural strength is required and more material can be tolerated.
The basic units are the CFC compartments (boxes) that house individual crystals. The CFC
boxes are combined into modules containing 3 7 crystals, each weighing approximately
80 kg. Figure 7-27 shows the basic module structure. The individual compartments are
sti ened at their rear (outer) ends with a ring that also corrects for tolerances in the crystal
placement and locks the crystal longitudinally into position in their boxes. The boxes
are joined to a strongback made from thin aluminum side plates (tabs) and an extruded
aluminum U-channel fastened to the rear of the modules. The module strongbacks transmit
loads to a large cylindrical aluminum tube with sti ening anges at each end, into which
the modules are inserted and aligned. Both the module strongback and the cylindrical tube
have access ports to allow for assembly, alignment, cabling, and repair prior to insertion into
Technical Design Report for the BABAR Detector
7.6 Mechanical Support Structure
289
,
,
,
,
,
,, ,,
,, ,
,
Fibertube Manufacturing
Mandrel
With Heater
and Spring
Tensioner
Temperature
Cure Under
Pressure
Release
Cloth
Removed
Mandrel
20 psi
Plastic
Cover
Epoxy
Blotter
Cored
Tube Ready
for Trimming
F iber Braided
Hose ± 45°
2-95
7857A12
Figure 7-25.
Autoclave
With
Expandable
Rubber Liner
Schematic of tube fabrication by a wet-layup technique.
the magnet coil. After insertion into the coil, the tube transmits its load to the coil and ux
return through a two-point mounting scheme on each end ange.
Module Fabrication
The choice of module size is made by considering ease of handling (without requiring large
cranes or special xtures within dry-rooms), access to elements within, size and complexity of
fabrication xtures, and the de ections and strengths of the elements. The delivery schedule
of each crystal type, and its in uence on the assembly schedule, is also considered. There
are 280 modules in the barrel, 40 in azimuth and 7 along the beam axis.
Each CFC box is fabricated using a conventional wet (epoxy-resin) layup technique, as
illustrated in Figure 7-25. This technique allows reproducible dimensions of the inside box.
The nal wall thickness is the only variable dimension. For each crystal size, an aluminum
master mandrel is precision machined to the size of a wrapped crystal (i.e., crystal dimension
plus 175 m). The mandrel contains a heating and thermostatic control element for curing
the epoxy. The mandrel blanks may be fabricated by casting.
To fabricate a CFC box, the mandrel is rst sprayed with a release agent such as Te on. A
woven tube, or sock, of CFC is stretched over the mandrel. The weave is chosen at 45{60
to accommodate the taper of the crystal. Socks will be custom fabricated to a nished
Technical Design Report for the BABAR Detector
290
Electromagnetic Calorimeter
thickness of 250 m. These contain 400 strands per inch of 25 m lament. The mandrel
will be covered with a 25 m aluminum foil electronic shield before placing the sock over the
mandrel. The sock is next uniformly impregnated with thermal-setting epoxy resin. The
mandrel and sock are wrapped together in peel-ply, bleeder cloth, and a protective polyvinyl
sheet. The layup is cured under 2 atm pressure applied uniformly by a simple xture at 40C
using a heater in the mandrel. After the 2 hour cure, the tube is removed from the mandrel
and trimmed to shape and length. The mandrel and tooling are reused. Tests performed
over the last three years (Figure 7-28) using CFC boxes of similar weave fabricated with this
technique and with several commercial epoxy systems indicate that the dimensional creep of
such boxes is acceptable. Small factories can be set up at various collaborating institutions
to fabricate the 5880 elements for the calorimeter barrel.
There are seven distinct module types along the length of the barrel, for which 40 modules in
(plus one or two spares each) are required. The boxes are joined into 3 7 arrays for each
of these module types by constructing an aluminum trough xture with inside dimensions
corresponding to the nished outside dimensions (symmetric wedge in and a three-faceted
wedge on each end (Figure 7-27) of the desired unit. The trough xture has stops along
each of the 21 crystal axes (Figure 7-26) to set the depth. To lay up a module, a mandrel
is inserted into each of the 21 boxes. The exterior surfaces of the boxes are roughened and
wetted with epoxy. The tubes and mandrels are then forced into the trough, in rows of
three, until all 21 spaces are occupied. By seating the ends at the depth stops, the epoxy is
forced out automatically, accommodating the irregular shapes of the outer walls of the boxes
while preserving the inside dimensions. The mandrel heaters cure the unit. Once cured, the
mandrels are removed, leaving the module in the trough. Any excess epoxy is cleaned o ,
and the perimeter edges of the two outer rows of boxes are trimmed to their nal shape.
The next module manufacturing step is the sti ening of the boxes and attachment of
aluminum support elements at the rear of the box array. The trough and un nished module
are moved to a dry-room area so that each compartment can be loaded with its wrapped
CsI crystal. An example of a suitable facility is Building 109 at SLAC, which contains an
existing, high-quality 2{3% humidity dry-room. This room will be enlarged and refurbished
with a high performance lithium-chloride-based dehumidi er to allow crystal and module
storage, assembly, and electronic testing in the same area. The original dehumidi er for the
building will be refurbished and used for the adjacent temporary assembly area, and as an
emergency backup system.
To complete a module, each crystal is pushed forward into its compartment. The taper of the
crystal module and the fabrication tolerance ensure that the crystals will sit loosely within the
box (0:5 cm from nominal radial position). An aluminum enclosure (Figure 7-29) is inserted
into the top and glued to the inside of the box with a small amount of conducting epoxy
so that the crystal's foil wrap remains in good electrical contact with the enclosure. When
all crystals are in place, the 1.27 cm-thick rear extruded aluminum U-channel strongback
Technical Design Report for the BABAR Detector
7.6 Mechanical Support Structure
291
Aluminum Strongback with
Optical Target Receptacles
Fiber Composite
Crystal Support
Tubes
Temporary
Optical Target
Receptacles
Trough xture for combining tubes
into module subassemblies.
Figure 7-26.
Figure 7-27.
barrel module.
3–95
7857A18
Schematic of a
Creep deflection (mils)
Epoxy system #1 (West)
Epoxy system #2 (Hysol)
Epoxy system #3 (Epolite)
Epoxy system #4 (RF)
100
Start at
300 hours
10
100
1000
10
4
Elapsed time (hours)
Creep tests of glass- ber
woven boxes prepared by a wet-layup
technique.
Figure 7-28.
Detailed side view of a
module showing crystal, shielded readout
package, cooling, and cabling.
Figure 7-29.
Technical Design Report for the BABAR Detector
292
Electromagnetic Calorimeter
assembly is screwed and bonded to the side supports. Bridging elements are bonded in place
to sti en the module.
A set of wire alignment target holders are epoxied to each module as shown in Figure 7-27.
Their position is set by the trough, allowing exact and reproducible placement of the targets.
Module Completion and Testing
When the module is completed and removed from its trough, it is placed into dry-room
storage until electronics installation. The electronics assembly involves the addition of a
wavelength shifter/diode package and preampli er board to each compartment in a module.
These elements attach to the aluminum enclosure and are spring loaded to press against
the back face of the crystal. A lip machined into the aluminum enclosure maintains a
uniform air gap between the WLS and the crystal. If the direct diode readout method is
selected, crystals have their PIN diodes pre-attached (and pre-tested) before being inserted
into the module. The preampli er is connected by soldering the diode leads through plated
vias in the preampli er board, and the preampli er card is strain-relieved to the enclosure
cover. The stamped cover of 1.5 mm thickness is screwed down to the aluminum enclosure,
making contact all around the perimeter. Cables enter through the top by means of a ribbon
connector. The nished module may be tested and returned to dry storage prior to insertion
onto the strongback cylinder.
Barrel Assembly
The assembly of the entire barrel and endcap calorimeter takes place adjacent to the dryroom in SLAC Building 109. A temporary dry-room (5 5 5 m) is constructed with
unistrut, wire mesh, and inner and outer plastic sheet walls. Walls are sealed to an epoxypainted concrete oor and dry air blown between them. This temporary assembly area is
connected to the storage/testing dry-room by a short interlocked passageway through which
modules can be carried from storage.
The assembly xture is shown in Figure 7-30. The large cylindrical strongback is rst
mounted to a support frame at the two points on each of the end- anges that will eventually
be used for support by the coil. By mounting it in this way, the strongback can be transferred
to the coil supports without a ecting the alignment of the modules. Alignment of individual
modules compensates for strongback and module distortions that occur during the loading
procedure.
A module is moved to the assembly area after electronics testing. The strongback of each
module is mounted vertically on a movable module insertion arm on an assembly xture
(Figure 7-30). The arm is rotated to the proper orientation and moved axially into
Technical Design Report for the BABAR Detector
7.6 Mechanical Support Structure
Figure
7-30.
strongback.
293
Barrel assembly xture for inserting and mounting modules in the
place within the barrel. The module is then rough-mounted to the strongback. Initially,
all modules are retracted radially about 2.5 cm from nominal, allowing space for adjacent
modules to be installed. When all modules are loaded, the strongback cylinder will be in its
fully deformed state. The alignment mechanism is then used to align each module into the
correct radius and orientation.
The alignment procedure is accomplished by placing an accurately machined template on
the ends of the strongback cylinder and iteratively aligning the pre-mounted wire alignment
targets with the template in each of the 40 module rows. Alignment telescopes are mounted
on the module installation arm for simplicity. The procedure is shown schematically in
Figure 7-31. Two or more crews can work simultaneously to align separate rows of modules,
and the alignment procedure itself does not add signi cant deformations to previously aligned
modules. Thus, a single alignment pass should be sucient.
Finite Element Analysis
A preliminary nite element analysis has been performed both on the strongback cylinder
loaded with dummy modules and on a few representative individual modules. This approach
Technical Design Report for the BABAR Detector
294
Electromagnetic Calorimeter
Figure 7-31.
Barrel alignment.
is suggested by the independent suspension of the modules in the cylinder, allowing separate
calculation of module de ections. The cylinder analysis determines an appropriate cylinder
thickness, taking into account the amount of material removed for module access ports, and
the required sizes of the forward and backward anges. The forward(backward) end ange
used in the cylinder analysis is 2.5 cm(5.0 cm)-thick and 36 cm(20 cm)-wide.
The strengthening tab thickness used in the model of a module is 0.64 cm. The analysis
assumes for simplicity that the crystal boxes are made of 250 m-thick aluminum sheets
(the modulus of elasticity of aluminum is about the same as the quasi-isotropic modulus of
CFC). While we will not have enough layers of carbon ber to make the composite behave
isotropically, the steep pitch of the sock bers makes the box sti er than aluminum in the
bending-strain direction. The loading of the crystal boxes is assumed to be concentrated
where the box walls meet at right angles because the box material is so thin that it will have
little mid-span bending sti ness. Under these conditions, the loading shifts to these corners
where the boxes are sti est.
The module model includes a matrix of 1.0 mm-thick aluminum electronics enclosures assembled behind the crystals. These enclosures also serve to keep each thin CFC crystal box
square at this end. Additional sti eners are included in modules near the end of the cylinder,
where the center of gravity of the crystals is farther from the cylinder wall. Sti er support
Technical Design Report for the BABAR Detector
7.6 Mechanical Support Structure
295
Element
De ection (mm)
Module in Horizontal (end)
0.40
Module in Horizontal (center)
0.42
Module in Vertical
0.04
Strongback Cylinder (unloaded)
0.04
Strongback Cylinder (loaded)
0.67
End Flanges (cylinder loaded)
0.04
Table 7-15.
Static de ections encountered in the barrel.
is required in this region, but there is also more room for material without obstructing
electronics and cable ways.
Static de ections of the system are summarized in Table 7-15. For example, when all
modules are installed, the 3.175 cm-thick strongback cylinder is deformed by 670 m at
the center of the cylinder. The inner faces of modules mounted in the horizontal plane
undergo the maximum de ection expected, about 420 m relative to the cylinder, or 1.1 mm
total vertical de ection relative to the cylinder end ange mounting points. However, this
1.1 mm de ection is e ectively removed during the nal alignment of the crystal modules.
Testing and Closure: Cabling, Cooling, Front and Back Panels
When modules are rst installed in the strongback cylinder, an in situ electronic test is
performed to ensure that there are no problems with individual crystals that require removal
and replacement of a module. Once the nal alignment is done, the cooling manifolds and
cooling tubes to each box are installed, followed by the nal cable runs. Plumbing for drynitrogen gas circulation is installed as is the inner rf/environmental cylinder. In order to
keep all plumbing electrically isolated from the cylinder, nonmetallic couplers are used on
the outer connections to manifolds.
A suciently long system test is then performed to nd any remaining problems in the
electronics. Next, the strongback cylinder access ports are closed with suciently thick
sheet metal to complete the Faraday cage and gas seal.
Barrel Installation into the Coil.
The fully aligned, cabled, and tested assembly is
carried to the assembly area for the coil and ux return in the detector hall. A system test
is once again performed to check that no damage occurred during transport to the hall.
Installation into the detector is discussed in Chapter 14.
Technical Design Report for the BABAR Detector
296
Electromagnetic Calorimeter
7.6.3 Endcap Fabrication, Assembly, and Installation
Endcap Overview
The design of the endcap is quite di erent conceptually from that of the barrel. Once
installed, the barrel will remain unopened, possibly for the lifetime or the experiment,
whereas the endcap must be capable of rapid mounting and demounting while maintaining
precision mating to the barrel. The mechanical structure is sophisticated, involving a high
precision, rigid outer shell with minimal material between the crystals.
The forward endcap is a conic section, with front and back surfaces tilted at 22.7 to the
vertical to match the drift chamber endplate, and is built in two monolithic pieces to enable
rapid demounting for access to inner parts of the detector. It is supported o the solenoid
coil and precisely aligned with the calorimeter barrel, and has been designed to minimize
both the material and the air gap between the two. The geometry is almost projective, with
the crystal axes pointing to an axial position 5 cm from the interaction point. The total
weight of the endcap is approximately 4 tonne.
In order to preserve optimum light collection and both spatial and energy resolution, similarsized crystals have been used throughout. This results in a layout of nine rings of trapezoidal
crystals, grouped in three super-rings of 120, 100 and 80 crystals. The total number of
crystals is 900, arranged in 20-fold symmetry.
Crystal Honeycomb Container
Each crystal package, consisting of crystal, readout electronics and cooling circuit, is held
rigidly in a separate compartment of the honeycomb structure, which is glued inside a loadbearing container, as shown in Figure 7-32. Structural rigidity is obtained from the solid
aluminum inner wall of the conic section, reinforced at the rear with a semicircular ange.
Bonded to this inner wall is the front face of the box, formed from two aluminum skins
of 1 mm thickness separated by a nomex core to give additional rigidity. These last two
surfaces also form part of the Faraday cage enclosing the barrel and endcap. The faces of
the container that mate to each other or to the barrel are made from 1 mm carbon ber
composite (CFC) and serve mainly to bond the honeycomb structure of crystal boxes. A
semicircular ange is bonded onto the rear of the thin outer wall of the conic section, both
to maintain its shape and form part of the structural support.
The honeycomb compartments are individually made in the form of tapered boxes of CFC of
wall thickness 250 m. The technique involves stretching a CFC sock over a mandrel, coating
it with epoxy, and curing it under pressure. The honeycomb is built up from 20 identical,
wedge-shaped modules, each consisting of 45 boxes. Assembly takes place in a precision-
Technical Design Report for the BABAR Detector
7.6 Mechanical Support Structure
297
dimensioned aluminum trough. Mandrels are inserted into the boxes, the outer surfaces of
which were previously roughened and coated with epoxy, and the whole cured under pressure.
Ten modules are epoxied into each segment of the endcap, using a combination of mandrels
and simple xtures to ensure bonding to the walls of the container.
Structural Assembly
Each of the two containers is prepared for transport, or assembly, by installing radial rails,
tted with cooling circuits and pre-cabled trays, between the inner and outer semicircular
anges, and clamping the walls of each crystal compartment in the outer super-ring to a rail
and tensioning them. The rails form a hub at the center, giving the necessary rigidity to
circumferential shearing.
During crystal loading, the container is held vertically in an assembly jig which supports it
from the front by means of the two rear anges and the solid aluminum inner wall, shown
in Figure 7-33. As each radial rail is removed in turn, crystal packages are loosely tted
into the compartments to rest on pre-measured styrofoam pads and secured with a square
ring. The readout and preampli er electronics unit for each crystal, pre-mounted with its
heatsink on a cap, is screwed to the ring.
Heat sinks, electrical shielding and connections are secured to the rail and the outer crystal
compartments clamped and tensioned. A nal electrical test is performed on each group of
crystals installed. The half ring support structure, with pre-mounted electrically shielded
ADC housings (Figure 7-34), is bolted onto the outer ange of the container, and electrical
connections made. The aluminum support plate is now bolted onto the support structure at
its outer radius and the hub formed by the rails at its inner radius, and cover plates bolted
over the access holes to complete the Faraday cage. Two assembly frames are required in
order that each segment of the endcap may be completely prepared for installation.
Prototype Analysis
Stress analysis has been performed for individual crystal compartments, a module of 45
such compartments, and one segment of the endcap comprising a honeycomb of ten modules, using nite-element modelling of the loaded structures. The results indicate a rather
uniform pattern of stresses and deformations with maximum values of 10 MPa and 0.1 mm
respectively.
Because of the precision required in mating the barrel and endcap, it may be necessary
to build a mechanical pre-production prototype consisting of a half-endcap, install it in
the assembly frame and load it with weighted blocks having similar properties to CsI.
Measurements taken with strain gauges might then be compared with predictions from the
Technical Design Report for the BABAR Detector
298
Figure
container.
Electromagnetic Calorimeter
7-32.
Endcap load-bearing
Section of endcap container showing assembly jig and rails.
Figure 7-33.
Technical Design Report for the BABAR Detector
7.7 Optimization and Prototype Studies
Figure 7-34.
299
Detail of endcap support structure showing ADC housings.
nite-element analysis, and di erences fed back into the model to enable modi cation of the
nal endcap design.
Endcap Installation into the Coil. The endcap halves are carried to the assembly area
in the detector hall. A system test is done to verify that the endcap has not been damaged
in transport. Installation of the endcap halves is discussed in Chapter 14. Figure 7-35 shows
the barrel and half of the endcap installed in the detector.
7.7 Optimization and Prototype Studies
Optimization and prototype studies consist of testing and re ning individual calorimeter
components as they are designed and built, as well as system tests. The system tests will
consist of two beam tests. The rst will primarily examine the optical elements and frontend electronics, while the second will feature a full system prototype. It is anticipated that
the rst beam test will occur in the fall of 1995, and that the second will occur prior to
full production, in the spring of 1996. Potential beam test sites in the US and Europe are
Technical Design Report for the BABAR Detector
300
Electromagnetic Calorimeter
Figure 7-35.
Side view of mated barrel and half of endcap.
being evaluated. Test-site selection criteria include the availability of low-energy beams to
measure the e ect of electronic noise on resolution and the availability of mixed beams of
electrons and pions to test e= separation and the e ect of hadronic split-o s.
In the rst beam test, it is anticipated that we will use an array of 5 5 crystals that have
been tuned and wrapped in the manner planned for the actual calorimeter crystals. The
optical elements to be tested consist of a photodiode and wavelength shifter readout, and
the front-end electronics consists of the preampli er/shaper ampli er package. To study
their performance, these components will be tested in a realistic environment. The boxes
used to hold each crystal will be made of the same materials as are planned for the actual
mechanical support structure. The front-end electronics and its shielding will be those
planned for the calorimeter. While not the nal production version, an electronics cooling
system will be implemented to obviate the e ects of heating on the photodiode and electronics
noise performance. Peak-sensing ADCs with a 12{13 bit resolution will most likely be used,
rather than a prototype ADC. We will also not attempt to prototype the trigger and data
acquisition system for this beam test. Aside from studying the front-end electronics, this
test is an opportunity to gain experience with handling crystals and with the e ects of the
mechanical support and shielding materials. We anticipate results showing the e ects of
electronics noise and interstitial material on position and energy resolution.
Technical Design Report for the BABAR Detector
7.8 Crystal Procurement Issues
301
The second test will consist of a nearly full system prototype, prior to production. One
or two prototype mechanical modules will be utilized. This will give us further experience
with mechanical fabrication techniques and with the installation of crystals and front-end
electronics into the modules. However, tests to verify the mechanical strength and viability
of the support structure will be conducted separately as part of the development of the
structure. Based on the results of the rst beam test, an improved, nal production version
of the front-end electronics and shielding will be implemented, as will a nal version of the
cooling system and cable layout. It is anticipated that prototype ADCs featuring a 17{18 bit
dynamic range, though not necessarily a production version, will be utilized along with DAQ
boards. The second beam test will give us a realistic check of the entire readout chain and
shielding and of the functionality of the support structure before production begins.
7.8 Crystal Procurement Issues
7.8.1
Radiation Hardness
The primary source of radiation for the CsI calorimeter is showers caused by beam-gas
interactions or o -axis particles striking machine elements at or near the IR, not synchrotron
radiation. The typical energies of the shower particles are in the MeV range. The expected
dose level for the barrel calorimeter is 1 krad/yr, and for the forward endcap 10 krad/yr.
At 1 MeV, the photon attenuation length in CsI is 5 cm, or 2.7X0. The radiation dose
is concentrated in the front 10{15 cm of the crystal. However, scintillation light produced
deep in the crystal is also a ected by radiation damage, as radiation damage results in the
formation of color centers or absorption bands [Hol88], and our readout system collects light
re ecting from the front and side crystal surfaces. Measurements of a small CsI(Tl) crystal
under radiation indicate that CsI(Tl) is relatively radiation resistant (Fig. 7-36). However,
larger samples from various vendors su er a degradation in light yield [Hit92] when irradiated,
perhaps due to impurities in large CsI(Tl) samples.
Since impurities in the CsI(Tl) salt may increase the likelihood for radiation damage in
CsI(Tl) crystals, we will attempt to establish a technique for growing radiation-hard CsI(Tl)
crystals by carefully controlling the purity of the salt. We will establish a radiation-hardness
speci cation for crystal procurement as a nal check of the nished crystals. We will
investigate the role of impurities in CsI(Tl) salt by measuring crystal samples doped with
known impurities. We will study the role of crystal processing in the development of
radiation-hard crystals. To determine trace elements and defects in the crystal, material
characterization studies and micro-structural analyses will be performed. We will also study
the light yield, scintillation, transmittance, and absorption of these CsI(Tl) crystals. If
Technical Design Report for the BABAR Detector
302
Electromagnetic Calorimeter
Relative Pulse Height
CsI(TI)
1.2
1st Growth
2166
2248
2nd Growth
2163
2252
3rd Growth
7380
7381
1.0
0.8
101
3-95
7857A20
102
105
103
104
Radiation Dose (rad)
106
Relative light yield of small (2.542.54 cm2 ) Q&S CsI(Tl) crystals vs.
radiation dose, measured with an avalanche photodiode with 1.275 MeV s from a 22 Na
source.
Figure 7-36.
correlations can be established between radiation hardness and an absence of impurities in
the salt and/or the crystal growing process, this understanding can be incorporated into the
growth of the crystals produced for the BABAR calorimeter.
Radiation hardness speci cations for crystal procurement will be established to check the
crystals supplied by vendors. A high dose test will be performed on small (2.54 cm cylinders)
samples from production boules. A sample will receive a short-term dose of 20 krad, after
which its light output will be measured. We require that the light output diminish by less
than 20%, and that it recover within ve days to within 5% of the light output measured
prior to irradiation. Large crystals may be tested for uniformity of transmission as well as
light output as a function of time.
In addition to the high-dose radiation measurements, we will carry out long-term (12
months), low-dose (1-3 rad/d) radiation studies using a modest number of production
crystals from each supplier, selected after a consistent production quality has been established. These crystals will be placed in an environment similar to the one planned for the
calorimeter (enclosure with dry-nitrogen ow, such that the relative humidity is 3% at
room temperature) and irradiated from the front with a 137 Cs source. The gain and resolution
of these crystals and a set of control crystals kept in the same type of environment but without
radiation exposure will be monitored throughout the irradiation period. This test will help
us distinguish radiation damage e ects from aging e ects.
Technical Design Report for the BABAR Detector
7.8 Crystal Procurement Issues
7.8.2
303
Quality Control and Testing
Obtaining the desired energy resolution from the calorimeter will require strict control
over the quality of the 6780 crystals, which will be of di erent types and from di erent
manufacturers.
As the crystals are produced by the manufacturer, the following tests will be applied, and
unsatisfactory crystals will be rejected.
The dimensions of the crystal will be measured to ensure that they lie within the
required mechanical tolerances;
Each crystal will be visually inspected for aws;
The light output from a standard source at several positions along the polished and
wrapped crystal will be measured using a standard readout system; this will measure
the overall light yield and the variation with position; and
Samples of the melt will be analyzed for trace impurities.
These tests will be carried out at the point of manufacture by the manufacturer working in
close liaison with representatives from the collaboration calorimeter group. The exact details
of the tests (which are still under study) will be speci ed as part of the purchase contract.
Immediately prior to assembly at the experiment, the crystals will be tested again, to con rm
that the properties have not changed due to damage in transit or other causes.
Each accepted crystal will be assigned a unique identi cation number and its properties
stored in a database. This will enable us to keep track of any batch-to-batch variations in
the properties of crystals that may emerge during the course of the experiment.
Technical Design Report for the BABAR Detector
304
REFERENCES
References
[Ake92] E. Aker et al., Nucl. Instr. Methods A321, 69 (1992).
[Ale94a] R. Aleksan et al., \Report of the Ad Hoc Calorimeter Committee,"
BABAR Note # 157 (1994).
[Ale94b] R. Aleksan et al., \Resolution Studies Using GEANT," BABAR Note # 152 (1994).
[BaB94] \Letter of Intent for the Study of CP Violation and Heavy Flavor Physics at
PEP-II," SLAC{443 (1994).
[BEL94] BELLE Collaboration, \Letter of Intent for a Study of CP Violation in B Meson
Decays" (1994).
[Blu86] E. Blucher et al., Nucl. Instr. Methods A249, 201 (1986).
[Eis93] A.M. Eisner, \The E ect of Material en route to the EM Calorimeter,"
BABAR Note # 110 (1993).
[Fis87] F. Fischer et al., Nucl. Instr. Methods A257, 512 (1987).
[Gro84] D. Groom, \Silicon Photodiode Detection of Bismuth Germanate Scintillation
Light," Nucl. Instr. Methods A219, 141 (1984).
[Hal95] G. Haller, D. Freytag, and J. Hoe ich, \Proposal for an Electronics System for
the BABAR Calorimeter," BABAR Note #184 (1994).
[Hea94] C. Hearty, \Performance Comparison between CsI and LKr Calorimeters for
PEP-II," BABAR Note # 141, (1994).
[HEL92] \HELENA, A Beauty Factory in Hamburg," DESY Report DESY 92/041 (1992).
[Hit92] D. Hitlin and G. Eigen, in Proceedings of the Int. Workshop on Heavy Scintillators
for Scienti c and Industrial Applications, edited by F. Nataristefani et al.
(Editions Frontieres, 1992), p. 467.
[Hol88] I. Holl et al., IEEE Trans. Nucl. Sci. 35, 105 (1988).
[Jes94] C. Jessop et al., \Report on the Performance of Hex vs. Trapezoidal Crystals,"
BABAR Note # 215 (1994).
[Jes95a] C. Jessop et al., \Development of the Front-End Readout for the BABAR CsI
Calorimeter," BABAR Note # 216 (1995).
Technical Design Report for the BABAR Detector
REFERENCES
305
[Jes95b] C. Jessop
\Reliability Issues for A AR CsI Calorimeter,"
A AR
# 217 (1995).
[Kam83] T. Kamon et al.,
A213, 261 (1983).
[Kei70] G. Keil et al.,
87, 111 (1970).
[Kub92] Y. Kubota et al.,
A320, 66 (1992).
[Lev94] M. Levi, \Impact of Backgrounds on the Calorimeter Resolution,"
A AR
# 151 (1994).
[Lor86] E. Lorenz et al.,
A239, 235 (1986).
[Mil90] U.S. Military Handbook, \Reliability Prediction of Electronic Equipment" (1990).
[Sch94] R. Schindler, \Experimental Highlights of B meson Production and Decay," in
L. Montanet et al., Physical Review D50, 1601 (1994).
[Spi94] A. Spitkovsky and F.C. Porter, \Study of Light Transport in Di erent CsI Crystal
Geometries," A AR
# 146 (1994).
[Zhu95] R. Zhu, \e= Separation in the A AR CsI Calorimeter," A AR
# 151
(1995).
[Zis91] \An Asymmetric B Factory Based on PEP," ed. by M. Zisman, SLAC{372 (1991).
B B
et
al.,
B B
Note
Nucl. Instr. Methods
Nucl. Instr. Methods
Nucl. Instr. Methods
B B
Note
Nucl. Instr. Methods
B B
Note
B B
Technical Design Report for the
B B
A AR
B B
Note
Detector
306
Technical Design Report for the BABAR Detector
REFERENCES
8
Muon and Neutral Hadron Detector
8.1 Physics Requirements and Performance Goals
M
uon identi cation and neutral hadron detection are provided by the Instrumented
Flux Return (IFR), which makes use of the large iron structure needed as the magnet
return yoke. Excellent muon identi cation over the widest possible momentum range and
the ability to detect KL0 s are distinctive features of the BABAR detector.
At the (4S ), muons are produced mostly in semileptonic decays, either directly from the
B mesons or from the cascade Ds. The sign of the charge determines the b or c avor of the
parent meson, thus providing a clean tag for the CP -asymmetry measurements.
For all particle species, the asymmetry of the machine results in a strong correlation between
momentum and angle. The absolute population and the shape of the muon momentum
spectra are quite di erent in the backward, central, and forward regions of the detector, as
can be seen in Figure 8-1.
2500
2000
1500
1000
500
0
0
2
p
4
GeV/c
7000
6000
5000
4000
3000
2000
1000
0
0
2
p
4
GeV/c
1400
1200
1000
800
600
400
200
0
0
2
p
4
GeV/c
Momentum distribution of direct and cascade muons in (a) the forward
endcap, (b) the barrel, and (c) the backward endcap regions. The shaded part of the
histogram refers to direct decays.
Figure 8-1.
Muon and Neutral Hadron Detector
p (GeV/c)
308
4
3.5
3
2.5
2
1.5
1
0.5
0
-1
-0.5
0
0.5
1
cos(theta)
Correlation between momentum and cos for particles reaching the IFR;
di erent momentum cut-o s for barrel and endcaps can be seen.
Figure 8-2.
The main goal for the IFR detector is to achieve the highest practical tagging eciency.
About 18% of all B decays contain at least one muon in the region covered by the BABAR
detector; these are roughly divided into 8% with a direct muon, 8% with a cascade muon, 1%
with a muon from a decay, and 1% dimuon events. High momentum muons (p 1:2 GeV=c)
are mostly direct decays, while muons from D decay peak at 500 MeV=c. A gain of almost a
factor of two in tagging eciency can be achieved if one can also detect low momentum (well
below 1 GeV=c) muons and correctly assign them to B or cascade D decays. Detection of low
momentum muons will also improve the measurement of jVbcj by reducing the systematic
error in the extrapolation to low momenta.
The muon momentum range to be covered by the IFR therefore extends up to a few GeV=c;
the lower end of the range is xed by magnetic bending in the barrel region (450 MeV=c)
and by energy losses in the inner detectors (250 MeV=c) in the two endcap regions, as shown
in Figure 8-2.
Charged tracks found in the central drift chamber will be matched to tracks in the IFR.
Identi cation as muons or hadrons will result from detailed analysis of the hit patterns in
the active detectors. The ratio of muons to hadrons in the three sections of the detectors is
shown in Figure 8-3. The maximum tolerable level of hadron misidenti cation will depend
on the physics topics under study; for tagging purposes, it has been shown that the hadron
contamination in the muon sample should not exceed the 15% level. This translates into an
upper limit of 5% for the probability that a hadron will be identi ed as a muon.
Technical Design Report for the BABAR Detector
8.1 Physics Requirements and Performance Goals
0.8
0.6
0.4
0.2
0
309
(a)
0
0.5
1
1.5
2
2.5
3
3.5
4
p (Gev/c)
0.8
0.6
0.4
0.2
0
(b)
0
0.5
1
1.5
2
2.5
3
3.5
4
p (Gev/c)
0.8
0.6
0.4
0.2
0
(c)
0
0.5
1
1.5
2
2.5
3
3.5
4
p (Gev/c)
Muon to hadron ratio as a function of momentum in the three sections of
the detectors: (a) forward endcap, (b) barrel, and (c) backward endcap.
Figure 8-3.
The IFR also constitutes a hadron calorimeter; it will allow the detection of KL0 s and
other neutral hadrons. A measurement of the KL0 energy is not foreseen due to the limited
calorimetric resolution. In any event, such a measurement produces no real improvement
in event reconstruction. The hits induced by the charged secondaries produced in the KL0
interaction will be used to detect KL0s. (About 70% of KL0 s interact before reaching the IFR.)
The KL0 direction is then inferred from the location of the energy deposition. The ability
to identify KL0s and reconstruct their directions creates the opportunity to use the decay
B 0 ! J= KL0 for CP -violation studies. This could provide an event sample comparable to
the benchmark decay B 0 ! J= KS0 and act as a systematic check on any observed CP
asymmetry, since KL0 and KS0 are opposite CP states. The parent B can be reconstructed
from the four-vector of the J= , measured in a leptonic decay, and the direction of the KL0 .
The momentum spectrum for KL0s produced in BB events peaks at 500 MeV=c (Figure 8-4).
KL0 s from B 0 ! J= KL0 are more energetic, as can be seen in Figure 8-5(a). Since this
case is a two-body decay, the momentum distribution is practically at, and extends from
1 to 3 GeV=c. The KS0 detection eciency averaged over this momentum range is 50%; this
is the benchmark with which to compare the KL0 detection eciency. The measurement of
the KL0 direction is enough to identify the B 0 ! J= KL0 decay, given the strong correlation
Technical Design Report for the BABAR Detector
310
Muon and Neutral Hadron Detector
400
350
300
250
200
150
100
50
0
0
0.5
1
1.5
2
2.5
3
3.5
4
p (GeV/c)
Figure 8-4.
Momentum distribution for all KL0 s produced in B 0 B 0 events.
between momentum and angle. The spread can be seen in Figure 8-5(b) and is due to
the B 's momentum distribution. However, a few signi cant physics backgrounds must also
be handled. Decays of both charged and neutral B mesons to J= K (892), in which the
K (892) decays to a KL0 pair, produce events that can directly mimic the desired signal
with branching ratios slightly larger than the signal. The momentum distribution and the
momentum angle correlation of the KL0 s produced in this process are shown in Figure 8-6.
These backgrounds can be suppressed by identifying the extra pion and reconstructing the
K (892) mass. Events in the charged case, B + ! J= K + ; K + ! KL0 + , may be further
suppressed by rejecting events in which the charged pion can be associated with an isolated
J= vertex.
To improve the signal to background ratio, it is essential to have good resolution in the measurement of the KL0 direction in addition to an ecient reconstruction of the KL0 . Figure 8-7
shows the dependence of the signal and of the background, before any other cut is applied,
on the angular resolution of the detector. The KL0 detection eciency improves steadily with
a ner iron segmentation, as shown later in this chapter.
Technical Design Report for the BABAR Detector
8.1 Physics Requirements and Performance Goals
311
250
200
150
100
50
0
0
0.5
1
1.5
2
2.5
3
3.5
4
p(GeV/c)
p(GeV/c)
4
3.5
3
2.5
2
1.5
1
0.5
0
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
cosθ
Figure 8-5.
The momentum distribution and the momentum vs. angle correlation for all
KL0 s produced in B 0 ! J= KL0 decays.
KL0
detection is useful in channels for which the KL0 direction alone is sucient to reconstruct
the momentum of the parent B . In particular, several D decays can be fully reconstructed
using charged tracks in a separate vertex and the direction of the KL0. For example, we have
tested the possibility of reconstructing the decay D, ! KL0 +,,. Decays of B 0 to D,+
have been generated in which the D, is then allowed to decay to KL0 +,,. The three-pion
D , decay should allow a good vertex reconstruction. The KL0 is then reconstructed by measuring its direction in the IFR system and recomputing its momentum under the assumption
that it comes from a D, decay. Figure 8-8(a) shows the resulting KL0 momentum distribution
together with the fraction interacting in the IFR (shaded area). Figure 8-8(b) shows the
resulting D,+ e ective mass for those events which can be reconstructed. Requiring in
addition a recoil B 0 ( Figure 8-8(c)) should allow this decay mode to be detected with little
background.
In addition to exclusive event reconstruction, the IFR can be used to veto events with missing
hadronic energy, in particular background events for the decay B ! ; when used as a
veto, it is obviously essential for the detector to have as complete a solid angle coverage as
practical and the highest possible eciency.
Technical Design Report for the BABAR Detector
312
Muon and Neutral Hadron Detector
350
300
250
200
150
100
50
0
0
0.5
1
1.5
2
2.5
3
3.5
4
p(GeV/c)
p(GeV/c)
4
3.5
3
2.5
2
1.5
1
0.5
0
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
cosθ
The momentum distribution and momentum vs. angle correlation of the KL0 s
produced in B ! J= K (892).
Figure 8-6.
8.2
8.2.1
Detector Overview
The Iron Structure
The muon and neutral hadron detector uses as an absorber the ux return iron of the 1.5 T
superconducting solenoid; this massive steel structure also serves as the support for the inner
detectors.
The design of the iron plates has to take into account both the requirements dictated by its
use as hadron absorber and muon lter and the complex mechanical problems caused by the
need to provide access to the inner detectors, all subject to the space constraints imposed
by the machine components and the experimental hall.
The IFR consists of three main components: the barrel and the backward and forward
endcaps. The barrel extends radially from 1.78 to 3.01 m and is divided into sextants;
the length of each sextant is 3.75 m, and the width varies from 1.88 to 3.23 m. Each endcap
Technical Design Report for the BABAR Detector
8.2 Detector Overview
313
0.5
ε
0.45
B →J/ψKL
0
0.4
0
B →J/ψK
0
0.35
*0
0.3
0.25
0.2
0.15
0.1
0.05
0
0
0.025
0.05
0.075
0.1
0.125
0.15
0.175
0.2
σ(rad)
Signal and background eciencies for reconstructing B 0
functions of the angular resolution in the KL0 direction.
Figure 8-7.
!
J= KL0
as
(a) Reconstructed KL0 momentum from D, ! KL0 + , , . The shaded
area represents the fraction interacting in the IFR. (b) D, + e ective mass distribution.
(c) Missing mass to the reconstructed B 0 ! D,+ .
Figure 8-8.
Technical Design Report for the BABAR Detector
314
Muon and Neutral Hadron Detector
Figure 8-9.
The IFR detector.
consists of hexagonal plates, divided vertically into two parts to allow opening of the detector
and has a central hole for the beam components and the magnetic shields.
The plates are tied together by side steel plates whose thickness (5 cm) reduces the area
available for the active detector. The dead area is concentrated at the joining of the hexagonal
blocks and accounts for a small loss in solid angle, approximately 7%. The endcap gaps are
divided into three by sti eners needed to prevent bending due to the magnetic forces and
to limit the stress due to mechanical and potential seismic loads. A view of the detector is
shown in Figure 8-9.
The iron segmentation is dictated by the physics requirements of identifying low-momentum
muons and detecting KL0 s. The plate thickness is graded from 2 cm for the inner plates to
5 cm for the outer ones, based on simulation studies which will be described below. The
structure consists of 11 plates 2 cm thick followed by six plates 3 cm thick and three plates
Technical Design Report for the BABAR Detector
8.2 Detector Overview
315
5 cm thick. The total thickness is 55 cm and is the result of a compromise between the cost,
which scales with the volume, and the need to reduce hadron punchthrough. Most of the
simulations, however, have been done with a total thickness of 60 cm of iron (this has been
the reference design).
The fraction of B -decay hadrons reaching the IFR that will fail to be absorbed in 55 cm
of iron is 1.3% overall: 0.2% in the backward region, 1.0% in the barrel, and 1.7% in the
forward endcap.
8.2.2 The Active Detector Choice
The IFR detectors will cover a surface of approximately 2500 m2 ; they will be inserted
in the gaps between the iron plates where access and replacement will be very dicult,
perhaps impossible. It is, therefore, necessary that the technique chosen be simple and
reliable as well as low in cost. Two technologies were considered, one using Plastic Streamer
Tubes (PSTs) and the other using Resistive Plate Counters (RPCs). RPC technology was
chosen for its superior exibility and performance and larger active area. It was originally
developed by R. Santonico and collaborators [San81, San88]; the technology is now well past
its development stage, having been successfully used in several experiments [Accel, Cosmi].
A large (600 m2) system is presently in use in the L3 [L3RPC] experiment at CERN, and
much larger systems are now being designed for the LHC experiments [LHC94].
As will be discussed in more detail in Section 8.4.1, the proposed RPC is essentially a gas
gap at atmospheric pressure enclosed between two 2 mm-thick Bakelite (phenolic polymer)
plates with bulk resistivity on the order of 1011 - cm. These electrodes are coated on their
outer faces, with thin graphite layers connected to high voltage (8 kV) and to ground,
respectively. A crossing charged particle produces a quenched spark that produces signals
on external pick-up electrodes.
Active R&D is ongoing in BABAR and other groups to select a non ammable and environmentally safe gas mixture.
Several RPC modules will be joined together to ll each IFR layer; on the two sides, planes
of orthogonal readout strips will be attached for signal readout, so that a space point will be
measured for each hit. The design of the strips is not nalized yet, pending further studies of
the granularity needed. The total number of digital readout channels calculated on the basis
of a 3 cm pitch is 40,000. Using data from the L3 chambers and Monte Carlo simulations
we will investigate whether a coarser granularity will suce and whether better performance
can be obtained by grading the pitch with the distance from the IP. The construction of the
strip planes, i.e., the gluing of pre-made aluminum strips on a PVC support, is considerably
Technical Design Report for the BABAR Detector
316
Muon and Neutral Hadron Detector
faster than that of the chambers, so this item is not on the critical path. We defer nal
decisions until after the completion of more sophisticated Monte Carlo simulations.
Each layer will be inserted into the iron gaps as a single object using appropriate tooling;
readout electronics cards will then be positioned near the gap border where they will be
readily accessible for testing and maintenance; the last 20 cm of each gap will therefore be
made 10 mm wider, to make room for the cards and cables.
8.3 Projected Performance
8.3.1 Muon Identi cation
Muons are positively identi ed if they penetrate all layers of the iron. Non-penetrating
muons (below 1.1{1.5 GeV=c, depending on the incidence angle) can be identi ed from the
ionization losses in the iron. Pion separation is achieved through a combination of range and
hit pattern cuts; its e ectiveness increases with segmentation.
Following the program outlined in the Letter of Intent, several simulation studies have been
performed in order to determine the IFR con guration with optimal performance for cost.
In order to study a large number of segmentation con gurations, as a starting point we
simulated, using the GEANT implementation of the BABAR detector geometry, an IFR
geometry consisting of 61 detector layers 1.2 cm thick alternating with 60 plates of iron
of 1 cm thickness. The ratio of iron to detector thickness was chosen to be the same as in the
Letter of Intent geometry, in which 24 iron planes of 2.5 cm were alternated with 25 detector
layers of 3 cm. Using this method, all con gurations with iron thicknesses that are multiples
of 1 cm can be reproduced by switching o the undesired planes at the analysis level. We
have studied many possible con gurations, including both uniform and graded, as a function
of radial distance from the beam, thicknesses. The minimum plate thickness studied is 2 cm;
plates thinner than 2 cm are not considered practical for mechanical reasons.
The BEGET generator [Wri94] is used to produce B 0B 0 events in which the B 0 can decay
into any channel (branching ratios from [PDG94] are used), and the B 0 becomes invisible
to the detector. In this way, B 0 tagging can be studied, assuming that the CP eigenstate
decay of the B 0 is completely reconstructed.
Figure 8-10 shows the distribution of the last detector layer reached as a function of momentum for muons and pions. Muons with momentum below 1.3 GeV=c stop in the IFR, and
most of the pions interact in the iron.
Technical Design Report for the BABAR Detector
317
Last plane reached
Last plane reached
8.3 Projected Performance
20
15
10
µ
5
0
0
1
2
3
for and .
15
10
π
5
4
p (GeV/c)
Figure 8-10.
20
0
0
1
2
3
4
p (GeV/c)
Distribution of the last detector plane reached as a function of momentum
A criterion which is particularly useful for distinguishing muons from hadrons is based on the
number of planes hit versus the number of planes traversed [Calc94]. A muon track typically
has one hit in each layer, while hadrons can interact strongly in the iron, generating a hadron
shower with several particles producing multiple hits per layer. It is also possible that only
neutrals are present, and this will cause the absence of hits in one or more planes, even for
an ideal detector. In general, for pion tracks the number of planes hit is smaller than the
number of planes traversed, while muon tracks have hits in almost all of the planes they
traverse. The di erence in the behavior of muons and pions in this respect can be seen in
Figure 8-11, which shows the number of planes hit versus number of planes traversed. A cut
on the di erence between these numbers gives a powerful pion rejection.
In a further development of this idea [Lis94], a maximum likelihood function can be built
from a detailed study of the number of hits in each plane for muons and for hadrons, as a
function of momentum. This can be used to determine a combined probability for a track
to be a muon or a hadron. We believe that this technique, still under development, will be
the one actually used to identify muons.
Energy losses in the drift chamber and CsI calorimeter have slightly di erent distributions
for muons and pions at low energy (Figure 8-12), and this di erence can be used to reduce
pion contamination at low momenta.
Technical Design Report for the BABAR Detector
Muon and Neutral Hadron Detector
Planes hit
318
30
µ
25
20
15
10
Planes hit
5
0
30
π
25
20
15
10
5
0
0
5
10
15
20
25
30
Last plane reached
Number of planes hit vs. number of planes spanned. The size of the boxes
is proportional to the log of the number of entries. 100% detector eciency is assumed.
Figure 8-11.
π
µ
10
10
-2
10
-3
10
0
2
4
6
8
10
π
µ
-2
-3
0
0.2
(dE/dx)/p
0.4
0.6
0.8
1
(CsI Eloss)/p
Figure 8-12. Energy loss in the inner tracker (average per hit) and in the CsI calorimeter.
Pions and muons are normalized to equal area.
Technical Design Report for the BABAR Detector
8.3 Projected Performance
319
Several Monte Carlo analyses have been performed with GHEISHA and FLUKA as hadron
shower generators. The variables used to discriminate muons are the following:
Number of planes hit;
Traversed material (Fe cm equivalent);
Number of planes spanned versus number of planes hit;
Number of consecutive missed planes along the track;
Likelihood function for number of hits in each layer;
Tracking residuals; and
Energy deposition in the inner detectors.
Several sets of cuts depending on momentum and polar angle have been used for these
variables, in order to increase eciency while keeping hadron contamination as low as
possible.
Misidenti ed hadrons assigned to the muon sample result in a dilution of the tagging
eciency, and introduce a background to the study of semileptonic decays and other channels
containing muons. Di erent analyses, however, will require di erent tuning of these cuts.
A typical example is the reconstruction of J= ! +, in which we can safely use loose
identi cation criteria because of strong kinematic constraints.
Flavor tagging, on the other hand, requires a clean muon sample because the e ectiveness
of the tag depends strongly on the tagging purity. In this case, the background to the
muon sample from misidenti ed hadrons can be further reduced by studying the kinematic
features of the two classes of tracks [Pia94]. A set of discriminant variables, such as those
shown in Figure 8-13, can be identi ed and a probability density function derived for each
variable. From the probability distributions associated with each discriminant variable, one
can calculate the probability that a track is a muon as follows [Jaf94]:
yeff
Q g
i
= 1,
;
(1 , ) Q g + Q g
i
i
(8.1)
where gi is the probability associated with the ith discriminant variable for the particle to
be a muon, gi is the similar probability for the particle to be a pion, and is the fraction
of pions in the total muon-pion sample. A simple cut in yeff
has been found to correctly
identify 95% of the muons, with a rejection ratio of 15 : 1 for pions.
Con gurations with uniform absorber segmentation have been studied with steel plate thicknesses ranging from 2 cm to 20 cm, and 31 to 4 detector layers, respectively. The eciency
Technical Design Report for the BABAR Detector
320
Muon and Neutral Hadron Detector
x 10 2
(a)
8000
6000
2000
4000
1000
2000
0
0
0.5
1
1.5
Muons
0
2
2.5
GeV/c
x 10
30000
3000
20000
2000
10000
1000
0
0
0.01 0.02 0.03 0.04 0.05
cm
100
50
0
degrees
degrees
(c)
1
1.5
Pions
0
2
2.5
GeV/c
(b)
0
Muons
150
0.5
2
4000
(b)
0
(a)
3000
0.01 0.02 0.03 0.04 0.05
cm
Pions
(c)
150
100
50
0
0.5
1
1.5
Muons
Figure 8-13.
2
2.5
GeV/c
0
0
0.5
1
1.5
Pions
2
2.5
GeV/c
= discriminant variables: (a) center-of-mass momentum, (b) distance
from the beam line, and (c) isolation, i.e., the opening angle from the closest track.
Technical Design Report for the BABAR Detector
321
1
eV
/c
0
700 MeV/c
650 MeV/c
75
0M
0.8
80
0.9
M
eV
/c
2 GeV/c
µ efficiency
8.3 Projected Performance
0.7
1 GeV/c
0.6
0.5
600 MeV/c
0.4
0.3
550 MeV/c
0.2
0.1
0
0
5
10
15
20
25
30
No. of planes
Figure 8-14. Muon eciency as a function of the number of detector planes for xed
momentum value. The simulation is based on FLUKA (total iron thickness of 60 cm).
Passage through a detector plane is simply counted with perfect eciency; no simulation
of RPC information is incorporated. Inner detector cuts are not applied.
for muon identi cation at a given hadron misidenti cation probability (4%) as a function of
the iron segmentation is shown in Figure 8-14 for several momentum ranges.
Reducing the number of planes clearly degrades the eciency and worsens the signal to
background ratio at low momenta and, to a lesser extent, at high momenta. It is also evident
from the gure that increasing the segmentation up to 30 iron plates does not improve the
muon eciency signi cantly for penetrating muons.
These results are obtained with a set of simplifying assumptions that detection eciency
(both geometric and for hits) is 100% and that each GEANT energy deposition gives exactly
one hit in the detector. Further studies to re ne this analysis are underway; it is, however,
safe to assume that for a realistic detector, we will have to rely more on likelihood methods
Technical Design Report for the BABAR Detector
322
Muon and Neutral Hadron Detector
than on imposing cuts in the hit pattern. The loss in performance for con gurations with a
smaller number of detector planes will therefore be signi cant.
8.3.2
Muon Tagging
The muon identi cation eciency and pion contamination obtained in these full Monte
Carlo studies have been parameterized and used in an ASLUND analysis to determine the
e ective tagging eciency as a function of the detector segmentation. The e ective eciency
is obtained after false tags have been taken into account. False tags may arise from:
Muons mislabeled as direct or cascade (100% wrong tag);
Muons coming from a source other than B or D decay (50% wrong tag); and
Hadrons misidenti ed as pions (50% wrong tag).
A sample of 1:2 106 Monte Carlo B decays was generated in order to study the characteristics of the direct vs. cascade muons and to choose discriminant variables. The kinematic
method used to distinguish muons from pions was applied to distinguish direct from cascade
muons. The variables found to provide the highest discriminant power were identi ed as the
center-of-mass momentum, the recoiling invariant mass, and the isolation; their distributions
are shown in Figure 8-15 for both classes of muons. The probabilities that a track belongs
to either class of muons can be expressed as:
direct
yeff
= 1 , (1 ,
cascade
yeff
Q
cascade gicascade
cascade ) gidirect + cascade
= 1 , (1 ,
Q
Q
Q gcascade ;
direct gidirect
direct ) gicascade + direct
Q
i
Q gdirect ;
i
(8.2)
(8.3)
where the gi are the probabilities associated with the ith discriminant variable belonging to
either class, and direct and cascade are the fractions of direct and cascade muons, respectively,
in the total muon sample. A simple cut in the yeff variables has been found to be very
e ective for assigning muons to the correct class.
The e ective tagging eciency reaches almost 9% for a con guration with uniform 2 cm
segmentation and decreases as the steel plate thickness increases; using the graded 20-plane
con guration that we propose, the gure is 8.7%; while for a 16-plane con guration, the
eciency drops below 8%.
Technical Design Report for the BABAR Detector
8.3 Projected Performance
323
(a)
1500
3000
1000
2000
500
1000
0
0
0.5
1
1.5
Direct µ
0
2
2.5
GeV/c
3000
(b)
(a)
0
0.5
1
1.5
Cascade µ
2
2.5
GeV/c
(b)
600
2000
400
1000
0
1
2
3
degrees
Direct µ
4
(c)
150
100
50
0
0
5
GeV
0
1
2
3
Cascade µ
degrees
0
200
4
5
GeV
(c)
150
100
50
0
0.5
1
1.5
Direct µ
2
2.5
GeV/c
0
0
0.5
1
1.5
Cascade µ
2
2.5
GeV/c
Figure 8-15. Direct and cascade discriminant variables: (a) center of mass momentum,
(b) recoiling invariant mass, and (c) isolation, i.e., the opening angle from the closest track.
Technical Design Report for the BABAR Detector
324
8.3.3
Muon and Neutral Hadron Detector
KL0
Detection
The two main issues for reconstructing the decay B 0 ! J= KL0 are the identi cation of
IFR hits belonging to the KL0 and the determination of the KL0 direction from those hits.
Since the KL0 s are detected via the charged secondaries from their interaction or decay, a full
Monte Carlo simulation that includes hadronic interactions is required to produce a realistic
hit pattern. The detector response has been simulated with BBSIM. The FLUKA hadronic
simulation package has been used because the KL0 cross sections below 1 GeV=c in GHEISHA
do not agree with the experimental data [Say68]. Full detection eciency in both barrel and
endcaps was assumed, and hit coordinates were generated assuming a 3 cm readout pitch.
A study of the e ect of iron segmentation on the detection eciency has been performed
by generating single particle KL0 events [Bal94]. Figure 8-16 shows results for the KL0
detection eciency as a function of momentum for four di erent con gurations of iron plate
thicknesses. Here, the KL0 signature is taken to be at least four IFR plane hits. Detection
eciency for KL0 s improves with segmentation; thin plates are required, especially in the
inner part of the detector, where most KL0s start or continue to interact. However, too
coarse a segmentation in the outer part of the detector also decreases the eciency for KL0 s
that start interacting in the IFR, for which the angular resolution is much better. Figure 8-16
shows that one can obtain a sizeable improvement in the eciency by adopting a nonuniform
iron segmentation; a 21-layer nonuniform con guration has much better performance than
a 20-layer uniform segmentation and is almost as good as a 30-layer one.
A study of the KL0 identi cation from the IFR hits, employing pattern recognition, has been
made [Wri94b]. A sample of B B events, in which one B decays to J= KL0 and the other to
any mode, has been generated. The KL0 identi cation process consists of clustering the hits
and eliminating clusters that are near the extrapolation of charged tracks from the central
drift chamber. The remaining clusters are due mostly to neutral particles (KL0 s, neutrons,
and photons from 0 s) and charged hadrons. The composition of the clustered hits in the
IFR is given in Table 8-1. About 80% of KL0s have at least one hit in the IFR. Requiring four
or more hits reduces the fraction to 68%. Neutral clusters that happen to be near a charged
track are sometimes eliminated. Currently, the best algorithm reduces the KL0 eciency from
68% to 56% due to these accidental overlaps.
While there is a correlation between the number of detected IFR hits and the momentum
of the KL0 (Figure 8-18), the distribution is quite wide and depends on whether the KL0
interacted before reaching the IFR. This KL0 momentum measurement is too crude to be
suitable for reconstructing and cutting on the B invariant mass.
The e ectiveness of background rejection cuts depends on the angular resolution of the
measured KL0 direction. For the KL0 s that interact or decay before reaching the IFR, which
make up 70% of the sample, the angular resolution is worsened, since the detectable
Technical Design Report for the BABAR Detector
8.3 Projected Performance
325
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
2.75
3
p (GeV/c)
Eciency for detecting KL0 s as a function of momentum. The di erent sets
of points represent di erent segmentations of the ux return iron. A solution with 21 active
planes with graded separation is compared.
Figure 8-16.
Sample Composition
Criteria
> 3 IFR Hits
After Cuts
CP Other Other
0
KL e . KL0 KL0 Neutral Charged
68% 21%
56% 77%
4%
11%
9%
4%
66%
8%
Composition of B 0 ! J= KL0 events after pattern recognition. The second
line indicates the composition after a cut that removes clusters near charged tracks.
Table 8-1.
Technical Design Report for the BABAR Detector
326
Muon and Neutral Hadron Detector
600
600
(b)
(a)
500
500
400
400
300
300
200
200
100
100
0
-40
-20
0
20
∆θ (degrees)
0
40
-40
-20
0
20
∆θ (degrees)
40
(a) Angular resolution for KL0 from B 0 ! J= KL0 determined by averaging
the unsmeared IFR hit positions of KL0 s that do not interact until reaching the IFR.
(b) Angular resolution with (black) and without (open) the inclusion of a detector layer
inside the coil for the KL0 s that interacted before the IFR.
Figure 8-17.
35
35
30
0
KL interacted in IFR
Number of Hits
Number of Hits
30
25
20
15
10
5
0
KL interacted before IFR
25
20
15
10
5
0
1
2
3
0
p (GeV/c)
1
2
3
p (GeV/c)
Average number of IFR hits as a function of momentum for KL0 in simulated
(4S ) ! (B ! J= KL0 )(B 0 ! X ) events. The error bars represent one standard deviation
of the spread in the data.
Figure 8-18.
0
Technical Design Report for the BABAR Detector
8.3 Projected Performance
327
400
350
300
250
200
150
100
50
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
prest
1
(GeV/c)
Figure 8-19. Reconstructed momentum of the B meson in the (4S ) rest frame (prest )
for B 0 ! J= KL0 . The narrower distribution includes only the smearing e ects due to the
beam energy spread and nite momentum resolution. The other two histograms show the
widening due to the KL0 angular resolution for the cases in which a detector layer inside the
coil is included (dashed) or not (open).
secondaries must travel an appreciable distance and pass through additional material before
reaching an active IFR detector. For example, the angular resolution for KL0 s that interact
before reaching the IFR is about a factor of 2 worse than for those that do not interact before
reaching the IFR, as shown in Figure 8-17. However, the inclusion of a detector layer inside
the coil signi cantly reduces this smearing e ect, as shown in Figure 8-17(b). Therefore,
such a detector has been included in our baseline design.
Technical Design Report for the BABAR Detector
328
Muon and Neutral Hadron Detector
Aluminum
Foam
X strips
H.V.
Bakelite
Insulator
Graphite
2 mm
Gas
2 mm
Bakelite
2 mm
Graphite
Insulator
PVC spacers
Y strips
Aluminum
Foam
1 mm
1 cm
Figure 8-20.
0
Schematic representation of RPC components.
8.4 Detector Design and R&D
8.4.1 Chamber Construction and Assembly
RPCs, shown schematically in Figure 8-20, have been chosen on the basis of their high intrinsic and geometrical eciency, low cost, robustness, and exibility in segmentation. We are
planning to use the technology [San81, San88] developed by R. Santonico et al., which is wellproven in various experiments, accelerators [Accel], and for cosmic ray physics [Cosmi], and
whose use is foreseen in the rst-level muon trigger systems designed for LHC experiments
at CERN [LHC94]. The knowledge gained in years of R&D has already been transferred to
industry, and the production capabilities are now well-matched to the large detector area
required by BABAR.
As shown in Figure 8-20, the electrode plates are made of 2 mm-thick Bakelite (phenolic
polymer) with a volume resistivity of 1011{1012 -cm, painted on the external surfaces
with graphite of high surface resistivity (100 k /square) and covered by two 300 m PVC
insulating lms. The two graphite layers are connected to high voltage (8 kV) and ground,
respectively. Gap surfaces are treated with linseed oil, which strongly enhances RPC performance (low noise, high eciency). Planarity of the two electrodes is assured by PVC spacers
(0.8 cm2 ) located on a 10 cm-square grid in the sensitive volume. The active volume is lled
with an argon-based gas mixture at atmospheric pressure.
Technical Design Report for the BABAR Detector
8.4 Detector Design and R&D
329
A charged particle crossing the chamber produces a quenched spark which produces signals
on external pickup electrodes. There is active R&D in BABAR and in other groups to select a
non ammable and environmentally safe gas mixture. The discharge generated by an ionizing
particle is quenched by:
UV photon absorption by hydrocarbon molecules;
Capture of outer electrons by electronegative additives, reducing the size of the dis-
charge; and
Switching o the electric eld around the discharge point.
In fact, the duration of the discharge is on the order of 10 ns, while the relaxation time
of the resistive electrodes is t = " ' 10,2 s, so during the discharge the electrode plates
behave like insulators, and a small area around the impact point shows a deadtime on
the order of t.
Induced pulses are collected on two pickup planes, 29 mm wide and separated by 2 mm, made
of aluminum strips glued onto plastic foils located on either side of the chamber. Strips run
in two orthogonal directions to provide two-dimensional information. The induced charge is
on the order of 100 pC, and the pulse has a rise time of 2 ns and a duration of 10 ns. The
low event rate in the IFR, the small size (3 mm2), and the short deadtime (10 ms) of the
area a ected by each quenched spark mean that there is no eciency problem. Stochastic
noise depends on Bakelite plate resistivity and on the operating voltage; a value well below
1 kHz/m2 can easily be reached.
In the upper part of Figure 8-21, the signal rate is shown as a function of high voltage for
a standard mixture (Ar 59%, Isobutane 38%, Freon 13B1 3%), and with Freon 14 (CF4)
instead of Freon 13B1. The Bakelite bulk resistivity is 1011 , cm in this test, and a clean
cosmic ray plateau with negligible background can be seen with Freon 13B1. In the case of
Freon 14, the plateau starts 300 V earlier, but the single rate goes up with high voltage.
In any case, the singles rate is not a problem for the IFR. Assuming, for example, a singles
rate of 1 kHz/ m2 due only to noise, we would have three random hits in a 1 s time window
around the event time for the full 2500 m2IFR. If the fast timing capability of the RPC is
used, by means of TDC measurements, a factor of 100 reduction can be achieved.
The eciency plateaus for the two gas mixtures cited above are shown in the lower part
of Figure 8-21. With the standard mixture, a timing resolution of ' 1 ns is reached. As
reported in Section 8.5, work is presently under way at LLNL and Naples to optimize RPC
performance, with the goal of using non ammable and environmentally acceptable gases.
Taking into account the dead areas of frames and spacers, the RPC eciency has been
measured to be higher than 96%; this value can be brought closer to 100% if, for some
Technical Design Report for the BABAR Detector
Muon and Neutral Hadron Detector
Rate (Hz)
330
1400
1200
59% Ar + 38% Isobutane % + 3% Freon 13B1
59% Ar + 38% Isobutane % + 3% CF4
1000
800
600
400
Efficiency (%)
200
0
100
90
80
70
60
50
40
30
20
10
0
5000
5500
6000
6500
7000
7500
8000
8500
9000
H.V. (V)
Figure 8-21. Singles counting rate (cosmic rays and random noise) and eciency plateaus
for the two gas mixtures 59% Ar, 38% Isobutane, and either 3% Freon 13B1 or 3% CF4 .
detector layer, a two-gap approach is chosen, as has been done in the L3 [L3RPC] experiment
at LEP, with the two spacer grids staggered by 5 cm.
8.4.2 System Layout
The production plant [Plant] that we plan to use, following the excellent experience with
the L3 RPCs, has developed construction tooling which allows a maximum chamber size of
1 m by 2.2 m. Within these limits, modules can be built in practically any shape. Although
it is possible to modify the tooling in order to increase the maximum width and/or length,
we have studied a layout of the chambers in the gaps within the present constraints. We
will carefully consider the cost e ectiveness of decreasing the number of modules versus
retooling. Most of the cost will be based on the number of modules built, not on the total
surface covered; on the other hand, major modi cations to the present tooling will increase
the unit cost of the chambers.
Technical Design Report for the BABAR Detector
y (cm)
8.4 Detector Design and R&D
331
300
280
260
240
220
200
180
-200
-100
0
100
200
x (cm)
Figure 8-22. Detailed layout of a barrel sextant. The angular coverage is in excess of
98% and is obtained with chambers of only ve di erent widths. The extra space at each
side will be used by steel plates that will hold the iron together.
Barrel
In the six barrel sections, all the layers have the same length. The iron plates are 375 cm
long, so all modules are 185 cm long and are joined in pairs; 2.5 cm is left at each end to allow
for connectors, cables, and gas piping. The width of the iron gaps varies from 188 cm for the
innermost layer to 323 cm for the outermost. The width of the modules therefore varies in
order to ensure maximum geometric coverage. If the maximum width of each module is xed
at the 100 cm that the present production facility allows, it is necessary to use two adjacent
modules in the inner layers and four in the outer ones. The total number of modules in each
gap is therefore four (Layers 1 to 3), six (Layers 4 to 19), or eight (Layers 20 to 21), given
the split in the z direction.
For each of the six barrel sections, 124 modules are needed, for a total of 744. Given such
a large number of modules, it is important to nd a layout which uses a minimum number
of di erent sizes. This will translate into a cost and time saving not only in the production
process but also in the subsequent steps. All of the quality control, testing, and assembly
procedures will be greatly simpli ed if only a few basic shapes are needed. On the other
hand, it is important to maximize the geometric coverage; as much as possible of the gap
surface in each layer has to be covered.
Technical Design Report for the BABAR Detector
332
Muon and Neutral Hadron Detector
Layer
#
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Radius
(cm)
175
198
203
208
213
218
223
228
233
238
243
248
256
262
268
274
280
286
294
302
310
Gap Chan. 1 Chan. 2 Chan. 3 Chan. 4 Total
width
(cm)
(cm)
( cm) ( cm) Width
188.4
100
86
186
194.1
96
96
192
199.9
100
96
196
205.7
68
68
66
202
211.5
74
68
66
208
217.2
74
74
66
214
223.0
86
68
66
220
228.0
86
74
66
226
234.6
100
66
66
232
240.3
96
74
68
238
246.1
96
74
74
244
251.9
100
74
74
248
254.0
100
86
66
252
260.9
96
96
66
258
275.0
100
86
86
272
281.9
100
96
86
278
288.8
96
100
86
286
295.8
100
96
96
292
305.0
100
100
100
300
314.2
96
74
74
68
312
316.3
96
86
66
66
314
Table 8-2.
Active module dimensions for barrel coverage.
For this optimization, we have relied on a computer calculation in which the gap widths
are introduced with the constraints that each should be lled at least to the 98% level with
2 cm free in order to allow some clearance for the installation. With these criteria and
a maximum module width of 100 cm, solutions are found starting with a minimum of six
di erent sizes. One of these solutions is shown in Figure 8-22, and the relevant parameters
listed in Table 8-2. In this layout, there are 744 barrel RPC modules in six di erent widths
(Table 8-3).
The iron is segmented into 20 planes of graded thickness, from 2 to 5 cm as discussed in the
previous subsections; the total iron thickness is 55 cm, and air gap between plates is 3 cm.
Each barrel section is split into two parts to keep the weight within the crane capability
available in the IR. Taking into account mechanical tolerances for joining the the two parts,
the total IFR thickness is 120 cm. The total RPC area in the barrel is 1320 m2. In each
layer, four, six, or eight modules are joined, and the strip planes glued on before insertion in
Technical Design Report for the BABAR Detector
8.4 Detector Design and R&D
333
Width ( cm)
Number of Modules
100
144
96
156
86
108
74
132
68
72
66
132
Number of RPCs:
744
Table 8-3.
Number of di erent modules needed for barrel coverage.
the iron; if the strip pitch is 3 cm, there will be a total of approximately 12,000 longitudinal
and 16,000 transverse strips.
Endcaps
The endcap geometry presents a bigger challenge for de ning an optimum RPC module
layout, since the chambers have irregular shapes, and the presence of sti eners between
plates imposes severe constraints on chamber size and shape. The plates all have the same
dimensions apart from the hole around the beam line, which di ers for each of the forward
endcap gaps.
Each endcap is divided vertically into halves to allow access to the inner detector without
interfering with the machine components; Figure 8-23 shows the typical layout of a half layer:
the maximum width and length of the modules are 100 and 220 cm, respectively, as for the
barrel layout. This scheme assumes that two horizontal sti eners 5 cm thick are sucient to
withstand the magnetic forces of a eld up to 1.5 T, as discussed in Chapter 9.
All the backward endcap plates are of the same size; the inner plates of the forward endcap
have a slightly smaller hole. The position of the horizontal sti eners is the same for all
plates, at 95 cm above and below the beam line, so that the layout of the chambers is the
same for all plates. A few specially shaped modules of slightly di erent sizes have to be used
in the innermost layers of the forward endcap to obtain maximum coverage.
The total number of endcap chambers is 720: 160 will be rectangular, 95 cm by 180 cm sides;
160 smaller trapezoids, 320 larger trapezoids, 80 special shapes, in three sizes. The area of
plate is approximately 27 m2, for a total of 1100 m2 for the total RPC area required for the
two endcaps.
Technical Design Report for the BABAR Detector
Muon and Neutral Hadron Detector
y (cm)
334
300
B
B
200
A
A
D
100
D
C
0
C
D
-100
D
D
A
-200
B
-300
0
100
200
300
400
500
600
x (cm)
Figure 8-23.
Layout of the RPC modules in the endcaps.
Inner Cylindrical Layer
A double-gap, resistive-plastic RPC, as shown in Figure 8-24, will be used to accommodate
the largest number of coordinate planes in the 3 cm alloted space inside the solenoid. Each
double-gap RPC has a single high-voltage plane made of aluminum foil laminated between
two pieces of resistive plastic. The ground planes on either side of each high voltage plane
have the readout strips attached to their back sides. The resulting four readout planes
encode z, , u, and v coordinates with 2 cm resolution. The thickness of the entire sandwich
is 2.5 cm, leaving 0.5 cm of insertion tolerance.
The inner RPC layers will be laminated together on a cylindrical mold of the nal radius.
The nal laminate will be self supporting in its cylindrical shape. The bottom of the cylinder
along the beam axis will have an open slit to accommodate the rail that supports the inner
detectors. Inner RPC strips that intersect this opening (z, u, and v strips) will be brought
out to the ends of the cylinder on ribbon cable. The inner RPC front-end electronics are
located on the end of the cylinder.
Technical Design Report for the BABAR Detector
8.5 Gas System
335
Figure 8-24.
Inner cylindrical RPC design.
Although double gap RPCs have been successfully operated, cylindrical RPCs have never
been constructed. Therefore, we will perform limited R&D on the cylindrical RPC prior
to construction. This R&D will involve constructing and operating a small planar version
of the inner RPC multi-gap con guration, followed by the construction and operation of a
partial segment of the cylindrical inner RPC.
8.5 Gas System
8.5.1 Gas Composition and Flow Rates
The traditional RPC gas mix of 59% argon, 37% isobutane, and 4% Freon 13B1 is undesirable
in two respects. Isobutane is ammable and Freon 13B1, an ozone-destroying substance,
may prove dicult to obtain over the lifetime of BABAR. We are currently performing R&D
at LLNL and Naples on non ammable RPC gas mixes that use quenching agents besides
F13B1. Both sites have precision gas mixers to prepare candidate mixtures that are fed
to multiple RPCs, which are tested for eciency, noise, cluster size, and time resolution
with cosmic ray triggers. Preliminary results are encouraging. We have obtained ecient
RPC operation with a mixture of 37% isobutane and no freon. We have also demonstrated
ecient RPC operation using much smaller concentrations of isobutane (less than 8%) with
F13B1. Preliminary tests show that the environmentally acceptable Freons 116 and 23, as
well as SF6, have promising quenching properties. More evaluation of these and other gases
is planned. Therefore, we plan to run the IFR RPCs with less than 10% isobutane and with
4{10% of the benign Freons 116, 23, 14, and SF6. The exact proportions of these gases
must be experimentally determined to provide the highest eciencies and maximal position
Technical Design Report for the BABAR Detector
336
Muon and Neutral Hadron Detector
resolution, at the possible expense of time resolution. Typical steady state operation of the
IFR will require about two volume changes of gas per week, corresponding to a ow rate
of 2 l/min, while startup operations are expected to require about one volume change per
day corresponding to a ow rate of 6 l/min.
8.5.2
Mixer
The three gases will be stored outside and transported to the mixing station through heated
metal tubing. The three gases ow continuously through the mixer, passing through a lter,
relief valve, pneumatic shut-o valve, regulator, and ow controller. The ow controller
settings determine the mixed gas composition. A separate ow meter monitors the resulting
ow rate for each gas.
8.5.3
Distribution
The mixed gas will be transported between the mixing station and the detector through a
one-inch diameter metal tubing. Supply manifolds on the detector will distribute the gas in
parallel to 14 separate groups of chambers (six barrel sextants plus four quadrants on each
endcap). Within each group, chambers in a given layer will be connected in series, while the
di erent layers will be connected in parallel. Matching return manifolds will collect the gas
for venting at a remote location. A blower will be used to reduce chamber over-pressure due
to the back pressure of the vent line.
8.6
Front-End Electronics and High Voltage
The front-end electronics for the IFR will be located directly in the iron gaps of the ux
return as close as possible to the detectors; the relatively high number of channels foreseen
(40,000) requires that a certain amount of on-board processing be done in order to minimize
the size of the outgoing cable plant. The electronic requirements for the detector are quite
easy to ful ll|no pulse height data acquisition is foreseen. A single bit per struck strip has
to be recorded and sent to the data acquisition system. The excellent timing properties of
RPCs could also be exploited, and given the very low occupancy this subsystem is expected
to have, a relatively small number of TDCs (one per layer) would be sucient. The frontend electronics boards will include a very simple discriminator. The pulses originating from
the RPC are quite large (200 mV) and do not have a large amplitude jitter, so a single
transistor would be enough to standardize them and convert to TTL levels. Connection of
Technical Design Report for the BABAR Detector
8.7 Final Assembly, Installation, and Monitoring
337
the strips to the boards will use twisted and at cables, at least one set of strips (either
longitudinal or transverse) will be connected to the electronics through a very short cable
path so that minimal deterioration of the timing characteristics will occur. High voltage will
be individually fed to each detector (a typical con guration in the barrel will include between
four and eight detectors per plane) through series resistors (20 M ); small additional series
resistors will be used as fuses so malfunctioning modules can be remotely disconnected from
the HV supplies. Provision will be made for measuring and recording individual currents
drawn from the detectors.
8.7 Final Assembly, Installation, and Monitoring
The major components of the IFR detectors (the Bakelite modules, readout strips, and the
front-end electronics boards) will be manufactured at various institutions and shipped to
SLAC for nal assembly. Approximately 750 Bakelite modules each for the barrel and
the endcaps will be manufactured in the RPC production plant in Italy. As they are
manufactured, the modules will be shipped to Naples, Genoa, and Frascati for acceptance
testing with cosmic ray muons with the accepted modules shipped to SLAC. The shipments
will be staggered over an 18-month period with approximately 140 modules being shipped
every three months. At SLAC, the modules will be processed in batches of 12 per week.
Each module will be out tted with a single, wide (20 cm) readout strip and inserted into
a 12-chamber cosmic ray tower where they will be tested for high voltage plateau and gas
integrity for approximately one day. After tests, the readout strip will be removed, and
the modules corresponding to a given IFR layer will be placed side by side on a pallet, then
joined at the seams with plastic lm to form a single IFR detector plane. Full-length readout
strips will be laminated across the modules of the detector plane. The front-end electronics
cards will be mounted on the edge of the detector, which will be accessible from the open
end of the barrel. For planes, the 4 m-long readout strips will be directly soldered to
the front-end electronics cards added to the ends. On the z planes, the z strips terminate at
the edge of the detector plane that is buried inside the IFR, so ribbon cable will be attached
to the ends of the strips and folded at right angles to bring the signals out to the front-end
electronics cards. The completed detector plane, on its pallet, will be inserted into a cosmic
ray tower for detailed tests of the eciency of each channel, using components of the BABAR
data acquisition system. After nal testing, the completed detector plane will be transported
to the IR on its pallet. Each plane will be hoisted into position and slid o the pallet into
the appropriate gap between the IFR plates. Spacers, wedged into the gap, will prevent the
detectors from sliding within the gap. During operation, the eciency of the IFR detector
planes will be monitored by reconstructing muon pairs.
Technical Design Report for the BABAR Detector
338
REFERENCES
References
[Accel]
A. Antonelli et al. (FENICE Collaboration),
A337, 34
(1993);
E. Petrolo et al. (WA 92 Collaboration),
A315, 45 (1992);
L. Antonazzi et al. (E 771 Collaboration),
A315, 92 (1992);
C. Bacci et al. (RD 5 Collaboration),
A315, 102 (1992).
R. Baldini-Ferroli, R. de Sangro, A. Palano, A. Zallo, \= Discrimination with
the Instrumented Flux Return," A AR
# 198 (1994).
A. Calcaterra, A AR
# 144 (1994).
F. D'Aquino et al. (MINI Collaboration),
A324, 330 (1993).
M. Ambrosio et al. (Cover Plastex Collaboration),
A344,
350 (1994).
D.E. Ja e et al., \Treatment of Weighted Events in a Likelihood Analysis of Bs
oscillations or CP Violations," A AR
# 132 (1994).
L. Lista and S. Mele, \Study of Iron Segmentation Optimization of the I.F.R.
Detector for separation," A AR
# 194 (1994).
ATLAS Collaboration,
A340, 466 (1994);
CMS Collaboration,
A334, 98 (1994).
A. Aloisio et al., Proceedings of the Sixth Pisa Meeting on Advanced Detectors,
Isola d'Elba, Italy (1994); submitted to
.
General Tecnica, Colli (FR) Italy.
M.G. Pia, \B Tagging with Muons," A AR
# 192 (1994).
L. Montanet et al. (Particle Data Group),
D50 (1994).
R. Santonico and R. Cardarelli,
A187, 377 (1981).
R. Santonico and R. Cardarelli,
A263, 20 (1988).
G.A. Sayer and E.F. Beall,
169, 1045 (1968).
D.M. Wright, \BEGET: The B -Factory Event Generator Version 21,"
A AR
# 149 (1994).
Nucl. Instr. Methods
Nucl. Instr. Methods
Nucl. Instr. Methods
Nucl. Instr. Methods
[Bal94]
B B
[Calc94]
[Cosmi]
B B
Note
Note
Nucl. Instr. Methods
Nucl. Instr. Methods
[Jaf94]
B B
[Lis94]
B B
[LHC94]
Note
Note
Nucl. Instr. Methods
Nucl. Instr. Methods
[L3RPC]
Nucl. Instr. Methods
[Plant]
[Pia94]
[PDG94]
[San81]
[San88]
[Say68]
[Wri94]
B B
Note
Phys. Rev.
Nucl. Instr. Methods
Nucl. Instr. Methods
Phys. Rev.
B B
Note
Technical Design Report for the
A AR
B B
Detector
REFERENCES
339
[Wri94b] D.M. Wright, \KL0 identi cation in B 0
(1994).
! J=
KL0 Decays," BABAR Note # 201
Technical Design Report for the BABAR Detector
340
Technical Design Report for the BABAR Detector
REFERENCES
9
Magnet Coil and Flux Return
9.1 Physics Requirements and Performance Goals
T
he BABAR magnet is a thin, 1.5 T superconducting solenoid within a hexagonal ux
return, as shown in Figure 9-1. Detector performance criteria and geometry considerations drive the design of the solenoid and the ux return. The magnitude and uniformity
speci cations for the magnetic eld are derived from drift chamber track nding and momentum resolution requirements. Studies of B 0 ! +, suggest that a magnetic eld of 1.5 T is
necessary to achieve a mass resolution of 21 MeV=c2 . The combined thickness of the vertex
detector, drift chamber, particle identi cation system, electromagnetic calorimeter, and
appropriate clearances set the solenoid inner diameter. Solenoid length is also determined by
the length of the nested subsystems. The solenoid thickness limits the momentum threshold
for detecting muons and the eciency of KL0 detection within the instrumented ux return.
The segmented geometry of the ux return allows tracking of muons and provides for
detection of KL0s with adequate angular resolution. The total thickness of the steel layers in
the barrel and end door is determined both by the minimum steel required to avoid magnetic
saturation and by the need for sucient thickness to ensure that most of the pions interact
in the steel. The minimal steel thickness to prevent pion punch-through is 55 cm (3.6 ).
Plate segmentation and thicknesses are speci ed both for ecient identi cation of KL0 s and
for distinguishing muons from pions based on range measurements. For more information
on the meson detection system refer to Chapter 8.
The overall thickness of the ux return is the sum of both the steel thickness and the number
and thickness of the RPC layers. Cost is also a factor in determining the number of RPC
layers. Separation and movement of the end doors are constrained by beam line components
and by the need to provide ready access to detector subsystems.
The physics performance and operational requirements for the solenoid and ux return
(Table 9-1) are similar to those of many operating detector magnets (Table 9-2).
int
342
Magnet Coil and Flux Return
Figure 9-1.
9.2
Geometry of the solenoid within the ux return.
Overview
The design of the superconducting solenoid for the BABAR detector is conservative and
within the state of the art [Des85, And82, Coils] for detector magnets. It is based on
the experience gained over the past 15 years with thin superconducting solenoids. Although
speci cally tailored to meet the requirements of BABAR (Table 9-1), this design is similar
to many operating detector magnets. A common feature of all these magnets is the use of
aluminum-stabilized conductors that are indirectly cooled by liquid helium pipes connected
to an aluminum alloy support structure. This technique was developed for CELLO, the
rst thin solenoid, and has been improved in subsequent designs. Table 9-2 shows the
main characteristics of some of these solenoids compared to the BABAR design. All of these
designs used a Rutherford-type cable made of NbTi superconductor encased in an aluminum
stabilizer that allows for adequate quench protection.
The BABAR detector schedule identi es the magnet as a critical procurement item. The threeand-one-half-year-long critical path is formed by: solenoid design and procurement; assembly
with the ux return; veri cation testing and mapping; and detector subsystem installation
and commissioning. While these task durations may be shortened, such reductions expose
the project to higher budget and schedule risks. The solenoid design and fabrication duration
Technical Design Report for the BABAR Detector
9.2 Overview
343
Solenoid Requirements
Central induction
Field uniformity in the tracking region
Nuclear interaction length
Cryostat inner radius
Cryostat outer radius
Minimize thermal cycling
Comply with ES&H requirements
1.5 T
2%
0.25{0.4
1400 mm
1730 mm
Flux Return Requirements
Provide an external ux path for a 1.5 T eld
Provide 3 cm spacing between the steel plates for IFR instrumentation
Provide the gravitational and seismic load path for the barrel detector
components to the concrete foundation
Fit in IR-2 (3.5 m radial distance from beam axis to the concrete oor)
Movable end doors to allow access inside the barrel
Comply with ES&H requirements
Table 9-1.
int
p
p
p
p
p
p
p
p
Physics performance and operational requirements.
of 24{26 months requires the contract to be awarded in the fall of 1995 to meet the overall
detector schedule.
The magnet cryostat will be designed, fabricated, and inspected according to the intent of
the ASME Boiler and Pressure-Vessel Code, Section VIII, Division 2 [ASME94], but will not
be code-stamped. The magnet will be subject to seismic design requirements described in the
SLAC Seismic Design Manual for mechanical systems [SDM91]. The magnet design will also
follow the requirements outlined in the Safety Analysis Document (SAD), which will address
ES&H issues. For steel structures, the allowable design stresses follow the standard guidelines
as speci ed in the AISC Manual of Steel Construction, 9th edition. Bolted connections and
fasteners will conform to their recommended torques and allowable stresses depending on
Technical Design Report for the BABAR Detector
344
Magnet Coil and Flux Return
CDF
ZEUS CLEO-II ALEPH BABAR
Location
FNAL DESY Cornell CERN SLAC
Manufacturer
Hitachi Ansaldo Oxford Saclay
?
Year Completed
1984
1988
1987
1986
1997
Central Field (T)
1.5
1.8
1.5
1.5
1.5
Inner Bore (m)
2.86
1.85
2.88
4.96
2.80
Length (m)
5
2.5
3.48
7
3.46
Stored Energy (MJ)
30
12.5
25
137
25
Current (A)
5000
5000
3300
5000
7110
Total Weight (t)
11
2.5
7.0
60
6.5
Radiation Length
0.85
0.9
n/a
1.6
1.4 max
Conductor Dimensions (mm) 3.8920 4.315, 516 3.635 3.230
5.5615
2
Current Density (A/mm )
64
78
42
40
74
Table 9-2.
Comparison of solenoids similar to BABAR.
the connection. The ux return is fabricated from ASTM A36 structural steel plates or a
material with similar mechanical and magnetic properties.
9.2.1
Description of Key Interfaces
The radial distance between the outer
diameter of the solenoid and the inner surface of the barrel ux return is 50 mm. The solenoid
weight and magnetic forces are transmitted to the inner and outer hexagonal rings of the
ux return as shown in Figure 9-2. This attachment, located at the vertical mid-plane of
the detector, also provides the load path of the inner detector components to the barrel ux
return.
The backward end doors provide a chase for the cryostat chimney. The chase is 400 mm
wide and extends 400 mm into the backward end doors.
Superconducting Solenoid and Flux Return.
Both ends of the barrel ux return have a 60% solid steel contact
area at the interface with the end doors. This area is composed of the 150 mm-thick inner
ring support plates, 150 mm-thick joint braces, and 150 mm-thick steel gap ller plates. The
remaining 40% open area on the barrel ends is reserved for cabling and utilities from the
inner detector components. The end doors are attached to the barrel with tie plates that
are bolted to the end door structure and to the barrel.
Barrel and End Door.
Technical Design Report for the BABAR Detector
9.2 Overview
Figure 9-2.
ux return.
345
Superconducting solenoid support bracket attached to the mid-plane of the
Particle Identi cation System. A vertical slot between the backward end doors permits
the support structure for the DIRC to penetrate into the detector. This structure also
supports the backward beam magnets Q2, Q4, Q5, and the backward ux return eld shaping
plug located physically inside the DIRC. The nal design details of the DIRC and the
mounting of the backward beam magnets are not yet fully resolved.
Forward Q2 Beam Magnet Shielding. The forward beam magnet Q2 is physically
located within the forward end doors. A specially designed, three-piece, conical magnetic
shield plug is mounted to the end doors to isolate Q2 from the detector magnet. The
shielding plugs are split along their vertical centerline, and each half is attached to a halfround mounting ange that is bolted to the face of each forward end door.
Inner RPC Detector and Solenoid. There is an RPC detector located between the
calorimeter and the solenoid. This RPC detector attaches directly to the inner diameter
of the solenoid cryostat with a 20 mm clearance gap between the RPC detector and the
calorimeter.
Technical Design Report for the BABAR Detector
346
Magnet Coil and Flux Return
Movable End Door Skids and the Beam Line. The end doors are mounted on skids
equipped with rollers so that they can be moved away from the barrel for maintenance access.
The end door skids move on tracks installed in the oor of IR-2. The end doors clear the
beam line magnets, vacuum pumps, magnet stands, and other beam line equipment during
door opening.
External Platforms, Stairways, and Walkways. The external platforms necessary to
install and service electronic racks and cryogenic equipment are supported from the ux
return. The requirements of these components have not yet been determined.
9.3 Summary of Projected Magnet Performance
9.3.1 Central Field Magnitude and Coil Performance
The magnetic eld of 1.5 T is obtained by energizing the solenoid with a constant current
of 7110 A. The conductor is operated at 45% of the critical current, with a peak eld in the
conductor of 2.5 T. This gives a large safety margin.
Magnetic uniformity is achieved by doubling the current density in regions at both ends
of the solenoid. This is done by adding more aluminum stabilizer to the central region
conductor, which reduces the current density there. Figure 9-3 shows the eld uniformity in
the central region. The areas in which the eld nonuniformity is greater than 2% are small
and are located in regions in which they do not a ect the performance of the drift chamber.
In addition, once the solenoid parameters are optimized, the corners of the drift chamber
should also be within 2% of 1.5 T.
The radial pressure on the conductor during operation is 1.5 MPa in the high current-density
regions and 0.78 MPa in the central region of the conductor. An aluminum support cylinder
surrounds the coiled conductor to react against these radial pressures and keep the conductor
from yielding.
The integrated axial force on the winding is 3.5 MN. The conductor winding and support
cylinder are mechanically coupled by an epoxy bond. This epoxy bond allows some of
the axial load to be transmitted in shear to the outer aluminum cylinder, which keeps the
conductor from yielding. There is an axial 18 kN de-centering force applied to the conductor
winding due to an asymmetry in the iron, mainly due to the di erences in the forward and
backward Q2 shielding.
Technical Design Report for the BABAR Detector
9.3 Summary of Projected Magnet Performance
347
Central Field
1.5 T
MX
II
H
I
G
C D E F G H
I I I I
H
H
H
C D E F GG H H
G
H
C D E F
HH
G
H
G
C D E F
H H H H
G
G
F
G
C D E
G
F
G
GG
E
F
G
C D
E
F
GG
G
G
D
E
F
C
G
GG G G GG
D
F
E
C
D
F
E
F
C
D
E
F
C
E
F
D
B
C
E
F
D
B
C
D
E
F
F
B
C
D
E
F
B
E
C
D
F
A B
E
C
A B C
D
E
F
MN
B
B
B
B
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
FF
F
F F F
F
F
F
Outline of the drift chamber
G
G
G
G
G
G
II
I
H
I I I
H GG
HH G F
H
G
F
G
F
G
G
F E
G
F
G
E
F
G
GG
G
F
E
G G G G G
F
E
F
E
F
E
F
F
E
F
E
F
E
F
E
F
F
E
F
E
H
H
H
HH
H
H H H
B0 = 1.4995 T
Bmin = 1.397 T
Bmax = 1.561 T
A = -6.2%
B = -5.0%
C = -3.8%
D = -2.6%
E = -1.4%
F = -0.2%
G = +1.0%
H = +2.2%
I = +3.5%
(Coil center at z = 370 mm)
Figure 9-3. Field uniformity inside the drift chamber. The central eld is within 2% of
1.5 T. Once the solenoid parameters are optimized, the corners of the drift chamber should
also be within 2% of 1.5 T. (The BABAR coordinate system is de ned in Section 14.2.)
9.3.2
Shielding of Forward Q2
The high luminosity of PEP-II requires that Q2, a non-superconducting septum quadrupole
magnet, be placed close to the interaction region. Consequently, Q2 is situated within
the forward end of the instrumented end door ux return (Figure 9-15). Q2 is subject to
induced multipole moments resulting from the magnetic eld in its vicinity, the octopole
moment being the major one. The luminosity is critically dependent upon the Q2 eld
quality. Hence, it is necessary to provide adequate shielding of the BABAR central eld to
ensure the quality of the quadrupole eld in Q2.
The present Q2 shield design is shown in Figure 9-4, where the three high-permeability
shields (dark gray) surrounding Q2 (light gray) are visible, along with the logarithms of the
magnetic equipotentials. The present design appears to shield Q2 from the detector magnet
but does not provide a safety margin, should actual parameters, e.g., steel permeability,
di er from those used in the magnetic modeling programs. Work, including a full threedimensional analysis, is continuing to improve the shield design.
9.3.3
Flux Return
The ux return assembly provides an external ux path for the magnetic eld of the
superconducting solenoid. Figure 9-5 shows the ux lines from the magnetic analysis. There
are large body forces in the rst few plates as a result of the magnetic eld. Figure 9-6 shows
Technical Design Report for the BABAR Detector
348
Magnet Coil and Flux Return
1.0
0.8
R (meter)
0.6
0.4
0.2
Q2
3.0
2.8
2.6
2.4
2.2
2.0
1.8
Z (meter)
Figure 9-4.
quadrupole.
Logarithms of magnetic equipotentials in the region of the Q2 septum
the force vectors in the barrel and end doors of the ux return. Preliminary results show
that sti eners are needed in the end door plates to resist these axial forces. The present
design has two sti eners in each end door. As the end door design is re ned, the locations
and number of the plate sti eners may change to keep the de ections and stresses in the
plates within acceptable levels.
Technical Design Report for the BABAR Detector
349
MX
9.3 Summary of Projected Magnet Performance
Vector potential lines in the full detector region. The potential lines are
horizontal in the region of the drift chamber, representing good eld uniformity.
Figure 9-5.
Forces on the ux return plates, solenoid, and end plugs as a result of the
magnetic eld. The forces are the highest in the rst few plates of the ux return.
Figure 9-6.
Technical Design Report for the BABAR Detector
Magnet Coil and Flux Return
R1400
350
Figure 9-7.
INNER RPC
NOT SHOWN
Overall view of the solenoid showing the cryostat, both conductor cross
sections, and radiation shielding.
9.4
9.4.1
Superconducting Solenoid
Magnetic Design
This section describes the main features of the superconducting solenoid. For a more detailed
description of the solenoid, refer to Reference [BF95]. A cross section of the solenoid is shown
in Figure 9-7, and parameters are given in Table 9-3.
The magnetic analysis is based on a two-dimensional axially symmetric model. This model
includes the solenoid, ux return plates, forward Q2 shield, backward shield, and the 150 mm
gap between barrel and end doors.
The backward shield is designed to accommodate the DIRC. Its main functions are to
improve the eld uniformity in the backward region of the drift chamber and to balance
the magnetic force on the solenoid due to the forward Q2 shield. A detailed design of this
shield is underway. The iron properties used for computation (ANSYS code [ANS95]|twodimensional magnetic element) are those of hot-rolled carbon steel.
The magnet design provides a magnetic eld of 1.5 T with a uniformity of 2% in the
tracking region. This is obtained by grading the current density of the solenoid in three
Technical Design Report for the BABAR Detector
9.4 Superconducting Solenoid
351
Parameter
Value
Central Induction
1.5 T
Conductor Peak Field
2.5 T
Uniformity in the Tracking Region
2%
Winding Length
3456 mm
Winding Mean Radius
1532 mm
Amp Turns
5 1192 106
Operating Current
7110 A
Inductance
0.985 H
Stored Energy
25 MJ
Total Length of Conductor
7000 m
:
Table 9-3.
Overall coil parameters.
regions connected in series. The central region is 1728 mm in length with 240 turns. Two
end regions are 864 mm in length with 240 turns each. The current density in the end regions
is twice that of the central part. A better eld uniformity may be obtained by reducing the
axial length of the two end regions and increasing the current to generate the same eld, but
this would cause a reduction in stability against thermal disturbance. For the initial design,
the maximum allowed current density in the conductor has been limited to the maximum
currently attainable for magnets of this kind, i.e., 80 A/ mm2 (ZEUS magnet). Thus, a
cross section of 90 mm2 for the smaller conductor corresponds to a maximum current of
7000 A.
Figure 9-5 shows the graph of the eld lines over the full detector region. Figure 9-3 shows
the eld uniformity in the central region de ned by
800 mm and ,1170
1910 mm
with respect to the IP. The magnetic eld is essentially symmetric. A eld uniformity of
2% is obtained. Field uniformity is required up to = 1670 mm in the forward region,
and the present design provides a uniform eld up to = 1910 mm, providing a factor of
safety. Further adjustment of the backward shield geometry may improve eld symmetry,
which would improve eld uniformity in the backward region.
r <
< z <
z
z
Technical Design Report for the BABAR Detector
352
Magnet Coil and Flux Return
9.4.2
Cold Mass Design
Aluminum-Stabilized Conductor
The conductor is composed of a superconducting Rutherford cable embedded in a very pure
aluminum matrix by a coextrusion process that ensures a good bond between aluminum and
superconductor. Table 9-4 shows the main parameters of the conductor.
The operating current for this conductor is 45% of the critical current at the peak eld,
giving a large safety margin. In the case of local heating up to 5.2 K, there is still a signi cant
margin on the critical current (I = 0 6Ic). At 2.5 T, the conductor-critical temperature is
Tc = 8 23 K, and the current sharing temperature is 6.5 K. A simple method to evaluate the
stability of the winding consists of considering the enthalpy margin per unit length between
the operating and the sharing temperature. This stability parameter for the BABAR solenoid
is 0.5 Joule/m, which is the same value obtained for the ALEPH magnet.
The cross section of the conductor is 3 2 30 mm2 for the higher current density regions
and 6 8 30 mm2 for the central region. The coil winding can be made using six 1200 m
lengths of conductor, requiring ve electrical joints. Each joint, made by either welding
of the aluminum matrix or soft soldering, has a resistance less than 5 10,10 after the
electro-deposition of copper, limiting the power dissipation to a few milliwatts.
:
:
:
:
Winding Support
The winding will be supported by an external aluminum alloy cylinder similar to other existing detector magnets. The winding support is designed for all aspects of force containment,
i.e., its weight and the radial and axial magnetic forces. Figure 9-6 shows these magnetic
forces on the solenoid and the ux return.
The highest radial pressure, 1.5 MPa, is generated in the high current density regions at the
ends of the coils. A pressure of 0.78 MPa is generated in the central region. An aluminum
alloy (5083) support cylinder surrounds the coiled conductor to react against these radial
pressures and prevent coil movement. The radial pressure is applied through the coil winding
to the outer support cylinder. Assuming that the pure aluminum coil winding can be loaded
to a stress of 20 MPa, the stress in the outer support cylinder can be estimated, using a
thickness of 30 mm, to be 65 MPa. This stress of 65 MPa is well below the 170 MPa yield
stress of the 5083 aluminum alloy. This simple analysis suggests that after the rst charge
the pure aluminum stabilizer will never be stressed beyond the elastic limit. This will help
prevent premature quenching during coil energizing. A more complete analysis including
both the Rutherford cable and the berglass epoxy insulation is not expected to signi cantly
change the calculated stresses.
Technical Design Report for the BABAR Detector
9.4 Superconducting Solenoid
353
Parameter
Conductor Type
Value
NbTi, Pure Al-stabilized,
Co-extruded
Aluminum RRR
> 500
Conductor Unit Length
1.2 km
Number of Lengths
6
Dimensions: Bare
3.2 and 6:8 30:0 mm2
Insulated
3.6 and 7:2 30:4 mm2
Superconducting Cable
Rutherford
Dimensions
9 1:23 mm2
Strands Diameter
0.84 mm
Number of Strands
20
Cu/Sc
1.8
Filament Diameter
20 m
Ic (B = 2.5 T, T = 4.5 K)
> 16 kA
Insulation Type
Fiberglass Tape
Insulation Thickness
0.4 mm
Table 9-4.
Conductor parameters.
An integrated compressive axial force of 3.5 MN is induced in the winding. The distribution
of the axial force within the coil is complex. The end regions, with higher current density,
compress the central part with 5.4 MN. The central part is axially stressed outward by a
force of 1.9 MN. For preliminary calculations of the axial stress, the maximum force (6 MN)
was considered. This would lead to an axial stress of 18 MPa on the pure aluminum, with
only the winding supporting the axial forces. However, if the axial force is transmitted to the
outer cylinder, the stress is lowered by a factor of two, with the pure aluminum working well
below its elastic limit. In this case, the shear stress between the winding and outer supporting
cylinder is less than 2 MPa. This low value of shear stress will allow the winding and support
cylinder to be mechanically coupled through an epoxy impregnation without applying any
axial prestress to the winding (as was done for the ZEUS magnet). Epoxy impregnation can
support a shear stress higher than 20 MPa, providing a high safety margin. This leads to a
simpli cation and cost saving in the winding fabrication.
The current design causes axial de-centering forces on the coil due to the iron asymmetry and
a residual force of 18 kN is applied to the winding. A more careful design of the backward
shield can help reduce the amount of this residual axial force by a factor of two or three.
O set forces have been calculated as follows. An axial displacement of the solenoid of 10 mm
causes an axial force of 98 kN in the direction of the displacement. A radial misalignment
Technical Design Report for the BABAR Detector
354
Magnet Coil and Flux Return
of 10 mm gives rise to a force of 89 kN. These values will be taken into consideration in
designing the support system and should not present any signi cant problems.
Table 9-5 shows the main features of the cold mass. The values are given at a temperature of
4.5 K. The dimensions at room temperature are higher by a factor of approximately 1.004.
Electrical Insulation
Electrical insulation is an important aspect of solenoid design and manufacture. Two
categories of insulation are required: ground plane insulation between the coil and support
cylinder, and turn-to-turn insulation.
The ground plane insulation must operate at relatively high voltages during quench
conditions and will be subjected to strict QA controls. The design of the quench
protection systems is based on a maximum voltage to ground of 250 V. The ground
plane insulation will be made by a 1 mm layer of berglass epoxy laminate that is
bonded to the support cylinder before winding. The insulation will be fully tested at
2 kV before winding.
The conductor will be insulated with a double wrap of 0.1 mm glass tape during
winding to give an insulation thickness of 0.2 mm. The resulting turn-to-turn insulation thickness will be 0.4 mm and will be fully impregnated in the bonding process.
Electrical tests will be carried out during winding to detect any failure of insulation.
The tests will include regular/continuous testing for turn-to-turn and turn-to-ground
insulation.
9.4.3
Quench Protection and Stability
Protection Concept
The solenoid will be protected by an external dump resistor which will determine the current
decay under quench conditions and allow extraction of 75% of the stored magnetic energy.
The quench protection concept is shown in Figure 9-8, and quench parameters are given in
Table 9-6. The protection concept is based on two main criteria.
A voltage limit of 500 V across the solenoid applies during fast discharge. Centertapping of the fast dump resistor to ground will limit the voltage to ground to 250 V.
The center-tapped resistor will also allow the measurement of ground leakage currents
as a safety and diagnostic tool.
Technical Design Report for the BABAR Detector
9.4 Superconducting Solenoid
355
Parameter
Winding:
ID
OD
Length
Weight
Value
3033.8 mm
3095.2 mm
3456.0 mm
2.4 tonne
Supporting Cylinder:
Material
ID
OD
Length
Weight
Ground Insulation:
Material
Thickness
Fiberglass epoxy
1.0 mm
Total Solenoid Weight:
4.9 tonne
Nuclear Interaction Length:
(Assuming Aluminum)
Maximum
Minimum
0.19
0.15
Table 9-5.
Al alloy 5083
3456.40 mm
3516.40 mm
3506.00 mm
2.5 tonne
int
int
Cold mass (4.5 K) parameters.
An upper temperature limit of 100 K applies during quench conditions. This limit will
give very good safety margins against peak temperature rise and thermally induced
stresses at quench.
Quench Analysis
A preliminary quench analysis of the BABAR solenoid has been made using a code developed
for the DELPHI solenoid design. The code models the thermal and inductive behavior of the
solenoid in order to take into account quench-back e ects and heat transfer to the support
cylinder. This analysis shows that quench-back is predicted about two seconds after opening
Technical Design Report for the BABAR Detector
356
Magnet Coil and Flux Return
Figure 9-8.
Solenoid power and quench protection concept.
Parameter
Value
Operating Current
7.11 kA
Stored Energy
25 MJ
Inductance
1.2 H
Quench Voltage
500 V
Protection Resistor
0.070
Time Constant
17.1 s
Adiabatic Peak Temperature
100 K
Overall Current Density: Conductor 1 74 A/mm2
Conductor 2 35 A/mm2
Aluminum Stabilizer RRR Zero Field
500
Table 9-6.
Quench parameters.
Technical Design Report for the BABAR Detector
9.4 Superconducting Solenoid
357
40.0
Temperature (K)
30.0
Coil (ThC(1) K)
Support Cylinder (ThS(1) K)
20.0
10.0
0.0
0.0
10.0
20.0
30.0
40.0
50.0
Time (s)
Temperature variation during quench. The temperature rise in the coil and
support cylinder during a quench should not exceed 40 K.
Figure 9-9.
the protection circuit breakers. Figure 9-9 shows that the temperature rise in the coil and
support cylinder during a quench should not exceed 40 K.
Stability
The BABAR solenoid coil will be indirectly cooled using the technology established for existing
detector magnets such as DELPHI and ALEPH. The reliable operation of those magnets has
demonstrated that safe stability margins can be achieved using high-purity, aluminum-clad
superconductors in a fully bonded, indirectly cooled coil structure.
Conductor stability has been estimated using an analysis code in order to establish the
minimum quench energy (MQE) for transient heat pulses. The computed MQE = 1.4J. The
computed minimum quench length (MQZ) is 0.6 m.
These margins are considered to be safe for the BABAR solenoid due to its low-stress design.
The stability margin will be optimized during the full design study.
Technical Design Report for the BABAR Detector
358
Magnet Coil and Flux Return
Cold mass cooling circuit. The cryogenic supply chimney passes through a
cut-out in the backward end of the barrel ux return.
Figure 9-10.
9.4.4
Cold Mass Cooling
Cold mass cooldown is accomplished by circulating cold helium gas either
directly from the refrigerator or from a storage dewar with gas mixing. A preliminary cold
mass cooldown analysis has been performed. A cooldown mass ow rate of 15 g/s will lead
to a cooldown time of ve days. The maximum temperature di erence across the cold mass
is limited to 40 K in order to minimize thermal stress during the cooldown from 300 to 100 K.
Cooldown.
Under operating conditions, the cold mass is cooled by circulating two-phase helium in circuits mounted on the coil support cylinder. The thermo-syphon
technique will be used to drive the cooling circuit. This technology is established and yields
the simplest operational mode. The thermo-syphon cooling circuit is designed for high ow
rates to ensure the correct quality factor for the helium. The conceptual layout of the
cold mass cooling circuit is shown in Figure 9-10. The circuit is fed through a manifold
at the bottom of the support cylinder. The cooling circuits are welded to the support
cylinder surface with a spacing of 0.3 m to limit the temperature rise to less than 0.1 K.
The cooling pipes terminate in an upper manifold. The circuit will be designed to provide
operation during quench conditions.
Operating Conditions.
Technical Design Report for the BABAR Detector
9.4 Superconducting Solenoid
359
Heat Loads. The estimated static heat loads for the solenoid are given in Table 9-7. Eddy
current heating in the support cylinder will cause additional heat loads during charging of
the solenoid. However, for a solenoid charging time of 30 minutes, the estimated transient
power is 2 W, which is small compared to static heat loads.
9.4.5
Cryostat Design
Vacuum Vessel. The cryostat consists of an annular vacuum vessel equipped with radiation shields and superinsulation (Figure 9-7). The vacuum vessel is designed to satisfy a
number of basic criteria:
1. Support vacuum loads in accordance with recognized pressure vessel codes;
2. Carry the cold mass and radiation shield weight through the insulating supports;
3. Operate with de ections of less than 2 mm under all loads when mounted in the ux
return barrel;
4. Carry the loading of the inner detectors; and
5. Operate under de ned seismic loadings.
The vacuum vessel is designed as two concentric cylinders with thick annular end plates, all
of aluminum alloy 5083; its basic parameters are given in Table 9-8. A preliminary nite
element (FE) structural analysis of the vessel has con rmed that design criteria (1){(4)
can be met with reasonable safety factors. Maximum vessel de ections are less than 2 mm,
and stress levels are generally lower than 40 MPa with all loads applied. De ections are
minimized when the vessel is supported on the horizontal centerline and detector loads are
also applied at that point. Performance of the vessel under seismic loadings (5) is still under
consideration.
Thermal Shielding. The cryostat is equipped with radiation shields, which operate at
40{80 K, and superinsulation. The shields are cooled by helium gas supplied directly from
the refrigerator. About 30 layers of superinsulation separate the vacuum vessel walls from
the radiation shields. Another ve layers will be installed between the shields.
Services. Cryogenic supplies and current supplies are connected from a services turret to
the cryostat through the service chimney in the backward end door. Current leads and local
control valves are mounted in the services turret. Cryogenic relief valves are also mounted
in the service turret for quench and refrigeration failure conditions.
Technical Design Report for the BABAR Detector
360
Magnet Coil and Flux Return
Item
Magnet Heat Loads at 4.2K
Parameter
Cold Mass
Total Surface Area
Radiation Heat Flux (Design)
Radiation Heat Load (Design)
Conduction Heat Load
Transient Heat Load (30 min)
Total 4.5 K
7000 kg
100 m2
0.4 W/m2
Load Liquifaction
(watts)
(l/h)
40 W
10 W
2W
52 W
73 l/h
Magnet Shield Heat Loads at 80K
Item
Parameter Load Liquifaction
(watts)
(l/h)
Shield Mass
1000 kg
Total Surface Area
100 m2
Radiation Heat Flux (Design)
3 W/m2
Radiation Heat Load (Design)
300 W
Conduction Heat Load
50 W
Total 80 K
350 W
2 leads 7 kA
Item
Current Leads
0.72 g/s
Cryoplant Heat Loads at 4.2K
Parameter
16 W
22 l/h
Load Liquifaction
(watts)
(l/h)
Valve Box and Valves
10 W
15 l/h
Transfer Lines (Lique er-dewar & Return) 4 m (2) 6 W
8.5 l/h
Transfer Line (Dewar-valve Box)
4m
3W
4 l/h
Coaxial Transfer Line (Valve Box-magnet)
60 m
3W
4 l/h
4000L Dewar
6W
8.5 l/h
Total 4.2K
28 W
40 l/h
Table 9-7.
Cryogenic heat loads.
Technical Design Report for the BABAR Detector
9.4 Superconducting Solenoid
361
Envelope Dimensions
Inner Radius
1400 mm
Outer Radius 1730 mm
Length
3850 mm
Materials
AL5083
Design Loads
Vessel Weight 6 tonne
Cold Mass
6 tonne
Calorimeter
50 tonne
Seismic Load Factors (Max)
Horizontal
1.2 g
Vertical
2.0 g
Table 9-8.
Vacuum vessel parameters.
9.4.6 Coil Assembly and Transportation
The coil will be assembled inside the cryostat at the manufacturer's plant. Electrical and
cryogen connections will be made at the chimney so that the coil can be tested before
shipping.
A complete cooldown will be carried out from room temperature to the operating temperature of 4.5 K. The cooldown will allow checking of cooldown time, temperature control, heat
loads, and full operation of sensors. A magnetic test will also be performed at low eld (30%
of the operating current) to check superconductor operation, the joint resistance, and the
additional losses due to the energization.
Before delivering the magnet, but after the tests at the factory, the end anges will be
dismounted to allow a hard connection of the cold mass to the cryostat walls. Depending
on the transport facilities, the chimney may also be dismounted. In this case, the electrical
and cryogen connections also must be dismounted and protected against breakage.
Technical Design Report for the BABAR Detector
362
Magnet Coil and Flux Return
9.5 Cryogenic Supply System and Instrumentation
Operation of the superconducting solenoid requires both liquid helium and cold helium gas
(20 K to 100 K) for cooldown and refrigeration of the thermal shields. Similar systems have
been used successfully throughout the HEP community. A summary of the cryogenic loads
is given in Table 9-7.
The helium plant, consisting of a helium lique er/refrigerator, a 4000 ` supply dewar, and
a distribution valve box (DVB), is located adjacent to the experimental hall, approximately
60 m from the magnet, as shown in Figure 9-11. It is possible that the DVB will be incorporated within the helium lique er cold box, depending upon the helium lique er selected.
This liquid helium plant meets all of SLAC's requirements and is sized conservatively at
150{200 `/hr. It will supply LHe to the BABAR superconducting solenoid, the two future
superconducting beam line focusing magnets (Q1), and an auxiliary dewar/trailer for all
other SLAC experiments. The detector solenoid is expected to consume less than 100 `/hr
of LHe.
Liquid nitrogen is required for the helium lique er, for the initial stages of coil cooldown,
and miscellaneous uses in IR-2. It will be supplied from a 20,000 ` tank located on the apron
above the experimental hall. This tank, which is an existing SLAC unit, will be refurbished
to serve all cryogenic system requirements. A second, similar vessel is also available if a
reserve LN2 supply is required. Vacuum-insulated transfer lines connect the LN2 tank to
the helium plant.
The 4000 ` liquid helium dewar is a refurbished SLAC unit fed directly from the lique er.
This volume provides approximately 30 hours of autonomous operation of the solenoid and
quadrupole magnet systems in the event of a minor lique er/compressor malfunction. LHe
from the dewar is supplied at 1.25 bar, via proportional control valves in the DVB, to the
appropriate magnet/auxiliary dewar system as required. These electro-pneumatic operated
valves are actuated by process controllers and superconducting LHe level gauges.
Liquid helium and cold helium gas are supplied to the detector solenoid and Q1 superconducting magnets in exible, vacuum-insulated, low-loss transfer lines. These transfer lines
provide the cold gas return path to the lique er/refrigerator. This type of transfer line has
been used successfully by SLD. The transfer lines are designed to be compatible with the
SLD lines so that the latter can serve as spares and/or Q1 transfer lines. LHe and cold shield
gas for the solenoid are routed to the magnet service chimney at the detector backward end
door penetration (north end). Magnet cooldown to 20 K is achieved with cold helium gas
via the DVB. Warm helium gas from magnet current leads is routed in uninsulated piping
to compressor suction.
Technical Design Report for the BABAR Detector
9.5 Cryogenic Supply System and Instrumentation
Figure 9-11.
363
Layout of the cryogenic system.
The solenoid is equipped with a full set of instrumentation sensors for monitoring, control,
and diagnostic purposes. Instrumentation includes temperature sensors for the cold mass,
shield cryogen ow monitoring, and strain gauges in the coil support cylinder. Voltage taps
monitor the electrical resistance of the conductor joints and quench detection. The quench
detection systems are hard-wired to interlocks. The solenoid instrumentation and controls
are integrated with the BABAR experiment and refrigeration controls.
The liquid helium plant, which is fully automatic, is furnished with a process control
system and all requisite logic and software necessary for all operational modes. Control
and monitoring of the cryogenic plant and the magnet coil, together with remote control
and monitoring of the compressor room, is carried out from a control room adjacent to
the plant room and IR-2. Main operating parameters are interfaced with the BABAR data
acquisition and monitoring systems.
Technical Design Report for the BABAR Detector
364
9.6
9.6.1
Magnet Coil and Flux Return
Flux Return
Overview
The ux return assembly provides the external ux path for the magnetic eld from the
superconducting solenoid. The steel plate segmentation enhances low momentum pion/muon
separation and KL0 detection eciency. The ux return also provides the gravitational and
seismic load path for the barrel detector components to the concrete foundation.
The ux return assembly consists of a barrel, four external support legs, two sets of end
doors, roller mechanisms, and vertical adjustment systems. The design of the ux return
and its components re ects the limitations of the IR-2 experimental hall.
End door components must be movable once the detector is assembled to allow maintenance
access to inner detector subsystems. The design of the ux return provides a means for
proper alignment of all ux return components with respect to the detector magnet and the
beam line. This alignment must be reestablished after maintenance is performed.
Analysis of the Flux Return
The ux return analysis will investigate the overall structural integrity, including that of the
supporting structures. This includes looking at all the components and connections involved
in the operation, assembly, and seismic loadings of the structure. Operating loads for the
barrel include both the gravitational and the magnetic forces. Assembly loads include any
additional xture weights that need to be attached during assembly. Seismic loadings assume
a site-speci c peak ground acceleration of 0.6 g. This seismic design standard exceeds that
of the minimum base shear requirements as formulated by the Uniform Building Code.
A nite element model program, ANSYS [ANS95], revision 5.1, is currently being used
to study the stresses and de ections of the barrel and its supporting structures under the
di erent loading conditions. It is also being used to model the structure's response to
oscillatory ground motion. The results from these studies will be used to verify that the
design requirements are met.
9.6.2
Barrel Flux Return Description
The barrel ux return assembly is shown in Figure 9-12. The barrel ux return assembly
extends 4050 mm in z, and the center is positioned 370 mm from the interaction point in
Technical Design Report for the BABAR Detector
9.6 Flux Return
365
1400
Figure 9-12. Barrel ux return assembly showing the support of the solenoid, gap ller
plates, and the external support legs.
the positive z direction. The barrel ux return extends radially from the detector axis from
1780 mm to 2920 mm and consists of inner and outer hexagonally-shaped concentric steel
rings. The radial gap between the inner and outer rings is 50 mm. The rings consist of
blocks with multiple steel plates oriented parallel to the axis of the solenoid. Two side plates
and the inner and outer parallel plates of each block are welded into a rigid box using full
penetration welds along their entire length in the z direction (parallel to the beam line).
The remaining parallel plates are held in place using skip welds at their connections with
the side plates. The use of double block construction for each of the hexagonal segments
of the barrel ux return provides a signi cant safety margin with respect to the 45-tonne
rated capacity of the overhead crane in IR-2. The blocks are designed to provide continuous
muon detection in azimuth. The weight of the barrel ux return is 312 tonne excluding the
external support legs.
Outer Ring Description
The outer ring is the primary structural support for the BABAR detector. This ring is
composed of six blocks. Each block consists of six 30 mm-thick and three 50 mm-thick parallel
Technical Design Report for the BABAR Detector
366
Magnet Coil and Flux Return
Figure 9-13.
Block-to-block interface in the ux return.
structural plates oriented so that the thicker plates are positioned towards the outermost
portion of the barrel hexagon. The plates are positioned in each block to provide 30 mm of
clear space between each plate and the next for the installation of Resistive Plate Chambers
(RPCs). The inner surface of the outer block is located 2350 mm radially from the interaction
point. The 30 mm spacing between plates is to be maintained during block fabrication.
Therefore, the tolerance build-up due to variations in the plate thickness and plate atness
extends outward in the radial direction. The spacing of the parallel plates is held xed by
60 mm-thick side plates located at the block-to-block interface, as shown in Figure 9-13.
Two of the blocks are provided with threaded holes in their outer plates for mounting of the
external support legs. Each block is 3750 mm in length, 570 mm in nominal height, 2714 mm
wide at the inner trapezoid base, and 3372 mm wide at the nominal outer trapezoid surface.
Each outer block weighs approximately 31.2 tonne.
Inner Ring Description
The inner ring is also composed of six blocks. Each block consists of eleven 20 mm-thick
parallel structural plates positioned to provide 30 mm of clear space between plates for the
installation of RPCs. The inner surface of the inner block is located 1780 mm radially
from the interaction point. Like the outer plates, the 30 mm spacing is held xed, and the
plate manufacturing tolerances are allowed to accumulate in the outward direction. A gap
Technical Design Report for the BABAR Detector
9.6 Flux Return
367
A 200 tonne capacity roller and a 250 tonne capacity jack are located at
the end of each support leg to support the barrel.
Figure 9-14.
of 20 mm is provided between the inner and outer hexagonal rings for this purpose. The
spacing of the parallel plates is held xed by 50 mm-thick side plates located at the blockto-block interface as shown in Figure 9-13. The side plates are provided with threaded holes
to mount a 150 mm-thick plate that supports the inner blocks from the outer ring. The
inner blocks are 5 mm shorter in z length than the outer blocks and will have to be shimmed
during assembly for a secure t. There is 10 mm of clearance provided at the block-to-block
connections for manufacturing and assembly tolerances. Each block is 3745 mm in length,
520 mm in nominal height, 2050 mm wide at the inner trapezoid base, and 2650 mm wide at
the nominal outer trapezoid surface. Each block weighs approximately 16.4 tonne.
External Support Legs
Two external support leg assemblies provide the gravitational and seismic load path from the
barrel ux return to the concrete foundation. Each assembly consists of two legs positioned
3200 mm apart in z. The legs are fabricated from 50 mm-thick structural steel plates.
Included at each support leg is a 200 tonne capacity roller and a 250 tonne capacity jack
as illustrated in Figure 9-14. A preformed fabric pad is positioned between the jack and
the roller to add compliance to the system. The horizontal spacing of the jack and roller
assemblies is nominally 8000 mm. Each support leg is approximately 3750 mm in length,
2400 mm in width, and 3300 mm in height. The weight of each support leg assembly is
approximately 20 tonne.
Technical Design Report for the BABAR Detector
368
9.6.3
Magnet Coil and Flux Return
End Door Description
The forward and backward end doors are an array of steel plates that form a regular hexagon
approximately 5840 mm across the ats and 1120 mm thick. The array consists of eleven
20 mm-thick plates, six 30 mm-thick plates, and three 50 mm-thick plates. There is a nominal
gap width of 30 mm between plates in the array for locating the RPCs. The plates for both
the forward and backward end doors weigh approximately 255 tonne. This weight does
not include plates and bars that tie the plates together for structural reasons, the support
members that attach the end doors to a movable skid, or the skid itself (Figures 8-23 and
9-15). Both the forward and backward end doors are split along the vertical centerline of
the detector forming right and left end doors.
Each end door is mounted on a skid that permits it to be raised onto high load capacity
rollers by hydraulic jacks and to be moved on tracks located in the IR-2 hall. This provides
a means to move the end doors into proper alignment with the barrel prior to being bolted
in place, and to be moved away from the barrel for maintenance access to the detector. The
center of gravity of the end door plates is high compared to the depth of the base. The skid
provides additional stability during the horizontal accelerations experienced during moving
or seismic events.
During operation, the magnet exerts an inward axial force that causes a signi cant bending
moment on the end door plates. The forward end doors also support the weight of the Q2
shielding plug and carry the axial magnetic load induced in the plug. To resist gravitational
and seismic loads, and to limit the plate stresses and de ections, all the end door plates are
joined together to form a single structural member.
The design of the end doors permits each door to be assembled and disassembled inside the
IR-2 hall with the existing facility crane. Since the capacity of the IR-2 crane is 45 tonne,
each separable part, together with all rigging and lifting xtures needed, cannot exceed this
limit. Each end door is therefore composed of two weldments that are fastened together at
installation. The inner weldment consists of eleven 20 mm plates and two 30 mm plates, with
each plate welded to a channel-shaped frame that extends along the top and bottom of the
hexagon shape and along the boundary between the right and left doors. Two horizontal
sti eners are positioned between the 20 mm plates. Additional sti eners are also required
to sti en the weldment and help maintain the necessary gaps for the RPCs. The outer
weldment consists of four 30 mm plates and three 50 mm plates welded together with similar
sti ening members. The detailed analysis of the response of the plates to the magnetic force
distribution is not yet available, nor has a detailed seismic analysis been done for the end
door plates. Several design options are being studied that can provide the necessary strength
and sti ness.
Technical Design Report for the BABAR Detector
9.6 Flux Return
369
Bolted
Tie Plate
Inner End
Door Weldment
Outer End
Door Weldment
Q2 Shielding
Plug Subassembly
Counter Weight
End Door
Skid
Figure 9-15.
Side view of the forward end door showing the Q2 shielding, counter weights,
and support rollers.
Technical Design Report for the BABAR Detector
370
Magnet Coil and Flux Return
The inner and outer weldments are joined together at installation by bolting each weldment
to the top of the skid along the bottom of the hex, and by bolting tie plates around the
remaining perimeter of the weldments, except where the shielding plugs are located. The
outside tie plates are bolted in place after the RPCs are installed, and provisions for cabling
are provided in these outside plates. These tie plates are also used to attach the end doors
to the barrel of the ux return.
End Door Skids
Each end door is mounted on a skid that is equipped with four 70 tonne capacity rollers, and
four 45 tonne hydraulic jacks, one in each corner, which allow each of the doors to be moved
relative to the barrel for maintenance access. The rollers ride on hardened steel tracks that
are permanently located in the oor of IR-2. Because the bases of the end doors are narrow
compared to their high centers of gravity, 30 tonne of existing scrap steel is bolted to each
skid to lower the center of gravity and move it toward the middle of the skid.
The end doors are bolted either to the barrel of the ux return when in the operating position
or to seismic restraint brackets when in the parked position. While the doors are being moved,
the 30 tonne counter weight provides lateral stability for a horizontal acceleration of 0.3 g.
Forward Q2 Shielding Plug Support
The preferred option for mounting the Q2 shielding plug is to fabricate and assemble the
shield in two sections split along the vertical axis. These assemblies are mounted to large halfround anges that are bolted to the back 50 mm plates of the forward end doors (Figure 9-15).
This simpli es access to the front portion of the detector when the end doors are moved. It
is likely that the halves of the end door will be tied together by two additional half-round
anges oriented at 90 to the shielding plug supports to sti en the assembly. This option
requires that each half of the shield be installed in the end doors at nal assembly since
each half of the shield assembly weighs approximately 20 tonne. Special lifting and assembly
xtures are required to accomplish this task.
The alignment of the shield and the Q2 magnet is done by adjusting the position of each halfround ange on the end door and installing dowel pins to maintain the alignment. Sucient
clearance between the shield and the forward Q2 magnet is provided to permit the shield to
move vertically with the stroke of the jack as the end door is raised to be rolled out of the
way for detector access.
Technical Design Report for the BABAR Detector
9.6 Flux Return
371
9.6.4 Options and Detailed Design Issues
A detailed stress and de ection analysis is proceeding for the nalized overall envelope
dimensions. A detailed magnetic eld and force analysis of the end doors ensures that they
will have adequate strength and sti ness to meet all the requirements imposed on them.
The tolerance on plate atness must be de ned together with the envelope dimensions of
the RPCs. Standard mill tolerances for plate atness do not meet our requirements for
the end door plates to permit reliable RPC installation; these tolerances exceed 15 mm,
half the nominal gap width. This issue, together with weld distortion in the plates during
fabrication, must be resolved with both the plate supplier and the barrel and end door
fabricator. Other manufacturing tolerances must be established as the system interface
dimensions are nalized, and the e ects of these tolerances on the physics performance of
the detector are reviewed.
9.6.5 Procurement, Fabrication, Assembly, and Schedule
The barrel and the end doors will be built by the same fabricator as part of the same procurement contract. This will eliminate some duplication in the review of vendor quali cations,
quality control plans and actions, and many contract administration issues. In addition,
control of other characteristics of the plate material, such as the chemistry of the plates as
it a ects weldability and machinability, the mechanical properties of the plate, etc., may be
more easily tracked by having one set of acceptance inspection criteria from one supplier for
all plates.
The barrel and end doors will be fully assembled and inspected at the fabricator's shop. In
this way, any problems that arise can be solved before nal assembly in the IR-2 hall. This
will also permit a thorough review of the assembly procedure by an experienced fabricator
and ensure that the necessary lifting and assembly xtures are functional and are provided
with the barrel and end doors. The details of the fabrication and assembly plan will be
developed by the fabricator subject to the review of the responsible design engineer in the
BABAR collaboration.
Technical Design Report for the BABAR Detector
372
REFERENCES
References
[And82] D. Andrews et al., \A Superconducting Solenoid for Colliding Beam Experiments," Ad. Cryogenic Eng. 27 (1982).
[ANS95] ANSYS, INC, \General Finite Element Code," Rev. 5.1 (1995).
[ASME94] ASME, \1992 ASME Boiler and Pressure Vessel Code" (1992).
[BF95] E. Baynham and P. Fabbricatore, \Superconducting Solenoid for the A AR
Detector," A AR
in preparation.
[Coils] A. Bonito Oliva et al., \ZEUS Magnets Construction Status Report," 11th Int.
Conference on Magnet Technology, MT{11, 229 (1989).
P. Clee et al., \Towards the Realization of Two 1.2 T Superconducting Solenoids
for Particle Physics Experiments," 11th Int. Conference on Magnet Technology
MT{11, 206 (1989).
H. Desportes et al., \Construction and Test of the CELLO Thin-Wall Solenoid,"
Ad. Cryogenics Eng. 25, 175 (1980).
H. Desportes et al., \General Design and Conductor Study for the ALEPH
Superconducting Solenoid," J. Phys. (Paris) C1-341, S1-T45 (1984).
Y. Doi et al., \A 3T Superconducting Magnet for the AMY Detector,"
A274, 95 (1989).
M.A. Green et al., \Construction and Testing of the Two-Meter-Diameter TPC
Thin Superconducting Solenoid," Ad. Cryogenics Eng. 25, 194 (1980).
M.A. Green et al., \A Large Superconducting Thin Solenoid for the STAR
Experiment at RHIC," IEEE Trans. App. Superconductivity, 104 (1993).
H. Hirabayashi, \Detector Magnet Technology for High Energy Accelerators,"
ICEC{11, 115 (1986).
H. Minemura et al., \Fabrication of a 3 m-dia. 5 m Superconducting Solenoid
for the Fermilab Collider Detector Facility," J. Phys. (Paris) C1-333, S1-T45,
(1984).
C.M. Monroe et al., \The CLEO-II Magnet{Design, Manufacture, and Tests,"
ICEC{12, 773 (1988).
M. Wake et al., \Excitation of a Superconducting Large Thin Solenoid Magnet,"
MAG{23, 1236 (1987).
F. Wittgenstein et al., \Construction of the L3 Magnet," 11th International
Conference on Magnet Technology, MT{11, 131, (1989).
A. Yamamoto et al., \Thin Superconducting Solenoid Wound with the Internal
Winding Method for Colliding Beam Experiments," J. Phys. (Paris) C1-337,
S1-T45 (1984).
B B
B B
Note
Nucl.
Instr. Methods
Technical Design Report for the
A AR
B B
Detector
REFERENCES
373
[Des85]
H. Desportes, \Recent Progress in the Design and Construction of Beam and
Detector Magnets," in Proceedings of the Int. Symposium on Magnet Technology
(Zurich, Switzerland), MT{9, 149 (1985).
[SDM91] Earthquake Safety Committee, \Seismic Design Manual - Mechanical," Rev. 1
Stanford Linear Accelerator Center, Stanford, California (1991).
Technical Design Report for the BABAR Detector
374
Technical Design Report for the BABAR Detector
REFERENCES
10
Electronics
10.1
10.1.1
T
Overview
Introduction
he electronics system encompasses all front-end, trigger, and data acquisition electronics.
The environment of the high-luminosity PEP-II machine poses certain unique challenges
to this system. PEP-II beam crossings occur at a rate of 238 MHz, which, in terms of the
response time of the electronics, is essentially continuous, unlike previous e+ e, colliders.
PEP-II may have severe backgrounds, producing high occupancies. These high occupancies
present a signi cant burden on the transport and recording of data. The rate of all physics
processes to be recorded at the design luminosity of 3 1033 cm,2s,1 is expected to be 100 Hz.
The electronics system will also need to accommodate the rate and background implications
of a luminosity upgrade to 1034 cm,2s,1 . These factors have resulted in the adoption of an
innovative system architecture that departs signi cantly from previous implementations at
conventional e+e, experiments. In this architecture, the electronics systems are capable of
collecting data from the detector in realtime, and processing and transporting the data in
parallel. The data are extensively bu ered to avoid deadtime losses while making trigger
decisions. In addition, the detector electronics, trigger logic, and data acquisition system
are capable of simultaneous trigger processing of multiple events.
As described in Chapter 12, the backgrounds expected in PEP-II include particles lost
near the IP, beam-gas interactions, and cosmic rays. The dominant source of occupancy
is predicted to be the lost particle background, which depends on the beam intensity and
the quality of the vacuum in the machine in the region from 3 to 50 m from the IP. At the
design pressure of 1 nTorr, simulations predict raw hit rates of several tens of megahertz per
layer in the silicon vertex detector, 100 kHz per layer in the drift chamber, and 15 kHz per
crystal in the calorimeter. The rst device to become unusable as background increases is
the silicon vertex detector. This will happen at approximately ten times the nominal rate.
Both long-term (ten-year) radiation damage and very high occupancies in the inner layers
provide limits to useful running conditions. The electronics system is fully operational under
background conditions up to ten times the nominal rate. At higher background conditions,
376
Electronics
System
Number of Channels
Vertex Detector
150,000
Muon System (IFR)
42,000
DIRC
13,400
Calorimeter
7,000
Drift Chamber
7,000
Aerogel (ATC)
144
Table 10-1.
Channel count for each detector system.
the system degrades gracefully, albeit with increasing deadtime, allowing the background
and detector performance to be studied.
The principal features of the electronics system are: a common architecture for all detector
electronic front-ends; common trigger, timing, and control interfaces; a common data acquisition interface; and a common online software environment from the lowest level within the
front-end signal processors to the highest level user interfaces and event processing farms.
By generalizing the architecture for all systems down to the point of contact with signal
digitization, the overall maintenance, software, and training needed to operate the system
are minimized, and engineering can properly be focused on detector-speci c requirements.
The BABAR detector has approximately 220,000 signal channels (summarized in Table 10-1)
distributed among six types of detectors, delivering signals with shaping times varying from
1 ns to 3 s. Since the accelerator operates essentially continuously, the fast trigger will be
unable to identify the speci c beam crossing associated with an event. In fact, the event
time will be uncertain to hundreds of nanoseconds. Consequently, a requirement of the
electronics systems is that they record the raw data in real time and extract the data within
the time uncertainty window. Providing sucient bandwidth to transmit all of the raw event
data within these time windows would be overwhelming, so as much signal processing of the
data as possible is performed prior to transmission. This requires storage of the data while
ltering is being performed and while the pertinent data (features) are being extracted.
All detector systems have a common interface for event readout, fast control, trigger, and
detector control. Yet, within this commonality, all detector front-ends have unique signal
processing requirements. While it is not possible to devise a uniform interface to the trigger
for each detector system or to provide a uniform implementation of the storage of data
awaiting a Level 1 decision, the same architecture of data storage and control has been
applied to the signal processing chains for all detector systems. A block diagram of the
major components of the electronics is shown in Figure 10-1. The following subsections give
overviews of the front-end electronics, trigger, and data acquisition systems.
Technical Design Report for the BABAR Detector
10.1 Overview
Processed
Detector
Signals
Event
Assembly
and Filtering
Readout
Crates
Front-End
Electronics
1-95
7857A4
Digital
Event Data
Clocks,
Triggers, etc.
Detector
Command
Data
Fast
Control
Detector
Control
Trigger Data
Raw
Detector
Signals
377
Archival
Storage
Monitoring
Data
Trigger
Processor
Figure 10-1. Block diagram of the electronics system showing the principal communication paths. The dark arrows show the event data path.
10.1.2
Front-End
In the BABAR nomenclature, the front-end is that portion of the electronics from the signal
source on the detector through the digitizer. For those systems which have their digitizers
mounted on the detector, the front-end also includes any additional detector-mounted parts.
As shown in Figure 10-2, the front-ends are customized to t the needs of the particular
detector system. Figure 10-2(a) shows a generic front-end. The raw detector signals are
appropriately processed and transmitted to the readout modules, which are housed in VME
crates in the electronics house. The electronics house (in contrast to the detector) is accessible
when the machine is running. Figure 10-2(b) depicts a front-end for the calorimeter. Since
the event time is not known at this point, the digitization is done periodically, at a rate
optimized for each system. Figure 10-2(c) is the front-end for the drift chamber. Here the
digitization is done on the readout module. Figure 10-2(d) shows the front-end for the vertex
detector, the instrumented ux return and the DIRC.
The processed detector signals are transmitted to the o -detector readout crates. For systems
in which the rst stage of bu ering is on the readout modules, processed signals constitute
a continuous data ow; selection of data by the triggers is accomplished on the readout
modules. Systems which have bu ers in the detector-mounted electronics send data selected
by a Level 1 strobe only upon the request of the readout module. Examples of the rst
case are the drift chamber and the calorimeter. Examples of the second case are the vertex
detector, the instrumented ux return and the DIRC.
Technical Design Report for the BABAR Detector
378
Electronics
On Detector
(a)
Raw Analog
Detector Signals
Cable Connections
to Readout Cards
Detector
“Generic”
Front-End
Electronics
Processed Detector
Signals to
Readout Cards
Flash
ADC
Raw time slice data
Trigger primitives and decisions
Processed Analog
Signal Stream
to Downstream
FADC
Signal Processor
Level 1
Data Storage
L1 Trigger
< 2 kHz L1 accept rate
Multi-Bit
DigitalSignal
Stream
Analog Signal
Processors
(c)
Fixed latency
9.5 ± 0.5 µsec
Circular
Buffer
Fast Clock
(b)
(d)
DC and EM
trigger primitives
Raw digitized data
Level 1
Filtered Data
Event
Buffer
Feature extracted data
L3 Farm
Reconstructed data
100 events/second
Clock
Level 1 “Accept”
1-95
7857A5
Figure 10-2. Examples of front-end electronics
for various detector technologies. Variations (a)
through (d) are explained in the text.
10.1.3
Mass Storage
10-3. Trigger structure and data acquisition data
ow, showing levels and bu ers.
See text for a discussion of
Level 2 options.
Figure
Trigger
De nition of Trigger Levels and their Functional Requirements
The structure of the trigger and data ow are shown in Figure 10-3. The trigger:
provides a front-end strobe (Level 1 strobe) after a latency between 9 and 10 s as
required by the nite depths of the front-end bu ers (Chapter 4);
initiates, in the readout modules, the extraction of hit times and magnitudes from the
sample streams (feature extraction);
initiates the collection of the data associated with an event in a single processor (event
assembly); and
Technical Design Report for the BABAR Detector
10.1 Overview
379
selects a sample of events to be written to mass storage at a rate of no more than
100 Hz, as required by the online computing system.
This architecture is the simplest robust solution that performs the trigger functions, given
the rate considerations.
The only practical method for delivering a Level 1 strobe in less than 10 s is to use a
hardware processor operating on a reduced representation of a subset of the data (trigger
primitives). The most powerful method for reducing the trigger accept rate to 100 Hz
with no loss to the physics program is to use the vertex detector to select events with
multiple tracks originating from a narrowly de ned primary vertex near the e+e, luminous
region. Simulation studies show that the required vertex detector information cannot be
made available to the Level 1 hardware trigger logic because the bandwidth required to
supply sucient position granularity to suppress background is comparable to the raw-data
bandwidth required after the Level 1 strobe [Lev94c]. Because the vertex information cannot
be included, the Level 1 accept rate will be considerably higher than the eventual rate of
writing events to archival storage. Event assembly could proceed comfortably at a rate of
2.0 kHz, based on experience with other experiments [Boz93, Dou94, Buo, Cul91, Pat94].
Once the events have been assembled, there will be ample information with which to lter
out the background events. Therefore, a two-level trigger containing a exible Level 1 lter
with a maximum output rate less than the maximum event assembly rate is necessary and
sucient.
To support the system architecture, dedicated data paths provide trigger data to the rst
level trigger hardware (Level 1). The maximum average acceptance rate for this trigger is
required to be less than 2.0 kHz (Table 10-11 in Section 10.9). A nal software lter (Level 3)
running in commercial processors analyzes the event data from all detector systems and
reduces the rate at which events are written to archival storage to less than 100 Hz. The
trigger protocol is expected to be highly ecient for a variety of physics modes in addition
to B physics events.
Each trigger level uses a unique set of data and algorithms, summarized in Table 10-2, that
limits the accepted events. Successive trigger levels use increasing amounts of time to perform
more intricate analyses of the data. Not all detectors contribute trigger data; those that do,
preprocess it, e.g., extract track segments from a collection of raw drift chamber hits. The
results of this preprocessing are referred to generically as primitives. The primitives used by
the Level 1 trigger are track segments from the drift chamber and energy sums from xed
sets of crystals in the calorimeter. The Level 1 trigger rate is dominated by the lost particle
background with current estimates showing 2.0 kHz at ten times the nominal beam-pipe
pressure. If needed, additional trigger criteria can reduce this rate with some small loss of
non-B -physics eciency.
Technical Design Report for the BABAR Detector
380
Electronics
Trigger
Level 1
New Information
Algorithms
DC Segments, EM Energy Sums Track Counting, Energy Cuts,
p Cuts, Matching
Level 2 or 3 Vertex System Occupancies
Primary Vertex Cuts
Level 3
Full Event Data
Final Event Selection
t
Table 10-2.
Trigger inputs and algorithms.
Additional robustness is designed into the system via hooks for an additional Level 2 lter
in case the actual Level 1 accept rate is higher than the maximum Level 3 event build rate,
or if the feature extraction function requires more computing power and/or bandwidth than
is currently estimated. The Level 2 trigger adds requirements on the track point-of-origin
to reject non-annihilation backgrounds. It might also simplify the Level 1 trigger by doing
the track transverse momentum cuts. Inputs to this trigger level would include all inputs to
Level 1 and any additional signals needed as the algorithms are developed.
All trigger levels have the exibility to examine event topology, count tracks, and match
charged tracks to energy clusters. All levels have some capability in setting adjustable
thresholds in energy or transverse momentum. This high degree of exibility is required to
reject known backgrounds under a range of conditions while maintaining very high eciency
for physics events.
10.1.4 Data Acquisition
The readout modules contain large memories that preserve the streams of digitized data.
For systems with detector-mounted bu ers, these memories store events after the Level 1
accept and are used to hold data that is ready for readout. The remaining systems also use
this memory to bu er data during the Level 1 trigger latency.
Within the data acquisition system, there are four principal communication paths:
An event data path by which all event data pass from the system electronics to the
event collection system;
A high speed clock, trigger, and control pathway for synchronizing the ow of data
and controlling reset, initialization, and throttling of data;
A relatively slow detector control interface between the system electronics and the
online system; and
Technical Design Report for the BABAR Detector
10.1 Overview
381
Detector
Bytes per Number of Bytes per Source Source Data Rates
System
Event
Sources
per Event
(MBytes/s)
Vertex
8400
5
1680
3.4
Drift Chamber
3600
18
216
0.4
DIRC
2700
1
2700
3.5
Aerogel
300
1
300
0.3
Calorimeter
7700
18
275
0.6
IFR
800
1
800
1.6
Trigger
800
3
400
0.8
Totals
24,300
47
Summary of event data contributions from each detector type at ten times
nominal occupancies and data rates per data source for each detector system for a 2.0 kHz
trigger rate.
Table 10-3.
A trigger data path that conducts detector data to the trigger processor.
The detector control interface is responsible for initializing the system and for monitoring
and debugging. Overall management, error logging, and databases are part of the online
system described in Chapter 11.
The data acquisition interfaces are envisioned to be very similar across the systems in the
following respects. The crates for each detector system will have
readout modules,
a fast control distribution module,
a readout controller, and
an event-data port card unit (data source).
For each event, the event-assembly network moves the approximately 25 kbytes of event
data, summarized in Table 10-3, from the data sources to the Level 3 processor farm. The
event-assembly network is implemented with commercial network technology. At a Level 1
trigger rate of 2.0 kHz, this technology is expected to be able to transport the approximately
50 Mbytes/s of data expected at 10 times the nominal design point. This type of technology
is easily upgradable and can accommodate the needs of luminosity, trigger processing, and
bandwidth upgrades.
Technical Design Report for the BABAR Detector
382
10.2
Electronics
Silicon Vertex Detector
This section describes the storage and transport of data from the silicon vertex detector
transition card into the data acquisition system. The readout chip electronics and the
transition card are described in Chapter 4.
10.2.1 Requirements
In the vertex detector readout, the data are digitized and sparsi ed at the detector frontend. Unlike the data from the drift chamber and calorimeter, the silicon vertex detector
data must be bu ered before being moved from the detector. This minimizes the deadtime
that can arise from multiple triggers occurring on a timescale which is short compared with
the readout time, or from events that generate large numbers of hits.
Because of the large number of signal channels in the silicon vertex detector (approximately
150,000), the electronics for this detector are required to produce sparsi ed data streams.
For each hit that occurs within a speci ed time window (e.g., 0:5 s) of a Level 1 trigger,
the system must generate three numbers: the number of the channel (strip) that was hit; a
rough measure of the time-over-threshold of the pulse produced by the hit; and the time at
which the pulse rst crossed the threshold.
Data fragments consisting of the above three numbers are generated by the silicon vertex
readout chips and placed into the chip's three-deep bu ers. Subsequently, the data in these
bu ers are gathered event by event, and merged into larger data fragments, each covering
a large number of hits associated with a particular trigger. Further numbers on particular
performance parameters like dynamic range, noise levels, and speeds are given in Chapter 4.
10.2.2 Overview
A block diagram depicting the organization of the vertex detector electronics is shown in
Figure 10-4. Each chip processes the signals from 128 strips as follows. A time-over-threshold
signal is created, which is a logarithmic function of the pulse height. For each clock period,
depending on whether the signal is above or below threshold, a binary digit with a value of
1 or 0, respectively, is input to a latency bu er, awaiting Level 1 decisions. These bits are
preserved for the latency period of the trigger, then erased. When a trigger is received, the
time history of the stored bits is examined, and the number of 1s within a certain period
before and after the trigger is determined. This number represents a measure of the pulse
height. In addition, the time with respect to the trigger of the start of the stored 1s is
Technical Design Report for the BABAR Detector
10.2 Silicon Vertex Detector
3 –5
Readout
chips
Silicon
Strips
128
383
DAQ Readout Card
3 Deep
Buffers
8 Bi–directional
Fiber Optic Links
Silicon
Vertex
"Personality"
Section
VRAM's
Config. Data
Clock
Trigger
Readout Strobe
1-95
7857A09
Figure 10-4.
Bit–serial Data
Transition
Card
Block diagram of the vertex detector electronics system.
found. This represents a measure of the time of arrival of the pulse from the strip. These
two numbers are then stored in the three-deep (FIFO) bu er for that channel. If there are
no 1s in the latency bu er close to the trigger, then a null-data indicator is stored in the
three-deep bu er.
During the readout of an event, the data in the last levels of the bu ers of all channels
containing valid data are read out in bit-serial mode, and channel serial numbers are added.
In this operation, several readout chips are connected in a daisy chain as shown in Figure 10-4.
The data travel a short distance to a transition card, which launches the bit-serial stream
on an optical ber link destined for a readout module (Section 10.10.4).
The optical ber link is bidirectional. In addition to the bit-serial data stream traveling
toward the readout card, a set of signals is carried in the opposite direction by a second
ber. These signals include: a system clock, Level 1 triggers, event data readout strobes and
controls, and several bits of con guration data. The con guration data can control certain
functions of the readout chips such as putting a chip into calibration mode.
A single bundle of data lines plus control lines is assumed for each readout section. A readout
section is either the z or side of a detector. To provide redundancy in case of a single point
failure, pairs of readout sections (z and ) on the same detector are grouped together such
that the data and control lines of one side can service chips on both sides, if necessary. There
are 208 readout sections servicing the silicon vertex detector.
The readout chips are daisy-chained on a serial communications pathway operating at
80 MHz. The data are passed through the chips rather than on a common bus. The advantage
of this is that a chip can only break the line in one place as opposed to taking down an entire
bus. The chips can pass their respective data in either direction on the pathway, and both
ends of the pathway are accessible to external electronics.
Technical Design Report for the BABAR Detector
384
Electronics
The data transmission control is implemented as a redundant set of readout strobes. A
pulse on the readout strobe gives the rst chip in the daisy chain priority to begin sending
its data. The packet of serialized data is always prefaced with a header and terminated by a
trailer. When this data has been transmitted, the token giving priority to the data pathway
is passed to the next chip in the daisy chain, and the process is repeated.
10.2.3 Data Format
The format of the bit-serial data travelling between the transition card and the readout card
contains three elds: a header, a body, and a trailer. The header has a unique code that
enables the personality card to locate unambiguously the start of an event data message.
The body contains the information from a number of hits. The number of hits will, of course,
vary from event to event. The information from each hit is contained in a sub eld containing
the three numbers|strip identi cation, pulse height, and time of arrival. These sub elds
are of constant length. The end of the sub elds is announced by a unique trailer bit pattern.
10.2.4 Description of the Readout Module
As shown in Figure 10-5, the signals on eight optical ber pairs are received by the personality
module, which is physically part of the readout card. A generic readout card is shown in
Figure 10-30 in Section 10.10.4. Here, the optical signals on the ber are converted to
bit-serial digital logic signals by the ber optic receiver. Next, the serial-parallel converter
converts the bit-serial data to word-serial data in which the words are 16 bits wide. The
words are then passed to the link controller.
The link controller on the vertex personality module must identify the start of data from
the header code and the end of data from the trailer code. Since each chip in the chain has
included a trailer after its data, the data acquisition system can count how many chips sent
data in response to a single read strobe. The system knows from a con guration database
how many chips should report data from a given readout section, so it can recognize the end
of the data for an event.
Relying on the format described above, the link controller in the personality module is able
to extract the hit data from the incoming data stream and prepare it for forwarding to the
VRAM on the DAQ readout card. When it senses the last trailer eld, the link controller
raises a \Done" ag. When all such ags are raised, an event-data-ready signal is strobed
to signal the onboard CPU that a new event has arrived and to enable the issuance of the
next readout strobe.
Technical Design Report for the BABAR Detector
10.3 Drift Chamber
Bit-serial
Data
Fibers
to/from
Transition
Card
385
Fiber-optic
Receiver
Bit–serial
Data
Fiber-optic
Transmitter
SerialParallel
Converter
16
Link
Controller
To
VRAM
Reception
Error(s)
Command
Data
From 7 Other
Link Controllers
Event Data
Ready
1-95
7857A10
Figure 10-5.
1
Word–serial
Data
Block diagram of the vertex detector speci c section of the readout module.
The readout strobe is not bu ered on the chips. The DAQ readout card must never send a
readout strobe until all of the data corresponding to the previous readout strobe have been
read out.
The serial data stream enters the VRAM bu ers synchronously with the 80 MHz serial
transmission clock (5 MHz after 16 bit deserialization). The VRAM bu ers are large enough
to store approximately 500 events.
Each of the readout modules also has a processor for signal processing and message handling.
Upon receipt of a Level 1 trigger, the processor copies all data from the relevant time slot
of the VRAM to conventional RAM. The processor extracts the physical quantities (times
and amplitudes) from the information stored and passes the result on for event assembly.
Additional bu ers are required to hold the post-processed events until they can be read out.
10.3 Drift Chamber
10.3.1 Requirements
The readout system for the drift chamber should not introduce any signi cant degradation of
the performance of the chamber. The timing resolution should be sucient to determine the
distance of closest approach of a track to each wire with a precision better than the inherent
chamber resolution. The charge deposition should be measured with an accuracy compatible
with the statistical uctuations in the formation of the primary clusters. Cell hit information
should be made available for the segment nder logic with an eciency approaching 100%
Technical Design Report for the BABAR Detector
386
Electronics
Cell wire
preamp
Figure 10-6.
c
c
line
receiver
-
Hit
detection
-
Time
measurement
-
Amplitude
Measurement
s
-
Segment
nding
--
Latency
bu er
?
Feature
extraction
To track nder
-
To DAQ bus
-
Functional diagram of the drift chamber readout system.
and with minimal contributions from electronic noise or neighboring channels. A functional
diagram of the readout chain is shown in Figure 10-6.
The position measurement should have a precision of 140 m. With an average drift velocity
of about 25 m/ns, this distance translates into a time precision of approximately 5 ns. A
timing resolution of about 2 ns would then degrade the measurement by 10%.
The energy loss in the gas is usually measured by integrating the charge collected on a
sense wire. The uctuation in the number of primary clusters is about 7%, so the charge
measurement requires a precision of a few percent.
10.3.2 Preampli ers
The preampli er is of the transimpedance type, giving an output voltage proportional to the
instantaneous current owing into the sense wire. The current shape from the drift chamber
wire has a rapid rise when the charge multiplication occurs, and then falls inversely with
time. For a shaping time of 6 ns and a gas gain of 2 104, the current pulse from one
produced electron is about 60 nA, with a dynamic range of about 250. Thus, a noise gure
of about 1000 electrons is required. The current cuts o when the last positive ion reaches a
eld wire 100 s later. In the inner wires of the chamber, where the count rates might reach
80 kHz under 10 background conditions, there would be pileup on the tail of the previous
event. The pileup would have little in uence on the dE/dx information quality, but would
cause some ineciency in the discriminator. Therefore, the preampli er includes a pole zero
cancellation circuit with a time constant on the order of the drift period (300{500 ns). The
preampli er risetime will be optimized to the digitizing scheme selected.
The use of an unsaturated gas in the drift chamber requires careful control of the parameters
that determine the gas density, e.g., the gas temperature. The power consumption of
the preampli ers which are all connected to one end of the drift chamber is of concern
Technical Design Report for the BABAR Detector
10.3 Drift Chamber
387
in this regard. The preampli er will be built as a bipolar semicustom ASIC, which allows a
reduction in power of a factor of 10 compared to previous designs. A goal of 4 mW/channel
is envisioned. The chip will have six channels, and several chips can be assembled on a
preampli er card to match the geometry of the drift chamber. Each preampli er chip is
individually protected by a polyfuse. In the event of a chip failure which shorts the power
input at the chip, the polyfuse will open and allow the other preampli er chips on the board
to remain powered. The fuse can recover from a temporary short or a surge by cycling the
power o and on. It should be noted that a loss of six cells on a layer does not prevent the
segment nder from working.
10.3.3
Digitizer Options
Two classes of implementation schemes have been studied for the digitizer: those in which
the time and amplitude measurements are done separately, and those in which the two
measurements are combined in a single device. In the latter class, the ash analog-todigital converter (FADC) is the natural candidate. It provides both time and amplitude
measurement capability, does not require external gating, can digitize continuously with no
deadtime, and is naturally compatible with a clocked dual-ported latency bu er. When
operated synchronously at a submultiple of the machine frequency, the FADC provides an
automatic rejection of the machine-related background noise. An FADC readout system can
be assembled from o the shelf components from multiple manufacturers. An FADC-based
readout system is also a formidable debugging tool|it is essentially a 7000-channel digital
storage oscilloscope.
The second class of implementation schemes uses a specialized device to measure the time.
For deadtimeless operation, a ash time to digital converter (FTDC), or time memory cell,
will ful ll the requirements. The amplitude measurement is still achieved with an FADC but
no time information is expected to be extracted from the data; thus the timing objectives
can be achieved at a substantially lower clock rate. It is slightly more complex, not immune
to machine related noise, and requires a low-time-walk discriminator to trigger the FTDC.
In the all-FADC approach, a discriminator is also present to generate the hit information for
the segment nder logic, but then the timing requirements are very modest.
An interesting variation of the FTDC/FADC approach is to build many such channels on a
single ASIC, including a shallow latency bu er. The advantages are lower power dissipation,
lower cost per channel for the components, and a considerable saving of space required on
the readout boards. A larger number of channels can then be packed on every readout board,
thus reducing the number of readout crates.
Technical Design Report for the BABAR Detector
388
Electronics
10.3.4 Implementing the All-FADC Scheme
An FADC readout system has been studied on the test drift chamber together with various
algorithms to extract the time information. The sampling frequency for the tests was
250 MHz, close to the 238 MHz machine frequency. The performance at lower frequencies
(250/2 or 250/3 MHz) could be deduced from the same measurements. Since commercial
FADCs have a very short time aperture, it is sucient to use one sample out of every two
or three to simulate the operation at reduced clock rate. At 83 MHz(250/3), the position
resolution achieved with the FADC readout channel was 135{142 m. That corresponds
to a degradation of 20% compared to the 115 m achieved with the TDC readout. At
125 MHz, the degradation reduces to 10%. It should be noted that the above gures are
somewhat unfair to the FADC method. The time-to-distance function is obtained by an
iterative process that uses only the TDC time measurement information. The same function
is applied to the FADC time measurement, though it has been optimized for the TDC system.
Expected improvements in the time extraction algorithms may lead to the rm conclusion
that one-third of the machine frequency (79 MHz) is a satisfactory sampling rate. But at
this time, the all-FADC system operating at 119 MHz (one-half machine frequency) remains
our baseline design. No technical uncertainty is involved in building such a system, and it
has been demonstrated to achieve acceptable performance with a drift chamber very similar
to the nal one. FADCs which operate at 119 MHz are expensive and power hungry. The
use of pairs of slower FADCs with interleaved sampling is being pursued. The uncertainties
of this approach are discussed in Section 10.3.6.
Time Measurement
One of the diculties in achieving a very good timing resolution with an FADC in the
BABAR drift chamber originates from the low-pressure, low-Z gas. This choice minimizes the
multiple scattering and yields very well-de ned track trajectories. However, it also reduces
the density of primary electron clusters along the track to about one cluster per millimeter.
For tracks that directly cross a wire, this represents a uctuation in drift time of about 15 ns
for the arrival of the rst cluster. Fortunately, this time di erence reduces rapidly as the
distance of the track from the wire increases and soon becomes a marginal e ect. However,
the coarse cluster granularity has a serious e ect on the pulse shape of the signal extracted
from the wire. It causes a succession of current pulses that can cover the full drift period
in the case of an anode-crossing track. As a result, the pulse shapes are very di erent from
event to event, making the time extraction a dicult task.
Various de-convolution methods have been tried. They succeed in separating the individual
clusters when they are suciently far apart in time, but at the leading edge, where the
geometrical e ect makes the clusters concentrate for tracks away from the sense wire, the
Technical Design Report for the BABAR Detector
10.3 Drift Chamber
389
situation is very dicult. So far, the so-called TDC-mimic algorithm, which is a linear
extrapolation of the rst two bins to the baseline, gives the best results. Other algorithms
are being studied. The timing performance might also bene t from tuning the preampli er
shaping. New, faster preampli ers with adjustable shaping are now being installed on the
test chamber to improve these measurements.
For the extrapolation algorithm to work accurately with small signals, it is necessary that the
smallest signals, corresponding to single clusters, have an amplitude on the order of 10 ADC
counts. This partially determines the dynamic range of the FADC. To prevent the digitizer
from over owing with large pulses, other multiplying factors have to be taken into account:
2 for track angle, 2.5 for highly ionizing particles, and 4 or 5 for cluster pileup, for a total
factor of 250, or 8 bits. An e ective dynamic range of 9 bits would be more comfortable.
Charge Collection
The evaluation of the primary charges deposited in the volume of a cell is made by summing
the FADC samples over the drift interval. The statistical uctuations in the formation of
the primary clusters already introduce an average variation of 7% along the full length of
the track. Thus, a measurement of the integrated charge with a precision of a few percent
will not degrade the dE/dx evaluation, and a precision of 6 bits is sucient for the charge
measurement. If this requirement is combined with that of time evaluation, the FADCs
should have an e ective dynamic range of 8 to 9 bits and an accuracy of 6 bits. It would
then be satisfactory in principle to use a 6-bit digitizer, and obtain the 8- to 9-bit dynamic
range with a nonlinear or a multislope conversion.
Readout Module
The readout module is composed of two distinct sections: a front-end speci c to the drift
chamberand a common data acquisition section. The common section contains the latency
bu ers, some local RAM, the processor that will extract the physical features from the raw
data, and an interface to the VME crate bus. The front-end section carries the FADC chips,
the signal conditioning electronics, the discriminators that provide the wire hit information,
and the segment nder logic, as shown in Figure 10-7. Each di erential signal line from the
preampli er arrives at the readout board at a line receiver featuring a large common mode
rejection. The output of the line receiver feeds a discriminator and a driver ampli er for the
FADC chip input. This driver also includes a low pass lter to optimize the signal shape for
the time extraction. An o set voltage is supplied to the driver from a 6-bit CMOS DAC, in
order to adjust the FADC pedestal.
Technical Design Report for the BABAR Detector
390
Electronics
From Neighbor (–1)
To Neighbor (+1)
24 Disc.
Segment Finder
4
Neighbor Disc.
Local Disc.
Segment Bit(s)
Control
4
24
Analog in
Reference
256 KBytes
VRAM Bank 2
Ped.
Clk
(24 Shielded
Twisted Pairs)
24 Line
Receivers
Filter
Preamp +
Preamp –
24 x
8 Bits
Clk
Clk
Sequential In
Clk
Contol
256 KBytes
VRAM Bank 1
Random Out
Buffer 1
24 FADCs
Ped.
Random Out
Buffer 2
24 FADCs
Analog Differential
Lines from Preamp
To Track Finder
24 x
8 Bits
Clk
Sequential In
Clk
Timing Generator
Buffer2 Clock (f/2)
Buffer1 Clock
FADC Clock (f=119Mhz)
Control
24 Voltages
DACs
ADC Pedestals
Disc. Thresholds
24 Voltages
Control
Local CPU
3-95
7857A19
VME Interface
Figure 10-7.
Drift chamber module block diagram.
Technical Design Report for the BABAR Detector
10.3 Drift Chamber
391
The FADC chips are 8-bit devices operated at a clock frequency of 119 MHz, which is almost
twice the maximum speed at which the VRAM bu er memories operate. Thus, it is necessary
to use two interleaved VRAM units for each FADC, interfaced through two bu er latches.
The chamber geometry will determine how many FADC/discriminator channels are put onto
each board. For the baseline design, the chamber is organized into sets of four layers each,
called superlayers, which are themselves organized into groups of 24 cells called supercells.
Since the data acquisition board supplies the information to the segment nder, and the
segment nder needs the information from one supercell, the number of channels on the
board will be equal to, or a multiple of, the number of cells in a supercell. The segment
nder logic also requires that some overlap be provided between adjacent supercells. An
interconnect is thus required between neighboring readout modules, forming a daisy-chained
closed loop. In the baseline con guration, three signals would be carried from one module
to the next.
10.3.5 Calibration
The gain calibration is accomplished by injecting pulses at the preampli er input. To limit
the number of pulser cables going to the preampli ers, a compromise is made between the
exibility in the trigger patterns and the size of the system. Each channel of a preampli er
integrated circuit can be pulsed independently. It is therefore possible to detect channels
shorted together and to measure crosstalk. Each preamp cable has two or four pulser lines,
for a total of about 1400 lines in the latter case. The pulser signals originate from the
readout module. A lesser number of pulser lines is required if the pulses are distributed
in a ring geometry. One pulser line is then connected to one of the four channels of each
preampli er chip belonging to that layer. This requires 320 pulser lines. Every line serves
at least 20 channels, but this granularity is still small enough to generate useful patterns
to exercise the trigger logic and the data acquisition system, in addition to the basic gain
calibration function. The pulser system only calibrates the electronic gain. It must be
combined with the information extracted from the data itself, such as the histograms of the
integrated charge for each cell, in order to account for the wire gain.
10.3.6 Research and Development
Operation at a Lower Sampling Frequency
The research program concerning the FADC time extraction algorithms will continue, since
even a slight improvement would justify reducing the sampling rate from 119 MHz to 79 MHz.
At this lower rate, a CMOS FADC may be used, reducing cost by a factor of 5, power
Technical Design Report for the BABAR Detector
392
Electronics
Signal
s
HH
HH
Disc. H
x
s
Stop
Start
Clock
Input
Figure 10-8.
FTDC
FADC
HH
HHH
Two stage
latch
MUX
VRAM
Flash TDC/FADC channel principle.
consumption by a factor of 4, and space on the the readout module by a factor of 10.
Current candidate CMOS FADCs have only 6 bits and thus would have to be operated in
a nonlinear mode. This has been done for other technologies but needs to be studied for
CMOS devices. A test board will be constructed to evaluate this possibility and measure
the maximum practical dynamic range expansion of the 6-bit CMOS FADC that is available
commercially. The LBL group also has designed a 6-bit 100 MHz CMOS FADC prototype
chip. It will be available for testing shortly. An 8-bit device could in principle be built
with the same technology, with a somewhat lower maximum sampling frequency yet to be
speci ed. It is also possible to use two slower FADC chips and operate them with split phase
clocks to increase the sampling rate. Two CMOS units could be operated at 59.5 MHz to
achieve an e ective 119 MHz sampling rate. However, this requires a good matching of the
transfer functions of the two FADC chips and might require a more expensive, more linear
version of the chip. A test board will be made with CMOS devices to check the dynamic
range compression method and the split phase technique.
Time Measurement with an FTDC
An alternate approach, aimed at lowering the sampling frequency, is to complement the
FADC readout with an FTDC time measurement. The FTDC measures the time with very
good resolution, and the FADC measures the charge only, while still providing a redundant
lower precision time measurement. This approach is interesting if the sampling frequency
is brought down to that of the VRAM bu ers (40{60 MHz). Only one VRAM chip would
then be required per channel. This is quite feasible since the FADC data would not be the
primary source of time information.
As shown in Figure 10-8, the FTDC produces an encoded measurement of the elapsed time
between a pulse of the sampling clock and the instant the discriminator res. There is only
one byte of TDC information for each hit, and it can be inserted in front of the FADC
byte stream. The two latch stages following the FADC provide the required digital delay
for that operation. This approach requires only one VRAM per channel instead of two, but
Technical Design Report for the BABAR Detector
10.3 Drift Chamber
393
the FTDC is an extra device. The FTDC function could also be achieved coarsely with
a 6-bit CMOS FADC sampling a voltage ramp synchronized with the sampling frequency.
In either case, the cost of this approach would be equivalent to that of the 79 MHz pure
FADC operation and be subject to the same uncertainties relative to the dynamic range
compression feasibility. However, since the FADC would no longer measure the time, the
dynamic range requirement could probably be relaxed by a factor of 2.
Readout ASIC
An integrated drift chamber readout circuit could reduce the cost and complexity of the
readout system. This readout IC operates like that for the silicon vertex detector, i.e.,
multiple channels of electronics are read out through a single circuit only after a Level 1
accept. This architecture reduces the processing time on a DAQ card since only channels
with hits are transferred from the readout IC. The architecture also minimizes the number
of readout (VRAM) channels required on the DAQ cards and reduces the total number of
readout crates required. Eight channels of drift chamber readout electronics are expected to
t on one integrated circuit.
The proposed drift chamber readout IC is composed of four main components: a time-todigital converter (TDC) marks the time of the threshold crossing to an accuracy of less than
1 ns; a 6-bit FADC samples the drift chamber output pulse at 100 MHz, a dual-port DRAM
stores the digitized drift chamber output pulse for the entire trigger latency of 10 s; and a
SRAM secondary bu er stores up to 3 s-long sets of drift chamber hits.
The TDC is crucial in determining the exact time of the threshold crossing. The TDC
operates by dividing the input clock period into 32 bins. The values of the bins are stored
during each threshold crossing, and these values determine the portion of the clock period
in which the threshold crossing occurred. The operation of the core of the TDC, the delay
elements and the phase detector, has already been tested with a previous IC. An updated
TDC currently in fabrication will test a new phase detector and decoding scheme.
An FADC built with auto-zeroed comparators has been designed. Simulations show that
6-bit operation at frequencies up to 100 MHz is feasible. The comparison scheme itself is
good up to 10 bits. A 6-bit test circuit has been submitted for fabrication. Veri cation of
operating parameters must be tested.
The DRAM and SRAM cores are designed and in fabrication. Some simple tests will
be undertaken to determine if the memories operate correctly. These memories are built
with memory generators. This allows adjustable size memories to be implemented for each
design. Some slight modi cations might be needed once the drift chamber IC channel pitch
is determined. There should be no e ect on the memory operation.
Technical Design Report for the BABAR Detector
394
Electronics
Once the above blocks have been tested, they can be connected at the board level to test the
system functionality. A complete drift chamber channel can be designed with these blocks
at the IC level concurrently with the system test. A four- or eight-channel prototype chip
must be built to study power, noise, and crosstalk issues. The die size of an eight-channel
chip is expected to be 16 mm2.
10.4
DIRC
10.4.1 Requirements
In the DIRC, the Cherenkov angle is determined from the position of hit photomultipliers
among the array of 13,400 tubes situated on the stando region backplane. The pulses are
the result of the impacts of single photons on the photocathodes. The times of arrival of the
photons from a given event are scattered within a 50 ns time window.
The electronics requirements are: to record single photoelectron hits quickly enough to avoid
being swamped by the accidental rates, to provide 1 ns time resolution, to discriminate
against background, to implement modest amplitude measurement (6 bits) in order to
monitor the gain of the tubes and maintain the threshold settings, and to provide adequate
bandwidth to safely transfer the physics data to the data acquisition system. Special
calibration runs are needed to determine pedestals and gains.
Possible Readout Implementations
Several options have been considered, and their cost and performance evaluated: (1) full
sampling, (2) full sampling with zero-suppression, (3) zero-suppression with local bu ering
with the Level 1 strobe starting the transmission of the data that arrived during the trigger
latency, (4) the same as (3) but with only the data which arrived during the trigger timing
window being transmitted. The above architectures can be implemented many ways. They
di er in location of the components (on the detector or on the readout modules), the cable
plant, number of readout modules and crates, amount of VRAM, and bandwidth of the
transmission medium between the detector and the readout modules. A comparison of
architectures is shown in Table 10-4.
At this time, options (3) and (4) are preferred because the number of cables is kept to a
minimum, the amount of data to be sent to the data acquisition system is suciently reduced
so that the whole DIRC readout resides in one or two crates, and custom integrated circuits
can be used. The price is modest, and the custom ICs are not too complex.
Technical Design Report for the BABAR Detector
10.4 DIRC
395
Option Number of
On
Cables Stando
1
13.4k
No
2
|
Yes
3
|
Yes
4
|
Yes
Table 10-4.
Mbytes/s
215k
6k
100
10
PMs per VRAM Readout Readout
VRAM Chip Chips Modules Crates
2
6700
210
14
27
500
65
4
1.34k
10
3
1
13.4k
1
1
1
Comparison of four DIRC readout options.
Choosing between options (3) and (4) requires additional detailed study to judge whether
the gain in bandwidth justi es the extra complexity of the digital front-end logic, since these
added features can be easily implemented on the readout module.
Although the proposed architectures make the use of custom chips attractive, it has been
decided to present, as the baseline implementation, a system in which discrete components
are used wherever possible. It is believed, however, that an optimal system will be built
mostly out of custom or semicustom integrated circuits for lower cost, improved performance,
and more reliability. Such an implementation is explored in Section 10.4.3. The baseline
solution implements option (4) as far as the functionalities are concerned, but the bandwidth
characteristics are derived assuming option (3). This allows option (3) to be a fall-back
solution for option (4).
10.4.2
Baseline Design
Overview
The plan is to install the front-end electronics as close as possible to the tubes. A data
driven scheme has been devised in which digitization occurs only when the pulse is above
a threshold. The time of each hit is measured to 1 ns with higher precision left as an
option. Amplitude measurements with 6-bit precision are foreseen for calibration events
only. Although the pulse height carries no physical information on a single hit, it would be
useful to record it on an event-by-event basis to sort out pileups and provide assistance with
debugging. This possibility is further explored in Section 10.4.3. An option is to add a high
threshold discriminator on each channel to ag hits with anomalously high pulse heights.
Digitized data are stored for the duration of the Level 1 accept latency (10 s) and are read
out upon receipt of a Level 1 strobe. An intermediate stage between the front-end and the
readout module is needed to synchronize the data ow and multiplex it to 15 optical links.
One readout crate houses the entire DIRC readout.
Technical Design Report for the BABAR Detector
396
Electronics
Front-end Card
Interface
Front-end
Channel
1 Card
Interface
Channel 1
ADC
Channel 64
ADC
Channel 64
ADC
Front-end Card
Interface
Front-end
Channel
1 Card
Interface
Channel 1
15 Cards
ADC
Channel 64
ADC
Channel 64
ADC
Figure 10-9.
Optical
Fiber
MUX
DAQ
Frontend
Controls
Front-end Crate
Layout of the DIRC front-end electronics.
The layout (Figure 10-9) calls for circuit boards located in crates situated near the tubes.
Each card would carry the front-end electronics for 64 tubes. The idea of integrating
the front-end electronics on boards shared with the high voltage system, attached to the
back of the tubes, has been abandoned for the present because of the complexity of the
interconnections and because the front-end electronics would get buried within the magnetic
shield of the stando box.
Short cables connect the bases to the circuit boards. A block diagram of the front-end is
given in Figure 10-10. The main features of each channel are: timing|a constant fraction
discriminator (CFD), TDC, and FIFO; the readout logic; and the slow analog calibration
chain. A preampli er with a gain of four and a bandwidth of 1 GHz may be needed at
the input, depending on the choice of photomultiplier tube. Most of these features can be
implemented with o -the-shelf components. A more detailed description of the items is given
below.
Timing Chain
After the CFD, the time is measured by a double-tiered TDC. A coarse range is obtained
by parallel loading 10 bits of a time register from the time counter, which records one pulse
out of 16 from the storage ring rf (1 tick every 33.6 ns). Time measurement over the total
Level 1 latency requires 14 bits. Sixteen bits have been allowed. The vernier range is
obtained by using a high-precision 6-bit TDC. Early versions of this device have been built
for the DELPHI experiment [Gen93]. The output of the 16-bit register is connected to a
FIFO deep enough to keep 16 hits. A time window comparator and interface, and readout
Technical Design Report for the BABAR Detector
10.4 DIRC
397
L1
32 MHz
Time counter
32 MHz
DAC
10
32 MHz
CFD
6
32 MHz
TDC
PMTs
TDC
Slow
peak
sensing
Slow
peak
sensing
Rd
16
10register
16 bit
16 bit register
“Delay”
“Delay”
Channel 1
Time window
comparator
16x16
Rd FIFO
16
6
CFD
Local
Bus
Interface
and
Readout
Sequencer
Output
Mux
&
Driver
wr
16x16
FIFO
wr
Analog
Mux
To Mux
Card
ADC
Channel 64
Figure 10-10.
Block diagram of the DIRC front-end electronics.
sequencer enable the transmission of the correct data. The trigger time is compared to the
hit times stored in the FIFO, and those which are in time are transmitted. The TDC and
the FIFOs will be implemented inside a 64-input standard cell CMOS digital chip. The
interface and readout sequencer adds the address of the photomultiplier within the card to
the data word, which becomes a three-byte string. The output MUX/driver sends the data
to the multiplexer boards.
Analog Chain
The pulse height will be measured during calibration. This requires an analog chain incorporating slow, charge-sensitive circuitry on each tube, which feeds a 6-bit ADC through a
16-to-1 analog multiplexer.
Multiplexing
From each front-end circuit board, the three bytes of data are sent at 400 kbytes/s to an
intermediate crate. Assuming 20 Mbytes/s for the bandwidth of the readout crate backplane
bus and a safety factor of about 4 to absorb the uctuations in event size and trigger timing
leads to a layout with 15 circuit boards per front-end crate. To minimize the cable plant,
it is advantageous to add a multiplexer board in the sixteenth slot of such a crate. Each
multiplexer board will concentrate the data of its front-end crate onto a 12 Mbytes/s optical
Technical Design Report for the BABAR Detector
398
Electronics
FADC
16 Channesl
ANALOG CHIP
4 Channels
Bipolar 10 GHz
PMT
Gated
Integrator
Leading edge
Discriminator
d
dt
12 bit
counter
6 bit Time vernier
16 Channels
Analog MUX
Delay
DIGITAL CHIP 16 Channels CMOS 0.7
6 bit
FADC
Zero
crossing
Discriminator
Time
window
comparator
6
L1
18
Write
L1
Latency
Clr
16 x 32
FIFO
Read
Clear
L1
Readout
controller
Control
Registers
FF
Threshold
DAC
Serial
Interface
DAC 8 Channels
Fast
Serial
Output
Serial
Link
Block diagram of the DIRC front-end electronics showing the suggested
implementation with ICs.
Figure 10-11.
ber serial link. This board will also add a 4-bit board address to make the data word
4 bytes. There are 16 front-end crates, each with its own readout ber.
10.4.3 Research and Development
Most of the front-end described above is assembled from commercial components. It is
well understood and buildable and has been used primarily for budgeting purposes. Only
the TDC and the FIFOs are ICs. It is believed that a somewhat improved architecture
can be built by using ICs throughout (Figure 10-11). Although such a design entails more
development work, it should cost less and lead to a more reliable system.
In this scheme, the basic granularity is 16, although a better value may emerge from
simulations. The philosophy of the timing measurement in this implementation is exactly
the same as in the baseline design. The di erence is that the amplitude will be measured
Technical Design Report for the BABAR Detector
10.4 DIRC
399
during data taking. There is one FADC (6-bit, 100 MHz) serving 16 photomultipliers. If the
rate is too high, not all tubes from a given readout section of 16 in an event will have their
amplitudes measured.
The most elegant layout calls for circuit boards, attached behind the tube bases, which carry
the front-end electronics for 16 tubes. There is enough room on these boards to consider
sharing them with the high voltage system, should we choose to do so. Short cables connect
the bases to the circuit boards. On each PC board, there are four analog chips, one or two
digital chips, an analog multiplexer, a 6-bit 100 MHz FADC, two eight-fold DACs, and the
output circuitry to send the data to the readout modules via digital multiplexers. More
detailed descriptions of these are given below.
Analog Chip
An analog chip services four photomultipliers. For each tube, there is a discriminator and
a shaping circuit to feed a charge-sensitive ADC. From the analog chip emerge the four
discriminated output pulses, which are fed to the digital chip, and the four analog outputs
of the shaper, which are directed to a 16-input/1-output analog multiplexer. In addition to
the four signal inputs, the analog chip receives four threshold inputs from DACs. The choice
among bipolar, CMOS, and BiCMOS processes is being studied.
Digital Chip
Each circuit board has one digital chip which accepts inputs from 16 tubes. The time
measurement is identical to that of the baseline design. The interface and readout sequencer
and the time window comparator could be made as a second separate chip or remain discrete
components. These decisions will be made as the design matures. The digital chip receives
16 signals from the analog chip, the master clock, the trigger signal, and a calibration line
used to measure the pedestals. The calibration proceeds as follows. A pulser controlled by
the data acquisition system generates a signal which enables the discriminator threshold to
be bypassed. The signal is also delayed to provide a pedestal trigger with the proper latency.
For calibration or noisy channel suppression, individual tubes can be masked using a slow
serial command line on the circuit board. The outputs of the digital chip are a 5-byte data
bus daisy-chained across six circuit boards towards the ber driver, and a 4-bit address bus
to command the analog multiplexer. Standard cell CMOS is the technology chosen for the
digital chip.
Technical Design Report for the BABAR Detector
400
Electronics
Front-End Layout
Nearly one thousand circuit boards will be used, each having four analog chips, a digital
chip, an analog multiplexer, an FADC, and two eight-fold DACs. On top of the input and
output signals already mentioned, cabling will carry the low power distribution lines. Slow
control will have to be implemented locally. These features have yet to be designed.
Data Transmission to the Readout Crates
The data rate from all the digital chips is 100 Mbytes/s. Assuming the front-end readout
modules accept data at 12 MHz, the ow must be shared among ten streams, each receiving
data from 84 front-end sections. This corresponds to ten multiplexers with 84 inputs and
one output, where a single address byte, that of the readout section, will be added to the
data. Optical bers similar to those considered for the vertex detector are appropriate.
Alternatively, two 1 Gbytes/s optical bers can be used.
10.4.4 High Voltage System
If a traditional resistor-divider photomultiplier base is chosen for the DIRC photodetector,
a photomultiplier with a maximum operating voltage of 2 kV will draw a current of approximately 0.3 mA or less. In the proposed system, the high voltage will be distributed
over long cables to a HV motherboard mounted on the photomultiplier support structure.
Each motherboard services 16 bases, for a total current load of approximately 5 mA per
HV channel. The motherboards will be connected to commercial multichannel high-voltage
power supplies.
The high-voltage power would be distributed via standard RG-59 HV coaxial cable, utilizing
a new connector scheme developed by LeCroy in collaboration with the AMP corporation.
This connector is an eight-fold coaxial block connector, fully rated and UL listed for 5000 V.
This results in signi cant savings over traditional SHV coaxial connectors. Connections to
the motherboards would be made using standard SHV connectors.
10.4.5 Low Voltage Power Supplies and Control Systems
The on-detector front-end electronics has been estimated to dissipate a total of about 20 kW.
The standard crates in the baseline design provide the needed low voltage and are water
cooled. Control systems have yet to be designed.
Technical Design Report for the BABAR Detector
10.5 Aerogel Threshold Cherenkov Counter (ATC)
401
10.5 Aerogel Threshold Cherenkov Counter (ATC)
10.5.1 Requirements and Overview
The signal processing provided by the o -detector electronics should be capable of fully
exploiting the intrinsic performance of the ATC. For good eciency, the threshold should be
set between 1 and 1.5 photoelectrons and the timing resolution should be less than a few tens
of nanoseconds in order to reduce the background from o -beam particles showering into the
detector. An approach based on discriminators and TDCs, or one based on ash ADCs, can
be used to achieve this goal. In the framework of the BABAR pipelined DAQ system, the most
natural choice is the second of these options. This also provides some additional advantages:
greater exibility in setting di erent thresholds depending on the expected signal
amplitude, which is a function of the particle momentum;
shorter possible deadtime in high background conditions; and
much more powerful diagnostic capability.
In the baseline design of the ATC detector, ne-mesh photomultiplier tubes (FM PMT) are
used as photon detectors. After a possible front-end ampli cation and shaping, these can
have an average pulse height and a shaping time similar to those expected from the drift
chamber preampli ers. Therefore, it may be convenient to use the drift chamber readout
modules. The dynamic range of the expected ATC may be easily accommodated in the 8-bit
e ective range of the drift chamber readout system.
10.5.2 Front-End Electronics
With the Hamamatsu ne mesh PMT available today, the gain in a magnetic eld of 1.5 T
is about 105 and a low noise current preampli er is required. To obtain a signal suitable for
a commercial ADC, a transresistance of 60 mV/ A is sucient. This would give a signal of
50 mV for a single photoelectron pulse at 105 PMT gain. This value is considered to be
the maximum requirement, since the PMT response still can be optimized with respect to
the orientation in the magnetic eld and improvements are expected in the performance of
the next generation of PMTs. Using the speci cations of a typical ampli er like the LeCroy
TRA1000, one obtains an estimate of the noise of 0.5 fC FWHM, which is equivalent to 0.03
photoelectrons. The preampli er is to be mounted directly on the PC board that supports
the PMT and can be connected to the o -detector electronics via twisted-pair shielded cables.
Technical Design Report for the BABAR Detector
402
Electronics
10.5.3 Readout
The details of the drift chamber readout board appear in Section 10.3. We plan to make
use of the processing power available on the board to perform the feature extraction after
the arrival of the rst level trigger. Once converted into amplitude and time information,
the amount of data to be shipped to the event builder would be negligible, on the order of
a fraction of a kilobyte. Provisions will be made to optionally send the sampled waveforms
for diagnostic and calibration purposes.
High Voltage
The ne mesh PMT will be operated at 2.5 kV, with a steady current of approximately 2 mA.
Existing commercial modular high voltage power supply systems are well-suited to the ATC
HV power requirements. The ATC requires 144 HV channels. In the proposed system, the
high voltage will be distributed over 12 cables to the PC boards where they will be split into
two 6-pin connectors, each carrying the HV power needed by a PC board.
10.6 Calorimeter
The electronics for the CsI(Tl) calorimeter must read out the 6780 crystals in the calorimeter
and provide information to the lowest level trigger as well as be able to respond to a trigger,
and fetch and process the data corresponding to that trigger. This section describes the
requirements of the calorimeter electronics and a proposed architecture which meets those
requirements [Hal94].
10.6.1 Requirements
The following items are the primary requirements [Req94] for the calorimeter electronics:
The calorimeter data acquisition system. The system should not make the resolution
for any photon with E 20 MeV signi cantly worse than that due to the intrinsic properties
of the detector alone. Speci cally, the data acquisition contribution to the error should be
less than half that of the CsI.
Technical Design Report for the BABAR Detector
10.6 Calorimeter
403
Dynamic Range. The calorimeter should have a dynamic range capable of reading out
showers in beam events with energies from 20 MeV to 13 GeV with a resolution of 100{200
keV at 20 MeV and also provide adequate resolution for reading out low-energy gamma rays
from radioactive sources for calibration. The dynamic range required for beam events is
17 bits. Allowing for a source calibration requires 18 bits of dynamic range, which is the
design goal.
Incoherent Noise. The rms incoherent noise for a single crystal should not exceed an
equivalent noise charge (ENC) of 350 electrons. This is the best that can be done at
acceptable cost. This number combined with the number of photoelectrons collected for
energy deposited (Chapter 7) determines the electronics contribution to the shower energy
resolution.
Digitization Contribution. The contribution to the resolution from the digitization
should be less than half the CsI(Tl) resolution. This requirement can be met with a 10bit ADC.
Reliability. The electronics components which are inaccessible must be provided in two
independent sets. Except for the crystals in the endcaps, access to components mounted on
the crystals will be impossible. Additional components are located inside the structure on
the end of the detector, but will be accessible by opening the detector endcaps.
Response to Background. The electronics should be designed to provide optimal per-
formance with the expected levels of background. At high background rates, soft photons
provide a source of noise which must be dealt with in the calorimeter (Chapter 12). This
requirement makes demands on the shaping time of the electronics and the signal processing
that must be performed on the output of the calorimeter.
10.6.2
Overview
The dynamic range requirement of the calorimeter combined with the large potential for
beam-related rf pickup at PEP-II make digitizing the signal close to the CsI(Tl) crystals
advantageous. The proposed design minimizes the cable length and dynamic range of analog
signals in the system, and places the digitization just outside the rf-shield of the calorimeter,
allowing access to the ADCs.
Technical Design Report for the BABAR Detector
404
Electronics
Clock
Gain
from
Amplifier
x1 Gain
12-Bit
Transmission
x1
Comparator
S/H
Vul
Digital
Range
Bits
Vul
x4
from
Amplifier
x32 Gain
13-Bit
Transmission
x1
x8
Vul
Range
Selection
Logic
+
Decoder
+
Analog
Multiplexer
Vul
Mantissa
ADC
10 Bit
Range Select
Overwrite
Custom Range-Selection Integrated Circuit (on ADC Board)
Figure 10-12.
Block diagram of gain stages and range encoding.
To minimize the distance the 18-bit dynamic range analog signal travels, an ampli er with
two gain stages is mounted within a shielded assembly on the rear of each crystal. By
choosing ampli ers with gains of unity and 32, only a 13-bit dynamic range is required on
the analog cables which go from the crystals to the digitizing board. At the digitizing board,
a custom IC receives the two analog signals and provides two additional gain ranges. The
outputs of this device are an analog signal, which is sent to an ADC at one of four possible
overall ampli cations, and two bits which specify which of the four gain ranges is being
used. The additional gain stages allow an 18-bit dynamic range oating-point number to be
constructed from a 10-bit digitization of the analog signal (the mantissa) and the two bits
specifying the range (the exponent). Figure 10-12 shows a block diagram of the layout of
the gain stages and range encoding.
Because it will not be possible to know which samples contain useful information until a
trigger decision is made, the digitized data ow down two paths. The rst path places the
digitized oating-point data in a VRAM bu er where they wait for a trigger decision. When
a trigger is received, the data acquisition system must use the trigger time to select the
correct range of stored data to analyze and process and then transfer these data to the next
level. In addition, the data from the calorimeter are used in the trigger decision, so digital
sums of energies must be formed and sent continuously to the central calorimeter trigger
logic. The trigger sums are more easily performed on the digital stream as the multiple
gains stages make analog sums more complicated.
Technical Design Report for the BABAR Detector
10.6 Calorimeter
405
from
other
crystals
Diode A
Diode B
shaping
Dual
Gain
~2 m
cable
avg.
shaping
Dual
Gain
from
other
crystals
25 Total
Amplifier Board
Average
A and B
Signals
Floating-Point
Conversion
Multiplexing
Parallel/Serial
Conversion
Serial/Parallel
Conversion
~30 m
FO
cable
Calculate
Charge
and Time
Level 1
Trigger
Primitives
Calibration
DAC
Voltage
Regulation
ADC Board
DAQ Board
Calorimeter system diagram showing preamp, ADC, and data acquisition
(DAQ) functions.
Figure 10-13.
Figure 10-13 shows a block diagram of the components of the proposed readout system. The
system has three main board types.
Preamp boards. One of these is mounted on the back of each CsI(Tl) crystal. They
are not accessible after the calorimeter is constructed, so they must be reliable and contain
a minimum of components. Each board provides an independent channel for each of the
two photodiodes on the crystal (outputs A and B). Since it contains the rst ampli cation
stage, it must be well shielded. In addition, the power consumption should be low, as the
temperature of the crystals needs to be regulated.
ADC boards. These are located at the end of the detector, just outside the rf-shield of
the calorimeter. They should be accessible on the timescale of one day. Each board receives
the signals from the A and B diodes on each of the approximately 25 crystals corresponding
to the trigger towers. It contains the custom range encoding chip which chooses the gain
then outputs two bits specifying the gain used. The board also includes the ADC chips, and
its output is a serial digital stream.
Readout modules. These boards each receive the serial digital data from one ADC board,
then de-serialize the data and store them in a VRAM bu er for later analysis. They also
form trigger sums and ship the trigger data to the central calorimeter trigger logic. It is
Technical Design Report for the BABAR Detector
406
Electronics
Printed-Circuit
Board
Amplifier Chip
Filter
Power A
Diff.
Rcver
Calibration
Pulse A
R
R
Diode A
x1 Signal A
R
Two-Channel
Amplifier Chip
FCC
JFETs
to ADC
Board A
Diodes
x32 Signal A
Test
Connector
R
x1 Signal B
x32 Signal B
Diode B
Diff.
Rcver
R
R
Filter
to ADC
Board B
Calibration
Pulse B
CsI
Crystal
Power B
Amplifier Board
Figure 10-14.
Block diagram of preampli er.
Calorimeter preamp/crystal shield
assembly.
Figure 10-15.
envisaged that there will be processing power on these boards to apply the gain and pedestal
corrections to each channel and to extract the total charge and time information for each
crystal when a trigger has been received.
Each of these components is discussed in more detail below.
10.6.3 The Preampli er Card
Figure 10-14 shows a block diagram of the preampli er board. It connects directly to the
photodiodes on the CsI(Tl) crystals and provides, on output twisted pair cables, signals with
1 and 32 gain for both the A and B photodiodes. The main components are a custom
two-channel ampli er chip (or perhaps two one-channel chips), and an external JFET. The
shaping time for a CsI(Tl) pulse is optimally near 3 s; however, the presence of background
in the form of a large number of low-energy photons may mean that shorter shaping times
are needed to give better background rejection. Shaping times less than 1 s mean a
signi cant loss of signal from the crystal. In the following, a shaping time of between 1 and
3 s is assumed.
Providing most of the circuitry for the preampli er on a single custom integrated circuit
minimizes possible breakdown from faulty connections between components. The preamp
IC contains completely separate paths for the power, signal, and calibration of both diodes.
The only element in common between the two pathways is the substrate. The preamp chip
Technical Design Report for the BABAR Detector
10.6 Calorimeter
407
provides charge-to-voltage conversion, shaping, and dual-gain di erential driver stages at 1
and 32 gain.
The 18-bit dynamic range input to the preamp board requires that the diode and preamp be
contained within a single shielded assembly. One possible layout of the crystal/diode/preamp
assembly is shown in Figure 10-15. In this design, the preamp board is mounted on a stand
attached to the crystal assembly. The diodes are then connected directly to the JFET
on the preamp board, minimizing the distance over which the 18-bit analog signal passes.
This crystal/diode/preamp package is enclosed in a metal shield, which can also be used
for cooling. The heat output of the preampli er board is estimated to be between 50 and
200 mW per channel.
Each preamp board requires many cables in order to provide independent power, calibration,
and signal paths for both photodiodes. A reasonable estimate of the cable plant is:
Power: Vdd , bias for the photodiode, ground, VFET ;
Signal: twisted pair for 1 and 32 signals; and
Calibration: di erential calibration pulse.
Thus, there are ten wires for each diode. Separate cables are provided for the A and B
diodes. These cables need to be approximately 2 m in length. Filtering for the power lines
and the receivers for the calibration pulses are provided on the preamp board.
10.6.4
Digitizing Board
The digitizing board collects the analog signals from groups of preamp boards, digitizes
the data, and ships the digital data in serial form to the readout boards. Approximately
280 of these are needed to read out the A and B diodes from all the calorimeter crystals.
Figure 10-16 shows a block diagram of the ADC board. A working assumption is that each
board serves both diodes from 25 preampli er boards. This board must provide the following
basic functions:
For each input channel, it must provide a oating-point number covering an 18-bit
dynamic range. Our solution is to have a 10-bit digital mantissa and 2-bit exponent.
This requires a 10-bit ADC.
It must sample with a period which is short compared to the shaping time after digital
ltering. This is needed to reduce background and improve the feature extraction. A
reasonable maximum rate at which the ADC board will be required to digitize a single
channel is 4 MHz ( ve samples over a peak for a 1.25 s shaping time).
Technical Design Report for the BABAR Detector
408
Electronics
Voltage
Regulators
2 Range Bits
Receivers,
~25 Signal
Channels
10-Bit
ADC
Post-Amp,
Range Selection
Custom IC
Commercial
10 Bit Analog Mantissa
Transmitter
Chip
FO
Link
(e.g. G-Link)
Power
for
Amplifier
Board
and
Diodes
Power
Clock/Timing
Generator
to
DAQ
Crates
Clock
Timing
Regulated
Power
Range Force
for Custom IC
Diagnostics
Temperature
Sensor
Calibration Pattern
Calibration
Pulse
Calibration
Signal
Generator
Cal Voltage
18-Bit
Micro
Controller
to Slow
Control
DAC
Calibration
Strobe
Figure 10-16.
Block diagram of ADC board functions.
It must provide the calibration signals to the preamp boards. In addition, the board
should be able to control the range encoding chip and perform diagnostics and monitoring on the preamp and ADC boards.
It must provide the power to the preamp boards.
One of the key components of the ADC board is the custom analog range encoding (CARE)
chip which takes the four analog input signals from the A and B diodes (1 and 32 for
each of A and B), averages the A and B signals on the appropriate range, and then passes
these through two more gain stages. The outputs of these four gain stages are sampled at
the digitization sampling frequency and are compared to threshold levels. The chip places
on its output the analog signal associated with the rst gain stage not above threshold (for
example 90% of the full range for that gain stage) and two bits signifying which of the four
gain stages were used. The CARE chip can select one of three modes for operation:
use input from only the A diode;
Technical Design Report for the BABAR Detector
10.6 Calorimeter
409
Energy Range (MeV) Bin Size (MeV) RMS Resolution (MeV)
0 to 46
0.050
0.029
46 to 366
0.40
0.229
366 to 2925
3.17
1.83
2925 to 13000
12.69
7.33
Table 10-5.
Calorimeter bin size and resolution as a function of energy.
use input from only the B diode; or
average the inputs from the A and B diodes.
Normally, the signals from the diodes are averaged. In the case of the failure of one diode,
the CARE chip can be switched to use only the input from the working diode, and the failed
diode can be disconnected.
Table 10-5 summarizes an example con guration and shows how the resolution and bin size
vary as functions of energy and range selection in the CARE chip. The numbers assume
that the overall gain ranges are 1 (1 1), 8 (1 8), 32 (32 1), and 256 (32 8), and that
a 10-bit
p ADC is used to digitize the data. The resolution is taken as the bin size divided
by 12 with an additional factor of 2 to account for losses in clustering. The results are
summarized in Figure 10-17 which shows the resolution from the digitization compared with
the speci ed resolution for the CsI(Tl) crystals.
Since the signal is presented to the ADC chip as a DC level, the ADC chips should be able
to deliver the speci ed number of bits of accuracy. We are exploring whether it is more
cost-e ective to assign a single ADC to each channel or to multiplex several channels into
fewer, higher-speed ADCs.
Assuming the highest rate of sampling (4 MHz) and multiplexing, the 12 bits from each of
25 channels into a single serial output requires a capability of 1.25 Gbits/s (1.5 Gbits/s with
protocol overhead). For this rate, the best technology is optical ber.
The ADC board also performs electronic calibrations and monitors the ADC board and
the preamp. A small microcontroller is placed on the board, with a low speed serial link
to the outside world. This controller can read the monitoring inputs and control the data
acquisition and calibration lines.
The location of the preamps and ADC boards is shown in Figure 10-18. Each group of 25
preamp cards has a cable plant of 500 wires, many of which are power and ground, and the
rest of which carry signals to the ADC board serving this group of cards. A possible solution
to the problem of handling this number of cables is to have a fanout board which connects
Technical Design Report for the BABAR Detector
410
Electronics
6
Energy Resolution (%)
5
CsI
4
3
Electronics
2
1
x256
x32
x4
0
10-3
10-2
x1
10-1
100
Energy (GeV)
101
102
Digitization resolution vs. CsI Resolution.
Figure 10-17.
to CALDAQ
Board in DAQ Crate
Fan-out Board
(a)
Fan-out Board
Amplifier
ADC-Board
ADC-Board
Crystals
ADC Board
(b)
(c)
Connector
Fan-out board
Cooling
Pipe Outside
Shielded
Cal Volume
ADC board
120 Phi Sectors
Figure 10-18.
ADC board located at the end of the barrel.
Technical Design Report for the BABAR Detector
Figure 10-19.
Block diagram of readout module functions.
the 500 cables from the preamps to a bus connector into which the ADC board then plugs.
Power lines are shared across this bus, while the signals are fed to the ADC cards. This
allows ADC boards to be easily replaced during shutdowns. The inputs to the ADC board
from the data acquisition system are: a sampling clock, a serial output link, a slow control
link, and a serial output optical ber.
The ADC boards are placed where they can be water cooled so that the heat generated on
the boards does not get transmitted to the crystals. The expected heat generation on each
ADC board is 5 W, leading to a total heat load of 1.5 kW.
10.6.5
Readout Module
The readout modules for the calorimeter are located in the electronics house. They receive
signals from the 280 ADC boards, provide trigger information to the Level 1 trigger, and
send processed signal information to the higher level data acquisition system. This section
discusses only the features unique to the calorimeter readout module. The features common
Technical Design Report for the BABAR Detector
412
Electronics
to the whole data acquisition system are discussed in Section 10.10. Figure 10-19 shows a
block diagram of the readout module functions. The basic functions of these modules are:
to receive serial data from the ADC boards;
to de-serialize the data and place it in the VRAM;
to contain the calibration look-up table for the trigger; and
to apply calibration corrections and to extract amplitude and time information from
the waveform samples.
For de-serialization of the input data, a clock is encoded on the optical ber signal. This
guarantees that the data owing into the VRAM are synchronized with the digitization
clock, which is synchronized with the overall system clock. Some information needs to be
provided to verify the channel encoding. The simplest implementation would be a single
bit to ag the rst channel in a stream of 25. At the cost of extra bandwidth, 5 bits could
be used to send individual channel numbers with the data. Data ow into the VRAM at
1.25 Gbits/s and can be stored in the VRAM for up to 6 ms before they must be moved to
more permanent memory.
For trigger purposes, the readout modules provide information about the amount of energy
deposited in the 5 5 arrays of crystals which are seen by each board. This part of the
module is discussed in Section 10.9.
A sophisticated algorithm to calculate the total charge and pulse height is needed to minimize
the e ect of the soft photon background on the resolution. Algorithms are being studied
using models of the signal plus expected background. The oating-point conversion and
calibration correction could be done continuously using a SRAM look-up table, but this is
likely to be costly compared to doing this on demand using an onboard processor. Thus, in
response to a trigger or a request containing better time information, the readout module
processor executes the following steps to nd the times and magnitudes of the showers:
calibration constants are applied to the raw data, the charge and time of the pulse in each
crystal are found, and if the magnitude is suciently large, this information is sent to the
next level data acquisition system. Provision is also made to occasionally send the raw data
to monitor the onboard algorithms.
The maximum data rate, allowing for online calculation of the time and pulse height in each
crystal at a 2 kHz trigger rate, is estimated by allowing for 20 bits of charge information and
12 bits of time information. This requires 0.2 Mbytes/s to send data for all crystals on an
ADC board, 3.2 Mbytes/s per readout crate, and 54 Mbytes/s for the entire calorimeter.
The design of the readout module is shared, as much as possible, with the other systems.
Technical Design Report for the BABAR Detector
10.6 Calorimeter
10.6.6
413
Electronic Calibration
For each channel, calibration corrections are applied on the readout module for both overall
gain (Chapter 7) and the response of the analog and digitizing electronics. The latter
information is provided by an electronic calibration subsystem.
In order to derive the calibration constants, the following tasks need to be performed:
Injection of charge on the preamp: This is accomplished by sending a signal derived
from an 18-bit digital-to-analog converter (DAC) on the ADC board through a capacitor onto the input of the preamp.
Range forcing on the CARE chip: In order to understand the transition region from
one range to another, the capability to force the CARE chip to remain in one gain
range is essential. In this way, the operation of the automatic range selection can be
veri ed.
Independent control of the phase of the calibration signal with respect to the clock.
Calibration of the CARE chip: it may be desirable to input a test signal to the CARE
chip directly, and thus separate e ects in the preamp from those in the CARE chip.
In addition to the calibration of the electronics, calibration with sources requires the readout
of the calorimeter in an untriggered mode. This requires dedicated running using only
the calorimeter and specialized software in the online processors to search for peaks in
individual crystals and to create clusters. Since this calibration is performed without the
beam, peak nding without a trigger should be feasible, as the design of the preamp ensures
reasonably large signal/noise for source events. Both the electronic calibration runs and
source calibration runs (Chapter 7) require special processing modes on the readout modules,
during which the calorimeter must be in local mode.
10.6.7
Monitoring and Control
Most of the monitoring tasks are performed by the onboard microcontrollers on the ADC
boards. These are interconnected with a slow low-cost link managed by an overall detectorcontrol processor. The main variables which are monitored by the system are:
Temperature: A tolerance of 1 C at the preamp boards and 2 C at the ADC board.
A sample of preamp boards and all ADC boards are monitored.
Technical Design Report for the BABAR Detector
414
Electronics
Bias Voltage Current: The bias voltage on and current from each photodiode are
monitored. There are also provisions for measuring the total current.
Supply Voltage: The low voltage supplies are likely to be shared across groups of ADC
boards. These are monitored and safety interlocks applied to the power supplies.
Support software interprets the slow control data and provides diagnostics which allow
problems to be isolated without opening the detector.
10.6.8 Research and Development
A preliminary version of the preamp IC has been constructed in a 1.2 m BiCMOS technology. This chip uses a previous design [Bag94] modi ed to suit the needs of the BABAR
calorimeter. The shaping time for this preliminary device can be modi ed using external
components. This chip will be used in conjunction with various possible JFET candidates
on the prototype calorimeter in order to evaluate the design, choice of shaping time, and
performance of the preamp.
The R&D program for the ADC and readout modules centers around the following: the
design and testing of the CARE chip; evaluation of candidate commercial ADCs; development of the electronic calibration system; development and testing of an appropriate VRAM
control system; testing proposed trigger schemes; and benchmarking signal processing algorithms and evaluating the processor power required. A rst version of the CARE chip
has been built in CMOS and tested at SLAC, and a BiCMOS version of the chip capable of
higher sampling rates is now in prototype fabrication. The remaining R&D issues will be the
subject of parallel e orts to verify the concept and resolution obtained with the proposed
methods. The new CARE chip and as many of the other components as possible will be
included in the rst tests of the calorimeter prototype with beam. Ongoing beam tests will
study in detail the proposed production versions of all boards and modules and address
system concerns which may arise only in the context of a realistic detector.
10.7 Muon System Electronics
10.7.1 Electronics Requirements
The functional requirements of the Instrumented Flux Return (IFR) electronics are to label
hit strips and to associate a hit with a triggered event.
Technical Design Report for the BABAR Detector
10.7 Muon System Electronics
415
The resistive plate chamber (RPC) single counting rate is on the order of 400 Hz/ m2 [Amb94],
which is comparable to the cosmic ray rate at sea level. Assuming the probability of recording
noise hits is on the order of 10,4, the number of accidental hits is below 20 per event. The
physics rate also is quite low, with an event generating at most a few tens of hits. Thus the
IFR is a very low occupancy detector.
The readout logic will utilize the Level 1 strobe which carries a xed latency of 10 s with
a maximum permitted jitter of 0.5 s. The average trigger rate is less than 2 kHz, and the
maximum rate the readout must tolerate is 10 kHz. The minimum time between successive
triggers is 1.5 s, and this is the time allotted to empty the front-end electronics. This small
time in which to read the 42,000 channels is the main constraint on the design.
The baseline architecture presented here represents a working model which takes into account
functional requirements and assumptions, and copes with constraints. The main goals of this
design is to reduce the electronics cost by minimizing the number of modules and connections
needed for the whole device, and to assure completely safe operation. This is accomplished
by reducing the overall complexity, and by using well-de ned existing technologies.
The rst module (the Front-End Card, or FEC) will be located close to the detector in the
iron gaps; the second module will be on the outside of the detector itself and accessible
without opening the endcaps, while the last one (readout module) will be on the electronics
house.
10.7.2 General Architecture
Because the RPC is practically a noiseless detector in which the rate is dominated by cosmic
rays, immediate rejection of unwanted data is not as important as very rapid data collection.
Bu ering during the trigger latency is unnecessary because the probability of having two hits
on the same strip during this time is negligible. Strip signals will be stored during the latency
of the trigger decision by simply stretching them and using a fast readout card to prepare
the front-end within the prescribed 1.5 s. The IFR electronics perform as a data-driven
system.
Generally, a serial readout of front-end cards is suitable when a large number of strips must
be handled and the acquisition speed is not a critical parameter. However, in this case, the
time allowed to empty the front-end cards is relatively short, and we plan to use a mixed
serial/parallel readout design.
Conceptually, the electronics system for the IFR is a sequence of blocks performing the
following functions:
record detected signals during Level 1 strobe latency;
Technical Design Report for the BABAR Detector
416
Electronics
Readout connector
Shift register
Readout
Clock
From previous
Board
Strobe
x16
M0
To next
Board
M1
V
th
+5
+7
-5.2
Test
V
V
V
ECL Fast-OR
Strips
Figure 10-20.
Schematic of the Front-End Card (FEC).
unload the data from the front-end within the allowed time (1.5 s), skipping empty
cards;
encode the addresses of hit strips; and
read hit data, assemble them into a full data stream, and transfer them to a storage
medium.
10.7.3 Implementation
Front-End Card (FEC)
Because of the inaccessibility of the front-end cards, they are designed with a minimum
number of components. A schematic is shown in Figure 10-20. Because RPCs produce
pulses consistently near 200 mV in pulse height, the discriminator stage of the FEC is not
critical. A single transistor will suce to the bring signal level to the stage of driving TTL
chips [Alo94c].
Technical Design Report for the BABAR Detector
10.7 Muon System Electronics
417
Strobe
Noise
Event
Noise
-15
Figure 10-21.
acquired.
-10
-5
0
µs
5
Double one-shot logic. Only strips red in the 1 s window will be
Each FEC connects directly to the ends of 16 adjacent readout strips. The output is
obtained from a logical AND of two one-shots (M0 and M1), which are 9 s and 10 s
wide, respectively. Stretching the strip signals acts both as bu ering for trigger latency and
for noise reduction by identifying the strip red within the 1 s time window (Figure 10-21).
The board also provides a Fast-OR signal of the 16 strips, which can be used for time
measurements, monitoring, diagnostics, and time calibration.
They are connected in a 16-card chain. Each chain is read serially as a 16-bit-wide shift
register, as shown in Figure 10-22. The FECs will be installed inside the IFR gaps, very
close to the RPC strips.
FIFO Bu ering, Zero Suppression, and Digital Encoding (FIB)
After the Level 1 strobe is received, input data, consisting of one bit per strip, are latched
into the FEC register. The chain is then read by the FIB, which is shown in Figure 10-23.
This module bu ers the data from the shift registers on 16 FECs and stores them in parallel
in the data FIFO. The FIB module handles four di erent chains (1024 strips in total) and
is located in a VME crate on the outside of the detector.
Each trigger has a time stamp which is read into the FIFO as a header to the data stream
referring to that particular trigger. The readout chains are clocked at 12.5 MHz so that the
FECs are ready for another trigger within 1.5 s.
Independently, FIFO data are pushed into the zero suppressor and digital encoding logic,
which encodes strip data into hit data. This is implemented with Xilinx LCAs and is a
Technical Design Report for the BABAR Detector
418
Electronics
DATA_IN_0
DATA_IN_15
FIB
FEC
#0
FEC
#1
FEC
#15
CK_0
CK_15
CLOCK_OUT
Figure 10-22.
Schematic of the parallel readout mechanism for each chain.
4 chains
Line receiver
and register
Counter
State
machine
control
logic
Control
bus
Data FIFO
A
u
x
B
u
s
Aux bus
control
Figure 10-23.
Zero
suppressor
VME Bus
control
Event
FIFO
V
M
E
B
u
s
A simpli ed scheme of the FIFO board.
Technical Design Report for the BABAR Detector
10.7 Muon System Electronics
419
copy of that used in the L3 experiment at CERN [Alo94a, Alo94b, Alo94c, Alo94d, Alo94e].
Another Xilinx LCA handles VME protocol, while the state machine for the arbitration is
implemented in a MACH gate array.
Readout Integration
The FIB boards are arranged in seven VME crates, three for the barrel and two for each
endcap. Each crate includes a data collector module (DCM) acting as interface between the
encoding module and the readout module. The DCM asks for a speci ed trigger number
from all FIBs plugged into the same crate. It uses an auxiliary bus connection and a tokenin/token-out protocol. Data ow from an FIB until no more data are available on that FIB
for that trigger number. The collected data are then ready to be transferred to the readout
module. This architecture is deadtimeless even under the worst background conditions.
10.7.4 Time Measurements
Time measurements can be achieved by using TDC channels with a granularity as coarse as
a single layer. The measurements will be performed via common stop, VME-based, TDC
modules located in the crates on the outside of the detector. The stop signal is supplied
with a (synchronous) Level 1 strobe. The TDCs have a range covering a 1.5 s time window
to ensure that the entire time jitter for the trigger is contained. Provisions have to be made
for the 10 s strobe latency. When the Level 1 strobe is asserted, the TDCs are stopped and
read into a FIFO bu er, associated to the right Level 1 strobe reference number, and sent
to the data acquisition system. The required TDC resolution is 0.5 ns.
10.7.5 Time Calibration
Since strip signals coming from di erent FECs have di erent path lengths, their arrival time
will be spread out. A time-zero measurement is thus needed in order to calibrate out di erent
cable lengths and electronics time response.
The time calibration is performed with an external pulser system. A synchronized test pulse
is injected directly in the input stage of all the FEC channels. The di erence in time between
the various TDC outputs provides the factor needed to calibrate the di erent signal delays.
Technical Design Report for the BABAR Detector
420
10.7.6
Electronics
Monitoring
The complete hardware paths can be monitored by sampling unprocessed data (spying)
during data taking. Furthermore, fast-OR signals directly collected by scalars provide frontend card rates for monitoring.
10.7.7
Diagnostics
The whole electronics paths will be checked in order to identify possible failures. A test
input provided on the front-end board is used to inject a test pulse on all 16 input channels.
This is the same pulse signal that is used in time calibration.
10.8 Trigger Requirements and Background Rates
10.8.1
Introduction
PEP-II may give rise to rather severe backgrounds, producing up to several kilohertz of
interactions with two or more charged tracks reaching the calorimeter. This rate is to be
contrasted with the desired logging rate of less than 100 Hz. The trigger and data acquisition
subsystems are designed to record data at no more than the latter rate under a variety of
background conditions and to accommodate luminosity upgrades up to 1034 cm,2s,1 .
The purpose of the trigger is to reject backgrounds while selecting a wide variety of physics
processes. This ltering function is performed through an increasingly sophisticated sequence
of algorithms applying mostly topological cuts on increasingly re ned data. Processes
selected with high eciency include nal states of e+e, interactions ((4S ), (5S ), cc,
and + , ) as well as nal states of and interactions. In addition, nal states such
as e+ e,, e+e, , and +, are selected for use with luminosity measurements, monitoring,
and calibration. Finally, for diagnostic purposes, cosmic rays, beam-induced backgrounds,
and random time slices are selected at a low rate.
In addition to providing high eciency for a large number of interesting physics processes
and rejecting most background events, the trigger is designed to allow precise measurements
of its eciency. The most powerful tool for this is the use of independent selection criteria.
Orthogonal selection methods are used as much as possible for those lters operating on
trigger primitives, i.e., on quickly derived reduced representations of the data.
Technical Design Report for the BABAR Detector
10.8 Trigger Requirements and Background Rates
421
Physics Data Samples
BB from the (4S ): CP Channels
BB from the (4S ): Non-CP Channels
B Physics at the (5S )
Charm Physics
Continuum qq
Physics: Rare Decays & Asymmetries
Physics: Branching Ratios
Requirements
High Eciency
Precise Eciency, High Eciency
High Eciency
Precise Eciency, High Eciency
Precise Eciency (Background Subtraction)
High Eciency
Precise Eciency
Precise Eciency
Other Data Samples
Bhabhas: Two Detected Prongs
Bhabhas: One Detected Prong
Purpose, Comments
Luminosity, Calibration, High Eciency
Calibration, Beam Monitor
Luminosity, Calibration
Calibration
Beam Monitor, Trigger Studies
Trigger Studies
+ ,
Cosmic Rays
Beam Backgrounds
Random Events
Table 10-6.
Principal trigger requirements for each type of data sample.
This section begins with a detailed discussion of the implications of BABAR's physics goals
for the trigger. It continues with a description of trigger algorithms, and concludes with the
results of simulations of the proposed trigger algorithms, including both the eciencies for
some benchmark processes and the expected trigger rates.
10.8.2 Requirements
The trigger needs to be ecient, selective, exible, measurable, and accurate. More details
are given in the BABAR Electronics Requirements [Req94]. The relative importance of these
attributes varies for di erent data sets. For asymmetry measurements and searches, it is more
important to have a high eciency than a well-known eciency. For branching ratio and
cross-section measurements, precisely measured eciencies are crucial. The principal trigger
requirements for each data set are listed in Table 10-6. In addition to the key physics events,
a number of other events are recorded for uses such as luminosity determination, detector
calibration, beam monitoring, and trigger performance studies. These impose additional
requirements on the trigger and are listed at the bottom of the table.
Technical Design Report for the BABAR Detector
422
Electronics
Because of their relatively large multiplicities and visible energies, it is straightforward to
achieve high trigger eciencies for and cc events. The challenge for the trigger designer
lies in detecting low multiplicity + , and events while keeping the rate of accepted
background events within the data acquisition limit, so as not to incur deadtime. To ensure
high eciencies in oine analysis and large background rejection power, trigger algorithms
need to have sharp eciency turn-on curves. For example, the eciency as a function of
cluster energy should approximate a unit step function as closely as possible to allow a highly
selective trigger.
Because PEP-II machine conditions will vary over the course of the experiment, the trigger
needs to be exible. Under temporarily severe background conditions, saturation of the
bandwidth should result in deadtime in such a way that a diminished number of whole
events are collected rather than a larger set of partial events. When backgrounds are severe
but data are useful for physics, prescaling factors can be applied to selection streams to ensure
that the available bandwidth is lled with interesting events. This can be done on a relatively
short timescale. It will be straightforward to change the global logic combinations of the
selection streams in each of the trigger levels, and it will also be possible to vary selection
parameters. Actual change of trigger selection algorithms is foreseen to occur rarely.
There are several ways to facilitate precise oine measurements of the trigger eciency.
First, the geometry of the detector and the readout electronics for each system contributing
to the trigger are designed with trigger considerations in mind. Second, the trigger and data
acquisition system is designed to:
use orthogonal selection criteria;
record all trigger primitives and decisions for accepted events;
accept prescaled numbers of events with looser criteria;
accept a prescaled number of rejected events;
have a sharp eciency turn-on as a function of the selection variables; and
be simple to model with simulations.
Finally, the online system will be designed so that the recorded trigger primitives and
decisions are readily available for oine analysis.
Use of the trigger results for precision physics measurements depends crucially on the trigger
system working as a whole. Like the other systems that deliver data to the data acquisition,
the trigger system should be designed with sucient monitoring, calibration, and diagnostic
testing capabilities so that trigger decisions and trigger data are known to be correct.
Technical Design Report for the BABAR Detector
10.8 Trigger Requirements and Background Rates
10.8.3
423
Backgrounds and Trigger Rates
The trigger architecture and selection algorithms designed to meet the requirements are
based on simulation studies of the physics, backgrounds, and the BABAR detector.
The total trigger output rate is the sum of the rates for the physics and diagnostic processes
listed in Table 10-6 and the rates for the background processes described in Chapter 12.
The main background processes that produce charged tracks traversing the drift chamber
and calorimeter are:
physics backgrounds from asymmetric Bhabhas with only one detected charged track
and two-photon events below energies of interest;
beam-induced backgrounds from hadronic interactions of lost beam particles, electromagnetic interactions of lost beam particles, and beam-gas collisions at the interaction
point; and
cosmic rays.
The main sources of occupancy in the vertex detector, drift chamber, and calorimeter are:
particles from small-angle radiative Bhabhas, and
beam-induced background from lost beam particles.
The rate of charged particles and the occupancy of physics backgrounds scale with the
luminosity, whereas those of beam-induced backgrounds scale with the beam current and
the pressure of the residual gas inside the PEP-II beam pipe.
Physics backgrounds are evaluated at the design luminosity, and the trigger and data acquisition system must be upgradable for higher luminosities. Because of their variable nature and
because of uncertainties in the predictions, beam-induced backgrounds pose more stringent
design requirements. The trigger and data acquisition system is required not to produce
substantial impact on the physics program as long as data are useful. Based on simulation
of the occupancies in the vertex detector as functions of lost-particle rate, pattern recognition
becomes seriously compromised at ten times the nominal background rate. The simulation
is described in Chapter 12.
Technical Design Report for the BABAR Detector
424
Electronics
10.8.4 De nitions of Trigger Filter Concepts
Event Selection
Event signatures are used to separate signal from background. Combinations of the following
global event properties are used in the trigger lters:
charged track multiplicity;
calorimeter cluster multiplicity;
event topology; and
primary vertex position.
These selection criteria have associated thresholds for the following parameters: chargedtrack transverse momentum (pt ), energy of calorimeter clusters (Eclus), solid angle separation ( and ), track match quality (for electrons and muons) and primary vertex impact
parameter (1 and z1 , respectively). The trigger de nition can contain selection criteria that
di er only by the values of thresholds, such as the pt or Eclus threshold values. A small
fraction of random beam crossings and events with failed triggers are selected for diagnostic
purposes.
Selection Filters
Simulation studies both of processes of interest in the areas of B, , and physics [Kra94b,
Kra94a, Lev94a, Lev94b, Lev94c, Lev94d, Sny94], and of background events in the BABAR
detector, have been used to guide the choice of trigger algorithms.
Table 10-7 lists \trigger objects" which are potentially de nable at Level 1 and which have
been determined to be useful in constructing global trigger de nitions. Vertex detector
information is available at higher levels. For a given trigger level, the global selection lter is
a logical OR of a number of speci c selection lters, each of which is the result of a boolean
operation involving any of the ingredients available at that level. The types of selection
lters may be arranged in the following hierarchy:
1. Multiplicity of drift chamber charged particles with implicit pt and cos cuts and
explicit -separation cuts;
2. Multiplicity of calorimeter clusters with explicit Eclus and implicit cos cuts, and
explicit - and -separation cuts;
Technical Design Report for the BABAR Detector
10.8 Trigger Requirements and Background Rates
A tracks
A0 tracks
B tracks
M clusters
G clusters
E clusters
425
Tracks in the Drift Chamber
Tracks that reach the outer layers of the drift chamber. For B = 1:5 T,
this implicitly leads to approximately pt > 0:210 GeV=c. These tracks
are contained in the solid angle of approximately ,0:81 < cos < 0:90.
A tracks that satisfy an explicit cut on the reconstructed pt.
Tracks that reach the middle layers of the drift chamber. This implicitly
leads to approximately pt > 0:130 GeV=c. The solid angle acceptance
for these tracks that can exit the middle of the drift chamber end cone
is substantially larger than for A tracks, ,0:92 < cos < 0:96.
Clusters in the Calorimeter
Clusters that have reconstructed energy deposits consistent with
minimum ionizing particles (MIP), or larger, in the calorimeter barrel
or endcap. The required value is near 150 MeV. A minimum ionizing
, deposits 181+20
,8 MeV, averaged over the solid angle of the barrel
calorimeter, where the limits indicate the variation in energy deposit
due to trajectories and varying crystal lengths [Kra94a]. For such
a muon to reach the calorimeter barrel, it is implicitly required to
have approximately pt > 0:250 GeV=c. The solid angle acceptance is
approximately ,0:78 < cos < 0:96. The cluster energy threshold is
adjustable.
Clusters with energies above those of M clusters; i.e., Eclus >
0:200 GeV. The cluster energy threshold is adjustable.
Clusters with energies large enough to be produced by e+ e, or ,
i.e., approximately Eclus > 1:0 GeV. The cluster energy threshold is
adjustable.
Matches Between the Drift Chamber and Calorimeter
A-M matches A tracks correlated in with barrel M clusters.
A0-M matches A0 tracks correlated in with barrel M clusters.
B-M matches B tracks correlated in with barrel or endcap M clusters.
Table 10-7.
Trigger objects.
Technical Design Report for the BABAR Detector
426
Electronics
3. Special triggers for calibration, monitoring, and diagnostics;
4. Multiplicity of charged particles correlated between the drift chamber and calorimeter
M clusters, with explicit matching in ;
5. Multiplicity of drift chamber charged particles with implicit cos cuts, and explicit pt
and -separation cuts; and
6. Multiplicity of charged particles correlated between the vertex detector and drift chamber with explicit matching in or explicit cuts on and z .
The list continues with more re ned selection cuts applied to the full data set available in
Level 3.
At each hierarchical step, there may be a number of selection lters (trigger lines), di ering
in whether certain cuts are applied or in the values of parameters used to make cuts. Let the
isolated symbol for an object represent the number of such objects, e.g., A is the number
of A tracks. Assume that counting and geometry cuts are such that no physical object will
ll a double role in the trigger. Then, using the notation of Table 10-7, category (2) might
include: M 2 with minimal separation cuts; one or more versions of G 1 M 1 with
minimal separation cuts; or G 1 M 1 with appreciable separation. Each of the trigger
lines will be subject to individual prescaling, which may have to be used to reduce rates for
the more open de nitions, while collecting all events satisfying more stringent de nitions.
The global selection OR at each level is formed after any such prescaling. Some speci c
lters will be evaluated for their e ects on signal and background events in the following
subsection.
10.8.5 Performance of Some Simple Triggers
Benchmark Physics Processes
In order to evaluate the eciency of selection algorithms, a set of physics processes has been
identi ed to serve as trigger benchmarks. As explained in Chapter 3, the BABAR physics
program focuses on studies using (4S ), (5S ), qq, + , , , and events. While the
primary mission of the BABAR experiment is the study of CP violation in B decays, the
trigger is designed to accept a large number of processes that may be important. The trigger
benchmarks represent processes that are dicult to distinguish from background. Other
similar processes have, or are expected to have, higher eciencies. Here we use Monte Carlo
information to study the e ectiveness of some simple triggers on these benchmarks.
For B 0 B 0 CP studies, the most dicult tag to trigger on is B 0 ! X because of its low
charged-track multiplicity and low cluster energy compared to the electron or kaon tags.
Technical Design Report for the BABAR Detector
10.8 Trigger Requirements and Background Rates
427
The low-multiplicity decay B 0 ! +, was chosen as a CP study decay mode to test the
eciency of our simple particle counting trigger algorithms. The high-multiplicity decay
B 0 ! D+D, was chosen to test algorithms that require fast tracks.
For + , studies, events in which both taus decay to one charged track are the most dicult
to select. To measure the muon branching ratio, events with , ! , accompanied by
+ ! e+ are used. This is a worst-case decay for the purpose of the benchmark table,
as the measurement is not expected to be statistics limited. The other + , benchmark
is an untagged event for which each tau decays to one charged pion, used to measure spin
correlations for V,A tests.
The benchmarks are straightforward: two-pion events that are either both charged or
both neutral. For two-charged-hadron nal states, one can normalize the hadronic cross
section to the better known muon-pair cross section. The benchmarks with two kaons
and one pion in the nal state, produced via an unspeci ed 1++ resonance, represent searches
for exotic mesons.
Benchmark Eciencies and Background Rates
An inclusive global trigger de nition that eciently selects each of the processes in Table 10-6 while rejecting some backgrounds requires a minimum of two charged tracks in
the drift chamber, two M clusters, or one E cluster in the backward direction [Bol91]. The
orthogonality of the two rst criteria makes it straightforward to measure the eciency of
the lter.
The following discussion focuses on de ning a exible open trigger with high eciency.
Based on the rst two criteria listed below, it is able to handle a large range of potential
beam-induced backgrounds.
There are three versions of the two-track trigger to consider: A 2, A 1 B 1, and B
2. The preferred de nition is A 1 B 1 (the so-called one and one-half track trigger)
because it has larger solid angle acceptance than A 2 and accepts fewer looping tracks
than B 2. Simulations show that A 2 is only 80% as ecient as A 1 B 1 for
inclusive tau decays [Bol91, Kra94b]. One soft charged particle can loop back, resulting in
two detected B tracks. This problem is less severe with the A 1 B 1 de nition.
Eciencies and output rates for the two-particle trigger were simulated for two global-trigger
input lines, D2 and C2 . The default two-track trigger D2 uses the A 1 B 1 de nition and
is simulated. The default two-cluster trigger C2 requires M 2, both with Eclus > 0:150 GeV,
where the chosen value for the cluster energy threshold is a compromise between eciency
and background rate. To suppress pile-up of soft photons due to electromagnetic interactions
Technical Design Report for the BABAR Detector
428
Electronics
Process
D2
B 0 ! + , , B ! X
1.00
0
0
+
,
B !D D ,B !X
1.00
+
+
,
,
! e , !
0.72
+
+
,
,
! , !
0.85
! +,, W > 2 GeV=c2 0.63
! 0 0, W > 2 GeV=c2
! X (1420) ! KK
0.48
! (550) ! + , 0
0.07
0
C2 C2 + D2
1.00 1.00
1.00 1.00
0.73 0.79
0.74 0.89
0.63 0.63
0.86 0.86
0.62 0.62
0.21 0.21
Table 10-8. Open trigger eciencies for benchmark physics processes. The eciency
denominator includes all events generated over the full 4 solid angle. The simulation used
1000 events. W is the mass of the system.
of lost beam particles, each calorimeter channel is required to have an energy deposit above
a threshold of 10 MeV. An open global trigger is de ned as C2 + D2.
Calculated eciencies for benchmark physics processes using the de ned trigger lines are
listed in Table 10-8. These eciencies are found to be very high for each of the benchmark
processes. Estimated open-trigger output rates due to the processes listed in Table 10-6 and
in Section 10.8.3 are shown in Table 10-9. The simulation is described in Section 10.9.5.
Using the open trigger, the maximum event-building rate goal is not satis ed at 10. Thus,
the open trigger would need to be prescaled and a tighter criterion would need be applied
when the total output rate exceeds 2.0 kHz. The best solution to this problem|i.e., with
the least impact on two-prong physics processes|is to require an A0 track instead of an
A track and a G cluster instead of one of the M clusters. These are the D02 and C02 trigger
lines. To suppress lost-particle electromagnetic backgrounds, the C02 line further required the
G cluster to be back-to-back with an M cluster. The eciency and rate for the two trigger
lines are given in Tables 10-10 and 10-11, respectively.
10.9
10.9.1
Level 1 Trigger
Overview
The rst level trigger architecture provides the Level 1 strobe for the front-end bu ers and
reduces the event output rate for feature extraction and event building. The trigger achieves
Technical Design Report for the BABAR Detector
10.9 Level 1 Trigger
429
Process
Physics:
(4S )
D2
qq
+ ,
+ ,
Bhabha
signal
background
Backgrounds:
Hadronic nominal
Electromag. nominal
Beam-gas 1 nTorr
Hadronic 10
Electromag. 10
Beam-gas 10 nTorr
Cosmic rays
Total nominal
Total 10
C2 C2 + D2
12 Hz
8 Hz
2 Hz
2 Hz
100 Hz
3 Hz
50 Hz
12 Hz
8 Hz
2 Hz
2 Hz
100 Hz
3 Hz
50 Hz
12 Hz
8 Hz
2 Hz
2 Hz
100 Hz
3 Hz
50 Hz
300 Hz
140 Hz
230 Hz
< 2 Hz
380 Hz
230 Hz
< 6 Hz
< 60 Hz
< 4 Hz
3000 Hz 1400 Hz 3800 Hz
450 Hz 3800 Hz 4000 Hz
< 40 Hz < 20 Hz < 60 Hz
80 Hz 410 Hz 440 Hz
620 Hz 970 Hz 1200 Hz
3700 Hz 5800 Hz 8500 Hz
Open trigger rates. The background rates are subject to large uncertainties.
The hadronic and electromagnetic backgrounds are due to lost beam particles. The rates
of the rst group of processes are dependent on the luminosity. The rates of the second and
third groups are dependent on the product of the beam current and the residual pressure
in the beam pipe. The cosmic rate is xed.
Table 10-9.
these goals through the use of a fast hardware trigger operating on primitives, provided via
point-to-point links by the drift chamber and calorimeter. The drift chamber and calorimeter
triggers feed information to the global trigger, which issues the Level 1 strobe.
The structure of the Level 1 trigger is shown in Figure 10-24. Data from the drift chamber
and calorimeter are converted into trigger primitives by local logic on the readout modules.
There are three steps of trigger logic between the formation of primitives and the global
Level 1 decision that produces the Level 1 accept signal: particle nders that identify drift
chamber tracks and calorimeter clusters, particle counters that identify the number of distinct
particles, and a track matcher that associates drift chamber tracks with calorimeter clusters.
Technical Design Report for the BABAR Detector
430
Electronics
Process
0
B 0 ! + , , B ! X
0
B 0 ! D+ D,, B ! X
+ ! e+ , , ! ,
+ ! + , , ! ,
! +,, W > 2 GeV=c2
! 0 0, W > 2 GeV=c2
! X (1420) ! KK
! (550) ! + , 0
D02
0.99
0.97
0.65
0.78
0.62
C02 C02 + D02
0.75 0.99
0.80 0.99
0.43 0.68
0.36 0.80
0.51 0.63
0.76 0.76
0.18 0.29 0.31
0.05 0.06 0.06
Safe trigger eciencies for benchmark physics processes. The eciency
denominator includes all events generated over the full 4 solid angle. The thresholds
1 > 0:600 GeV for C0 . Clusters were required to be
were pAt > 0:600 GeV=c for D02 and Eclus
2
separated by at least 150 in . W is the mass of the system.
Table 10-10.
Process
Physics Signal
Background
Hadronic nominal
Electromag. nominal
Beam-gas 1 nTorr
Hadronic 10
Electromag. 10
Beam-gas 10 nTorr
Cosmic Rays
Total nominal
Total 10
D02
47 Hz
50 Hz
C02 C02 + D02
124 Hz 127 Hz
20 Hz 50 Hz
90 Hz
< 60 Hz
0 Hz
10 Hz
< 60 Hz
< 1 Hz
100 Hz
< 60 Hz
< 1 Hz
900 Hz
120 Hz
980 Hz
< 220 Hz
220 Hz 220 Hz
< 2 Hz < 10 Hz < 12 Hz
50 Hz
300 Hz
1300 Hz
2 Hz 50 Hz
220 Hz 390 Hz
500 Hz 1400 Hz
Safe trigger rates. The background rates are subject to large uncertainties.
1 > 0:600 GeV. Clusters
The D02 line required pAt > 0:600 GeV=c and the C02 line required Eclus
were required to be separated by at least 150 in . The total output rate is the sum of the
safe trigger rate, the prescaled open trigger rate, 10 Hz of prescaled asymmetric Bhabha
triggers, and 5% of special triggers.
Table 10-11.
Technical Design Report for the BABAR Detector
10.9 Level 1 Trigger
Segment
Finders
DC
431
∆φ Cuts
Positions
Track
Linker
N(A)
φ(A)
Multiplicity
Counter
φ(B)
N(B)
N(A′)
DC
Tower
Summers
EM
Positions
Curvature Thresholds
φ(M)
Energies
EM
Energy Thresholds
∆θ ∆φ Cuts
Cluster
Finder
Curvature
Finder φ(A′)
Time Aligner
∆φ Cuts
θφ(M)
θφ(G)
θφ(E)
Track
Match
N(A-M)
N(A′-M)
Multiplicity N(B-M)
Counter
N(M)
Multiplicity
Counter
Global
L1
L1
Accept
(to
fast
control)
N(G)
N(E)
θ(E)
Global Cuts
Figure 10-24.
LG 087
Level 1 trigger structure, data ow, and decision logic.
The di erent timing structure of the drift chamber and calorimeter primitives and the delay
associated with the inputs to the track matcher require one stage of time alignment.
A number of variable selection cuts are available to these trigger logic steps. They could
be tuned during initial running but should be stable during subsequent running so as not
to complicate oine analysis. The global decision parameters are also exible in the same
way, to accommodate changing data-taking conditions. During more rapidly changing beam
conditions, the Level 1 accept rate is controlled by prescaling certain components of the
selected events. Information from the Level 1 global trigger is used for detailed accounting
of prescaling factors and deadtimes.
In addition, there are important interfaces of the Level 1 trigger to the data acquisition, fast
control, and online systems. These interfaces are discussed after a detailed description of the
trigger decision implementation.
Technical Design Report for the BABAR Detector
432
Electronics
3
Discriminator
1
Digital 1shot
Discriminator
24
Digital 1shot
1 ea.
1
10
Serializer
Transmitter
Segment
Finder
Cable
3
17
1 of 18
Receiver
DC
Trigger
3
10
16
Serial Clock
3
LG 085
Figure 10-25.
Level 1 drift chamber trigger electronics for primitives.
10.9.2 Drift Chamber Trigger
The drift chamber trigger performs three functions: local primitive formation, transmission
and gathering of primitives, and global track nding. The structure of the drift chamber
trigger is shown in Figure 10-24 with details relating to the primitives shown in Figure 10-25.
The trigger algorithms consist of two track nders. The rst identi es local track segments
using information from readout cards that receive signals from 20 neighboring wire-chamber
cells in the same superlayer. The second links these segments together to form track
candidates. This section describes the algorithms and a preliminary electronics design for
the baseline axial-stereo drift chamber [Kra94b].
Primitive Finders: Local Segment Finding
The segment nder algorithm is closely tied to the electronics implementation [Bol91, SLA93].
As is shown in Figure 10-7, trigger primitives are formed on the readout modules, which
receive analog signals from 24 channels comprising a 6 4 supercell. To accommodate the
maximum drift time, these signals are discriminated and converted to 600 ns-long signals by
digital one-shots. Such signals from the 24 channels on the card are input, together with
four signals from the neighboring readout modules, to a eld-programmable gate array that
nds track segments at a rate of one decision per 250 ns. Note that drift time information is
not used.
Possible track-segment patterns consist of a hit in each of the four layers, with each hit being
an immediate neighbor of the hit in the layer inside. An example of such a pattern is shown in
Technical Design Report for the BABAR Detector
,
,
(a)
1 2 3 4 5
25
,,
,,
433
XBL 9412-4268
10.9 Level 1 Trigger
(b)
Figure 10-26. Level 1 trigger primitives for the (a) drift chamber and (b) calorimeter. A
drift chamber track produces patterns like the one shown and uses majority logic to allow
three of four layers hit. For the calorimeter, energies from 25 blocks are summed locally to
form towers. The cluster sums shown for the example energy deposit are formed from the
primitives in a central location.
Figure 10-26(a). The elimination of sideways tracks provides background rejection without
signi cant loss in physics eciency. Majority logic is implemented to guard the algorithm
against ineciencies. Patterns containing three of the four hits required for a pattern also
produce valid segments. The segment nder generates 12 output bits, representing position
in the 6 4 supercell. The local angle of the segment is not reported.
Transmission Links: Segment Gathering
Transmission of drift chamber primitives is via a dedicated point-to-point serial link, as
outlined in one implementation of the calorimeter trigger [Wun94]. Using a serial link reduces
the number of wires needed and makes it possible to gather the signals in one location.
Particle Finders: Global Track Linking
A simple and ecient track linking algorithm is the Binary Link Track nder (BLT),
developed for the CLEO II trigger [CLE91] by K. Kinoshita [Kin89]. The method is to
start at the second superlayer and move radially outward. At each superlayer, a segment is
scored only if it is hit and any of its three immediate neighbors in the previous superlayer
have been hit. Here, hit refers to a segment- nder hit. At the last superlayer, denoted a
counting layer, any surviving scored segments represent tracks that traveled continuously
from the rst superlayer to the counting layer. The requirement of three neighbors imposes
loose curvature cuts on tracks, while being forgiving to superlayer stereo stagger e ects.
For BABAR, majority logic is incorporated by allowing combinations of four out of ve
superlayers to contribute segments. The B-track linker uses all ve permutations of four
Technical Design Report for the BABAR Detector
434
Electronics
or ve of the the inner ve superlayers, while the A-track linker uses all 36 permutations of
four or ve of the inner superlayers together with four or ve of the outer superlayers.
The BLT algorithm was chosen because it prevents double counting and because it avoids
searching for tracks in \roads." In the current drift chamber geometry, the latter method
would require a sector with = 74 (one- fth of 2) in order to just barely contain a track
with p = 180 MeV=c. To contain any track, one would need to search in twice that region
and then account for overlaps. Furthermore, unlike most road-based algorithms, the BLT is
insensitive to the sign of the electric charge of the track. Because it can be implemented in
combinational logic hardware, the BLT is also fast and inexpensive.
t
Particle Counters: Global Track Multiplicity Finding
The BLT outputs are used to count distinct tracks using simple multiplicity logic. The
output of the BLT is two digital strings with positions of segments in the fth and tenth
superlayers, representing tracks passing through these points. The logic checks to see if there
are at least two B tracks separated by an adjustable number (taken to be two) of supercells
in the fth superlayer (for a isolation of at least 22). For events with two or more A tracks,
isolation is checked in superlayer ten. The multiplicities of distinct A tracks and B tracks
are reported as 3-bit numbers to the global trigger.
Transverse Momentum Determination: Curvature Finding
Using positions of the axial layers of the drift chamber, it is straightforward to design a
Level 1 drift chamber trigger that reports multiplicities above crude p thresholds [CLE91].
The thresholds can be used to suppress the most severe trigger backgrounds from lost-beamparticle interactions, as shown in Section 10.8.
t
Drift Chamber Trigger for the All-Stereo Option
An all-stereo chamber with fully interleaved stereo layers is also being considered. This would
require elaboration of the tracking trigger. Logical superlayers would have to be formed from
alternating layers. A possible choice is six three-layer superlayers in each stereo view. The
de nition of logical supercells then leads to more than 300 unique segment- nding algorithms
with increased interconnectivity of the readout modules.
The apparent curvature of tracks in one view varies with the track's dip angle. This would
be largely corrected by summing the curvatures of tracks found in each view. For events with
small multiplicities, all pairs would be summed in parallel. Events with large multiplicities
would be accepted without the p requirement.
t
Technical Design Report for the BABAR Detector
10.9 Level 1 Trigger
435
10.9.3 Calorimeter Trigger
A calorimeter trigger based on towers (groups of contiguous crystals) will be more ecient
than a trigger which uses only the energy deposits in individual crystals [Kra94a] because
the center crystal of a shower will contain at most about 80% of the shower energy [SLA93].
Besides the improvement in eciency, the better estimate of the total energy deposited in a
tower compared with a crystal means that any energy threshold cut applied in the trigger
is more precise, which is not only better for background rejection, but is also important for
understanding and modeling the trigger response accurately.
The algorithm for the calorimeter Level 1 trigger [Kra94a] is summarized in this section. It is
based on xed towers of mostly 55 crystals and performs three functions: local tower energy
summing to form primitives, transmission of the primitives and cluster reconstruction, and
central particle nding and counting. The formation of primitives occurs on the calorimeter
readout modules (Section 10.6). The primitives are then transmitted to a central trigger
crate where cluster reconstruction is carried out. Finally, the results of the clustering are
compared and physical particles are found.
Primitive Finders: Local Tower Summing
Since the Level 1 calorimeter trigger is based on towers, the most straightforward arrangement is to organize the electronics by tower at the board level. Each tower, consisting of the
signals of a 5 5 group of crystals [Bol91, Kra94a], will be read out by a single calorimeter
readout module (Section 10.6), and the calorimeter trigger primitives will be constructed
from tower sums formed on the board. With the numbers of crystals listed in Chapter 7 and
a 5 5 tower structure, there will be 24 towers in and ten towers in along the barrel,
with the towers at the backward edge of the barrel containing only 4 5 crystals. For the
endcap, there is a 20-fold symmetry in the three superlayers, with each sector containing
45 crystals. Two towers will be formed from each sector, with ve unused channels shared
between each pair of boards. This gives a total of 24 10 + 20 2 = 280 towers. The trigger
sampling rate is xed to be the same as the data digitization rate (4 MHz).
For each time sample, the 25 channels of data are calibrated and converted to a cruder energy
scale of 8 or 16 bits using a look-up table in SRAM. This table allows a large degree of
exibility; an individual channel threshold cut can be trivially applied to remove low-energy
backgrounds. The 25 outputs are summed using a programmable logic device (PLD) to give
a total tower energy for each sample, again in 8 or 16 bits, and the sums are transmitted
to a central trigger crate. These sums form the calorimeter trigger primitives, although
it is possible that, to obtain a suciently accurate time and energy for the tower, further
processing might be performed on the readout modules rather than in the trigger crate.
However, assuming the raw sums are sent, the data rate is 4 MHz 8{16 bits or 4{8 Mbytes/s.
Technical Design Report for the BABAR Detector
436
Electronics
Transmission Links: Tower Gathering
The transmission of the calorimeter primitives is done via a dedicated point-to-point serial
link. The 4{8 Mbytes/s rate is reasonably low and is straightforward for a TAXI or similar
link [Wun94]. The primitives are collected together either on several cards in one trigger
crate, or possibly all on a single card.
Particle Finders: Global Cluster Formation
Due to energy sharing between towers, the towers are considered in neighboring pairs, called
dominos [Ell92], as indicated in Figure 10-26(b). All possible dominos are formed, and the
total energy of each domino cluster is used for the trigger. There will be three separate
energy thresholds for clusters in the trigger: an M cluster (above a low energy consistent
with a minimum ionizing particle, e.g., 150 MeV), a G cluster (higher than a MIP, e.g.,
600 MeV), and an E cluster (consistent with a Bhabha electron). These correspond to the
trigger objects in Table 10-7. The multiplicities of each object required for a trigger are
programmable in the central trigger logic. Due to the di ering crystal lengths in di erent
regions of the calorimeter (Table 7-3), the MIP signal size will depend on , and di erent
energy thresholds can be used in di erent towers if necessary.
Particle Counters: Global Cluster Multiplicity Finding
To make isolation requirements on the clusters, the cluster pairs are grouped together in
superclusters of between 9 and 12 towers each. For example, with eight superclusters in
and three in in the barrel, and eight in and one in in the endcap, there will be a total of
32 superclusters. By counting the number of clusters above threshold in each supercluster,
and comparing which superclusters have been hit, simple isolation requirements such as two
non-neighboring superclusters can be easily set.
One subtlety is that low-angle Bhabhas with both the electron and positron hitting the
calorimeter have a cross section on the order of 30 nb [Eis90], giving a rate of 100 Hz at
design luminosity. These are prescaled.
10.9.4
Global Trigger
The global Level 1 trigger consists of two pieces: a track matcher and a global decider.
Technical Design Report for the BABAR Detector
10.9 Level 1 Trigger
437
Track Matcher
The track matcher uses maps from the drift chamber and calorimeter triggers to nd the
match multiplicity. There are two types of matches: a track to barrel M cluster, and B track
to barrel or endcap M cluster. The track matcher takes the di erent granularities in the
drift chamber (42 bins) and calorimeter (24 bins) into account.
For the drift chamber and calorimeter, the multiplicity logic de nes distinct particles using
variable minimum cuts. The multiplicity of each type of match is reported as 3-bit
numbers to the global decider.
The track matcher reduces the orthogonality of the two independent Level 1 triggers and is
intended for use only during very severe beam backgrounds.
Global Decider
The global decider uses the multiplicity lines from the drift chamber, calorimeter triggers,
and the track matcher to issue the Level 1 strobe. It also has access to the calorimeter
-slice number for single-prong Bhabha events that need to be prescaled as a function of .
For a diagram of the the trigger input lines, see Figure 10-24. A completely exible trigger
is implemented through the use of a RAM-based look-up table with 24 input bits (each of
which can be masked o ) and one output bit. This implementation is upgradable.
10.9.5 Simulation
The Level 1 trigger was simulated using the GEANT model of the BABAR detector, BBSIM.
For the drift chamber trigger, the scored space-point hits in each layer of the GEANT model
were converted to wire hits that were fed to segment nders. The BLT algorithm [Kin89] was
used to nd tracks, and global trigger criteria were constructed based on counting distinct
A tracks and B tracks [Kra94b]. A display of an example B 0 event is shown in Figure 10-27.
A curvature nder was simulated using axial segment positions and patterns as inputs.
For the calorimeter trigger, the scored energy deposits in each CsI(Tl) block were summed
into xed 5 5 towers that were paired to form two-tower (50-block) clusters. Isolation
criteria in and were used to count clusters [Kra94a].
The implicit transverse-momentum and solid-angle trigger acceptances for single muon tracks,
identi ed as the A-track, B-track, and M-cluster objects of Table 10-7, are shown in Figure 10-28. Note that the ineciency due to tracks hitting adjacent corners of four towers,
Technical Design Report for the BABAR Detector
438
Electronics
Example of trigger response to a simulated B 0 event. The dots represent
wires that were hit. Boundaries of segment- nder supercells that were hit are drawn.
Crosses mark locations of A tracks and B tracks in counting layers. The event satis ed
the open trigger requirements.
Figure 10-27.
thus depositing only half their energy in one domino, is neglible. Event eciencies and rates
using this simulation are given in Section 10.8.5.
10.9.6 Trigger System
Timing Considerations
The Level 1 trigger has a latency of 9:5 s and a jitter window of 0:5 s. These constraints
are dictated by the characteristics of the front-end systems.
The latency time was set to a number that would not change and that would provide the
maximum time to perform the Level 1 trigger while not degrading the physics of the vertex
Technical Design Report for the BABAR Detector
(per 0.025 GeV/c)
dε/dpt
10.9 Level 1 Trigger
439
1.0
(a)
0.8
0.6
0.4
B
A’
0.0
1.0
(per 0.02)
M
0.2
0
dε/dcosθ
A
0.2
0.4
0.6
0.8
pt (GeV/c)
(b)
A
0.8
M
B
0.6
B
0.4
A
M
0.2
0.0
-1
-0.9
-0.8
-0.7
0.7
0.8
0.9
cosθ
1
Di erential eciencies for single muons to be identi ed as trigger objects
A, A', B or M (see Table 10-7), as a function of (a) pt with cos < 0:7 and (b) cos with
pt > 0:4 GeV=c.
Figure 10-28.
j
k
detector. At full luminosity and 10 times background, there is little impact on the physics
by the latency. At full luminosity and 10 times background, the time uncertainty will result
in 30% occupancy in some sections of the vertex detector due to lost beam particles.
The pipeline length is composed of several elements, listed in Table 10-12. The trigger
contains circuitry to align the track linker output to the late-arriving cluster nder output
before these reach either the track matcher or global decider. The estimated minimum
pipeline length of 6 s is well within the xed latency of 9:5 s. Part of the safety margin
is expected to be consumed as detailed designs are developed for the trigger. To make
up any di erence or account for potentially shorter calorimeter shaping times during highbackground running, the output of the global decider passes through an adjustable-delay
unit before producing the Level 1 strobe.
The jitter is set by the characteristics of the drift chamber and calorimeter. Calorimeter
signals will have a 1{2 s peaking time. These signals are sampled every 0:25 s to form
the trigger primitives. Simple peak nding is used in the Level 1 trigger. This results in
Technical Design Report for the BABAR Detector
440
Electronics
Path
Latency
Detector Signal Development
Drift chamber Drift Time (600 ns)
Calorimeter Peaking Time (4000 ns)
4000 ns
Transmission to readout modules
100 ns
Logic to Form Local Trigger Primitives
Drift chamber Segment Finder (250 ns)
Calorimeter Tower Summer (250 ns)
250 ns
Transmission to Particle Finders
Drift chamber Track Linker (300 ns)
Calorimeter Cluster Finder (250 ns to 450 ns)
450 ns
Logic to Form Trigger Decisions
500 ns
Transmission to Fast Control Distribution
100 ns
Transmission to Readout Modules
100 ns
Transmission to Vertex Detector Front-End Cards 100 ns
Total Minimum Pipeline Length
5600 ns
Table 10-12.
Level 1 trigger pipeline length.
nding the peak to within one time bucket or a resulting jitter of 0:25 s. Drift chamber
signals are sampled at a faster rate. They are extended by digital one-shots of 600 ns (the
maximum drift time). Decisions are made every 250 ns by the segment nders. The resultant
jitter window is 850 ns. A trigger jitter shorter than 0:5 s would reduce the storage and
bandwidth requirements on the front-end systems, which could be signi cant for the vertex
detector.
Globally, these triggers from the drift chamber and calorimeter are processed to form a global
trigger by the track matcher and global decision maker every 0:5 s. This speci cation is set
to be consistent with the Level 1 trigger jitter requirement of 0:5 s.
System Functions
The Level 1 trigger system interacts with almost all other components of the BABAR electronics and online computing systems.
Global triggers will have prescaling of the selected trigger types. This prescaling is included
to allow for optimization of the desired rate for physics data and for monitoring functions
such as Bhabhas.
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
441
It is necessary to have timing and control interfaces to the fast control system. The trigger
delivers the Level 1 accept signal to the fast control system. Fast control produces the Level 1
strobe for the front-end systems. The trigger is a consumer of fast control signals like all
other front-end systems.
Monitoring of deadtime and eciency must be part of the functionality of the electronics
system. Deadtime monitoring is accomplished in the fast control system by measuring the
physical time the trigger is not accepting new triggers and monitoring of scaled-down triggers
such as Bhabhas. The eciency of the trigger will be determined by the online system that
is monitoring the primitives and resulting triggers. The scheme to perform this functionality
must be included in the online design.
Readout by the data acquisition system will include all trigger primitives, drift chamber and
calorimeter trigger output, and global decisions. This is needed to monitor the eciency
and operation of the trigger for both online and oine analysis.
The trigger will be partitionable such that it can run at the crate level, trigger system level,
and BABAR system level. This functionality is included to allow for installation, debugging,
and system testing. As part of the testing, the system will have the capability of injecting a
calibration signal into the front-end electronics that allows for the determination of trigger
primitive thresholds. Provisions will also be made for memories that can be downloaded
with patterns which can be played through the trigger. These functions will facilitate the
characterization of the trigger and help determine whether it is operating as desired.
10.10 Data Acquisition
10.10.1
Introduction
The key function of the BABAR data acquisition system is to collect the event data selected
by the Level 1 trigger from the front-end and trigger systems, and transfer these data into
a processor farm for packaging and ltering. Both hardware and software are required to
concatenate the event fragments from all the data sources into complete events. Hardware
and software protocols are needed to provide error-free transmission of the data, detect
invalid operating conditions, prevent bu er over ows, and throttle the trigger rate. The
data acquisition system also provides calibration, monitoring, and debugging facilities. The
specialized equipment needed to calibrate the system electronics, including fast strobes,
software processes, and hardware veri cation is a natural extension of data acquisition
activities. Monitoring of front-end and trigger components, providing instantaneous status
Technical Design Report for the BABAR Detector
442
Electronics
information, localization of errors, and error recovery tasks such as the ushing of queues
are a joint responsibility of the data acquisition system and the online computing system.
The data acquisition system interfaces to all detector front-end systems for event data
collection, to the trigger systems for both data collection (of trigger information) and event
data ow control, and to the online system for setup and runtime control as well as monitoring
of the data acquisition system.
Data are transferred downstream within the data acquisition system in response to the
identi cation of events by the trigger system. Under normal conditions, the Level 1 trigger
rate is expected to be less than 1.5 kHz with the open trigger shown in Table 10-9. The
largest component in this rate is cosmic ray events. Under severe machine background
conditions, the Level 1 trigger rate can be kept below 2.0 kHz using the safe trigger shown
in Table 10-11. In subsequent trigger levels, the drift chamber information and the silicon
vertex detector information are used to restrict the tracks to those that originate near the
interaction point. The nal Level 3 trigger is expected to send approximately 100 events
per second to be logged to tape. Contributions to this data rate are produced in many
crates and ow through the event-assembly network over many individual data paths. Thus
the total bandwidth can be easily achieved using any one of a variety of modern transport
mechanisms. The BABAR detector trigger and data acquisition system is fully pipelined in
both the trigger and the data ow paths. As a result, the system is nearly deadtime free
at background levels up to ten times the rate given by machine simulations. Commercial
equipment is used whenever suitable modules are available. The architecture is such that the
capacity of the system is easily upgradable with the addition of data paths and processors.
10.10.2 Requirements
The requirements for the BABAR data acquisition system fall into four broad categories. First
are quantitative requirements such as those on the bandwidths of data paths and the number
of data sources to be accommodated. Next are architectural requirements that allow frontend and trigger systems to be operated as a complete system, or allow portions of each to be
operated as separate partitions. Then there are requirements for initializing, controlling, and
monitoring the system. Last are quality control requirements for items such as reliability,
error rates, and maintainability. For a more complete discussion, see the BABAR electronics
requirements document [Req94].
Quantitative Requirements
The data acquisition system must satisfy the requirements for event rates, bandwidths,
latencies, and deadtimes. The fundamental requirement of the data acquisition system
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
443
Number of
Number of Number of
Detector System
Readout modules Crates Data Sources
Vertex Detector
26
2
5
Drift Chamber
285
18
18
DIRC
3
1
1
Aerogel (ATC)
8
1
1
Calorimeter
280
18
18
Muon System (IFR)
7
2
1
Trigger
3
3
3
Total
612
45
47
Table 10-13.
system.
Numbers of readout modules, crates, and data sources for each detector
Maximum Trigger Rate
2 kHz
Max. Level 1 Latency
10 s
Max. Average Event Size
25 kbytes
Max. Event Data Rate
50 Mbytes/s
Max. Deadtime
10%
Number Data Sources
47
Number Level 3 Processors in Farm
10
Table 10-14.
Requirements of the data acquisition system.
is that it be capable of collecting all event data for those events selected by the trigger,
while imposing less than a speci ed amount of deadtime. Table 10-13 shows a list of all the
detector systems that generate event data. The middle column of the table gives the number
of readout crates required to house the modules associated with each detector system; the
number of data sources for each detector system is shown in the last column. Most crates
contain only one data source; the crates for the vertex detector each contain two or three
data sources.
Together, all sources are expected to generate approximately 25 kbytes of data per event,
with a combined event data rate of 50 Mbytes/s. Other requirements are derived from the
expected physics and trigger rates, and the deadtime requirement of less than 10% at design
luminosity. A list of important requirements is given in Table 10-14.
Technical Design Report for the BABAR Detector
444
Electronics
Architectural Requirements
The system must scale modularly to accommodate upgrades, since it is inevitable that the
data acquisition requirements will change during the life of the BABAR detector. A key
element in this scalability is the design of the event-assembly network.
Furthermore, the data acquisition system must be partitionable. Partitioning means logically
dividing the electronics of the detector into two or more independent subsystems, each
with independent control. This allows multiple detector systems to perform calibration,
diagnostic testing, and data acquisition tasks simultaneously and without interference. With
this design, it will be possible to assign individual crates or entire subsystems to a partition
under computer control without the need for recabling.
Control Services Requirements
The data acquisition system must also provide a set of services to control and utilize the
calibration and monitoring facilities built into the various front-end and trigger subsystems.
It must recognize and handle error conditions|i.e., errors in the transmission of data or
malfunctions of components of the system. It must record non-event data streams such as
magnet currents, high voltages, and accelerator parameters. It must initialize, download,
and control the various processors and modules distributed throughout the system. Finally,
it must deliver event data to consumer processes reliably.
Quality Control Requirements
The data acquisition system must ensure the integrity of the event data stream. It must
transport data from the front-end and trigger systems without errors in either data content
or the identi cation of origin.
Care is being taken to ensure a high level of reliability. Proven industry standards and wellsupported commercial software and hardware will be used wherever possible. Diagnostic
features such as error detection and correction schemes will be implemented. To minimize
downtime, the data acquisition system will be able to detect operational errors or malfunctioning modules and correct or report these conditions. Most system hardware will be
implemented with modular components to permit the rapid replacement of malfunctioning
units.
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
445
10.10.3 Architectural Overview
A block diagram showing the main components and interfaces of the data acquisition system
is shown in Figure 10-29. In this gure, the data ow from left to right. The main components
are the readout crates (VME), the event data ow control system (fast control), the eventassembly network, and the Level 3 event processing farm. The main interfaces shown are the
front-end interface, the Level 1 trigger interface, and the slow control interface. Within each
readout crate are the components of the various detector front-ends, trigger interfaces, and
data acquisition system. These components collect analog data from the detector front-ends,
process them in response to a trigger signal, and deliver digitized data to the downstream
parts of the data acquisition system. The main components of the data acquisition system
are described in the following sections.
Crate-Based Electronics Overview
Within the readout crates, there are four principal components or modules:
readout modules that interface to the detector-mounted electronics and, for the drift
chamber and calorimeter, provide speci c information to the prompt trigger;
fast control distribution modules that interface the crate to the fast control system;
a readout controller that connects to the detector control system, provides monitoring
and diagnostics, reads the event data fragments in the crate, and controls the ow of
data through the bu ers; and
an event-data port card (data source) that interfaces the crate to the event-assembly
network.
In each crate, the readout module connects the detector-mounted electronics to the detectorspeci c trigger, fast control system, and readout controller. Since most of these interfaces are
identical for all detector systems, it is useful to look at the common features. Emphasizing
the commonality of the systems creates a uni ed data acquisition structure. Although there
are unique interfaces to the on-detector electronics and to the trigger, there is a common
interface to the fast control system and another to the readout controller through the crate
bus.
Event-Data Control Overview
The ow of the event data through the data acquisition system is controlled by the eventdata control system. Its primary function is to keep data owing eciently from the
Technical Design Report for the BABAR Detector
446
Electronics
(Optional)
Level 1
Trigger
Processor
Front-End
Electronics
Trigger
System
DAQ
System
Trigger Data Path
Drift
Chamber
Readout
Crates
To
Readout
Crate
Level 2
Trigger
Processor
(2 Rings)
Calorimeter
(5 Rings)
Level 3
Processor
Switch
Vertex
DAQ
System
(2 Rings)
Fast
Control
Online
System
1-95
7857A06
Figure 10-29.
Level 3
Processor
Event Data Rings
Triggers
PID, IFR
Trigger
Fast Control Path
Detector Control Path
(5 Rings)
Detector
Control
Run
Control
Block diagram of the Data Acquisition system.
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
447
front-end electronics, through the readout and trigger-system crates, through the eventassembly network, and to the Level 3 processor farm. The data acquisition system has three
separate mechanisms for providing the event-data control: high speed control and trigger
throttling (the fast control system); VME-based data collection and ow control algorithms;
and network control and resource allocation.
The high-speed ow control uses a \backpressure" model to guarantee the orderly delivery
of data. The mechanism works as follows. Ten s after a particular beam interaction occurs,
the Level 1 trigger declares a Level 1 accept. The fast control system fans out this Level 1
accept signal to all crates currently accessible to the main trigger partition. The signal is
received synchronously at these readout crates by the fast control distribution modules. If
this module is con gured to respond to signals from the main partition, then the Level 1
accept is propagated though the crate's fast control bus to each readout module. When
input queues on the readout modules become full, a busy signal is transmitted back through
the fast-control system to the trigger.
When a readout controller processor has completed processing an event fragment, the data
will be located in a region of memory accessible by VME. The processor will also place the
location of this event fragment and the length of the fragment in a separate queue, also
accessible by VME. If either the data queue or the pointer queue becomes full the processor
stops and the VRAM lls, resulting in a throttle condition. The locations of the heads and
tails of both queues are accessible to both VME and the processor. The read-modify-write
protocol of both the VME and processor ensure that the queues are properly maintained.
Each readout controller within a partition is assigned a unique position in the readout
sequence. When the readout controller in the rst crate in the sequence nds that it has
a block of events ready and receives an event-request token, it transmits the event-data
fragment to a designated Level 3 processor. When the transmission is complete, the token
is passed to the next readout controller in the sequence. This process continues until the
last crate in the sequence is reached. At that point, the token is sent back to the Level 3
processor. When the complete event has been successfully read, a message is broadcast
to all readout crates signaling the successful completion of the event assembly. A suitable
algorithm will distribute events among the Level 3 processors.
Event Assembly Overview
The task of the event-assembly network is to send the event fragments from the 47 event-data
port cards to the designated Level 3 processor.
Some of the requirements of the event-assembly network are listed in Table 10-14; these
numbers will change as the design progresses. For example, the number of destination
processors will depend on the computational speed of the processors, re ned estimates of
Technical Design Report for the BABAR Detector
448
Electronics
the computational capability required to process a single event, and re ned estimates of the
event rates generated by the trigger processing.
The event-assembly network will be based on a commercial network similar to those presently
being used or under development for the computer and telecommunications industries.
Such a network, combined with the use of standard communication protocols, has many
advantages. The hardware is readily available from multiple vendors. It is reliable and
has a good price/performance ratio, especially when development costs are considered. The
communication protocols are already implemented and include extensive error checking, ow
control, and collision management. The data are usually self-directing and do not require
any central intelligence to control the data ow. The amount of programming labor that is
saved by employing this solution instead of a home-grown network is enormous.
The bene ts of an interconnect based on commercial networks are not free. The key disadvantages are constraints imposed on the architecture by network hardware and protocols.
The advantages, however, include the possibility of implementing exible load-balancing
algorithms in software without changes to the hardware con guration and the capability of
scaling the system incrementally.
The ow control and error recovery procedures of the network protocol can be used to
implement a simple data- ow architecture. A distributed ow control system allows high
link utilization without special hardware to implement control functions. Readout synchronization in the trigger system and front-end electronics is required to achieve this bene t.
A Level 1 trigger inhibit must be asserted if any readout crate lacks sucient bu ering to
read out its portion of an event. This prevents partial readout of an event.
The network protocols and interfaces have been designed for ecient transmission of large
messages and modest transaction rates. Thus, a message size of about 4 kbytes or more is
required for ecient use of the network. This will require fragments containing data from
several events to be sent in each message originating at a data source in a readout crate. This
has the advantage of decoupling the real-time event rate from the rest of the data acquisition
system. Sucient memory must be provided at appropriate points in the system to bu er
the large messages.
Online Interface Overview
Software development for the readout controllers and readout module processors requires a
commercial development environment and a commercial real-time operating system. The
BABAR computing group is considering the VxWorks real-time operating system and the
EPICS control system software tool kit for use in the control systems. Broad use of these
tools throughout the computing, online, data acquisition, and trigger systems permits shared
expertise and seamless integration of experiment control and data acquisition. The online
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
449
interface provides the physical means whereby operators and applications software have the
ability to control and monitor the many processors in the front-end and data acquisition
systems.
10.10.4
Readout Crates
The readout crates provide housing for several types of modules required for the electronics
system. These include the readout modules, the fast control distribution module, the readout
controller, and the event-data port card module. The readout crates will be standard VME
crates. All readout data, messages, and monitoring are to be transferred between modules
on the VME backplane bus. Lines on the VME P2 bus are to be used to transmit high-speed
control, trigger, and timing signals. Each crate will have spaces for 21 modules. These will
be able to accommodate the readout controller, the event-data port card, the fast control
distribution module, a temporary bus monitor module (for debugging), one space reserved
for future use, and the 16 readout modules.
Readout Modules
All detector systems are able to present a common hardware and software interface to
the central electronics system because of the design of the readout modules. As shown
in Figure 10-30, each card contains a standard set of functions, which are shown as solid
blocks. The \personality" section, shown in the gure as a dashed block, is designed to
accommodate each detector system's special characteristics. It is expected to make provision
for bi-directional data ow between a readout module and its associated front-end electronics.
This will permit, for example, dummy events or con guration data to be downloaded into
the front-end electronics for tests of the overall system.
The readout module receives triggers, clocks, fast resets, etc., via the crate's fast control
bus. Triggers are time stamped and bu ered in the FIFO before being presented to the
card's CPU. This feature bu ers the CPU from the rate uctuations. The FIFO almost
full condition results in trigger throttling; receipt of a trigger in the full condition generates
an error signal. The interface between the personality section and the readout module will
convey all necessary signals to permit any special processing needs of the detector systems.
These may include analog signal processing, fast digitization, and data bu ering. A bitparallel digital data bus will convey data over the interface. A serial clock signal strobes the
data into or out of the VRAM depending on the state of a direction signal. The clock line
stops whenever there are no incoming data. The numbers of DAQ readout cards required
by the various detector systems are given in Table 10-13.
Technical Design Report for the BABAR Detector
450
Electronics
VME
Crate Bus
From
Front-End
Electronics
“Personality”
Section
DAQ Readout Card
VRAM
RAM
Program
Memory
CPU
VME
Interface
Trigger
Time Stamp
Store
FIFO
Clocks, Reset, etc.
Fast
Control Bus
1-95
7857A7
Figure 10-30.
Block diagram of a DAQ readout card.
To achieve event synchronization across the interface, all data streams and event serial
numbers are initialized by a system-wide reset signal. The read strobe then ensures that the
multiple data streams remain synchronized. The personality module sends an error signal
when it detects transmission errors relative to the incoming data stream.
Fast Control Distribution Module
The fast control system provides all of the high speed timing signals required for the selection,
synchronization, and identi cation of data. These functions include fast reset, initialization,
synchronization, clocking, calibration, triggering, and busy indications. The system also
conveys those signals needed for rapidly enabling and disabling the triggers as required by
conditions of lling and emptying of data bu ers. A crate fast control bus will augment the
VME backplane, and many of these signals will be transmitted over the unassigned pins of
this bus.
The fast control distribution module is a VME card that receives the high-speed signals from
the fast control system and repeats the signals on the crate fast control bus. This card is
able either to drive directly the signals received from the clock and control interface or to
generate these signals under program control through VME. This latter capability is needed
to perform stand-alone operation and calibration of systems created by forming partitions.
For example, the timing card can be told to issue a trigger or a calibration strobe at a speci c
moment, or select a particular partition with which to communicate. In addition, this card
performs various control functions within the crate, including receiving a trigger throttle
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
451
signal and an error signal from each readout card. These signals are then made available to
the central fast control system, to be used in the trigger throttling algorithm.
Readout Controller
The readout controller is the only VME master contained within each readout crate and
so is responsible for many activities, the most important of which is the event readout. In
addition, the processor on this board is responsible for downloading software to all processors
within the crate, responding to crate interrupts, performing detector control, monitoring,
calibration, debugging, and operating in stand-alone mode. Since EPICS is likely to be
used in the detector control subsystem, this processor must operate under VxWorks. While
minimal memory requirements are needed to support the small event sizes of the BABAR data
stream, excellent processor throughput will be required to assist in data formatting, queue
management, and interrupt servicing.
The processor on the readout controller receives a trigger as an interrupt. After the processor
services the interrupt, it executes user-de ned codes to allocate bu ers and read in the data.
It builds a list of commands to transfer these data and then performs them as a sequence of
high-speed input-output operations. When the readout controller determines that a complete
event has been processed within the crate, it waits for a signal (token) before requesting
transmission of the event fragment to the event-assembly network. Due to occasional long
latencies in the readout, there may be several events ready for readout.
Event-Data Port Card
The event-data port card (data source) in each crate interfaces to the transport medium for
all event data streams and is responsible for transmitting the crate's event fragments via this
medium to the event-assembly network. Since none of the expected data streams exceeds
5 Mbytes/s, many data transmission technologies and logical arrangements are possible
candidates for this task. BABAR uses switched FDDI.
The data sources associated with the various detector systems output data at widely di erent
rates. This makes it quite reasonable to arrange the data sources into a topology of several
rings, thereby reducing the number of virtual sources seen by the event-assembly network.
This arrangement is discussed in detail in Section 10.10.6.
Technical Design Report for the BABAR Detector
452
Electronics
Trigger
Processor
Strobes
Throttle
Error
Type Bits
2
2
2
2
ROC
Fast Control
Source Modules
To Detector
Type B
Fast Control
Distribution Modules
1-95
7857A3
Readout Crates
for Detector Type A
Figure 10-31.
10.10.5
Readout Crates
for Detector Type K
Partitioning logistics.
Event-Data Flow Control
Partitioning
It is important to be able to partition the electronics systems for the various detectors
into independently operating systems for commissioning, calibration, and repairs. It is also
useful to be able to run the detector in normal data-taking mode even though some parts
are unavailable for various reasons, including being independently operated. Counting the
trigger, BABAR contains seven independent systems. It will be possible to create at least seven
partitions, allowing the systems to be initially commissioned independently. Figure 10-31
illustrates the partitioning scheme.
The lowest level of granularity in partitioning is the crate. A crate is either oine or online in
a partition. A crate in a partition is under the direct control of the data acquisition system.
An oine crate can still communicate with the online software for purposes of reading or
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
453
setting voltages, running dedicated software diagnostics, etc., but it is not included in the
normal data acquisition operation.
The partition containing the trigger system is referred to as the main data-taking partition.
From the trigger system, triggers are distributed to other systems via the fast control
distribution system. We do not envision the trigger system itself as partitionable, although it
may be possible to take some of its crates oine for limited repairs. A crate which is oine,
or which is not in the main partition, may not be generating valid trigger primitives and will
be ignored by the trigger system. This will be set up as part of the partition initialization
process.
The crates in a partition other than the main partition still communicate using the fast
control distribution system and the data-transfer protocols. A system within that partition
may be responsible for generating strobe signals for distribution in lieu of triggers, for
example, to control calibration and diagnostic activities. It is also possible to de ne a
partition that does not contain a module to drive the fast-control system; this partition is
then fully functional from a data-transfer perspective but will require that all processing be
initiated by software messages instead of by hardware strobes.
Crates can be read out via the main data transfer path or at a much slower rate via the
controls local-area network. Depending on the nal architecture for the main path, it may
be possible to divide it logically among partitions or to share it on a rst-come rst-served
basis. Either of these options will provide acceptable functionality for commissioning and
calibrating the detector.
Fast Control System
Clocks, strobes, type-bits, throttle signals, and error indications must be distributed among
the data acquisition crates. These must be distributed in a exible way to allow systems to
use them for their own commissioning needs within special purpose partitions.
The free-running nature and small physical size of the BABAR electronics system makes timing
and control distribution relatively easy. We plan to have a few central modules (fast control
source modules) generating the required timing and control signals. The timing signals will
be derived from the PEP-II rf clock and will maintain a stable phase relationship at the
nanosecond level on the module output.
The set of strobe signals is used to carry time-speci c information such as the Level 1 trigger
and calibration strobes. Type-bits are used to indicate special purpose strobes, for example,
an occasional diagnostic trigger which requires custom processing by the data acquisition
system. Throttle signals are used to hold o further strobes when a system or crate is unable
to process further requests for a short time, or they can be used to indicate that processing
Technical Design Report for the BABAR Detector
454
Electronics
of a given strobe is complete. Error indications are used to inform the online system that
some condition has been detected that requires higher level intervention.
The main partition will have type-bits, to classify events as, for example, normal, Bhabha,
uncompacted, and snapshot. The last two are typical of several types used for occasional
checks of the data acquisition logic. They will carry the Level 1 accept on a strobe and may
use a second strobe to inform data acquisition modules in advance of a diagnostic trigger.
Strobes in the main partition, including any that need interlocking to the throttle signal and
special intertrigger time requirements. The fast control system monitors trigger deadtime.
During system-speci c calibration, one might use several strobes to command the crates to
generate local timing signals, and then trigger data readout on the result. It is the responsibility of the particular system to generate the required strobe signals, with interlocking to
the throttle control as appropriate. If a system has only one set of electronics for generating
these signals, it will not be possible to split the system into two fully functional partitions.
Fast Control Distribution Module
We currently envision eight strobes, eight busy signals, eight error signals, and eight sets of
type-bits. This is enough to allow all seven major systems of the BABAR detector to run in
independent partitions when required.
To distribute these signals, each data acquisition crate will contain a fast control distribution
module. It receives the entire set of signals and provides a standard interface to the crate.
It can be programmed via the VME interface to route particular strobes and/or type-bits to
either front panel connections or a set of reserved backplane signals, or both. Additionally,
VME interrupts can be generated from strobes. Throttle and error signals can be driven from
front panel connectors or from backplane signals. All signals can also be interrogated via
VME read operations for diagnostic purposes. VME-readable scalars and LEDs are provided
for most functions to help in commissioning and debugging.
Fast Control Source Module
The fast control source module is used to drive the distribution system. It is con gured
via VME to drive strobes and type-bits from dedicated backplane signals or front panel
connections, and to break out throttle and error signals onto the backplane, front panel, or
into VME interrupts as required. Typically, the error signal will interrupt the local readout
controller, which will poll the appropriate modules to nd the cause of the problem. All
signals can also be interrogated via VME read operations for diagnostic purposes. One or
more of these fast control source modules is then used by the trigger or calibration control
electronics in each system to distribute the required strobes, and other signals to the other
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
455
crates in the system for operation in a partition. It may be possible to combine both the
fast control source module and fast control distribution module onto a single VME card to
reduce expense.
Using this system, the online software can exibly allocate signals to particular partitions
by controlling the settings of the control module(s). This, in combination with informing
the trigger system of which crates to monitor and setting up the data-transfer network
appropriately, allows each partition to function independently.
Throttling and Bu er Management
The basic hardware throttle for the BABAR data acquisition system is provided by the throttle
lines of the fast-control system. Asserting the relevant line for a partition inhibits its strobes.
In the data-taking partition, this means that distribution of the trigger is inhibited, causing
deadtime. To allow an accurate determination of deadtime, there will be pairs of scalers
counting system clock pulses|one gated by throttle and the other not. One pair of scalers
will be assigned to each possible partition. In addition, the trigger will count Level 1 strobes
internally, even when the data acquisition system is inhibited. All these scalers will be
accessible as part of the normal event stream.
There are two sets of bu ers that the fast control system must protect from over ow. The
rst of these are contained in the readout chips of the vertex detector electronics. These
contain on-chip back-end bu ers which are of xed length and three events deep. They
bu er against the variations in transmission time along the ber to the readout modules.
The fast- ow-control system must promptly assert throttle to prevent any overwriting of
data if all three on-chip bu ers on any readout chip are occupied.
Due to the limited space available, this portion of the fast control system will reside in the
link controller at the data acquisition end of the line. The link controller will keep track of
the chip bu er occupancy by monitoring a local model of the chip bu er states.
The other bu ers controlled by the fast control system are the VRAM bu ers on the readout
modules. These are operated as circular bu ers, in some cases (such as the drift chamber
subsystem) with addresses linear in time, and in other cases, (e.g., the vertex detector
system) with addresses linear in event number. In the event-linear case, the throttle will be
applied if there is too little space to accommodate two maximum-size events.
There is a third source of fast control signals. When a Level 1 strobe is presented to the
readout module, the address in the VRAM of the corresponding data needs to be noted
for later digestion by the readout module's CPU. This work queue will be maintained in
a hardware FIFO. This FIFO will also be able to generate throttle if it is at risk of lling
completely.
Technical Design Report for the BABAR Detector
456
Electronics
The throttle signals from each readout module will be presented individually to the fast
control distribution module in its crate. This will allow disabling of these signals from
broken hardware.
Event Data Flow Control
The fast ow control system is the only form of event data ow control that directly throttles
the trigger. All later stages of control operate by a backpressure mechanism.
In the backpressure scheme, any stage in the event ow|starting with a processor in
the Level 3 farm|which lls up simply refuses to release resources used by its upstream
neighbors. This eventually causes the upstream neighbor to ll. The process cascades until
the upstream neighbor is the memory on the readout module. When that memory lls,
throttle is asserted and deadtime is introduced.
These algorithms become more complex in the case in which events are bundled into super
events to amortize the cost of interrupts. Note that bundling events can also reduce the
interrupt load within the crates; the fast control system could be programmed to provide
interrupts to the readout controller only after a designated number of events is available in
the readout modules.
10.10.6 Event Assembly
After the delivery of the Level 1 accept, the readout modules and readout controllers proceed
autonomously to the point at which all useful event data from a crate are collected into that
crate's readout controller. At this point, the data represent the crate's event fragment.
The event assembly process collects all the crate-level fragments for a particular event into a
single Level 3 node for further ltering, reconstruction, and storage. The baseline technology
choice is switched FDDI, a commercially available high-speed network.
Such an assembly network must provide enough throughput to handle the worst-case average
event size at the Level 1 accept rate with a reasonable latency. Instantaneous rates above
the network throughput will have to be bu ered before the network or deadtime will result.
Since chance uctuations may ll any bu er, the goal is to provide enough bu ering to keep
this deadtime below the few percent level at 10 times nominal background.
The pre-network bu ers imply a latency for assembling events that is uctuating and potentially large compared to the event rate. Since it is easy to provide this bu ering on the
readout controllers, and there are no hard real-time restrictions after the data have been
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
457
Readout Crates
Event Data
Readout Ports
1
2
16
FDDI Switch
Processors
in the Farm
1-95
7857A8
Figure 10-32.
Crate readout ring topology.
rescued from the trigger-latency bu ers, this latency is not important. The uctuations
introduced by the Level 3 nodes are likely to be much greater.
In a switched network with multiple destinations, only the switch needs to see the full
data bandwidth. Typically, switches are designed with very high bandwidth backplanes.
For example, one currently available FDDI switch has a capacity of over 3 Gbits/s (about
400 Mbytes/s). Real utilization approaching 80% of this gure has been observed at LBL,
whereas the BABAR maximum demand is only 50 Mbytes/s. Hence, the switch itself should
not be a limit.
After the switch, the next potential limit is the capacity of an optical ber. At an FDDI
data transfer rate of 100 Mbits/s (12.5 Mbytes/s), it is clear that a single ber cannot cope
with the load. On the other hand, assigning one ber per data source would result in a large
number of input ports on the event-assembly network, with many of the bers signi cantly
under-utilized in terms of data rate. A reasonable alternative is then to assign several data
sources to each ber data channel, with the number of sources per channel to be determined
by the data rates of the sources.
Technical Design Report for the BABAR Detector
458
Electronics
System
# of Sources # of Sources/Ring # of Rings Data Rate/Ring
Calorimeter
18
4
5
3.2 Mbytes/s
Drift Chamber
18
9
2
3.6 Mbytes/s
DIRC & ATC
2
2
1
3.8 Mbytes/s
Vertex
5
1
5
3.3 Mbytes/s
IFR & Trigger
4
4
1
3.2 Mbytes/s
Totals
47
14
Table 10-15.
Readout ring topology and bandwidth at 2.0 kHz.
Figure 10-32 shows a ring topology for combining several data sources or sinks onto a single
ber channel. Here, it is assumed that the event-assembly network is a 32-port device.
Sixteen of the ports are assigned as input ports, with each port connected to a ring connection
of data sources. The remaining ports are used as output ports to the processors in the Level 3
farm, which are also connected in rings. None of the expected maximum data rates per input
ring exceed 25% of the 12.5 Mbytes/s bandwidth of switched FDDI.
These rings could be upgraded with a future higher performance technology at modest cost
in order to take advantage of higher throughput. Technologies such as FCS ( ber channel
standard), or SCI, can be used in a similar ring topology and provide signi cantly greater
throughput. Technologies such as HIPPI and ATM do not have ring topologies, but suitable
substitutes exist with similar performance improvements.
The sequencing of data onto the input rings is controlled by token passing. To gather event
fragments, an input port of the event-assembly network will launch a token. The rst data
source on the ring will receive the token and will transmit its event fragment(s) via the ber.
When it is nished, or if it has no fragments to send, it immediately forward the token to
the next source.
Tokens can be used to convey instructions to the data sources, e.g., the Level 3 farm processor
to which the current event data fragments are to be addressed.
The numbers used to assign data sources to particular rings are given in Table 10-3, where the
last column shows the data rates for data sources assigned to each detector type. Table 10-15
shows how the 47 sources can be connected into 14 rings, with the rings having roughly
equivalent data rates easily within the capacity of FDDI bers.
On the output side of the switch, there needs to be at least six rings to keep the load within
bounds. The number of farm nodes per output ring is essentially unlimited; BABAR will use
whatever is needed to supply sucient computing power.
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
459
The next potential bottleneck is the data rate capacity of the event-data port cards. These
are the cards that interface a data source or a farm processor node to an FDDI optical ber.
Of course, every event-data port card will put bits on the ber at the full 100 Mbits/s for the
duration of its packet. The rst question is how quickly it can or needs to supply packets
so that the occupancy of the ber is as high as possible. The second question is how much
local processing power in the event-data port card is required to support a TCP/IP protocol
at high rates. There are reports of event-data port card/processor combinations which can
achieve full ber loads even while running TCP/IP over the ber, but these gures are likely
to be reached only under ideal circumstances. Fortunately, doubling the number of nodes per
ring directly cuts the event-data port card load in half, so there is a simple cure within this
architecture. Also, this is a rapidly developing technology. BABAR will monitor developments
and take advantage of performance improvements and cost reductions.
Note that, for the above minimal output con guration, the trac of events out of Level 3,
which are recorded on tape, represents less than 7 Mbits/s (800 kbytes/s). This is a suciently small load that it will not be discussed further.
The data rate per data source listed in Table 10-3 is the rate each crate on the input rings
must handle. For some cases, such as the drift chamber, the load per event-data port card is
quite modest. On these rings, the eciency with which the FDDI protocol will multiplex the
packets is of primary importance. On other rings, such as the trigger and vertex detector,
the load per event-data port card is a signi cant fraction of the ber bandwidth. Again,
published reports indicate that such a throughput is attainable, but much depends on the
event-data port card/processor combination; the BABAR readout controllers need to do more
than simply feed the FDDI event-data port cards. Measurements on prototype con gurations
and modeling of more complete con gurations will be done to decide if this is a serious
problem.
Because the maximum throughput of the above con guration is determined by the input
bers, there is a simple chain of consequences. First, the input bers must be used as
eciently as possible. Second, to use the ber eciently, each event-data port card should
be ready at any time to send useful data. Third, for data to have a useful data density, the
ratio of data to overhead should be high.
The second consequence suggests maximizing concurrence in the data gathering process both
within events and across events. In particular, one would like each crate to send its portion
of a given event to the destination farm node without waiting for other crates. In the FDDI
protocol, if two nodes try to send to the same destination, one of them will be held o
until the other nishes. Within a given input ring, this is not necessarily bad. If all crates
tried to send at once, they would indeed serialize, but the ber would be fully occupied.
Unfortunately, however, the entire drift chamber ber, for example, would have to be idle
while the vertex detector crates on another ring are transmitting data. The solution is to
have one ring's crates transmit a di erent event to a di erent destination while another
Technical Design Report for the BABAR Detector
460
Electronics
ring's crates are active. Hence, multiple events will be gathered simultaneously. Other data
transmission technologies may not have this problem.
Because of contention for the output ber bandwidth, some form of trac shaping or control
of fragment assembly order will improve event gathering throughput. An example of such
an algorithm has been presented in Reference [Dou94]. In this scheme, the data acquisition
crates are logically arranged in a ring. A farm node volunteers to handle an upcoming event
by sending a token to the rst crate in the ring. When the speci ed event arrives, the rst
crate sends its data to the farm node and forwards the token to the next crate in the ring.
This crate sends its data and passes on the token. When the token leaves the last crate, the
event is completely assembled. At no time is more than one crate talking to the destination
node. Since each available farm node has registered a node with the rst crate, that crate
can begin delivering data for the next event before the rst token has fully traversed the
ring. Since the destinations of data for di erent tokens are di erent nodes, there is no
contention for output bandwidth. Many variations on this idea are possible, and some may
have better performance. For example, in the version outlined here, tokens do not overtake
each other. This means that several events can be stalled waiting for the delivery of data
from a particularly full crate. We intend to do simulation studies to evaluate the impact of
such e ects and ways around them.
Step three in the consequence chain is relevant since some crates, such as the drift chamber
crates, have a low data rate. There is a risk that the useful data on the ber will be swamped
with headers and protocol messages. Fortunately, it is a simple matter to cluster multiple
events together into bundles which are delivered as a unit into a single destination farm node.
This amortizes message overhead over many events. The cost is more complex software in
the farm nodes, which must deal with noncontiguous event fragments. Since the events are
in random-access memory, a properly designed algorithm should minimize the performance
loss. Any high rejection lters that run in Level 3 will be designed to run before the events
are unscrambled. The cost of memory is such that bundles of 64 or even more events are
practical.
10.10.7
Event Distribution
Once events have been fully ltered and reconstructed in the Level 3 farm nodes, they
must be collected and stored on permanent media. The BABAR baseline design allows for
both multiple partitions running simultaneously and multiple data streams within a single
partition. This leads to several issues regarding event distribution.
Because crates are not partitionable, event distribution in a running partition involves little
more than being sure that the participant at each stage|readout controller, Level 2, Level 3,
and event consumer|knows where to forward its events. A resource-manager database
Technical Design Report for the BABAR Detector
10.10 Data Acquisition
461
will be used to control con guration of partitions to prevent stages from being assigned to
more than one partition. The resource-manager database will also be a source of currentcon guration information for the control-source and distribution modules to ensure sensible
con gurations of strobes. This database can be queried by the event-recorder consumer at
the beginning of a run to record the detailed con guration of the partition for use in later
analysis.
Level 3 is made up of identical autonomous processors, which could easily be distributed
across partitions. While BABAR will retain this ability, it is not likely to be used often; the
full Level 3 farm will generally be assigned to the main partition. Partitions running in local
mode will typically contain a few crates directly feeding a conventional workstation running
an event consumer for monitoring and analysis. Such a workstation will run a single-node
version of the Level 3 software to gather data from the partition's crates and supply events
to the event consumer as described below. These event consumers could be located at the
institutions responsible for the systems.
Partitions will be de ned and con gured through an interactive interface to the resourcemanager database, with some common groupings of partitionable components such as the
whole of Level 3 being prede ned. Running partitions may not be recon gured. Resolution
of contention for resources, such as Level 3 nodes, will be achieved by discussion among the
people on shift.
Farm Node Allocation
Closely related to the problem of assigning bers on the output side of the switch is the
problem of assigning Level 3 processors to events. In CDF, for example, this decision is made
by a central bu er manager which keeps track, via message exchange, of the availability
of each farm node. The nodes are then queued for reassignment using a FIFO queue
discipline. The disadvantage of this scheme is the high message load placed on the single
bu er manager. An intrinsically distributed mechanism is needed to control the message
load on any particular node.
The algorithm of Reference [Dou94], described above, may apply. This node allocation
algorithm is presented as a near-round-robin queuing discipline for the farm nodes under the
assumption that most of the time most of the nodes claim an event as the token passes it.
Unfortunately, this will not be the case in general. First, note that the token must continue
to circulate at all times. Under a lightly loaded system, most nodes will have already claimed
an event which has not yet occurred and will therefore have no resources available; hence
they will all pass the token. Under a heavy load, most nodes will be busy with their events
and will again pass the token. In both cases, there is no speed control on the rotation of
the token, and the token-related messages will increase rapidly in number. By recasting
Technical Design Report for the BABAR Detector
462
Electronics
this allocation problem in slightly di erent terms, one can choose from a rich literature of
distributed algorithms [Lam74, Ric81, Dol82, Pet82].
10.10.8 Data Integrity
The data acquisition scheme is quite complex, with data fragments being bundled and sent
through a network, and the bundles of fragments assembled into events. To be sure that
events are being assembled correctly, each fragment will need an event label and content
descriptor. As the events are assembled, these will need to be checked to ensure that all
pieces of the event are completely and correctly included. A global process will need to track
all triggers and verify that they are either rejected for cause or archived.
10.10.9 Research and Development
System Modeling
Simulations are being used to re ne our knowledge of the data acquisition requirements. The
trigger data- ows have been studied using our most detailed simulations of the expected
backgrounds. The bu ering required in the vertex detector [Lev94b] and drift chamber
front-ends has also been examined using these background simulations.
The task for the immediate future is to simulate the data acquisition system as a whole to
investigate signaling and protocol overheads, and to ensure that the bandwidth and latency
requirements are understood. We intend to do this using commercially available simulation
packages and to build up a family of detailed simulations of the various parts of the BABAR
detector and its response to the expected backgrounds. The detailed simulations of the
trigger and individual systems will also continue to be improved as more detail about their
functions becomes available.
Alternative Network Technologies
Asynchronous Transfer Mode (ATM) networks will allow for data signaling rates per ber of
155 Mbits/s (120 Mbits/s usable) in the near future and have the potential for 620 Mbits/s
(480 Mbits/s usable). There is very widespread industrial support for ATM, which should
ensure the availability of components at reasonable cost. There are some possible problems
with use of ATM for data acquisition. The Fiber Channel runs at data signalling rates
from 132 Mbits/s up to 1 Gbits/s. It has fairly widespread support in industry, especially in
connection with high speed peripherals. At present, there are only a few manufacturers of
Technical Design Report for the BABAR Detector
10.11 Level 2 Trigger
463
Fiber Channel switches, and cost per node is relatively high. Scalable Coherent Interface
(SCI) networks operate at a data signaling rate of up to 1 Gbits/s. Industry support for SCI
is modest.
10.11
Level 2 Trigger
The baseline design omits a Level 2 trigger stage and builds all events accepted by Level 1.
Since having a Level 2 may prove to be desirable and even necessary, we describe this option
here. The Level 2 process for an event begins with the Level 1 accept. It receives input
data from the Level 1 trigger and some of the readout modules. A positive decision results
in an event being built. Level 2 provides an opportunity for intermediate event lters which
reduce the number of events built, and thus diminishes the burden on the readout module
processors, the event assembly network, and the Level 3 processors. If all envisioned cuts
are made, the rate of event building could drop from the 2.0 kHz of the Level 1 accept to
200 Hz.
10.11.1
Filter
The primary candidate for a Level 2 lter is a cut on the longitudinal component of the
distance of closest approach of tracks to the interaction point (a z cut), for which the primary
source of information for determining this quantity is the vertex detector. The volume of
data needed for a vertex detector based z cut makes gathering the required information
impractical before the Level 1 strobe.
A Level 2 trigger might also implement the cut on track transverse momenta (a p cut), thus
simplifying Level 1. At a minimum, the p cut requires the Level 1 trigger information from
the drift chamber. It could be accomplished in hardware or software, although for some of
the proposed drift chamber geometries, a hardware implementation of the p cut may be
complex and require more latency than the 10 s allowed for the Level 1 accept. A software
implementation could bene t from additional information from the drift chamber.
Simulations of background processes indicate that appropriate cuts in z and p can reduce
the trigger rate to near the logging rate with no substantial reduction in eciencies, even
for low-multiplicity tau decays [Lev94d].
t
t
t
t
Technical Design Report for the BABAR Detector
464
Electronics
10.11.2 Implementation
Logical
The Level 2 design should have minimum impact on the design of the rest of the system
and should require no substantial hardware developments. As an example of a possible
implementation of Level 2, a network with a low-overhead fast protocol can be used to send
one message per Level 1 accept to the crates serving the vertex detector, the drift chamber,
and the Level 1 trigger. These crates return information to a Level 2 processor, which makes
a decision before feature extraction takes place.
Using multiple processors for Level 2 decouples throughput from latency. The Level 1 accept
rate (between 1 and 10 kHz), combined with a reasonable limit on the number of processors
devoted to the Level 2 decision making (such as 10), implies an average latency of no more
than 1{10 ms. This is the same order of magnitude as the length of bu ers implemented
with VRAM. If the Level 2 decision can be made in time to copy the relevant data from the
VRAM, movement of the data upon a Level 1 accept can be avoided, thus reducing the load
on the readout module processors.
Once a Level 2 accept has been issued, event building proceeds as described in Section 10.10,
but at a much reduced rate. Events could be bundled if desired. The leading engineering
constraints on implementations of Level 2 include:
the rate of Level 1 triggers;
the network latencies for message passing;
the network bandwidths for data transport;
the processing overhead incurred by the network protocol;
the number of processors sharing the loads of ltering and event building;
the execution times of their tasks; and
the worst-case latencies for response to asynchronous messages (typically much di erent
for workstations and embedded processors).
Hardware
An example, based on two recent products, illustrates how a Level 2 trigger could be realized.
Commercial re ective memory modules with interrupt capability have appeared in the PCI
Technical Design Report for the BABAR Detector
10.11 Level 2 Trigger
465
Mezzanine Card (PMC) format. Re ective memories modules are connected together by
a dedicated network. A message-passing mechanism, transparent to the user, keeps the
contents of appropriately mapped locations in all modules the same. From the processors'
point of view (PMC), they are accessing the same memory. VME PowerPC processors with
PCI mezzanine slots are now on the market. This example uses several of these processors,
each with a PMC re ective memory node. These appear as the readout controllers in the
20 drift chamber crates, the four vertex detector crates, and the Level 1 trigger crate, as
ten Level 2 processors, and as the global controller. All re ective memory nodes form one
logical network.
In response to the Level 1 accept signal, the global controller moves the Level 1 trigger
primitives and time stamp into the next available bu er, then sends a message over the
re ective memory network to the next available Level 2 processor, pointing to this bu er.
The Level 2 processor then sends a request message to all the readout controllers on the
re ective memory network for the Level 2 trigger symbols for the time interval speci ed by
the stamp. Each readout controller then manages this request in its crate and responds, also
over re ective memory. When all readout controllers have responded, the Level 2 processor
completes the p and z cuts on Level 1 events.
Level 2 noti es the global controller of its decision and makes itself available for another
event. For accepted events, the global controller noti es all readout modules, via their
readout controllers, to save all data from the event, and then manages the remainder of the
event processing in the same manner as in the baseline design.
t
10.11.3
Conclusion
A data acquisition system with a Level 2 lter would have greater architectural and data ow
complexity than the baseline option. This may be o set by having a simpler Level 1 trigger
and a reduced load on the event-assembly network and the processors. Furthermore, it may
be required by the running conditions. We will continue to investigate its design, including
prototyping the communication protocols and simulating the overall system, and we will
keep its requirements in mind as we construct the event-assembly network. Simultaneously,
we will track developments in networking technology, identify promising options, and select
the most cost-e ective low-risk implementation.
Technical Design Report for the BABAR Detector
466
Electronics
10.12
Global Support Electronics
10.12.1
Detector Monitoring and Control
Purpose
A primary tool for assuring the quality of data collected by BABAR is the careful control and
monitoring of the detector hardware. This includes such traditional items as setting and
monitoring power supplies and monitoring environmental factors such as temperature and
humidity. It also includes monitoring to ensure the safety of both personnel and equipment,
and communicating with the run control system to prevent it from taking actions which are
not allowed by the current hardware status.
Design Strategy
The detector control includes areas of responsibility belonging both to the individual detector
systems and to the central data acquisition group. The central group will provide a standard
framework consisting of a set of displays, a set of procedures for setting and reading values
from the hardware devices, and a messaging system for communication. The detector systems
have the responsibility to identify their needs early in the design cycle and to provide the
necessary interfaces to the standard hardware and software. The design process will use an
object-oriented approach to abstract those parts of the implementation that are common
to many systems, and from this, the central data acquisition group will provide a common
framework. An important point is that the system designers think about these functions
from the start so that the elements of the common framework can be identi ed as early as
possible.
Among the quantities that will be routinely monitored for all detector systems are:
all power supply voltages at the supply and as close to the point of use as practical;
power supply currents where appropriate;
power supply and crate temperatures;
temperatures at sensitive points in the detector and in the electronics house;
all gas pressures; and
many additional items particular to individual detector systems.
Technical Design Report for the BABAR Detector
10.12 Global Support Electronics
467
The distribution of which functions are provided by software and which are provided by
hardware will depend largely on the time-critical nature of that function. This will emerge
as the design progresses.
Implementation
The detector control system will be implemented using EPICS/VxWorks. This modern,
modular approach frees the designer to choose the hardware technology which is most
appropriate to the problem and most cost e ective. For example, devices which communicate
via GPIB, VME, RS232, or other means can all be accommodated easily by using the
appropriate software driver. The e ects of changing a hardware module are thus highly
localized.
To eliminate interference with data gathering, the communication path for detector control
will be kept separate from that used to collect events. An Ethernet-based system using
the TCP/IP protocol will be used for the backbone. Processors on this backbone in turn
communicate with hardware modules via the appropriate bus, e.g., GPIB, VME, or RS232.
These backbone processors will run VxWorks and are seen by EPICS as I/O Controllers
(IOCs).
A number of software tools will be needed. These are discussed more fully in Chapter 11.
Some of the tools needed for this task are:
A set of standard displays such as strip chart plots, histograms, and status displays.
These allow the operator to keep track of the hardware and are provided by the Display
Manager in EPICS.
A messaging system for communicating with the operator and the run control system.
Messages will be classi ed according to hazard level as alarms which need immediate
action, warnings, or information. Communication with the run control system will
allow it to take appropriate automatic action for alarms and warnings and to allow it
to check the status of appropriate hardware before allowing a change of the state of
the experiment; e.g., the high voltages must be on in order for data to be taken. The
run control system is discussed more fully in Chapter 11.
An interface to the online database, which is needed in order to store target values,
tolerances, and measured values.
Technical Design Report for the BABAR Detector
468
10.12.2
Electronics
Data Monitoring Support
Monitoring is the continuous watching of the incoming data and the hardware to spot
problems as soon as possible. It is a front-line tool to ensure data quality. Examples
include synchronization checks, comparing histograms of raw data to expected distributions,
and monitoring event sizes. Monitoring of the detector control hardware is discussed in
Section 11.
Monitoring of the physics event stream is primarily a software task and is discussed in
Chapter 11. An electronics implication is that data acquisition hardware may need to supply
a bu er in which sampled events are stored in such a way that they do not interfere with the
primary data gathering path. Individual monitoring processes could then access this data
for analysis.
10.12.3
Diagnostic Support
Purpose
Diagnostic actions are taken to isolate a problem once it has been detected by monitoring or
to perform routine checks to verify that the hardware is functioning properly. They involve
operations that are not normally part of data taking. Examples include executing scope
loops and measuring response to known stimuli such as test pulses.
Design
The choice of diagnostic operations for a particular detector system is best left to the
designers of the system, but it is important that they address the issue early in the design
process. Global support from the electronics and online groups will be in the form of a
framework upon which the system designers can build. This framework will come from a
global, object-oriented design of the needs of the systems and will consider hardware and
software together. The design methodology will be iterative to allow information learned in
late stages to be fed back into earlier stages. In many cases, the choice of using hardware or
software for a given function is primarily a matter of the response time needed. By deferring
this decision, we can optimize before making costly decisions.
Technical Design Report for the BABAR Detector
10.12 Global Support Electronics
469
Implementation
Diagnostic runs can take place in either global or local mode. In global mode, there is
central management for run control, trigger, data gathering, and timing. There can be only
one such run in progress at a time. Global mode is used when the nature of the problem is
such that two or more systems must run together. An example of a global diagnostic run
is a pattern load of data into the system bu ers to verify that the data are being read out
correctly. In local mode, the system is in a partition that has its own local run control, local
readout strobe, local data gathering, and local timing signals. This will be implemented
by providing a local timing controller and a local readout controller for each partition. In
global running, these units act under the control of a central partition. In local running,
the timing controller can either pass on signals from the central partition, e.g., clocks and
synchronization pulses, or use locally generated ones. Similarly, the local readout controller
can either pass the data on to the central partition or handle it locally. Many partitions can
be running independently and concurrently, and within a given partition, it may be possible
to have multiple concurrent activities. The local partitions must have access to the central
database.
10.12.4
Calibration Support
Purpose
Calibration involves determining the parameters needed to turn digitized data like ADC
and TDC counts into physical data such as energies and coordinates. This involves applying
known stimuli to the analog portion of the data acquisition chain and observing the response
through the digitization process.
Design and Implementation
In terms of data acquisition properties, calibration runs are very similar to diagnostic runs.
As such, the general framework for detector diagnostics described above applies. The choices
of how to calibrate and how often to do it are primarily up to each detector system and are
discussed in their respective sections of this chapter.
Some general principles of calibration runs are: they should use the same analog and
digitization electronics as the physics data; they should strive to measure one number instead
of piecing together multiple measurements; and the absolute calibration may in many cases
have to come from the physics data, e.g., getting the absolute energy scale for the calorimeter
from the Bhabha events.
Technical Design Report for the BABAR Detector
470
Electronics
Calibration runs can be in either global or local mode. An example of a global calibration
run is using cosmic rays to do tracking alignments and to measure the dE/dx values from
muons. An example of a local calibration run would be using calibrated electronics to inject
known signals into the detector data acquisition system.
A special case of calibration involves automatic steps that need to be taken at speci c times;
e.g., pedestal measurements which are taken at the start of each run. These processes will
be fully automated.
Technical Design Report for the BABAR Detector
REFERENCES
471
References
[Amb94] M. Ambrosio et al., \Performances of RPC Detectors with Tracking & Timing
Electronics,"
A344, 350 (1994).
[Alo94a] A. Aloisio et al., \High Density Zero Suppressor Card for the Muon Trigger with
RPC for the L3 Experiment," IEEE{NS 41, 225 (1994).
[Alo94b] A. Aloisio et al., \The Suppressor Card for the Muon Trigger with RPC for the
L3 Experiment," INFN{TC{94{11 (1994).
[Alo94c] A. Aloisio et al., \The RPC Trigger System for the L3 Forward-Backward Muon
Detector," in Proceedings of the CHEP94 Conference, San Francisco, California,
LBL{35822 (1994).
[Alo94d] A. Aloisio et al., \The L3 Forward-Backward Muon RPC Trigger System,"
submitted to
(1994).
[Alo94e] A. Aloisio et al., \Performances of the RPC Trigger System of the L3 ForwardBackward Muon Spectrometer," submitted to
B (1994).
[Bab94]
, SLAC{443 (1994).
[Bag94] R. Baggs et al.,
A344, 547{557 (1994).
[Bau92] D. Bauer, \The Impact of Two-photon Physics on a B Factory Detector,"
A AR
# 77 (1994).
[Bol91] T. Bolton et al., \Report of the Trigger and Data Acquisition Group," in
Proceedings of the Workshop on Physics and Detector Issues for a High-luminosity
Asymmetric B Factory at SLAC, SLAC{373, 597 (1991).
[Boz93] W. Bozzoli et al., \Data Transfer and Distribution at 70 Mbytes/s,"
CERN/ECP 93{7, (1993).
[Buo]
S. Buono et al., \Studies on RIO/HiPPI Based Event Building," RD13 Note N.70
(1993).
[CLE91] C. Bebek et al. (CLEO Collaboration), \CLEO-II Trigger System,"
A302, 261{276 (1991).
[Cul91] D. Cullen-Vidal et al., \D0 Level-2/Data Acquisition; the New Generation," in
Proceedings of the 1991 International Conference on Computing in High-Energy
Physics, 659 (1992).
Nucl. Instr. Methods
Nucl. Instr. Methods
Nucl. Phys.
Letter of Intent
Nucl. Instr. Methods
B B
Note
Nucl. Instr.
Methods
Technical Design Report for the
A AR
B B
Detector
472
[Dol82]
[Dou94]
[Eis90]
[Ell92]
[Gen93]
[Hal94]
REFERENCES
D. Dolev, M. Klawe, and M. Rodeh, \An O(n log n) Unidirectional Distributed
Algorithm for Extrema Finding in a Circle," J. Algorithms 3, 245{260 (1982).
D. Doughty Jr., D. Game, L. Mitchell, G. Heyes, and W.A. Watson III, \Event
Building Using an ATM Switching Network in the CLAS Detector at CEBAF,"
Presented at the International Data Acquisition Conference on Event Building
and Event Data Readout, Batavia, Illinois (1994).
A. Eisner, \Bhabhas at an Asymmetric B Factory," A AR
# 32 (1990).
N. Ellis et al., \A Calorimeter-Based Level-One Electromagnetic Cluster Trigger
for LHC," in Proceedings of the 1992 International Conference on Computing
in High-Energy Physics, edited by C. Verkerk and W. Wojcik (Annecy, France,
1992), 210, (1992).
P. Bailly and J.-F. Genat, \A 100 Picosecond Resolution, 6 Microsecond Full
Scale Multihit Time Encoder in CMOS Technology," in Proceedings of the Third
Int. Conf. on Electronics for Future Colliders, Chestnut Ridge, p. 57 (1993).
See also: J.-F. Genat, \High Resolution Digital Time to Digital Converters," in
Proceedings of the First Int. Conf. on Electronics for Future Colliders, Chestnut
Ridge (1991).
G. Haller, D. Freytag, and J. Hoe ich, \Proposal for an Electronics System for
the A AR CsI Calorimeter," A AR
# 184 (1994).
K. Kinoshita, \A Fast Hardware Trigger for Charged Particles in Multilayer
Tracking Devices,"
A276, 242 (1989).
F. Kral, \GEANT Simulation Results for a Fast Calorimeter Trigger,"
A AR
# 133 (1994).
F. Kral and D. Worledge, \Initial Design and GEANT Simulation of a Fast Drift
Chamber Trigger," A AR
# 162 (1994).
F. Kral, \Simulation of Trigger Rates due to Beam-Gas Backgrounds,"
A AR
# 165 (1994).
F. Kral, \Trigger Rates due to Cosmic Ray Muons," A AR
# 164 (1994).
L. Lamport, \A New Solution of Dijkstra's Concurrent Programming Problem,"
Comm. ACM 17(8), 453 (1974).
L. Lamport, \Time, Clocks, and the Ordering of Events in a Distributed System,"
Comm. ACM 21(7), 558 (1978).
B B
B B
[Kin89]
B B
Note
Note
Nucl. Instr. Methods
[Kra94a]
B B
[Kra94b]
Note
B B
[Kra94c]
B B
[Kra94d]
[Lam74]
[Lam78]
Note
Note
Technical Design Report for the
B B
A AR
B B
Detector
Note
REFERENCES
473
[Lev94a] M. Levi and F. Kral, \Simulated Background Rates in Calorimeter Trigger
Towers," BABAR Note # 134 (1994).
[Lev94b] M. Levi, \Impact of Backgrounds on Silicon Vertex Detector Architecture and
Detector Trigger," BABAR Note # 136 (1994).
[Lev94c] M. Levi, \Obituary for the Prompt Silicon Trigger," Netnews Item
slac.b-factory.vertex #54, (1994).
[Lev94d] M. Levi, \Ecacy of a Level 2 Trigger Algorithm," Netnews Item
slac.b-factory.vertex #55, (1994).
[Lev94e] M. Levi, \Impact of the Electroproduction of Hadrons on the Trigger,"
BABAR Note # 191 (1994).
[Pat94] J. Patrick et al., \The CDF Ultranet-Based Data Acquisition System," in
Proceedings of the 1994 Int. Conference on Computing in High-Energy Physics,
in preparation.
[Pet82] G.L. Peterson, \An O(n log n) Unidirectional Algorithm for the Circular Extrema
Problem," ACM Trans. on Programming, Languages and Systems 4(4), 758
(1982).
[Req94] \BABAR Electronics Requirements," in preparation.
[Ric81] G. Richart, and A.K. Agrawala, \An Optimal Algorithm for Mutual Exclusion in
Computer Networks," Comm. ACM 24(1), 9, (1981).
[SLA93] \Status Report on the Design of a Detector for the Study of CP Violation at
PEP-II at SLAC," SLAC{419 (1993).
[Sny92] A. Snyder, \Backgrounds in the BABAR Drift Chamber," BABAR Note # 88 (1992).
[Sny94] A. Snyder, \Electro-Production Triggers in BABAR," BABAR Note # 180 (1994).
[Wun94] S. Wunduke, M. Levi, and F. Kral, \Architecture for Readout Electronics,"
BABAR Note # 138 (1994).
Technical Design Report for the BABAR Detector
474
Technical Design Report for the BABAR Detector
REFERENCES
11
Computing
T
here are a number of computing challenges for the proposed BABAR detector. The
data rates are high for an e+e, collider, and there is a large quantity of useful data
which must be stored and analyzed. Convenient interactive access must be available for the
ecient extraction of physics results. In addition, the wide geographical distribution of the
collaboration implies signi cant issues for remote access to the data and code, collaboration
communication, and for remote computing.
The BABAR Computing System, as de ned here, includes several aspects: (1) the online
data acquisition, control, and monitoring hardware and software; (2) the oine computing
hardware|CPUs (\farm" and desktop), storage, X servers, and networking|whether local
to SLAC or at remote locations; (3) the software environment in which the computing work
is done; and (4) the code itself used by the collaboration. All of these are important items
for BABAR computing, although parts of these, including most of (2), are properly considered
institutional infrastructure.
The next section reviews the computing requirements, and Section 11.2 gives an overview of
the proposed computing system. Following these are more detailed discussions of the computing model, software environment, online system, reconstruction and analysis framework,
computing support plan, integration issues, system responsibilities and management, and
cost and schedule.
11.1 Requirements
In this section, the basic computing requirements for BABAR are summarized. These are
divided into operational requirements and technical requirements. Operational requirements
are broad capabilities that must be present in order for the system to respond adequately
to the environment within which it must operate. Technical requirements are speci c
capabilities that the system must exhibit in order to meet the operational requirements,
for example, the CPU power and bandwidth. Estimates are given for the early years of
the experiment. It is recognized that the requirements may increase as the experiment
progresses; hence, a general further requirement exists that the system be scalable to meet
future needs.
476
Computing
11.1.1 Operational Requirements
The BABAR Computing System must support several activities which compete for available
resources. These include program development by a distributed developer community, data
acquisition, control and monitoring, bulk data reduction and Monte Carlo simulations, and
interactive physics analysis.
As the collaboration is worldwide, a working scheme which in no way reduces the ability of
the remote collaborators to contribute to the experiment must be implemented. The remote
computing issues cover the way in which the collaborators exchange information (at the
moment WWW, ftp, netnews, and mailing lists), how software development is integrated,
and how to implement updates of software production releases, data analysis, Monte Carlo
production, database access, and software licensing. Access to the data will be provided at
regional centers, which will keep some terabytes of data.
The computing system must provide a real-time response adequate to perform its many roles.
This impacts several areas. The most obvious is that of data acquisition. To bene t fully
from the available luminosity at PEP-II, it is important to reduce as much as possible the
e ective deadtime for the online system. The time for routine operations that can interfere
with data taking, such as run start and close-down or system reboot, should be small. Finally,
the latency between the acquisition of the data and its analysis for physics results should be
short compared to the duration of an accelerator scheduling cycle.
The computing system should display a uniform face to the users that hides any intrinsic
heterogeneity and allows them to be productive at their home institutions, at regional centers,
or at SLAC. It must be robust, both in the sense of exhibiting a high degree of reliability
and high availability, and also in allowing short recovery time from a failure. This in turn
implies a high degree of maintainability. Given the extended time over which the experiment
will be active and the probability of unforeseen demands and changes in the requirements,
the computing system must also be highly exible and extensible, this being achieved with
a minimum of e ort and in time intervals dictated by the accelerator scheduling cycle.
Finally, the computer system must balance ease of use with security so that developers can
be productive, and also ensure that high quality, well-understood, and reproducible physics
results can be achieved.
Technical Design Report for the BABAR Detector
11.1 Requirements
477
Detector
(raw data)
TRIG
DAQ
Offline
Reconstruction
(bulk
processing)
Triggers
TRIG
Online
Reconstruction
event
DAQ
event
playback
playback
Mass
Storage
event
n-tuples
DST
Interactive
& Batch
Analyses
Figure 11-1.
DST
DST
Creation
Schematic of the data ow for the BABAR experiment.
11.1.2 Technical Requirements
CPU Requirements
There are several substantial computing tasks that require signi cant CPU resources: the
online system, including the Level 3 trigger and data acquisition, event sampling, detector
monitoring, calibration, and control; oine reconstruction; Monte Carlo simulation; creation
of DSTs from the reconstructed data, involving a reduction in event size and/or a reduction
in the number of events; analysis of large DST datasets; and interactive analysis tasks. These
tasks are represented in a data- ow diagram in Figure 11-1. The CPU power1 required to
handle this data ow is summarized in Table 11-1.
The familiar term MIPS is used to describe CPU processing power. In terms of the current industry
standard speci cations, one MIPS represents a power corresponding to approximately one SPECint92
(integer performance) and one SPECfp92 ( oating point performance).
1
Technical Design Report for the BABAR Detector
478
Computing
Online (Level 3)
Reconstruction
MIPS
MIPSs/event
events/year
passes/year
MIPS
Monte Carlo
MIPSs/event
events/year
MIPS
DST Creation
MIPS
DST Analysis
MIPS
Interactive Analysis MIPS
Table 11-1.
3000
25
109
2
5000
500
108
5000
500
2000
5000
Estimated CPU power requirements.
The online CPU requirement is estimated assuming a 2000 events/s input rate to the Level 3
farm. This rate is higher than the average anticipated rate, but is used to ensure that the
system is able to keep up with the data ow under extreme conditions. This is quickly
reduced to the order of 200 events/s using t and information from the trigger. More
complete event information is then used to reduce the rate to tape to 100 events/s. A full
reconstruction will be done on approximately 30 events/s (of which 10 events/s are hadronic
via single photon) at L = 3 1033 cm,2s,1 .
Based in part on the CLEO-II experience [CLE94], similar gures are derived for the CPU
requirements relevant to event reconstruction and Monte Carlo simulation (5000 MIPS
each). The CPU requirement for Monte Carlo simulation, as for the data acquisition and
reconstruction, is based on a peak rate requirement, namely the ability to keep up with the
peak hadronic event data rate. This corresponds to a rate of 108 Monte Carlo events per 107
seconds.
DST creation refers to a major pass through the data to create a selected dataset (in which
the output event size may or may not be reduced. The generic term DST is used here to
include di ering de nitions of DST, mini-DST, etc.). The CPU power required is computed
according to:
p
z
MIPS = 0 25(MIPSs/event) 109(events) 2(users) 106(s),
:
=
which assumes that it is acceptable to have this task take less than a month, hence the
106 second time period; the two users allow for up to two such major passes through the
data to be simultaneous. This estimate, of course, is a particular, and rather conventional,
model for how an analysis progresses into selected subsets of the data. Alternatives will be
investigated, such as using a direct access database technology [Bad90].
Technical Design Report for the BABAR Detector
11.1 Requirements
479
DST analysis is a substantial processing of a selected DST dataset for physics analysis. The
formula used to estimate the CPU power in that case is
MIPS = 1(MIPSs/event) 108(events) 20(users)=106(s).
This assumes 20 such analyses occur once per year, each requiring one month of CPU time
to process the full dataset of interest. This estimate is in addition to the resources required
to design and debug such analyses.
Interactive analysis involves debugging, exploratory analysis, graphics, and program development. One hundred active users, each employing a 50 MIPS workstation, are anticipated.
In summary, if the numbers above are summed, the total CPU requirement is 20,000 MIPS.
At least 10,000 MIPS have to be located at SLAC for data acquisition, reconstruction, and
other tasks. However, the total installed MIPS may be lower depending on the model (Section 11.2.2); the requirements are based on keeping up with the data acquisition, e ectively
realtime, at the peak luminosity rate, rather than at the rate averaged over a year.
Throughput
Transformation of the raw data into the nal reconstructed event information, and subsequent physics analysis, requires several data transfer steps (Figure 11-1). The most signi cant
steps (in terms of required bandwidth) include:
Transfer of data into the Level 3 online farm;
Transfer of data from the Level 3 farm to mass storage (tape);
Reading and writing the data in the bulk reconstruction;
Writing of Monte Carlo events to tape;
Reading through the data to create DSTs;
Passes through the DSTs for physics analysis; and
Interactive analysis on small datasets.
The estimated requirements are summarized in Table 11-2. The upper-limit requirement of
2000 events/s is assumed going into the Level 3 data acquisition farm, and the event size is
estimated to be 25 kbytes [Por93], yielding a 50 Mbyte/s rate. The event size is assumed to
double to 50 kbytes once the reconstruction information is added, but the rate decreases to
100 events/s. Hence, the reconstruction data rate is 7.5 Mbytes/s, 2.5 Mbytes/s input plus
5 Mbytes/s output. The Monte Carlo event size is also estimated at 50 kbytes, which, at the
simulation rate of 10 events/s, implies a data rate of 0.5 Mbyte/s.
Technical Design Report for the BABAR Detector
480
Computing
Task
Input
Output Bandwidth
(events/s) (events/s) ( Mbytes/s)
DAQ to Level 3
2000
50
Level 3 to Tape
100
2.5
Reconstruction
100
100
7.5
Monte Carlo
10
10
0.5
DST Creation
50(1 + )
DST Analysis
5/task
Interactive Analysis
0 2/user
Wide Area
80 Sessions
05
Data Transfer
0.5
f
:
>
Table 11-2.
:
Estimated bandwidth requirements. The f factor is explained in the text.
DST creation and analysis represent a large network load. Substantial DST creation may
occur simultaneously with the reconstruction via multiple output streams. However, additional passes through the data should be anticipated here. In DST creation, is the product
of the sample fraction times the average sampled event size divided by 50 kbytes. If is
small, and two such tasks are running, 100 Mbytes/s is required. The task of DST creation
will typically be heavily I/O bound. DST analysis, with 20 analysis tasks simultaneously
running, requires 100 Mbytes/s.
The interactive analysis number is estimated assuming a 2 kbytes/event reduced DST, and a
throughput of 1000 events in ten seconds, appropriate for small datasets which are assumed
to be analyzed at a desktop workstation. There may be an occasional need for interactive
analyses on larger DSTs, with larger bandwidth requirements. It is anticipated that such
analyses would take place on the same machines as the DST analysis tasks, supported with
higher bandwidth.
Based on this analysis, aggregate network capacity for oine tasks in excess of 200 Mbytes/s
is required, dominated by the reconstruction and the multiple reads through the data required
for DST creation and analysis. However, the DST creation task, which is a potentially large
burden speci cally on tape access, may be greatly mitigated by having this process proceed
largely via multiple output streams during reconstruction.
The wide-area network will be used for code development, interactive analysis, user communications, and transfer of small datasets (less than 10 Gbytes). A rule of thumb is that an
X Window session requires 50 kbits/s. Thus, supporting an estimated 80 sessions nominally
requires 0.5 Mbyte/s, except that the scaling is sublinear with number of sessions over a
f
f
Technical Design Report for the BABAR Detector
11.1 Requirements
481
Tape Storage:
Tbytes/yr
Raw Data
25
Reconstruction Output
50
Monte Carlo
5
DSTs
5
Disk Storage:
DSTs
Databases
Table 11-3.
Tbytes/yr
2
0.03
Estimated storage requirements.
network. To transfer a small dataset in an acceptable time (a few hours) requires a similar
bandwidth.
Storage
Table 11-3 summarizes the estimated storage requirements for the BABAR experiment. Approximately 100 Tbytes/yr of tape storage is required. If necessary, the 25 Tbytes of raw data
could be moved oine after reconstruction. The 5 Tbytes/yr for DST storage is estimated
by assuming 10 DSTs produced per year, with 107 events in each (sampling fraction of 0.01)
and 50 kbytes/event.
Current experience at SLAC is that a (disk) staging space of 2% of the active tape storage
is required. Alternatively, 2 Tbytes may be sucient for 40% of the DSTs produced to be
on disk. However, it is probably appropriate to regard this space as staging space, with the
details of whether a particular job's data are already staged to disk transparent to the user.
The database space requirement is implementation dependent but should be small compared
with the event data requirement.
Disk space is also required for group code (and documentation, etc.) and for user space.
The group code space will be relatively small, probably in the tens of gigabytes. The user
space required will also be comparatively minor but cannot be neglected because of the large
number of users. From developing experience, it appears that at least 100 Mbytes/user of
disk space will be needed. If there are 300 users, this means at least 30 Gbytes.
Technical Design Report for the BABAR Detector
482
11.2
Computing
Overview
This section gives a summary of the basic design choices which have been made and an
overall description of the planned BABAR computing system.
11.2.1
Chosen Technologies
Choices for key computing technologies depend largely on capability and economics. Additional technology choices are then driven by those initial selections. Finally, some technologies are selected re ecting a judgment of how best to solve a particular problem. Given that
the computing industry has become highly competitive and that technologies will evolve
quickly on the timescale of the BABAR experiment, it is important to identify available and
exible solutions whenever possible. Thus, some of the initial selections may be considered
working decisions; the initial design is expected to evolve between this design and the nal
implementation and nal uses (a period of approximately 15 years).
The magnitude of the computing need overwhelms the traditional mainframe approach and
dictates the use of a distributed, multiprocessor environment. In the following discussion, it
is assumed that all computing, both online and oine, will share a common environment.
It is recognized that to accommodate the diversity within a widespread collaboration and
changes occurring within the computing marketplace, special care must be taken to plan
for multiple platform support. This is both a hardware and a software issue; thus, generic
tools which are platform independent have been and continue to be sought. At present, this
implies the use of workstations utilizing inexpensive (RISC) technology and operating system
software within the Unix family. While it is too early to decide upon a speci c networking
medium, the industry-standard Transmission Control Protocol/Internet Protocol (TCP/IP)
is expected to survive and will, therefore, play a major role both in local-area and wide-area
network communications. The use of an object-oriented (OO) approach will be encouraged
for all software development within the collaboration. The primary programming language
will be C++. Fortran 90 will be an acceptable alternative. Windowing technology will be the
X11 client-server system with the Motif library and window manager. Printing will support
both text and the PostScript page description language. The management of program code
releases will be, in part, based upon the Concurrent Versions System (CVS). World Wide
Web (WWW) will serve as the primary mechanism for information access. The online
system will be built around the Experimental Physics and Industrial Control System (EPICS)
(Section 11.5.5). This, in turn, relies upon the VxWorks real-time executive; although it is
Technical Design Report for the BABAR Detector
11.2 Overview
483
expected that this requirement will be relaxed within a few years as EPICS becomes POSIX2
compliant.
11.2.2
Overall Description
Computing Model
The BABAR computing model is based on systems of Unix processors. Detailed choices will
be deferred until need actually arises in order to take advantage of the rapid developments
in computing technology. With the technology of today, a satisfactory solution can be built
with a combination of Ethernet and FDDI local networking with nonblocking Ethernet-FDDI
hubs and FDDI switches, SCSI-2 disk farms, and helical-scan tape technology in existing
tape silos. A small number of remote computing centers, based on similar technology choices,
are anticipated, with wide-area communications achieved by a combination of Internet
connections and 3490 silo cartridge or other tape (e.g., 8 mm) transfers.
Software Environment
A consistent, robust, and easy-to-maintain software environment will be created by resorting to common facilities for writing, managing, and distributing code relevant to all the
computing tasks. Recent advances in software engineering, such as the object-oriented
methodology and programming languages, will be exploited. An approach has been chosen
which includes the rapid creation of guidelines and standards documents, code templates,
and sample programs, together with the establishment of a code development environment
that supports the needs of the software developers while also ensuring that the end users'
need for stability is also addressed.
Online System
The online system has not only to support the steady state operation of the detector for
physics data acquisition, but also has to deal with the con icting demands of detector
commissioning, calibration, and diagnostics. A system that accommodates these demands
is described in terms of multiple viewpoints, from the most abstract which deals with the
need to control multiple logical experiments or \partitions" during commissioning, to the
most concrete which deals with the detailed control protocols among modules within the
data acquisition system.
2
IEEE Portable Operating System Interface for Computing Environments.
Technical Design Report for the BABAR Detector
484
Computing
Reconstruction and Analysis Framework
A common framework for applications is being developed that will allow physicist-developed
code to be used for online triggering and monitoring, for bulk reconstruction and Monte Carlo
simulations, and for physics analysis. This framework will shield the application programs
from the di erent operating conditions in these areas and allow single applications to operate
in all areas, accepting event data from real-time event servers in the online area, or from
disk or tape les oine, and supporting output of analyzed data to a variety of output
destinations including the real-time event server and disk or tape les. This framework must
furthermore provide for both interactive and batch styles of operation and accommodate the
con icting needs of experts and novices. Both text-based and graphics-based user interfaces
are under development in order to accommodate these needs, addressing also the needs for
bulk job submission and bookkeeping.
Computing Support
Computing support refers to both internal (within the collaboration) and external (university
and laboratory support organizations) activities. Internally, this support will be coordinated
directly by the computing system group. Designated members (Code Coordinators) of the
collaboration will be assigned to look after the various software packages, to manage the data
production and distribution tasks, to maintain the di erent databases, and to provide system
management support for the online computers. The collaboration will see that appropriate
documentation is generated, training needs are identi ed and covered, and that speci c tools
and utilities for the experiment are written. These support roles are expected to consume a
signi cant amount of manpower within the collaboration.
External support will be provided by support sta at the collaborating institutions and
is expected to include a wide variety of services. The overall computing system must be
designed with the active participation and cooperation of these groups. Hardware elements
must be planned, purchased, installed, operated, and maintained. This includes networking,
mass storage, and CPUs. These heterogeneous and distributed computer systems must be
managed with respect to user accounts, disk and other resources, backups, system tuning, etc.
Wide-area networking operations will require continual monitoring with occasional attention.
Software must be licensed, installed, and maintained. The details of data storage and access
must be designed and implemented. A system for eciently utilizing distributed computing
cycles must be installed and maintained. Finally, the support sta must spend some fraction
of their time in system analysis, technology tracking, and strategic planning.
Technical Design Report for the BABAR Detector
11.3 Computing Model
485
11.3 Computing Model
A comprehensive computing model must satisfy the functional requirements outlined in
Section 11.1. In addition, such a model must address practical considerations, for example:
the \abilities" (usability, exibility, scalability, reliability, maintainability, and a ordability);
smooth integration of all hardware and software components; ability to evolve in response
to commercial developments; and compatibility of SLAC on-site computing with that of
o -site collaborators. In the sections that follow, an architecture which addresses these
requirements, a model to describe its overall operation, and a cost model based upon present
and near-future technologies are presented.
11.3.1 Architectural Model
The overall computing system architecture includes a number of basic hardware and software
components. Necessary hardware components include the elements of processing power (e.g.,
computers), networking, and mass storage. The main software component is the operating
system. A schematic model containing these components is given in Figure 11-2.
CPU Complex
The BABAR Collaboration has concluded that, currently, the most cost-e ective way to satisfy
the basic CPU requirements is through the use of multiple microprocessor-based computers.
Microprocessors utilizing a RISC architecture, as found in Unix workstations, represent the
optimum in today's market. A continued high rate of growth in computing power of this
kind is expected for at least the next ve years. Alternative computing architectures are not
excluded in this model, and the possibility of a transition sometime during the period of this
experiment has been considered.
A commitment to clusters of distributed workstations for an online system, reconstruction,
Monte Carlo simulation, and interactive data analysis, along with general software development and debugging, requires a sophisticated and mature, multitasking, networked operating
system. This requirement is met by using Unix, o ered by all major workstation vendors.
Such a solution also tends to capitalize on relatively high-volume, inexpensive, components
(e.g., video monitors, SCSI disks, and RAM), thus o ering a degree of vendor independence
both for hardware and software.
Commercial development of workstation-based computing is proceeding in several directions.
For example, workstation clusters currently tend to use single-processor implementations.
Technical Design Report for the BABAR Detector
486
Computing
Detector
Data Acquisition
Accelerator
Monitor/Slow Control
Detector
Slow Controls
2000 evt/s
50 Mbyte/s
OPERATOR CONSOLES
TRIGGER
CONTROL
RUN CONTROL
SWITCH
MONITORING
HUB
3000
MIPS
LEVEL 3 TRIGGER &
100 evt/s
2.5 Mbyte/s
online
ANALYSIS FARM
RECONSTRUCTION & MC
FARM
offline
100 Tbyte/year
MASS
STORAGE
DISK SERVERS
2 Tbytes
HUB
10000 MIPS
SWITCH
TRANSPORTABLE MEDIA
DST ANALYSIS
FARM
DESKTOP WORKSTATIONS
ROUTER
2500 MIPS
HUB
WIDE AREA
~5000 MIPS
NETWORK
AIR FREIGHT
REMOTE
FARM
Figure 11-2.
MASS
STORAGE
Conceptual computing architecture.
Technical Design Report for the BABAR Detector
11.3 Computing Model
487
There is a trend towards multiprocessor cluster implementations that improves processing
performance by virtue of a high-speed backplane interconnecting the individual CPUs.
Current examples of this architecture include the SGI Power Challenge, the IBM SP-2,
and the Convex Exemplar. Such multiprocessor implementations could be used to optimize
the design, but only if the costs become competitive.
The Unix operating system is also undergoing a series of changes. The direction of change
appears to be toward standardization of both the basic command set and system libraries.
By carefully adhering to such standards (e.g., POSIX), one can attempt to minimize the
work required to make a transition to another computer environment should that become
necessary. Given the rapid pace of developments within the computing industry, the e ort
to abide by such standards within the collaboration is deemed essential.
Networking
There are two levels at which networking is crucial to this experiment: local networking
at SLAC, because of the large number of machines and high data rates involved both
online and oine; and wide area networking, due to the large, widespread international
collaboration. Local area networking refers to the links between the various components of
the experiment, including those to desktop machines and the central computing complex.
Wide area networking refers to those links between the various collaborating institutions and
to SLAC. While the SLAC LAN is completely under lab control, the WAN is not. BABAR
may hope to in uence various funding agencies to accommodate WAN needs but cannot
expect to sponsor signi cant upgrades due to the high costs, as shall be discussed later.
A networking system consists of various components and subsystems, including: workstation
and other device interfaces; switches and hubs; various diagnostic devices; and the cable
plant. The basic LAN architecture consists of a switched network of point-to-point links.
In some cases, the I/O demands on a particular system may allow point-to-point links
to be replaced by a short series of daisy-chained devices, or for a relatively expensive
technology to be replaced by a cheaper (and lower performance) technology. The latter
concept is illustrated in Figure 11-2 in the distinction between the Reconstruction & MC
farm, characterized by a relatively high CPU:I/O ratio, and the DST Analysis farm, in
which condensed data is read at a high rate with relatively little processing per event. The
networking protocol will be TCP/IP, an industry standard available on all Unix machines.
Mass Storage
A large amount of reliable, inexpensive storage with fast, automated access from the Unix
environment is needed. Mass storage technologies, like those in other areas of computing,
Technical Design Report for the BABAR Detector
488
Computing
are rapidly developing within the industry. Current mass storage candidates are typically
expensive and unique to a particular vendor. While it seems unlikely that on BABAR's
timescale a de facto industry standard will emerge, several near-future products are expected
to address these needs satisfactorily. Unlike the computing and network environments,
however, the mass storage system is likely to be rather specialized.
The transfer of data between SLAC and remote collaborator sites may be complicated if
the supported media at the various institutions are incompatible. For transfers to such
institutions, an a ordable and transportable data medium will be essential. Currently,
8 mm Exabyte tape cartridges are supported at SLAC and are in widespread use, although
alternative formats (e.g., 4 mm and DLT) are gaining in popularity. In general, capacity and
performance are increasing while the cost per stored byte is decreasing. Hence, the decision
as to which technology to use will be postponed until it is needed.
Prototype Compute Farm
SLAC has been pursuing R&D towards this model in the form of a prototype compute farm,
shown in Figure 11-3, designed to exercise these ideas. Two batch environments are available
for testing: LoadLeveler from IBM and Load Sharing Facility from Platform Computing.
SLAC Computing Services (SCS) is working with CERN on the latest data staging software
being developed as part of CERN's SHIFT. It is anticipated that in the next several years,
there will be investigations into commercial products which support the IEEE Mass Storage
Model to provide data staging and hierarchical storage management.
11.3.2
Operational Model
Hierarchy
A hierarchy of sites exists in the computing for the BABAR experiment. The data will be
recorded at SLAC, giving SLAC a unique position. Demands of detector debugging and
monitoring, physics analysis by collaborators resident at SLAC, and simpli cations in data
handling will mean that much, if not all, of the bulk data processing will be performed at
SLAC. Next, there exists, within the collaboration several sites with the potential to provide
computing facilities (data storage, computing capacity, and support) similar to those at
SLAC. These regional centers will serve as analysis centers, providing convenient data access
to institutes with poor network connections to SLAC and to those which cannot provide the
levels of support required to maintain copies of the data locally. These centers could also
be used for data processing, DST production, etc., as performed at SLAC. Finally, there
are the computing facilities in all of the other collaborating institutions. This nal grouping
Technical Design Report for the BABAR Detector
11.3 Computing Model
489
Rtr Ring
3 Interactive Compute Servers
298SpecInt92
SLD
SLAC
LAN
ESA
60GB
4 Tape Servers
4x2.5MB/s
60GB
2x5MB/s
2 File Servers
8mm/4mm
DLT Server
5MB/s
5MB/s
4 Robotic Tape Silos
10x0.5MB/s
10x0.5MB/s
24TB
20 Batch Compute Servers
1414SpecInt92
Legend
FDDI
Switch
Tape Silo
Ethernet
Bridging
Hub
Fiber
Concentrator
Figure 11-3.
FDDI
Tape Silo
Control Unit
Bus
UNIX
cpu
Prototype Unix compute farm at SLAC.
Technical Design Report for the BABAR Detector
490
Computing
of home institutions obviously includes a wide range of capabilities. They are characterized
here, however, by the fact that they will not be used to perform bulk data processing, and are
unlikely to want copies of signi cant fractions of the raw or processed data. It is envisaged
that data analysis in the home institutions will be based (in terms of access to bulk data)
around the data store at SLAC or a regional center, whichever is most convenient. Home
institutions may want copies of DSTs. These again could come from SLAC or a regional
center. It is quite possible that some of these institutions will contribute to Monte Carlo
generation.
SLAC Site
It is essential to have only one master copy of production software and databases. The SLAC
site is an obvious choice for these, especially during the installation and early running of the
experiment when updates and changes may be required frequently and quickly. This model
of organization has already been implemented by BABAR for software development. The use
of CVS/rCVS places central control of the master copies of all code at SLAC while allowing
users at remote sites access to this code with a mechanism for including changes back into
the master. The user's view of this process is essentially the same whether they are based
at SLAC or at a collaborating institution.
The bene ts of performing bulk processing at the site of the experiment have been proven
many times. Most large experiments now do this task routinely (ALEPH, DELPHI, OPAL,
L3, H1, ZEUS, CDF, D0). Though other models do exist (E791), the quasi-online processing
possible with local facilities is an ecient use of resources and has signi cant advantages
in producing results more quickly. Clearly, the ability to perform bulk data processing is
required at SLAC. A batch system is needed to facilitate this processing.
The presence of the raw and processed data, as well as subsystem experts, at SLAC makes the
SLAC site an obvious place for data analysis. To make the most e ective use of this requires
that an ecient analysis framework be maintained at SLAC, in particular, one which allows
users not actually based at SLAC simple and ecient access. This system should not require
that remote users log on to the SLAC computers. Remote users should have e ective access
to SLAC's batch system allowing creation of small DSTs or n-tuples which can easily be
transferred back to the home institution for interactive analysis. Current WAN technology
allows n-tuples of many tens of megabytes to be eciently transferred in this way. For larger
datasets, high-density tape technology provides a method of transport. The ability to ship
data in this way requires additional copying facilities at SLAC.
Technical Design Report for the BABAR Detector
11.3 Computing Model
491
Regional Centers
Support for regional centers has signi cant manpower and cost implications both at the
regional centers and at SLAC. However, there are distinct advantages to having access to
a local data store. The regional centers will serve as data store, processing, and analysis
centers. A complete data sample and support structure could be maintained at these sites.
Regional centers will help aliated remote users by providing easy access to the data and
enabling them to build up the strength of a community.
The concentration of computing and data handling capabilities found at regional computing
centers also makes them suitable for large scale Monte Carlo productions. Experience has
shown that this is an e ective use of these centers and an ecient use of collaboration
resources. This also ts well with the regional center's role in general computing support for
smaller collaborating institutions.
Regional centers already act as concentrations of support e ort in BABAR. For example,
rather than many institutions independently attempting to keep all software and libraries
up to date over poor network connections, this e ort is concentrated in a few places. Other
institutions then make use of this local infrastructure directly via NFS/AFS and le transfer
using the much better network connectivity that they have to the regional centers.
There will be varying forms of regional centers de ned by the actual level of service and
support they provide. This will be in uenced by many factors. International networking
and the costs of data transport may change signi cantly over the next ve to ten years. The
regional centers could play a role analogous to SLAC or any other home institution, or, more
likely, somewhere in between.
Two sites have already expressed strong interest in the Regional Center model; CCIN2P3
in France and DRAL in the UK. Possible further sites in Europe or North America are not
excluded, but the cost and e ort associated with the copying of large quantities of data are
likely to be such that the number of such centers will always be limited. Both CCIN2P3
and DRAL already provide large scale computing facilities for HEP experiments (including
large Unix workstation farms) on a similar scale to SLAC and are equipped with large tape
robot systems compatible with the one at SLAC. Both the UK and France already make use
of good national wide area networks, with BABAR software being maintained centrally and
accessed by collaborating institutions via the network/NFS.
Home Institutions
Physicists in the home institutions should be able to contribute to code development remotely. The mechanisms for this are already in place. The long term success requires that
the collaboration computing model remains broad based in terms of supported hardware,
Technical Design Report for the BABAR Detector
492
Computing
application of coding standards, and support for remote access and transport of small
datasets.
It is anticipated that the analysis work of remote users will be centered at SLAC or a regional
center in terms of access to the bulk data or full DST. Physicists will then transfer n-tuples
or small DSTs to their own institutions for intensive interactive work.
Many home institutions have signi cant computing capability, and we may wish to make
use of this for Monte Carlo generation, for example. The ability to generate Monte Carlo or
reprocess small data samples locally, will almost certainly be required by many physicists.
Code, Data, and Database Access
The access to code and data is covered elsewhere in this chapter. Providing convenient
access to code for remote sites also has repercussions, however. For example, facilitating the
running of BABAR code on essentially any home institution (Unix) computer entails the risk
of potential problems with compatibility of results from di erent machines, architectures,
or avors of operating system. This problem could become particularly important when
considering reprocessing the data at one or more regional centers or home institutions. The
collaboration is aware of these problems and aims to organize the production and analysis
e orts so as to minimize their e ects.
Implicit in having good access to the software and data samples is the requirement of access
to the collaboration databases. Any institution processing or reprocessing data, generating
Monte Carlo, monitoring detector performance, etc., will require access to one or more
databases. This access could be achieved by real-time queries to a master database at SLAC,
over the WAN. This seems impractical today, especially when good real-time performance is
required, for example, during a batch job. Most likely, copies of all or part of the database
must be made available to collaborating institutions. These copies will almost certainly
require frequent, automatic updates. The ability to do this will be important when choosing
a suitable database technology for the experiment.
License Issues
It is likely that at least some of the software used by the BABAR collaboration will be licensed
commercially. Given that some of this will be duplicated in collaborating institutions, the
collaboration should try to enter into blanket license agreements, covering SLAC and remote
sites, in order to reduce the total cost of these licenses to the collaboration. For example,
the possibility of purchasing AFS client licenses which are extensions of the Stanford AFS
server license, as CERN has done with theirs, will be investigated.
Technical Design Report for the BABAR Detector
11.3 Computing Model
11.3.3
493
Baseline Cost Model
It is too early to purchase the bulk of computing equipment needed for BABAR for many
reasons: it is not needed immediately; decisions about inter-institution compatibility have
not yet been made; technology is changing rapidly; and costs are decreasing. However, it
is both useful (as a reality check) and necessary (to provide cost estimates for the funding
agencies) to detail the design of a computing system using current and near-future technologies. In this section, such a system following the basic architectural plan set forth earlier
(Figure 11-2) is described.
Networking
The baseline design speci es components available today (December 1994). The expectation
is that the actual implementation will evolve with experience and product availability as
items are required. The baseline design of the network should allow relatively easy migration
to emerging technologies as they become cost e ective. Based on the data ow, the various
network links in the schematic diagram are addressed below.
Online Networking
The online networking is used for data acquisition, trigger, and online monitor/control. It
is based on point-to-point FDDI links connecting the Level 1 single board computers to a
cross-bar switch (Digital's GIGAswitch [GIG93]) and nally to the Level 3 CPU farm. A
single GIGAswitch will provide 16 connections to the FDDI rings containing the front-end
data acquisition crates; direct connections to 14 Level 3 farm CPUs; a link to a local Ethernet
hub; and a single long-haul link from the IR hall to the computer center's mass storage silo.
All other online networking for program development and monitor/control will use lower
cost Ethernet links.
Local Area Networking
The design for local area networking is based on switched 10 Mbits/s Ethernet and 100 Mbits/s
FDDI technologies. In particular, Digital's GIGAswitch and Alantec's Powerhub 3000
Ethernet switch [HUB94] are used to interconnect workstation farms, mass storage servers,
and desktop machines as described below. The link between online and oine Gigaswitches
will be an FDDI link running over existing multimode ber between IR-2 and the computer
center, a distance of about 5000 ft.
Technical Design Report for the BABAR Detector
494
Computing
The need to follow activities in ATM (Asynchronous Transfer Mode) and the 100 Mbits/s
Ethernet switching and interface markets as they evolve, is recognized. To gain practical
experience and skills, it will be necessary to make serious on-site pilot studies of ATM and
100 Mbits/s Ethernet in 1995 or early 1996.
Wide Area Networking
All of the computing tasks (other than data acquisition) can occur at the sites of remote
collaborators. The degree to which this may be done is regulated by the computing investment at the remote sites and by the development of the WAN connectivity between SLAC
and the collaborating institutions in the following years.
Today, SLAC has two T1 (1.544 Mbits/s) connections to ESnet, one to Caltech, the second
to LBL. In addition, there is a 10 Mbits/s Ethernet microwave link to Stanford University
that is used to access Stanford and to provide backup access to the rest of the world if the
ESnet links are down. Wide area networking support to rst order is outsourced to ESnet
and Stanford University.
Data rates and response times between SLAC and other sites depend on the distance, the
number of network hops, the speed of the links between hops on the route between SLAC
and the other site, implementation details such as window and packet sizes, the error rates,
and how busy the link is. Typical values for US sites vary from 90 kbytes/s and 12 ms ping
response time, to 12 kbytes/s and 120 ms. For European nodes, the numbers are 10 kbytes/s
and 200{360 ms.
As a minimum, all collaborators will need reasonable WAN connectivity to SLAC for the
purposes of transferring code and performing interactive work using the X Window system.
For X sessions, one study [Aba94] shows that the number of concurrent X sessions between
adjacent T1 ESnet nodes is on the order of 24. Other studies [NCD89] indicate that the load
imposed by an X terminal is approximately 50 kbits/s.
In spring of 1995, SLAC plans to upgrade its ESnet links to T3 links (45 Mbits/s). Recent
measurements made by ESnet [Bos94] for one hop on unloaded T3 links between LLNL,
LANL, and PPNL gave TCP data rates of between 0.58 and 1.72 Mbytes/s for optimized
settings of window, packet, and internal write bu er sizes. Depending on budgets and
demand, further upgrades of ESnet links to OC3 speeds (155 Mbits/s) may be initiated in
1995 for one or two sites, followed by other major sites a year later. The next step after OC3
is OC12 at 622 Mbits/s. Technically, upgrading to such links should be feasible in 1996/1997.
International links are more expensive than mainland US links. Today's cost for the ESnet
T1 (1.54 Mbits/s) link from Princeton to Germany is about $12K/month for the US half
Technical Design Report for the BABAR Detector
11.3 Computing Model
495
circuit. This is one- fth of the total cost of the link. One might expect a T3 link to be four
to six times as expensive [Bos94].
Connectivity between the wide area network and the on-site network will be provided through
a rewall FDDI router with connections to the SLAC FDDI backbone and the FDDI ring
to which the o -site routers will be attached.
Compute Farms and Other Processors
Online System. Software development, run control, and monitoring tasks will use gen-
eral purpose workstations interconnected with each other and with the computer center
as described above. The Level 3 farm, which functions as a part of the data trigger,
online reconstruction, event tagging, and event monitoring, will also use general purpose
workstations.
Reconstruction and Monte Carlo Farm. Event reconstruction is done in the oine
reconstruction and Monte Carlo farm. Data from tape are rst staged to a series of disk
servers via FDDI connections through the oine switch. Each reconstruction farm machine
is directly connected to a network hub, which in turn connects to the disk server via the
switch. Both reconstruction and Monte Carlo tasks tend to be CPU bound rather than I/O
bound, thus allowing the use of slower and less-expensive Ethernet network links to these
machines.
The data aggregate rates for this farm are 8 Mbytes/s (Table 11-2). Assuming the nodes
in the farm are rated at an average of 200 MIPS each, there will be about 50 nodes to
provide the 10,000 MIPS required (Table 11-1). Using existing technology, the networking
requirements can be met by connecting the nodes to two Alantec Powerhub 3000s. Each
of the Powerhubs will support 12 switched full speed (10 Mbits/s) Ethernet connections
and an FDDI connection. Each Ethernet connection will have two to three nodes attached
via Unshielded Twisted Pair (UTP) connectors, for up to 36 nodes per Powerhub. Each
of the two FDDI connections (one for each Powerhub) will attach to an FDDI port on a
GIGAswitch.
The DST Analysis Farm. Creation and analysis of DSTs occur in a second farm of
machines with direct FDDI connections to the oine switch. These jobs are I/O intensive,
thus requiring the higher performance and more costly FDDI network links. Data ow is
similar to that for the reconstruction and Monte Carlo, from the mass storage to disk servers
and nally to/from the DST farm.
Technical Design Report for the BABAR Detector
496
Computing
The aggregate network capacity required for the DST analysis farm, as estimated in the
preceding requirements section, is 200 Mbytes/s. The compute requirements (Table 11-1)
of 2500 MIPS could be met with a farm of thirteen 200 MIPS processors. To handle the
bandwidth requirement of 200 Mbytes/s, however, about 20 nodes will be needed, assuming
each node can support a maximum of 11 Mbytes/s FDDI transfer rate. In this model,
each processor node will have its own dedicated FDDI port on an oine GIGAswitch.
Measurements must be made to see what FDDI transfer rates can be supported by these
nodes while still allowing them to deliver the requisite compute cycles for DST creation and
analysis. One estimate for an ATM switching network [Dou94] indicates that approximately
8 MIPS of CPU power are required to deliver each megabyte per second of data. If this
turns out to be a limitation, then an alternative is to double the number of farm nodes
(hence, halving the required bandwidth for each node) and for pairs of nodes to share a
single GIGAswitch port.
Desktop Workstations. Interactive analysis involves the manipulation (and sometimes
creation) of n-tuple datasets, and other event data in the highly interactive manner of PAW
or HippoDraw. This is also an I/O intensive task, but in general, it involves data sets of
modest size. This type of exploratory analysis occurs on desktop workstations. The data
ow is identical to that of the farm machines.
The desktop workstations are expected to be mainly located in the Central Lab, the Central
Lab Annex, the Warehouse, IR-2, and the new B Factory building. The data rates for
individual workstations are 0.2 Mbyte/s (Table 11-2). This is well within the capability of the
10 Mbits/s Ethernet interfaces of today's workstations. To ensure that multiple workstations
do not require more aggregate bandwidth than can be sustained on a single 10 Mbits/s
Ethernet segment, each segment will support only a few workstations. As more bandwidth
is required for a particular desktop, the number of desktops on a segment can be reduced to
one. The individual segments will be connected to an Ethernet switching hub located in the
same building as the desktop, each with up to 36 nodes attached. In buildings with large
(> 30) concentrations of workstations, multiple switching hubs will be deployed, each hub
being located close to a workgroup, and in turn, connected via a multimode ber to a FDDI
port on the oine GIGAswitch.
Mass Storage
Tape Mass Storage System. A relatively small number of candidate mass storage
technologies available today might serve BABAR's needs. One seems especially well suited for
use at the SLAC site, where four STK robotic silos with 3490 tape technology are currently
in use. Each tape cartridge in this system holds 800 Mbytes (before hardware compression),
while each tape controller (connected to four tape drives) has a maximum throughput of
Technical Design Report for the BABAR Detector
11.3 Computing Model
497
about 2.5 Mbytes/s. The total storage capacity is thus 24 Tbytes. An attractive upgrade
option involves migrating to STK's Redwood helical scan technology. The expectation is
that this will initially provide 25 Gbytes of data per cartridge, about 15 Mbytes/s maximum
throughput per tape drive/controller pair for a total storage capacity of 600 Tbytes. Cartridge capacity is expected to quickly rise to 40 Gbytes and eventually to 80 Gbytes. Such a
system satis es the BABAR storage requirements while reusing a signi cant amount of existing
equipment. This solution may also be attractive to remote regional computing centers within
the collaboration where there is already a large investment in STK silo equipment.
Mass Storage Server. The data rate from the online to the mass storage is 2.5 Mbytes/s.
In addition, there will be 7.5 Mbytes/s from the reconstruction and 0.5 Mbyte/s from the
Monte Carlo generation to tape (Table 11-2). To this must be added the requirements for
staging analyzed data from tape to disk. This requirement is assumed to add an extra
5 Mbytes/s which would be sucient to allow the 2 Tbytes in the disk farm to be staged
from mass storage twice per week. The aggregate data transfer rates for the mass storage
are thus about 15.5 Mbytes/s. To this must be added the requirements for the creation of
major DSTs, which is forseen to take place mainly during the reconstruction. To the extent
that additional major passes must be accomplished in short time periods, the model may
require scaling to additional bandwidth.
The baseline design calls for the mass storage to be served by four nodes. Each of these nodes
will have its own dedicated FDDI connection to the oine GIGAswitch. The transfer rates
achievable on the FDDI are on the order of 11 Mbytes/s [Che93]. Initially, the performance
is expected to be limited by the workstation servers and the interfaces.
Disk Mass Storage System. The disk servers have bandwidth requirements similar to
the DST analysis farm since that is where most of the data are needed. Using the GIGAswitch
model, a need of about 20 disk server nodes each with a dedicated FDDI connection to the
GIGAswitch is again anticipated. Today's model of GIGAswitch can only accommodate
32 FDDI connections, so at least two GIGAswitches will be needed to support the DST
analysis farm and the disk servers. This will require segmenting the DST analysis farm and
disk servers, each segment having connections to the reconstruction farm, the mass storage,
the desktop workstations, the transportable media, and the external router. The connections
between the two segments will be made by one or two FDDI links. This segmentation will
need further evaluation to ensure a reasonable balance can be achieved. A new model of
GIGAswitch is expected to have 52 ports with 155 Mbits/s ATM, and will be evaluated as
part of the on-going design study.
Technical Design Report for the BABAR Detector
498
Computing
Transportable Media. Standard 8 mm helical scan magnetic tapes will serve as the
basis for data transfer to institutions without STK silo compatible drives. The data rate
requirements for the transportable media are expected to be modest and easily handled by
a single FDDI connection from the server to a GIGAswitch. If more than one server is
required, then the servers will share a single FDDI port on the GIGAswitch.
The decision to use 8 mm tape in the baseline design stems from its widespread use within
the physics community and current support at SLAC. Other technologies, such as 4 mm or
DLT, are also possible for comparable costs. However, at most, one medium besides the silo
medium would be supported by the collaboration.
11.4 Software Environment
The basic requirements for the software infrastructure are that it provide an environment in
which developers are highly productive, provide the necessary stability to ensure that highquality, reproducible physics results may be achieved, and also be exible so that changes in
both the detector hardware and physics goals may be accommodated quickly and eciently.
Applying the principles of commonality across all areas of software development, from the
real-time systems within the data acquisition and trigger systems through the online and
monitoring systems, bulk data reduction and Monte Carlo generation, and nally physics
analysis, can bring about important bene ts. However, application of these principles must
not be at the expense of productivity and must take into account that there are di erent
demands in the various application areas.
11.4.1
Software Methodologies and Languages
Types of Developers
Several types of software development exist within the collaboration, each type having
di erent requirements in its support levels. Individual physicists developing their own
analysis programs should not be overburdened with rules and regulations, whereas software
developed in the real-time environment of the data acquisition and trigger systems must
be subject to rigorous quality tests to ensure that physics data are not lost. Intermediate
between these extremes is the software for the bulk reduction of the raw data and the online
and monitoring software that ensures that the data from the experiment are of high quality.
Within each of these levels, an appropriate level of quality control will be applied as described
later.
Technical Design Report for the BABAR Detector
11.4 Software Environment
499
Software Methodology
Our overall approach to software development is based upon the object-oriented (OO)
methodology. This methodology allows for a direct, intuitive association of physical entities
(e.g., detectors and particles) and abstract entities (e.g., detector status and run control)
with software objects. Commonalities between objects are also identi ed and exploited via
the concepts of inheritance and overloading. Mapping of software objects into lines of code
is most naturally done with an object-oriented programming language such as C++. The
debugging and maintenance phases of development are well supported by the modularity
imposed by the use of software objects. Many aspects of the object-oriented methodology
can be automated through the use of computer aided software engineering (CASE) tools.
Several particular OO methodologies and notations have been investigated, including Booch,
Schlaer-Mellor, OMT, and Objecteering [OOM94]. Two CASE tools, one supporting the
Booch methodology, the other supporting the Objecteering methodology, are under detailed
consideration. Each provides automatic C++ code generation as well as some reverse
engineering capabilities for existing C++ code.
Programming Languages
After much consideration and experimentation, it has been decided to base most of the
software development on the C++ programming language while continuing to support the
Fortran 90 language. Most of the real-time and online systems, together with the core of
the oine system, will thus be implemented in C++, while adequate functionality will still
be provided for the Fortran 90 user. Such an approach is believed to produce the most
robust applications, taking advantage both of the power of the OO approach and C++ and
of the large existing Fortran code base and many man-years of experience with Fortran.
This approach implies a mixed language environment and therefore carries the burden of
additional support, but it is believed to be the most realistic approach, given the schedule
demands and learning curve of the new OO paradigm.
Cross-Language Issues
Ecient access to software modules and data from both of the supported programming
languages is obviously of crucial importance. Signi cant R&D has been performed in
understanding how to accomplish this. Of central importance is the fact that the client
programmer's view of the data format is established not by the data itself, which might
be highly compressed in order to minimize the disk or tape space requirements, but by the
interface that the corresponding object or module provides. Transforming the data between
its compressed, internal form, to that seen by the user programmer, can be achieved in an
Technical Design Report for the BABAR Detector
500
Computing
ecient manner for both C++ and Fortran 90. Further R&D is underway to see whether
both representations can be derived from a common source. The Interface De nition Language (IDL), a component of the Common Object Request Broker Architecture (CORBA)
[OMG91], allows this and has the additional advantage that it supports distributed object
applications. An incomplete prototype IDL compiler that generates both C++ and Fortran
90 stubs and skeletons has been demonstrated and shows some promise, but more work in
a real application environment will be necessary in order to decide whether to continue this
project. This approach seems to be well matched to fairly coarse-grained clients and servers,
such as the packages described in the code management section.
11.4.2 Development Environment
Approach
In recommending that most program development be performed using the OO paradigm
and implemented using C++, it is recognized that ease of use will have to be maintained,
and therefore, support tools will be necessary in order to minimize the learning curve.
The approach is aimed at establishing an environment in which newcomers can quickly
be productive. Tools that are already in place, or shortly will be established, include code
guidelines, code templates, sample programs, and tutorials.
Code Guidelines, Templates, and Sample Programs
Code guidelines based on previously published documents [Ell92] gather rules and recommendations as well as lists of common pitfalls and suggestions for avoiding them. Code
templates exist for both supported programming languages.
Some introductory tutorials have already been developed within the collaboration, but these
will be extended to cover a wider range of topics including the selected OO methodology
and notation, the CASE Tool, and C++ language tutorials for more pro cient programmers.
Other vendor tutorials will be combined with those written within the collaboration.
Quality Assurance
Our software quality assurance (QA) strategy is based on a stable software release mechanism with responsibility for quality assurance being distributed among several software
coordinators, as detailed in Section 11.4.4.
Technical Design Report for the BABAR Detector
11.4 Software Environment
501
One tool in the overall QA strategy is code reviews, in which one model under consideration
is that of a mentor who can informally review new code and advise the developer of improvements, rather than a formal review process with its large manpower overheads. A bug
tracking system is being evaluated that should ensure that bugs are given sensible priorities
and xed in a timely manner.
A crucial component of software QA is adequate documentation, both at the level of the user
and at the level of the maintainer. There is a commitment to the creation of documentation
as an integral part of the development process, not as an afterthought. Word processors
have been used to create documents (the Code Guidelines being an example of this) that
can be viewed both in printable form (i.e., PostScript documents) and also as browseable
WWW documents for online use.
Compilers, Debuggers, Application Builders, and CASE Tools
It is unfortunate that the C++ language has only a draft standard and that ecient Fortran
90 compilers are only just becoming widely available. The GNU C++ compiler has been
identi ed as providing a suitable baseline in functionality, while excluding any extensions
that are not part of the draft C++ language standard. Compilers from other vendors
that conform to this minimum functionality are available on all hardware platforms being
considered for use within BABAR. The situation for Fortran 90 is that one compiler (NAG)
is available on many platforms and adheres strictly to the standards, and native compilers
from the major vendors are either already released or shortly will be.
As mentioned in a previous section, some CASE tools are being evaluated, though it is
realized that their use will probably be limited to a core set of developers. However, the
more widespread use of the methodology and notation for which the CASE tools provide
support is being promoted.
Application builders exist, although their use within HEP is not widespread. Several are
being evaluated, some of which are commercial and therefore raise licensing issues. A
prototype application framework that will support both batch and interactive environments
is under development.
11.4.3
Data Model
The use of an object-oriented methodology for the software development environment leads
naturally to the description of the data model for the experiment in terms of objects. This
bypasses many of the solutions that have arisen throughout the FORTRAN era such as BOS
and ZEBRA and integrates access to the data naturally into the programming language.
Technical Design Report for the BABAR Detector
502
Computing
Event
version
run
record
Raw
version
Hits
version
date
time
type
VcTracks
version
DcTracks
Tracks
Vertices
version
version
nCharge
nNeutral
version
nVertex
VcTrk
version
DcHits
version
nHits
each hit:
layer
wire
phiMid
driftDistMid
dEdxPH
rawHit
status
DcTrk
version
phi
VcTrk
version
VcTrk
version
DcTrk
version
phi
tanDip
curvature
x
y
z
hitsNode
tanDip
curvature
x
y
z
hitsNode
nHits
hit1
hit2
...
nHits
hit1
hit2
...
Figure 11-4.
Portion of a Farfalla event tree for the BABAR experiment.
A uniform data model will be applied throughout the software. Thus, the same model will
apply to the access to raw, reconstructed, Monte Carlo, or calibration data, and to geometrical constants. This model will also allow for a smooth schema evolution. Advantages of this
approach are that it makes possible access to the new generation of database management
systems that manage objects, and that the implementation of the data storage is decoupled
from its access.
An investigation of various approaches to the data model for BABAR has been carried out,
including a number of products based on C and C++ [Sax94]. The Farfalla [Wal94]
package is the most promising existing object-oriented package on which to base the BABAR
data model. A prototype e ort using Farfalla is underway.
The Farfalla package provides a means to store an experiment's data on disk or tape, with
convenient access in the user's program. It is based on the idea of nodes in a tree structure.
Technical Design Report for the BABAR Detector
11.4 Software Environment
503
Some of the nodes in the BABAR prototype event tree are illustrated in Figure 11-4. As
Farfalla is written in C++, it provides a match with object-oriented design in C++
for other software. The access to the data from C++ programs is straightforward. The
data itself is encapsulated (hidden), and basic access is provided via inline accessor member
functions. This decouples the means of storage, which can be optimized for space, from
the user interface, which can be optimized for convenience. Additional capabilities are
provided by more complicated member functions. Because the detector simulation currently
employs GEANT, written in Fortran, and some prototype reconstruction e orts are also
being undertaken in Fortran, one of the objectives of the data model prototype is to provide
access to this data. It has already been demonstrated that this can work, and current concern
is in making this access in Fortran and C++ look similar.
11.4.4 Code Management and Distribution
Introduction
The code management environment must satisfy con icting demands from the code developers, who want to be able to make changes rapidly with a minimum of controls and overheads,
and the end users who wish to see a stable and bug-free code base. The situation is further
complicated by the need to accommodate distributed program development throughout the
collaboration and to make the resulting software rapidly available to the distributed end
users. Finally, various machine architectures and dialects of Unix must be accommodated.
Implementing a system that resolves these con icting demands involves not only robust
computer tools, but also policies on their use, and perhaps most important, acceptance by
the developer and user communities that some controls need to be applied in order to ensure
that the experiment produces reliable, reproducible, and defensible physics results.
Policy
The adopted policy is based on the concept of a code package. A package is a well-de ned
piece of code that has an internal coherency and well-de ned interfaces to other packages.
Examples of packages are track nding, vertex tting, histogramming, etc. Many of these
packages will come from within the BABAR collaboration, but others will come from outside.
Individual applications are also treated as packages in this model. Each package is assigned
a Package Coordinator, a responsibility that might be shared among several individuals.
The coordinator is responsible for the development of stable snap-shots or versions of the
package, ensuring that a minimum set of test procedures has been performed. Di erent
versions for each package might be generated quite rapidly during periods of intense program
Technical Design Report for the BABAR Detector
504
Computing
development, but the development pace is dictated by the developers themselves, not outside
constraints.
BABAR software releases are created at intervals dictated by stability and scheduling constraints. The release mechanism is controlled by a Code Coordinator. A release is formed
from the most recent stable versions of every package, where stable is de ned by the package
coordinators. Testing focuses on interdependencies between the packages on the assumption
that each package has already received some internal testing. As bugs are detected and
corrected, the corrections are propagated forward to the most recent development version in
a manner that does not interfere with the ongoing development. Software tools aid in the
coalescing process.
Quality assurance is therefore a distributed activity, with all developers bearing some responsibility, although the main burden lies on the overall Code Coordinator and the Package
Coordinators.
File Organization
A simpli ed le organization is shown in Figure 11-5. The directory tree is split into
package and release subtrees. All software packages reside under the package tree, with the
primary code repository and subtrees corresponding to individual versions. Each package
has directories for source code, libraries, executables, documentation, etc., for each version.
Under the release subtree are stored all the complete releases, although space limitations
will probably cause only the more recent releases to be maintained at any one site.
Tools
The primary code management tool is the CVS package, which allows multiple developers
ecient access to a code repository while supporting tracking of code changes and their
authors. The utility rCVS, developed at SLAC, extends the use of CVS to distributed
developers by maintaining proxy code repositories at remote sites, via which changes are
automatically propagated to and from the primary repository.
One weakness of CVS and the Unix le system is the lack of Access Control Lists (ACLs), by
which individual developers may be granted access to a subset of the overall code repository.
Since it is expected that most of the collaboration will act as code developers for at least one
package, accidental corruption of the repository is of some concern, and ACLs could play
an important role in this. Advanced distributed le systems such as AFS and DFS support
ACLs and would address these concerns, while simultaneously making the code repository
available throughout the collaboration without the need for rCVS. More work remains to be
Technical Design Report for the BABAR Detector
11.4 Software Environment
$BFDIST/
packages/
505
pkgA/
1.0.1/
Makefile
1.0.2/
doc/
pkgA/
1.0.3/
(include files)
pkgB/
releases/
1.0.7/
1.0.9/
1.1.2/
pkgA/
pkgB/
include/
1.1.4/
pkgA/
pkgB/
current/
production/
lib/
AIX/
HPUX/
bin/
Figure 11-5.
structure.
libpkgA.a
libpkgB.a
OSF1/
Simpli ed le organization, showing the release and package subtree
done on understanding the performance issues of these le systems, and R&D is underway
at SLAC with a small user community on AFS.
The various dialects of Unix each have their own syntax for the make utility that builds
the code dependencies and automates the building of application programs from the various
packages or the libraries that might be part of a package. Two alternatives are being investigated in order to standardize on a version of make: imake (part of the X11 distribution)
and GNU make.
Finally, a tool that automatically creates standardized version and release numbers and
performs the functions for creation of a new release would greatly reduce the burden on the
code coordinator. The Production Code Manager (PCM) prototype under development at
SLAC is designed to perform these functions and is under evaluation.
Code Packages
As has been mentioned previously, a package is a well-de ned body of code. Any one package
is likely to be referenced by other packages. One area that needs more work is ensuring that
circular dependencies do not develop between packages, since this greatly impacts the ease
Technical Design Report for the BABAR Detector
506
Computing
with which applications can be built. This implies a hierarchy in which a package can
reference classes or functions from packages lower in the hierarchy, but not from its peers.
Software Release Mechanisms
A release of the BABAR software will consist of a self-consistent and tested set of all software
packages. It will be entirely self contained, applications having been built for all supported
hardware platforms, thus relieving remote sites of the burden of building applications from
source code. This implies a certain minimal level of consistency in operating system versions
across the collaboration with only a small set of operating system versions being supported
for each hardware platform. Remote sites can either have direct access to a release using a
distributed le system such as AFS, or the directory tree corresponding to each release can
be made available in compact (tar) form suitable for shipping across a network or via tape.
The goal is to have reasonably stable releases (public releases) available at fairly frequent
intervals (a few weeks to months) with more stable production releases having longer lifetimes
for bulk production requirements.
An area that needs further work is that of integrating the bug reporting and tracking from
remote sites in such a manner that xes or patches are made available in a timely and
coherent manner across the complete collaboration.
11.4.5
Databases
The BABAR project faces a formidable set of information management requirements. The
varieties of data that must be managed range from conventional institutional information
to highly technical engineering and physics data. The computing model in which it must
be managed is distributed, heterogeneous, and geographically dispersed. Modern database
management systems (DBMS) will be essential to satisfy these requirements. A partial list
of databases required for BABAR includes the following:
Detector information: geometry constants; calibration and alignment constants; hotand dead-channel maps; monitor logs.
Acquisition system information: constants for front-end electronics.
Run information: run start and stop times; trigger and readout con gurations; event
and record counts; tapes written; online logbook.
Oine information: location and status of datasets; status of jobs for each run (real
data and Monte Carlo).
Technical Design Report for the BABAR Detector
11.4 Software Environment
507
Tape management system.
Document catalog: BABAR notes, manuals, papers, drawings.
Engineering data: component and subsystem test and calibration records.
Personnel information: collaboration directory, distribution lists, author list, shift list,
ES&H training, human resource estimates.
Management information: budgets, schedules, equipment inventories, software license
tracking.
The possibility of storing information about interesting physics events in a database will also
be investigated.
Two di erent database technologies are relevant to BABAR requirements. Relational database
management systems (RDBMS) are the current choice for managing institutional data in
most large organizations, including other current HEP experiments. This is a mature
technology with several major vendors and many high-quality third-party products on the
market, a well-established theoretical foundation, and an expressive and standardized query
language (SQL) to provide portability. On the other hand, RDBMS have not been integrated
smoothly into programming languages, and their simple data model (rows and columns) is
poorly matched to the complex data structures needed in a HEP environment and supported
by an object-oriented language such as C++.
Object-oriented database management systems (OODBMS) provide an increasingly popular
alternative to RDBMS. These systems generally map database elements directly to software
objects. Most of the complexities of dealing with the database management system are
hidden from the programmer. Unfortunately, these products are less mature and lack both a
rm theoretical foundation and (as a consequence) satisfactory ad hoc query tools. Moreover,
though several candidate interoperability standards exist, there is not yet strong, crossvendor support within the industry for any of these speci cations.
Oracle is already in place at SLAC and is in use by both BABAR and PEP-II for a variety
of di erent types of institutional and engineering data. Most such data will likely remain in
Oracle at least until OODBMS with equally expressive query languages become available.
BABAR will investigate the use of OODBMS for those databases accessed heavily from physics
programs, e.g., calibration and alignment constants, channel maps, etc.
Some types of data, such as the run database, are likely to be frequently accessed both from
within the physics code and from external query tools. The choice of technology for such
databases will be deferred until more experience with OODBMS is gained and a better idea
of the access requirements of the physics code is obtained. It is likely that the physics code
Technical Design Report for the BABAR Detector
508
Computing
will require access to at least some relational databases, however, so BABAR will also select,
or if necessary, develop a class library for accessing a generic RDBMS.
Large scale databases represent potentially major investments, both for licensing and administration. It is, therefore, unlikely that all collaborators will have local access to all database
software. Various models can be imagined whereby data is shared with other collaborating
institutions. For cases in which database software must be present, site licensing will be
investigated, and a mechanism for periodically replicating and distributing certain databases
will be developed. Where licensing is not essential or is impractical due to expense, services
will be provided to access data over the network either via BABAR-developed client tools or
via low-cost, third-party \data-dipper" applications.
11.4.6
Graphics
Computer graphics serve an important role in various stages of an experiment. During
detector design, they aid the physics simulation tasks and CAD. In the course of code
development, graphical detector representations and interactive event displays are necessary
for pattern recognition and event reconstruction studies. For physics analysis of data, it
is essential to have tools for event viewing, parameter representation, multidimensional
statistical analysis, and histogramming. Graphics displays with graphic editors are useful
for producing high-quality hard copies for presentation and publication. Also, graphical user
interfaces (GUIs) are important for human interfaces in a variety of contexts, such as basic
computer interaction (the desktop), software engineering tools, job submission, and detector
control and monitoring. In this section, the available computer graphics and GUI options
are summarized.
Graphics Hardware
Compared to the industry standards, the hardware requirements for graphics in high-energy
experiments are moderate. The major factors are compute performance, variety of color
display, screen resolution, and the ability to produce quality hard copies. For screen display,
there exists a range of workstations with sucient performance for most applications. Their
price/performance ratio is improving with time. For hard copies, color printers and color
copiers are becoming inexpensive. These trends are expected to continue.
Technical Design Report for the BABAR Detector
11.4 Software Environment
509
Software Options
For histogramming and plotting curves, there are several packages already available such
as Handypak and Topdraw, along with the CERN graphics packages HBOOK, HPLOT,
PAW++, etc. For event and detector displays, an assortment of approaches ranging from
simple two-dimensional X11 or PostScript cartoons to complex three-dimensional structures
can be useful. Most three-dimensional packages have high manpower overhead, and options
should be examined carefully. One could also adopt a package like that of ALEPH's V-plot
for representing three-dimensional data of events [Dre91].
In the Unix environment, there are a number of commercially supported three-dimensional
graphics packages available such as Graphics Kernel System (GKS), Programmer's Hierarchical Interactive Graphics System (PHIGS and PHIGS PLUS), Silicon Graphics Library (Open
GL), and Renderman. While X11 and PHIGS can be obtained free of charge from the Open
Software Foundation, other packages must be purchased from commercial vendors. PEX
(PHIGS Extensions to X) also supports three-dimensional graphics within the X-Window
environment. Some of these packages have ISO and/or ANSI standards de ned. They also
provide device-independent interfaces. In the near future, a choice among these packages
will be made based on relative performance, ease of use, and portability.
There are also a number of visualization packages available on the market such as AVS, apE,
and PV. With these packages, the user can develop graphical displays interactively using the
mouse and keyboard at a workstation.
Graphical User Interfaces
A graphical user interface (GUI) is characterized by pointing, clicking, and dragging with a
mouse; pull-down menus; dynamically resizable windows; and other features. GUIs o er the
user a high degree of interaction and assistance in navigating through unfamiliar programs.
Each component of the program can be associated with graphical icons illustrating the
intuitive connection with its associated function. A properly designed GUI can dramatically
aid the training to operate a program, reduce errors, and increase operator eciency.
There can be a considerable cost associated with these user interfaces in terms of programming development. In fact, there are reports that 80{90% of a program's development may
be required to design and produce a good GUI in the absence of a good interface builder.
The availability of commercial packages called interface builders signi cantly eases the work
load in developing GUIs. Products producing X11-based GUIs are numerous, including
XVT-Power++, TAE+, UIM/X, Garnet, and TeleUSE.
Technical Design Report for the BABAR Detector
510
Computing
11.5 Online System
11.5.1
Introduction
The online system provides for the operation of the BABAR detector while it is preparing for
or engaged in data taking. The system's primary responsibility is for collecting and storing
the experiment data. It will also monitor and maintain the integrity and quality of that
data. The online system must maintain a high level of availability during collider operation.
This system can be organized into several main areas as shown in Figure 11-6 and listed
below:
Experiment Control. Controls and monitors the environment of the detector.
Run Control. Organizes, controls, and monitors the detector components, hardware, and
software for data collection.
Data Flow Control. Controls and monitors (in real-time) the passage of data from
the detector subsystem back-end electronics to the nal data repository and other data
consumers.
Trigger Control. Controls and monitors the decision making processes that determine
whether data will be recorded.
Clock and Control. Provides a variety of precise timing signals and strobes to the
subsystems.
The online system must interact with and link together the various components of the BABAR
detector as illustrated in Figure 11-7. Some of the systems illustrated, such as the PEP-II
control system, require only a communication protocol speci cation from the online system
since they are external to the detector system's responsibilities. Other subsystems, such
as the detector front-end electronics, require detailed speci cation so a complete range of
operations is allowed. The online system will share much of its architecture and software
support with the oine system.
The online system will use an object-oriented design methodology for many of its components. This approach stresses the reuse of computer code, logical constructs, and ideas.
Technical Design Report for the BABAR Detector
11.5 Online System
511
Electronics
Computing
Online System
Clock &
Control
Trigger
Control
Figure 11-6.
Data Flow
Control
Run Control
Expt.
Control
Schematic organization of the online system.
Operator / Physicist
PEP-II
Control
System
BABAR
On-line
System
Detector
Front-end
Electronics
Detector
Support
System
Figure 11-7.
Data Store
Off-line
Computing
Context diagram of the online system.
Technical Design Report for the BABAR Detector
512
Computing
Within this methodology, the system will develop iteratively with each major release usable
as a test system, but without complete functionality. Later versions will build on the existing
system by lling in missing functionality. This model furnishes concrete deliverables for
test stands, early beam tests, the high-energy ring commissioning, and the colliding beam
commissioning. It also provides for adaptive improvement of and evolution in the system.
The description of the online system that follows presents di erent views of the system. The
logical view focuses on the functional description of the system. The subsystem view details
the interface to the detector subsystems and their common logical and physical constructs.
The operator view describes the system's appearance to and allowed interactions with the
physicist/operator. The implementation view describes the working decisions and speci c
implementations of the system. This approach aids in coping with some of the system's
intrinsic complexity by focusing on only one aspect of it at a time.
11.5.2 Logical View of the Online System
The online system will use a client-server, distributed-object model for its architecture in
which high-level control of the system is arbitrated by a series of console, manager, and interface processes (Figure 11-8). A console process such as a GUI program or a script always acts
in the client role and allows the user to interact with and exert direct external control over the
online system. A manager process provides resource allocation and control for the system's
consoles, device interfaces, and other managers, and is composed of objects using Common
Object Request Broker Architecture (CORBA) [OMG91] techniques to implement clientserver communications. A device interface layer with well-de ned functionality controls the
system's low-level distributed devices, encapsulates the associated electronics, databases,
and processes, and shields the rest of the online system from these complexities. In this
architecture, multiple clients may need the services of a speci c manager, but only one client
at a time is allowed control. Token passing controlled by specialized managers called brokers
will be used in this role.
Experiment Control. Experimental Control processes ensure correct functioning and
ecient performance of the BABAR detector by continually monitoring its components. This
involves periodic readback and analysis of the detector electronics and diagnostic procedures
on the detector components. Experiment Control is not responsible for the data ow, trigger,
or clock and control areas, which have their own control processes. Since the functionality of
this system is common to most other high-energy physics experiments, the EPICS tool kit
will form the basis of the Experiment Control framework. EPICS is freely available software
developed originally at LANL and later at ANL and CEBAF. It was originally designed as
an accelerator control system, although its current use includes several experiment control
Technical Design Report for the BABAR Detector
11.5 Online System
513
ownership of
token for yy
system
(obtained from
yyBroker)
Client
zzConsole
Server
xxManager
yyConsole
zzManager
Client
Server
yyManager
yyInterface
Application Program
Interface ( API )
(corresponds to Object
Request Broker ( ORB ))
yy Device
Figure 11-8.
Interaction of multiple consoles/managers, logical layers, and interfaces.
systems. Because EPICS is intrinsically a procedurally designed and implemented tool kit,
a manager and interface software layer will be built above it.
Run Control. Run Control will guide the operator along the proper operational sequence
by which the detector is readied for data collection, calibration procedures, or diagnostic
procedures. Crucial hardware and software components will be automatically tested. During
data taking, Run Control will continually monitor the detector and experiment data to detect
failures and malfunctions. If an error is detected, an expert system will be used to help
diagnose and x the problem. This system is expected to evolve so that it can solve some
routine problems on its own without involving the operator.
Run Control organizes and controls the detector for data taking using a con guration
manager and a partition manager. It also has an expert system manager which aids the
Technical Design Report for the BABAR Detector
514
Computing
operator in diagnosing and xing problems. Con guration management provides the mapping among the physical hardware/software components and the appropriate control and
monitor managers. All persistent attributes of these components will be stored in centralized
object-oriented databases. For example, the device interface will build its EPICS databases
from the central database according to the hardware needs. In particular, each component
will be described by a state transition diagram. Partition management provides dynamic
reorganization of the system's components into independently operating logical subsets. For
example, the main partition could include the trigger and other system-wide resources, while
several subsystem-level and individual readout crate partitions also exist. Con guration
of a particular partition requires the instantiation of the objects representing the various
hardware and software components on a workstation and a selective loading of data les and
code to the associated embedded CPUs.
Trigger, Data Flow and Clock and Control. Trigger, data ow and clock and control
are described in the Chapter 10. From the perspective of the online system, the trigger
controls the data ow in the experiment. Trigger signals in the form of accepts and rejects,
busy and ready, success and failure, are distributed as strobes via Clock and Control.
11.5.3 Subsystem View
Subsystems have several interfaces with the online system. Front-end electronics provide
input to the trigger and data ow processes as described in Chapter 10. General purpose
hardware and software perform routine monitoring and control functions. A set of special
purpose hardware and software performs diagnostic testing for each subsystem. All devices
are associated with a standard readback identi cation and status either directly or through
the controlling processor. The identi cation is used for inventory, maintenance of repair
records, and con guration details.
Most of the controlled and monitored components, including some purely logical components,
can be described by a state transition diagram which has a direct EPICS implementation.
The status is a state-dependent quali er that describes the device's readback in terms of the
device's expected behavior in that state.
The device manager drives the device interface through its transitions, enforcing an internal
logical consistency and conformity with external constraints. For example, the magnet power
supply in the state OFF is expected to have a current output of zero, within the tolerances
of the readback. If the current monitored is not zero, then the returned status would be
OVER CURRENT and color-coded red to signify a possible malfunction. The set of status
values will depend on the allowed states of the device.
Technical Design Report for the BABAR Detector
11.5 Online System
515
Data retrieved from the monitoring processes will be archived and error and message logs
maintained. Status and state changes will be monitored by Run Control to maintain the
integrity of the experiment data. During normal data taking, special readout options will
be available. Nonsparsi ed data from a subsystem can be collected and monitored. This
might check the validity of deleting below-threshold data from the normal data stream.
A snapshot mode can monitor data from the VRAMs. This mode might check feature
extraction algorithms by allowing comparison with full wave forms. An analysis mode can
save all data bu ers associated with a given trigger, so that the history of an event in the
data acquisition can be fully reconstructed.
The process of controlling a device involves changing its state or the values of its functional
parameters. A manager acquires a control token for the device and then takes the desired
action. During the transition, a graphical representation of the action and its progress will
be continually updated and displayed. For complex collection objects, the manager will
enforce any necessary synchronization and present the most relevant progress display. Any
error messages generated are sent to the operator and Run Control and are archived.
Most detector components will have associated calibration and diagnostic procedures to ensure their correct performance during operation. As required by partitioning, all subsystems
must be capable of independent and concurrent calibration. In addition, individual readout
controllers must be allowed to run in diagnostic mode while the others are participating in
normal data-taking operations or calibration. As a result, a controller or set of data- ow
strobe signals may either be replicated, or be made part of a diagnostics partition to allow
independent operation.
A parallel load test will also be provided, using the controller to issue data pulses and triggers,
which exercises the entire normal data-taking stream of the main partition.
11.5.4
Operator View
The online system presents an operator of the detector with a standardized, hierarchically
arranged set of control, diagnostic, and monitoring panels in which the most important
parameters and controllers are also those most prominently displayed. Control panels allow
an operator to perform and monitor normal operations on a particular subsystem. Diagnostic
panels allow an operator to perform normal operations as well as the capability to test and
exercise individual subsystems. Monitor panels only display information about a subsystem
and do not allow an operator to perform any control functions. Detailed information and
control of each element in a subsystem will also be available, but only when requested
explicitly by the operator. The system will rely on the standard graphical user interface
libraries and tools of the X-Window/Motif system.
Technical Design Report for the BABAR Detector
516
Computing
For example, the calorimeter subsystem might have a main control panel that includes
the status of the bias voltage power supplies with an overall gain knob, the most recently
performed diagnostics, a summary of the average hit rates in the crystals, the status of the
calorimeter's front-end electronics and DAQ modules and a set of histograms re ecting the
overall performance and recent history of the calorimeter's functioning. The bias voltage
power supply status would, for example, allow access to each of the individual power supply
crates, which would in turn allow access to individual bias voltage channels through a
standard point-and-click interface.
11.5.5 Implementation View
The ingredients for implementation of a successful online system must not only meet the
technical and operational requirements, but must, in addition, contribute to a coherent
overall architecture. Options available today are considerably di erent from those available
only a few years ago. Rapid advances in electronics and computer technologies are likely to
continue changing the landscape over the design and implementation period for the BABAR
detector systems. Thus, these ingredients must exhibit suciently general properties so
that they and other ingredients may be changed relatively freely and without requiring
large architectural changes to the system. To this end, a preliminary selection of online
components has been identi ed.
Industry standard VME will provide the backbone of the online system. VME is a mature
technology in terms of reliability and availability of products. Its backplane performance is
satisfactory for control applications and many data applications. Enhancements to a VME
system are possible via VME-64 and/or VXI. Various auxiliary buses (e.g., CAMAC) and
cable interconnects (e.g., MXI) can be used in conjunction with VME.
Single board computers residing in VME crates will assume the role of intelligent crate
controllers. The need to provide portable online code implies the use of coding standards
to avoid system-speci c extensions, and the use of a real-time kernel to prevent code from
being dependent upon the hardware speci cs of a particular single board computer. The
VxWorks kernel satis es these requirements with excellent real-time performance, ANSI
C/C++ language support, and compliance with the POSIX standards.
A framework within which to build an online system, perform overall control functions, and
collect and analyze monitor data will be based on the EPICS package. EPICS is based
upon an architecture of Unix workstations and single-board computers running VxWorks
connected by a network running the TCP/IP protocol. In its basic form, EPICS includes
a communication system, a database, periodic monitors, a sequencer, and various device
drivers for speci c VME data acquisition cards. The toolkit also includes a number of applications for handling alarms, archiving data, editing the database, building new applications,
Technical Design Report for the BABAR Detector
11.6 Reconstruction and Analysis Framework
517
Various third party software packages (e.g., Mathematica and MatLab) interface to
EPICS.
Another aspect of software development concerns the quality of design and maintainability.
Consistent with overall computing guidelines, the online system will be based upon the
object-oriented philosophy, utilizing the C and C++ programming languages. The ANSI C
language will be used for compatibility with the EPICS package, while C++ will be used for
higher level software components. CASE tools will be employed where applicable to assure
sound software design.
Various networking technologies are candidates for the online system. The TCP/IP protocol
is necessary for compatibility with the EPICS system and the rest of the SLAC site. Currently, both Ethernet and FDDI o er acceptable solutions for control and slow monitoring
applications. It is expected that emerging technologies such as ATM (Asynchronous Transfer
Mode), FCS (Fiber Channel Standard), and SCI (Scalable Coherent Interface) may become
realistic candidates in the timescale of BABAR; such a solution might address both data
acquisition and slow control needs. Bandwidth and latency are the primary issues facing the
network system; these are discussed further in Chapter 10.
Generally, solutions for the online system will be chosen so as to be compatible with or
identical to those for the oine computing system whenever reasonable. Mass storage, for
example, will need to be equally available to both; thus a common network solution is
desirable. Real-time event reconstruction will require executing the full oine code in the
online environment, thus a standard workstation environment is needed.
etc.
11.6 Reconstruction and Analysis Framework
11.6.1
Scope
A common framework for application programs has to encompass several areas of deployment. For the data acquisition, trigger, and online systems, it must provide support for the
monitoring processes and for the trigger algorithms within the online trigger farm. Within
the oine environment, it must provide support for Monte Carlo simulations, the bulk
processing and reduction of data, and physics analysis. This implies that the framework
must provide support both for input of data from a variety of sources including disk and
tape les and the online event server and for output to a similar set of permanent and
transient destinations. Furthermore, it must provide an interface that is suitable for use in
both interactive and batch environments, and ideally support both expert and novice users.
Technical Design Report for the BABAR Detector
518
Computing
User Interface Toolkit
Application Framework
User Modules
Database Toolkit
Other
Toolkits
Application
Toolkit
Data Access Toolkit
Data
Figure 11-9.
Software framework.
The question arises whether a framework is just a class library (using object-oriented terminology) or whether it encompasses greater functionality. The key distinction between
a framework and an arbitrary collection of classes, however closely related those classes
might be functionally, is that a framework describes not only the objects, but also their
interactions with each other in a domain-speci c manner. An application may therefore
consist of several di erent frameworks, one providing access to the event data, another
providing histogramming support, all under the overall control of the application framework.
A framework that is available purely as a class library is called a toolkit.
11.6.2
Conceptual Overview
Our concept of an application framework is illustrated in Figure 11-9. User-written code, in
the form of modules, can be incorporated with other modules under the overall control of the
application framework, but with access to lower level frameworks and toolkits. Modules can
be associated into paths, multiple paths being supported, where each path could correspond
to a particular physics process, for example. The framework allows the user to select an
input module and will then execute the modules in the various paths in the appropriate
sequence. Filter modules may terminate or redirect processing of events in a particular path
if the data do not meet the speci ed selection criteria. Events that reach the end of each path
may be directed to a set of output streams, allowing for simultaneous stripping of selected
data samples in the same pass through the data.
Technical Design Report for the BABAR Detector
11.6 Reconstruction and Analysis Framework
519
11.6.3 Toolkits
Several toolkits have been identi ed, including the application, data access, database, user
interface, and visualization toolkits. The application toolkit is part of the framework itself
and provides mechanisms whereby user code can nd out attributes of the environment such
as the run number and the presence of other toolkits. The data access toolkit provides access
to data within an event from both C++ and Fortran, and the database toolkit provides access
to the calibration, geometry, and other databases that specify the detector environment. The
user interface toolkit allows users to prompt for operator input and to display quantities
of interest, and the visualization toolkit provides a method whereby event data may be
displayed graphically in a variety of modes. Thus, a track may be displayed in a Cartesian
coordinate system, or as residuals from the tted space points.
11.6.4 Framework
A framework that addresses the requirements is under development. It is written in C++
and uses the Tool Command Language (Tcl) [Ous94] as its primary user interface. It thus
provides a exible text-based interface that supports scripts and is therefore suitable for
the expert user and for a batch mode of operation. It is intended that a graphical frontend to this framework based on the Tk package be created. Expect [Lib91] is a suitable
candidate for this. The main focus of the development for this framework prototype is
the data model prototype discussed elsewhere. In parallel with this development, several
third-party application frameworks are being evaluated, including both public domain and
commercial products. One problem appears to be their use in a batch environment, but they
o er sophisticated user interfaces when used interactively.
11.6.5 Application Builder
The overall concept of the framework is that an application can be built that contains a
large set of modules, only a few of which may be selected if so desired at run time. Thus, the
same application may be used for reconstruction, diagnostics, or analysis depending on which
modules are enabled. This model depends on there being dictionaries of standard modules
that may be combined with user-written modules. It is possible that shared libraries may
also be used to this end. Some of the application frameworks under investigation support
the graphical creation of applications, and this capability has been shown to greatly improve
the user friendliness of such systems.
Technical Design Report for the BABAR Detector
520
Computing
11.6.6 Standard Applications
Several standard applications will be developed. One of these will be the stand-alone event
display, although it is expected that an event display module for incorporation into other
applications will also be available. Another application will support the dumping of event
data in a variety of formats as a diagnostic aid. The applications running in the online
trigger farm and for the main oine data processing may themselves be considered to be
standard applications, and therefore available, in order that tests of the trigger eciency
and other systematic studies may be performed.
11.6.7 Job Submission
Much of the more time-consuming data processing will take place in a batch environment,
and the application framework should interact seamlessly with the supported batch systems.
Basing the framework prototype on Tcl was motivated by its availability as a scripting
language within the Unix shell environment as well as its use as a command parsing toolkit
within applications. The generation of job submission scripts, spanning multiple jobs as well
as single job submission, should therefore be a well-supported operation, capable of good
integration into the overall batch environment.
11.6.8 Bulk Production
The submission of jobs for the bulk reduction of the raw data and generation of DSTs
requires a level of support much greater than for casual job submission. Databases will need
to track which data have reached which stages of the reduction, and which jobs failed and
need further investigation. It might be possible to provide some tools to aid in quality checks
over and above scanning of many histograms. Rapid feedback and high quality assurance
are crucial to ensure that problems are detected quickly and do not a ect the overall physics
results.
Technical Design Report for the BABAR Detector
11.7 Computing Support
521
11.7 Computing Support
11.7.1
Introduction
Support of the BABAR computing environment will be split essentially two ways. The
collaboration will provide support for experiment-speci c tools and the online system, as
well as for oine analysis software. A substantial part of this e ort will be handled using
computing resources at remote institutions.
Infrastructure support including computing hardware, networking, and system software will
be provided by the local computing center and/or departmental sta . SLAC Computing
Services, for example, provides this support at the SLAC site.
11.7.2
Collaboration Support
The collaboration must design, maintain, and document the online and oine software
framework, data model, analysis environment, and user interfaces to the various tasks. The
personnel model is similar to recent typical approaches in HEP, in which physicists write
most of the analysis and reconstruction code, including many of the peripheral tools and
scripts, and software engineers address the other aspects where necessary, using appropriate
commercial design tools. Examples include database technology and the design of the
software framework. This model has the bene t that the physicists will have close contact
with the physics aspects of the code, while the computer science and systems aspects are
handled by people with the requisite expertise.
It must be stressed that the software for the BABAR experiment represents a very large
system, and there are several major tasks that require special attention. The support
for these activities is expected to consume a signi cant amount of manpower within the
collaboration. Most of this manpower will consist of physicists and students, and each
collaborating institution is expected to contribute a signi cant portion of its manpower
to software development. The use of software engineers will have to be limited to those
tasks for which they are essential, in order to minimize costs. Five full-time-equivalent
(FTE) engineers are thought to be required. The following is a list of some of the major
support tasks, most of which include some amount of infrastructure development, database
technology, and coordination.
Project Computing Engineer. One of the FTEs mentioned is at a senior level to serve
as the Project Engineer, responsible for the entire computing system.
Technical Design Report for the BABAR Detector
522
Computing
UNIX System Management. There will be a Unix system manager, responsible for the
smooth running of the computing hardware, particularly online.
Code Management. Code coordinators will be appointed for all software packages, such
as the online system, event reconstruction, and Monte Carlo. Their responsibilities include
supervision, quality control, and management of production releases.
Data Processing. In the course of the experiment, a very large amount of data will be
collected. The production of DSTs is therefore a major undertaking that needs careful
management by data aides under the supervision of a physicist.
Data Distribution. Large amounts of data in DST form will need to be distributed to and
received from various remote sites as discussed earlier. Data aides monitored by a physicist
will perform this task.
Database Management. The BABAR experiment will have a large number and variety
of databases. Database management is not a trivial task, and this will require at least one
dedicated expert.
Documentation and Training. Documentation methods will be developed and enforced.
Various topics may require training, such as C++ and databases. Identi cation of these
topics and organizing training sessions is necessary.
Tools and Utilities. The experiment will require a great variety of tools and utilities.
Some will be designed and written. All must be subsequently maintained.
11.7.3
Infrastructure Support
Infrastructure activities are expected to take place within all collaborating institutions,
and these activities must be coordinated between institutions and with the collaboration.
This coordination will be facilitated through a regional representative committee within the
collaboration and through regular meetings between the collaboration and local support
organizations. An example of the latter is already in place at SLAC where the majority of
computing resources will reside. It is expected that similar arrangements will also be put
into place at other regional centers.
Technical Design Report for the BABAR Detector
11.8 Integration Issues
523
An active discussion with SCS has been underway to identify the support that they will
provide to the BABAR collaboration. A detailed proposal de ning possible infrastructure
support is given in Reference [SCS94] and is summarized below.
SCS will support the computing hardware, including CPUs, networking, and storage. This
includes much of the oine environment and the network components at the detector. It
also includes certain aspects of the Level 3 farm itself, which is envisioned to consist of
general-purpose workstations. Hardware support includes planning, purchasing, installation,
operation, tuning, and maintenance. Software aspects supported by SCS include system
management of the Unix workstations and the X/Motif graphical environment. In addition,
a number of general programming and database tools will be managed by SCS, including
compilers, editors, graphics, and Oracle. Software support also includes management of user
accounts, disk and other resources, backups, a batch system, etc.
11.8
Integration Issues
The online computing system will be an important tool in the commissioning of the BABAR
detector, and even of the PEP-II accelerator. The high-energy ring is scheduled to begin commissioning approximately a year prior to the completion of the low-energy ring.
Background conditions will need to be studied during this period. The interaction region
commissioning will involve simple detectors in advance of BABAR which will require some
data acquisition. The essential components must, therefore, be ready relatively early, and
this also has implications for the schedule for the control room and electronics areas.
The Level 3 data acquisition farm will include peripherals and CPUs, as well as a data
link to the SLAC computer center. Operator interfaces will be discussed below. There
already exists a ber optic cable between the IR hall and the computer center that meets
the requirements. The CPUs and peripherals should be installed in a clean, air conditioned
environment with only occasional access necessary. The space for approximately four 19-inch
racks is required|two for the CPU farm itself, and two for the peripherals.
Another crucial aspect for the e ective running of an experiment is placement and number
of computers and displays for control, monitoring, and debugging. Approximately ten
monitors will be required in the control room to permit simultaneous development and
control activities, and to dedicate screens to displays from the accelerator, environmental
monitoring, event display, and run status. Several additional displays will be required in the
electronics area to permit e ective debugging and maintenance of the apparatus.
An external system that must be linked to experiment control is the PEP-II accelerator.
The PEP-II Control System will be an extension of the SLC Control System. However, new
developments such as longitudinal feedback and rf control will be based on EPICS, with
Technical Design Report for the BABAR Detector
524
Computing
the IOCs (Input/Output Controllers) integrated and supported by PEP-II Controls. At this
time, the role of the PEP-II Controls VAX is seen as limited to that of channel access client;
data from the IOCs of the BABAR detector are thus directly available to PEP-II Controls.
Conversely, information stored in the VAX database would be sent via the network to the
BABAR Online System on a periodic basis (on a likely timescale on the order of minutes).
This information would include beam position monitor data, beam energy, lifetime, current
and emittance data, local pressure gauge readings from the ring vacuum system, etc. This
data will be entered into the BABAR PEP-II database and be available for history plots and
other display options.
Circumstances that may require a more immediate response than is possible with the network
link to an EPICS controller are envisioned. One requirement is to assist in the process of
beam tuning, both to reduce backgrounds and to increase luminosity. In response to a
request, the online system will provide fast digital signals from detector chambers for noise
metering, as well as signals from the detector luminosity monitor.
For proper control of both the detector and the accelerator, the state and status of many
components must be monitored and the information shared between the online and the
PEP-II Control Systems. Control over energizing the BABAR solenoid, over allowing access
to the interaction region, and over allowing injection of beam into the ring, will be allocated
on the basis of permissives exchanged between BABAR and PEP-II Controls. Information
about the presence of colliding beams, the active logging of data, the value of the solenoidal
eld, etc., will be made available to both control rooms. During injection, radiation damage
to detector components will be minimized by reducing the injection rate if a threshold is
exceeded in a PEP-II ionization detector at the IR.
11.9 System Responsibilities and Management
The overall responsibility for the BABAR computing system rests with two system co-managers
and the computing project engineer. There will be designated people responsible for the
various elements of the system, including, for example, code coordinators for the various
packages, an online coordinator, etc.
Broad policy, communication, and planning issues are addressed by three committees:
The detector subsystem (vertex chamber, drift chamber, etc.) reconstruction, simulation, calibration, and diagnostic code will largely be written by people closely involved
in the design and construction of the subsystem. A committee of representatives from
each of the subsystems will be responsible for ensuring a uniform approach within the
context of the computing system and to allow communication of requirements.
Technical Design Report for the BABAR Detector
11.9 System Responsibilities and Management
525
To ensure close and e ective integration of computing e orts throughout the collaboration, a Regional Computer Representative Committee has been formed. This group
brings together the Computing System Managers and Engineers plus a representative
of each socio-geographical region in the collaboration. The group provides a forum
for raising and solving problems of particular concern to collaborating institutions. It
also provides a channel of communication from the Computing System to the widely
spread computing e ort of the collaboration.
Finally, a planning committee exists with membership from both SCS and BABAR.
This group is considering the support issues and resource requirements and is an
important input to SCS for their planning and budget requests, and for coordinating
which requirements will be the responsibility of the collaboration and which of the
computing center.
Technical Design Report for the BABAR Detector
526
REFERENCES
References
[Aba94]
F. Abar and W. Lidinsky, \The E ects of X-Window HEP Graphics Applications
on ESnet," presented at CHEP94, San Francisco, California (1994).
[Ast92] D. Aston, A. Boyarski, A. Breakstone, D. Cords, T. Glanzman, F. Harris, T. Hung, P. Kunz, and F. Porter, \Computing for a B Factory,"
BABAR Note # 82, (1992).
[BaB94] The BABAR Collaboration, \Letter of Intent for the Study of CP Violation and
Heavy Flavor Physics at PEP-II," SLAC{443 (1994).
[Bad90] A. Baden and R. Grossman, \A Model for Computing at the SSC," Superconducting Super Collider Laboratory Technical Report No. 288 (1990)
C. T. Day et al., \Database Computing in HEP|Progress Report," in Proceedings of Computing in High-Energy Physics (1992)
C. T. Day et al., \The PASS Project Architectural Model," in Proceedings of
Computing in High-Energy Physics (1994).
[Boe94] C. Boeheim, K. Gounder, and R. Melen, \Computing Model Options for B
Factory," BABAR Note # 154 (1994).
[Bos94] Talk presented by Rebecca Bostwick at the ESnet Site Coordinating Committee
at CEBAF (1994).
[Boy90] A. Boyarski, T. Glanzman, F. A. Harris, and F. C. Porter, \Report of the
Computing Group," in Proceedings of the Workshop on Physics and Detector
Issues for a High-Luminosity Asymmetric B Factory at SLAC, ed. by D. Hitlin,
SLAC{373, LBL{30097, CALT{68{1697, 575 (1990).
[Dou94] D. Doughty Jr., D. Game, L. Mitchell, G. Heyes, and W.A. Watson III, \Event
Building Using an ATM Switching Network in the CLAS Detector at CEBAF,"
presented at the International Data Acquisition Conference on Event Building
and Event Data Readout, Batavia, Illinois (1994).
[Che93] H. Chen, J. Hutchins, and J. Brandt, \Evaluation of DEC's Gigaswitch for
Distributed Parallel Computing," SAND93{8013B, Sandia National Laboratories,
California, (1993).
[CLE94] Information concerning CLEO-II computing was kindly provided by M. Ogg,
J. Urheim, and A. Weinstein.
[Dre91] H. Drevermann et al., in \Computing in High-Energy Physics '91," ed. by
Y. Watase and F. Abe, Universal Academy Press, Tokyo, Japan (1991).
[Ell92] Ellemtel Telecommunication Systems Laboratories, \Programming in C++:
Rules and Recommendations," Sweden (1992).
Technical Design Report for the BABAR Detector
REFERENCES
[EPI92]
[GIG93]
[Gou94]
[HUB94]
[Lib91]
[NCD89]
[OMG91]
[OOM94]
[Ous94]
[Por93]
[Sax94]
[SCS94]
[Wal94]
527
\Experimental Physics and Industrial Control System (EPICS)," Argonne National Laboratory and Los Alamos National Laboratory (1992).
Digital Equipment Corporation, \GIGAswitch Systems Manager's Guide," Part
No. EK{GGMGA{MG.A01 (1993).
K. Gounder, S. Bracker, K. Hendrix, and D. Summers, \A Simple Multiprocessor
Management System for Event-Parallel Computing," BABAR Note # 128, (1994).
Alantec Corporation, \PowerHub Reference Manual," vol. 2.2, Rev. C, Issue 1,
Part No. 400{0336{000 (1994).
\Expect: Scripts for Controlling Interactive Programs," Don Libes, Computing
Systems, Vol. 4, No. 2, University of California Press Journals (1991).
\Network Loading and the NCD Network Display Station," Note 3, available
from NCD, (1989).
F. Abar, C. O'Reilly, and E. Wicklund, \Network Load of X Terminals at CDF,"
FNAL Report, (1992).
R. Cottrell, \X Terminal White Paper," SLAC (1993).
Object Management Group, The Common Object Request Broker: Architecture
and Speci cation, Revision 1.1, OMG TC Document 91.12.1 (1991).
G. Booch, \Object Oriented Analysis and Design with Applications," 2nd ed.
(Benjamin Cummings, 1994) ISBN 0-8053-5340-2.
D. Coleman, \Object Oriented Development, The Fusion Method" (1994)
P. Desfray, \Object Oriented Modeling," (Addison Wesley 1994).
J.K. Ousterhout, \Tcl and the Tk Toolkit" (Addison-Wesley, 1994).
F. Porter, \BABAR Event Size Estimate," BABAR Note # 108 (1993).
S. Saxena and T. Glanzman, BABAR Note (in preparation).
\SCS Support of the B Factory," BABAR Note # 145 (1994).
C. Walter and B. Nolty, \The FARFALLA User Guide, FARFALLA; V. 1.5"
(1994).
Technical Design Report for the BABAR Detector
528
Technical Design Report for the BABAR Detector
REFERENCES
12
Interaction Region and Backgrounds
T
he PEP-II requirements of high beam currents, asymmetric energies, and head-on
collisions conspire to make machine-induced backgrounds a signi cant challenge in the
design of PEP-II and the BABAR detector. To achieve head-on collisions with asymmetric
energies and 1.26 m bunch spacing, beam line elements must be positioned very close to
the interaction point. The detector must be well protected from excessive occupancies and
radiation damage, while maintaining a large solid-angle acceptance for (4S ) decays. The
expected sources of detector backgrounds are:
synchrotron radiation photons produced in the machine magnetic elements;
lost beam particles (bremsstrahlung and Coulomb scattering o residual gas molecules);
luminosity-related backgrounds such as elastic and radiative Bhabha scattering; and
other sources such as inelastic beam-gas interactions, cosmics, muons from collimators,
etc.
These background sources can give rise to primary particles that either enter the detector
directly or generate secondary debris that ultimately reaches the detector.
This chapter describes the various background sources and the results of simulations aimed at
estimating the rates and impacts on the various detector components. Section 12.1 describes
the interaction region design and the e ect of background considerations on that design.
Also in this section is a discussion of PEP-II operating modes, including issues relating to
injection. Sections 12.2 and 12.3 describe the tools used to study the above backgrounds
and estimates rates for each. Section 12.4 describes in detail the e ect of the backgrounds
on each of the major detector components in BABAR.
530
Interaction Region and Backgrounds
Parameter
HER
LER
Units
Center-of-Mass Energy
10.580
GeV
Peak Luminosity
3:0 1033
cm,2s,1
Luminosity Lifetime
1.55
hr
Bunch Spacing
1.26
m
Circumference
2219.3
m
Beam Energy
9.000
3.109
GeV
Bunch Length
1.0
1.0
cm
Number of Bunches
1658
1658
10
Particles per Bunch
2:73 10 5:91 1010
Beam Current
0.986
2.140
Amps
Transverse Damping Time
36.8
54.0
msec
Horizontal Emittance
48.2
64.3
nm-rad
Horizontal IP Beta
50.0
37.5
cm
Horizontal IP Spot Size
155
155
m
Horizontal Tune Shift
0.03
0.03
Vertical Emittance
1.93
2.57
nm-rad
Vertical IP Beta
2.00
1.50
cm
Vertical IP Spot Size
6.2
6.2
m
Vertical Tune Shift
0.03
0.03
Table 12-1.
12.1
PEP-II parameters.
PEP-II Design
12.1.1 Parameters
The PEP-II project is an e+ e, colliding beam storage ring complex designed to produce a
luminosity of at least 3 1033 cm,2s,1 at Ecm = 10:58 GeV, equivalent to the mass of the
(4S ) resonance, with unequal energy beams of 3.1 and 9.0 GeV. The parameters of the
accelerator are listed in Table 12-1. Both the high luminosity and the energy asymmetry
require signi cant advances in storage ring design and construction. A full description of the
PEP-II project may be found in the PEP-II Conceptual Design Report [PEP93].
Technical Design Report for the BABAR Detector
12.1 PEP-II Design
531
30
1
20
Q5
Q2
3.
Q4
G
eV
B1
B1
Q1
Q1
10
x (cm)
9 GeV
0
9 GeV
–10
Q1
Q1
B1
B1
3.
1
Q4
–20
Q2
G
eV
Q5
–30
–7.5
–5.0
–2.5
0
2.5
5.0
7.5
z (m)
Figure 12-1. Plan view of the interaction region. The low-energy beam enters from the
lower right and exits from the upper left. The high-energy beam enters from the left and
exits on the right. The vertical scale is highly exaggerated. The dashed lines represent the
beam stay-clear envelopes, and the 300 mr detector acceptance cuto .
12.1.2 Interaction Region Components
The high luminosity of PEP-II is achieved through high beam currents and strong focusing.
The high current must be divided into a large number of low-charge bunches to avoid the
beam-beam tune-shift limit, and the beams must collide only at the IP to avoid additional
tune shift. PEP-II collides the beams head on but separates the beams horizontally before
the next collision at 62 cm from the interaction point (IP), utilizing the energy asymmetry
and separation dipoles (B1) within 20 cm of the IP. Strong focusing requires quadrupoles
located close to the IP. The nal quadrupole Q1 is common to both beams and starts at
90 cm from the IP, partially penetrating the central detector. The low-energy beam (LEB)
is o -axis in Q1 to maximize the beam separation. The next quadrupole Q2 focuses only the
LEB, with the high-energy beam (HEB) passing through a eld-free region. The following
pair of quadrupoles, Q4 and Q5, focus only the HEB, with the LEB passing through a
eld-free region. The layout of the PEP-II interaction region is shown in Figure 12-1. It is
described in more detail in the PEP-II Conceptual Design Report [PEP93, pages 55{58 and
66{74].
Technical Design Report for the BABAR Detector
532
Interaction Region and Backgrounds
The B1 and Q1 magnets are inside the 1.5 T detector solenoid eld. Conventional iron
magnets do not function in this environment, so the only possibilities are superconducting
and permanent magnets. There is not enough room between the beam pipe and the 300 mr
detector acceptance envelope at the location of B1 to allow a cryostat, so the only option
is a permanent magnet. In the PEP-II baseline design, Q1 is also a permanent magnet. (A
superconducting Q1 design is presently under consideration. It would not change the IR
optics in a fundamental way, nor would it change the luminosity, but it would allow greater
operational exibility at a signi cantly higher cost.)
B1 and Q1 are constructed of rings of SmCo permanent magnet material. Each ring is
constructed of blocks with their magnetic moments normal to the beam direction and varying
in azimuth around the beam in the manner described by Halbach [Hal81]. Permanent
quadrupole magnets very similar to Q1 have been used at the Cornell Electron Storage
Ring for many years [Her87a, Her87b].
The beam pipe at the interaction point is a double-wall structure, with an inner tube 50 mm
in diameter and 800 m thick, separated by 2 mm from an outer tube of 400 m thickness,
with helium gas owing between the tubes for cooling. It is made of beryllium to minimize
multiple scattering but is coated inside with 10 microns of gold to attenuate synchrotron
radiation, totaling 0.6% of a radiation length. (The background studies have primarily
assumed 25 m of copper rather than 10 m of gold.)
There are water-cooled masks inside the B1 magnet to prevent synchrotron radiation from
striking the beam pipe. There are additional masks to protect the septum region of the Q2
quads from synchrotron radiation. The IP beam pipe and synchrotron radiation masks are
described in the PEP-II Conceptual Design Report [PEP93, pages 357{371].
The two Q1 magnets, two B1 magnets, IP beam pipe, and vertex detector are assembled into
a single rigid support barrel of 43 cm outside diameter, shown in Figure 12-2. The designs
for B1, Q1, Q2, and the support tube are described in Reference [PEP93, pages 276{301].
The support barrel is assembled and internally aligned outside the detector and installed by
sliding it through the drift chamber. The two end sections of the support barrel are made
from stainless steel and carry the Q1 magnet assemblies. The middle barrel section is a
carbon- ber composite equivalent to 0.005X0 in thickness. The total assembled weight is
approximately 2500 kg, concentrated in the end sections.
12.1.3 Background Implications
The bending near the IP required to separate the beams generates a large synchrotron
radiation ux that is not present in more conventional e+e, colliders. The bending also
sweeps o -energy beam particles into the detector. There are also many more bunch crossings
Technical Design Report for the BABAR Detector
12.1 PEP-II Design
533
,
,
,
,
,
,
,
,,,,,,,,,
,,,,,,,,,,,
,,,,,,,,,
,,,,
,
,,,,,,,,,,
,,,,,,,,,
,,,,
,,,,,,,
Vertex
detector
IP
2.5 cm radius Be
beam pipe (cooled)
,
,
,
,,,
Carbon-fiber
support tube
0.5% Xo
B1 trim
B1 dipole
Distributed
ion pump
Q1
,
,
,
Q1 trim
Inboard and outboard LEB masks
(water cooled)
Magnet support
43 cm OD
Z5-6 3:48
7379A144
Figure 12-2. Support barrel for interaction-region components inside the detector. Only
one end is shown.
within the detector resolving time than in previous colliders. Machine-induced detector
backgrounds have been considered in the PEP-II design optimization process from the start.
The synchrotron radiation and lost-particle backgrounds have been calculated during the
accelerator design iteration process, and the design has been altered to minimize the backgrounds. For example, the Q4 quadrupole is o set from the HEB orbit so that synchrotron
radiation from Q5 that would otherwise hit the IP beam pipe is directed into a mask.
The B1 magnets have a dipole eld of 0.75 T and produce an intense synchrotron radiation
fan, but it is generated so close to the IP that most of it travels harmlessly through the
beam pipe and is absorbed far away. Masks have been designed to prevent other sources of
synchrotron radiation from shining directly onto the IP beam pipe. The B1 and o set Q1
magnets bend in opposite directions and are on opposite sides of the IP, which allows the
masks to be on only one side of the beam pipe minimizing backscattering problems. The
synchrotron radiation from the LEB has a low critical energy, so it is easily absorbed by
the copper beam pipe. The physical aperture of the machine vacuum chamber has been
designed to increase relative to the local beam size when traveling from the machine arcs to
the IP. Adjustable beam collimators will absorb distant and local beam-gas bremsstrahlung
and Coulomb scattering. Special attention has been paid to the vacuum system near the IP
to minimize beam-gas scattering. The B1 and Q1 magnets serve as e ective shielding for
lost particles.
Technical Design Report for the BABAR Detector
534
12.1.4
Interaction Region and Backgrounds
Operating Modes
Most background calculations are based on stable colliding beams. The dominant beam loss
mechanism is then beam-gas scattering distributed around the rings, although beam-beam
bremsstrahlung at the IP can become comparable at high luminosity. The high dispersion
in the bending arcs prevents damaging products of distant beam-gas bremsstrahlung from
reaching the detector. Beam-gas bremsstrahlung in the upstream part of the IR straight
section is stopped by collimators, but bremsstrahlung close to the detector produces irreducible backgrounds. Beam-gas coulomb scattering far from the detector is also absorbed
by collimators and in general is less of a problem than bremsstrahlung. Most synchrotron
radiation is directed away from the detector, and the remaining radiation is largely absorbed
by masks. There is an irreducible background from scattering o the tips of the masks.
On the order of once per hour, it will be necessary to inject into PEP-II. The injection process
and components are described in Reference [PEP93, pages 483{560]. Normal injection will
be in top-o mode, i.e., charge will be injected without dumping the stored beam. This
mode and re-injecting after a total beam loss are projected to require on the order of six
minutes to complete. The beams will be injected at collision energy into collision optics, with
no necessity to ramp or squeeze the beams. The beam emittance provided by the injector
(the Stanford Linear Collider with some modi cations) will be smaller than the equilibrium
emittance of PEP-II. Therefore, injection will be possible with all collimators at the settings
assumed for background calculations.
A beam particle damps from its injected orbit to the stored orbit in a small fraction of a
second, compared to the expected lifetime of many thousands of seconds. Thus, any radiation
dose from the capture process is expected to be much less important than radiation from
the normal stored condition. There is no known reason for the injection process not to be
highly ecient, or for injection losses to produce larger detector radiation doses than losses
from the stored beam. Since injection will take place when a few tens of percent of the
stored beam has been lost, the dose from injection will not dominate unless the injection
ineciency is worse than tens of percent. There is a sucient safety margin in the radiation
damage thresholds that, if injection resulted in a dose comparable to normal running, this
would be acceptable.
Signals from the detector or local radiation monitors will be used as inputs to the PEP-II
beam-dumping system. If a rapid increase in radiation is observed, the stored beam will
be dumped into a distant absorber. Such systems are in use already by the LEP detectors.
The injection process will also be interrupted if radiation is observed to be abnormally high.
It is assumed that particularly sensitive detector components like the drift chamber will be
ramped down during injection to minimize radiation damage, and injection will be dependent
on these components being ready.
Technical Design Report for the BABAR Detector
12.2 Tools
12.2
535
Tools
12.2.1 QSRAD
The program QSRAD was built to study synchrotron radiation generated by the beam as
it traveled through the last two focusing magnets prior to the collision point in PEP-I. The
program traces weighted rays from a Gaussian beam pro le through the speci ed magnetic
optics and produces a geometric fan of synchrotron radiation with equal power density and
constant critical energy for each magnetic element. These fans are then traced, and a tally
is made of the fraction of each fan that strikes de ned apertures. Apertures are speci ed in
planes transverse to the beam. An aperture can be a circle, an ellipse, a horizontal slit, or
a vertical slit. The distribution of photon critical energies is accumulated for each aperture
and converted to a photon energy distribution. In addition, the beam rays that strike each
aperture are recorded, and a table of
locations is produced, indicating the origin in the
beam pro le for those particles.
An enhanced version of QSRAD (called SYNC BKG) is used to calculate Factory synchrotron radiation rates. The following improvements to SYNC BKG have been implemented: a non-Gaussian beam tail distribution can be added to the beam pro le. The
beam tail distribution is modeled as a lower atter Gaussian beam pro le; tilted and o set
quadrupoles are allowed; horizontal and vertical bend magnets are recognized as magnetic
elements; and individual magnetic elements can be subdivided to more accurately simulate
long quadrupoles.
The information from SYNC BKG on photon rates striking various surfaces near the IP is
used in a second stage to study backscattering and tip scattering from the mask surfaces.
EGS4 [Nel85] (which includes Rayleigh scattering and K-shell uorescence) is used to calculate the number and energy of photons that scatter from a mask tip, penetrate the detector
beam pipe, and are absorbed in various layers of the silicon vertex detector.
x; y
B
12.2.2 Decay TURTLE
Decay TURTLE [Car91] is a program for tracking particles and decay secondaries through
beam line components. A version of this package has been modi ed to do bremsstrahlung
and Coulomb scattering. A scattering point is picked randomly along the beam line, a
beam particle transported to that point, and the scattered secondary propagated down the
remaining beam line until it hits an aperture or the end of the beam line. For bremsstrahlung
scattering the photon is also propagated. Scoring planes are used to simulate collimators,
masks, and other apertures. Selected particles, particularly ones hitting an aperture near
Technical Design Report for the BABAR Detector
536
Interaction Region and Backgrounds
the IP, can be written to a disk le for later input to the GEANT or OBJEGS detector
simulations described below.
In the studies presented here, the PEP-II machine lattice is simulated from the middle
of the preceding arc to several meters downstream of the IP (185 m). Placement of
collimators in the interaction region reduces the rate of upstream scatters reaching the IP. A
graded aperture is employed in this region of the machine lattice (60 m from the IP), i.e.,
progressively larger apertures are used as one approaches the IP. Multiturn e ects have not
yet been studied. Optimization and design of upstream collimation is in progress subject
to space constraints and the avoidance of signi cant synchrotron power dissipation on the
collimator.
12.2.3 OBJEGS
OBJEGS [Hea91] interfaces EGS4 [Nel85] to simple cylindrical geometries. The user speci es
beam line and detector components in terms of the inner and outer radius, lower and upper
longitudinal limits, and material type. Files of lost particles from TURTLE or other sources
of background such as radiative Bhabhas can be input from a le one particle at a time.
Integration over time must be done as a separate step operating on the output of OBJEGS.
This can be as simple as adding up total energy deposits to obtain the radiation dose or as
complex as combining hits produced by multiple rays to make events for a trigger simulation.
The restriction to cylindrical geometries in OBJEGS makes it impossible to describe the
BABAR detector perfectly, especially as the detector design has grown more sophisticated and
realistic with features such as conical drift chamber end plates and variable crystal lengths.
The biggest problem for OBJEGS geometries is the elliptical shape of the synchrotron
radiation mask; to approximate this, two versions of OBJEGS geometries are used|one
that reproduces the longitudinal pro le of the low-energy beam synchrotron mask, and one
that reproduces the high-energy synchrotron mask. The geometry corresponding to the side
of the detector which is struck by a particle is used for the simulation of the shower produced
by that particle. Due to these complications, BABAR has been moving toward a GEANTbased simulation (Section 12.2.4) as its primary detector simulation tool. A number of the
studies in this document were performed using OBJEGS, but will be redone with GEANT
as time permits.
12.2.4 GEANT
An extensive GEANT [GEA94] simulation of the BABAR detector, called BBSIM, is currently
the primary tool for detector performance studies. As part of this implementation, there is
Technical Design Report for the BABAR Detector
12.2 Tools
537
Drift Chamber
Support Tube
Q1
Si
SR Mask
Q2
Septum
5 cm
B1
50 cm
Figure 12-3. GEANT simulation of beam line components. The interaction of a LEB
lost beam particle is also shown (with a high shower cut-o energy of 10 MeV).
a rather detailed simulation of the beam line components. Figure 12-3 shows an expanded
view of this central part of the BABAR detector with a lost beam particle generated with the
TURTLE package described above.
The simulation includes an approximation to the synchrotron radiation masks, which have
an elliptical aperture, and the B1 and Q1 beam line magnet elements. The dipole and
quadrupole elds of the B1 and Q1 (both SmCo permanent magnets), respectively, are also
included. At the ends of the beam pipe are the Q2 septum masks that shield the Q2 coil.
Files of lost particles or radiative Bhabhas can be input to the simulation, either one particle
scatter at a time or integrated over some detector time window and superimposed on a physics
event.
Technical Design Report for the BABAR Detector
538
Interaction Region and Backgrounds
12.2.5 GELHAD
GELHAD [Sny94] is a program for simulating the electro-production of hadrons via N
interactions. The interactions of electrons with nuclei can be added using a virtual photon
approximation; however, hadronic interactions in electromagnetic showers are dominated by
real N interactions.
GELHAD is based on subroutines from the hadronic shower simulation GHEISHA [Fes85].
Two basic models are available: (1) the photon is absorbed by one nucleon that is then
forced to interact with the residual nucleus; and (2) the photon is replaced by a pion of
the same energy that interacts with the nucleus. In both models, energy conservation and
charge conservation are respected, but in general, momentum conservation is violated. The
default used in most of the studies for this document uses model (1) below E = 150 MeV
and model (2) above.
GELHAD has been implemented both in OBJEGS and GEANT. Trigger studies with the
OBJEGS version use the supplementary program TRACK [Sny94] to simulate the behavior
of hadrons in OBJEGS geometries. GEANT-based studies have, of course, the full power of
GEANT and the detailed BBSIM geometry available.
12.3
Background Sources
12.3.1 Synchrotron Radiation
There are several sources of synchrotron radiation backgrounds:
direct synchrotron radiation;
photons that scatter o a mask tip;
synchrotron radiation from elements far upstream of the interaction point; and
sources of backscattered photons from downstream surfaces.
Separating the unequal-energy beams by the use of bending magnets and o set quadrupoles
generates several fans of synchrotron radiation. The geometry of the interaction region
optics, however, is designed to minimize the amount of synchrotron radiation that strikes
nearby surfaces. In particular, the S-bend geometry of the beam lines allows most of the
synchrotron radiation generated by magnetic elements upstream of the interaction region
Technical Design Report for the BABAR Detector
12.3 Background Sources
539
to pass through the detector region without hitting local surfaces. Primary masks near the
collision point are used to prevent direct radiation from hitting the detector beam pipe, while
at the same time keeping the number of photons that strike the mask tips to an acceptable
level. The synchrotron radiation fans from both beams are shown in Figure 12-4.
The LEB generates synchrotron radiation fans as it passes through the Q1 and B1 magnets on
its way to the collision point. The LEB mask is designed to prevent the synchrotron radiation
(either fan or quadrupole) generated by the upstream magnets from directly striking the
detector beam pipe. The surfaces of the LEB mask are sloped such that scattered photons
cannot travel directly to the detector beam pipe. The LEB mask absorbs about 3.6 kW of
synchrotron radiation power. Fans generated by the LEB in the two B1 magnets and in the
downstream Q1 magnet pass through the interaction region without striking nearby surfaces.
The rst surface that intercepts the B1 fans is the septum mask in front of the Q2 septum
quadrupole, located 2.8 m from the IP. This septum mask absorbs 3.5 kW of power.
Synchrotron radiation fans generated by the HEB in upstream bend magnets and o set
quadrupole magnets deposit nearly 1 kW of power on the HEB mask. These upstream HEB
elements are positioned to ensure that radiation generated by beam particles even 10 o axis in these elements does not strike the LEB mask. The fans of radiation generated by
the HEB as it passes through the two B1 magnets do not strike any surfaces in the detector
region and are absorbed in a dump about 17 m downstream of the detector.
The geometry of the masking near the IP is such that no synchrotron radiation can hit the IP
beam pipe directly, nor can it scatter o the face of a mask onto the IP beam pipe. Scattering
from the tips of the LEB and HEB masks is the dominant source of photons striking the
detector beam pipe. This process has been simulated in great detail. The individual photons
that strike the LEB and HEB mask tips are generated from an energy spectrum given by
a program that traces the photons from sources to surfaces of interest. The photons that
end up striking the detector beam pipe are followed through the detector components by
the EGS Monte Carlo program [Nel85]. Further details of this simulation package can be
found in Reference [PEP93, pages 108{112]. The results of the simulations are the photon
spectra shown in Figure 12-5; backgrounds resulting from synchrotron radiation are shown
in Table 12-2.
Other sources of synchrotron radiation, such as magnets farther upstream, and backscatter
from the Q2 septum masks, produce negligible contributions to the detector background
compared to primary mask tip scattering. While a high photon ux strikes the dump for
HEB radiation from the B1 magnets, the small solid angle of the detector beam pipe as seen
from this source, the photon's small angle of incidence on the beam pipe, and the fact that
most of the beam pipe is shielded by the LEB mask, together reduce the background from
the 17 m dump to a low level.
Technical Design Report for the BABAR Detector
540
Interaction Region and Backgrounds
20
300 mrad
Q2
Q1
10
Centimeters
(a)
Q1
B1
B1
B1
B1
0
–10
Q1
Q1
Q2
–20
20
300 mrad
Q2
Q1
10
Centimeters
(b)
Q1
B1
B1
B1
B1
0
–10
Q1
Q1
–20
–3
–2
Q2
–1
0
1
2
3
Meters
Synchrotron radiation fans from the low-energy (a) and high-energy (b)
beams. The density of shading gives an indication of the relative photon intensity from the
various radiation fans.
Figure 12-4.
Technical Design Report for the BABAR Detector
12.3 Background Sources
Photons per keV per crossing
10
Photons from LEB incident on
1. LEB mask tip
2. Detector beam pipe
3. First layer of Si
4. Second layer of Si
5. Third layer of Si
6. Drift chamber
Photons from HEB incident on
1. HEB mask tip
2. Detector beam pipe
3. First layer of Si
4. Second layer of Si
5. Third layer of Si
6. Drift chamber
3
1
10
541
0
1
2
–3
2
10
3
–6
10
3
10–3
4
5
–6
10
4
6
5
10–9
6
0
50
100
Photon energy (keV)
150
50
Photon energy (keV)
100
Photon spectra from synchrotron radiation. Left plot is for high-energy
beam; right plot is for low-energy beam.
Figure 12-5.
The e ect of magnet misalignments on detector backgrounds has also been evaluated. Of
the 24 di erent cases studied, two cases, a 5 mm displacement in x of Q1, Q4, and Q5,
produced a threefold increase in the background rate. The rest of the misalignment checks
produced small (< 50%) increases in backgrounds, with some con gurations producing rates
that are actually below the nominal background rate. None of these misalignment checks
resulted in synchrotron radiation photons directly striking the detector beam pipe.
12.3.2
Lost Particles
Bremsstrahlung and Coulomb scattering of beam particles from residual gas molecules in
the beam pipe can result in high-energy beam particles and photons striking masks and the
beam pipe near the IP. The resulting electromagnetic showers can cause excessive detector
Technical Design Report for the BABAR Detector
542
Interaction Region and Backgrounds
4 < E < 100 keV
# of Photons
Energy ( keV)
4 < E < 20 keV
# of Photons
Energy ( keV)
Detector Limits
# of Photons
Energy ( keV)
Incident Absorbed Absorbed Absorbed
on Be in First in Second in Third
Pipe Si Layer Si Layer Si Layer
Incident
on Drift
Chamber
1.24
17.6
0.028
0.43
6:9 10,4 5:6 10,4 2:4 10,5
2:2 10,2 2:0 10,2 3:2 10,3
0.96
9.34
0.022
0.17
3:8 10,5 8:6 10,6 2:1 10,7
9:2 10,4 1:3 10,4 3:7 10,6
|
|
2.3
95
> 2: 3
> 95
> 2:3
> 95
125
5000
Table 12-2. Synchrotron radiation background simulation results. The numbers are for
each crossing. Multiply by 238 to get photons per s. In the simulation, the beam pipe
consists of 25 m of Cu and 1 mm of Be.
occupancy and/or radiation damage. Hadronic showers from photoproduction can also occur
in these showers and are discussed in Section 12.3.3.
Section 12.2.2 describes Decay TURTLE, the tool used to establish the rate of particles
incident on the masks and beam pipe near the IP. The rate is reduced signi cantly with
upstream collimation. Recent studies assume a uniform beam line pressure of 1 nTorr (N2
equivalent) to determine regions of sensitivity to pressure. Additional pumping in these
regions can then be employed to further reduce this background. Figure 12-6(a) shows the
location of the beam particle scatter for particles that end up striking near the IP. Rates are
quoted per s, a typical integration time for detector elements. Since higher energy particles
deposit more energy in the detector, Figure 12-6(b) shows the same distribution weighted
with the energy of the scattered particle.
Studies are in progress to determine a realistic estimate of the IR vacuum pressure pro le.
Preliminary results suggest that, on average, the assumed uniform 1 nTorr distribution is a
reasonable baseline for these studies. It appears possible to reach pressures below 1 nTorr
(0.2{0.5 nTorr) in the regions from 3 to 40 meters from the IP in both the incoming HEB and
LEB beam lines. Due to space constraints at the IP that preclude placement of pumps very
close to the IP and the high heat load on the synchrotron radiation masks, the pressure within
three meters of the IP will likely exceed 1 nTorr. Scatters very close to the IP (1 m) will
tend to make it out of the detector region before striking an aperture. Average background
rates for these preliminary pressure pro les tend to be comparable to the rates found with
the uniform 1 nTorr assumption, so this is taken to be the nominal level. Assuming this
baseline, the HEB deposits 12.9 GeV per s from bremsstrahlung scattering and 1.0 GeV
Technical Design Report for the BABAR Detector
12.3 Background Sources
number / µsec/ 2 m
1.4
543
HEB
1.2
LEB
(a)
1
0.8
0.6
0.4
0.2
Energy (GeV) / µsec / 2 m
0
-80
-60
-40
-20
0
20
z scatter point (m from IP)
40
60
80
1.4
1.2
(b)
1
0.8
0.6
0.4
0.2
0
-80
-60
-40
-20
0
20
40
60
80
z scatter point weighted with particle energy (m from IP)
Distribution of the scatter point of beam particles that hit near the IP
(a) unweighted and (b) weighted with the energy of the particle. A uniform 1 nTorr vacuum
pressure is assumed.
Figure 12-6.
from Coulomb scattering within 1:5 m of the IP. The LEB contribution is 9.2 GeV from
bremsstrahlung scattering and 1.4 GeV from Coulomb scattering.
GEANT is used in the detector simulation package BBSIM to evaluate the impact of this
energy deposition near the IP. Secondaries from electromagnetic showers in the beam line
components are followed into the detector elements. The showers are cut o for electron and
photon kinetic energies below 50 keV. Figure 12-7 shows the resulting energy, azimuthal,
and z distributions for secondary electrons and photons incident on Layer 1 of the silicon
vertex detector. The peak at 0.5 GeV in the energy distribution is due to e+e, annihilation.
The strong peaks in the distribution at 0 and 180 are from bremsstrahlung scattering in
which the B1 dipoles bend the energy-degraded beam particles horizontally into the beam
pipe. The other peaks in this distribution are from double counting the ux in regions of
overlapping silicon detectors. Figure 12-8 shows the same distributions for Layer 1 of the
central drift chamber. Since few photons interact in the drift chamber, the ux into the
particle identi cation system and CsI calorimeter is nearly identical to the ux entering
Technical Design Report for the BABAR Detector
Interaction Region and Backgrounds
10
# of particles / µsec / .02 MeV
# of particles / µsec / .2 MeV
544
1
10
-1
0
1
10
2
4
6
8
-1
10
0
0.2
0.4
0.6
0.8
1
E (MeV)
E (MeV)
3
# of particles / µsec / 4
o
# of particles / µsec / cm
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-100
0
φ (degrees)
100
2.5
2
1.5
1
0.5
0
-20
-10
0
10
20
z (cm)
Figure 12-7. Energy, azimuthal, and z distributions of photons (solid) and electrons
(dashed) incident on Layer 1 of the silicon detector for lost beam particle backgrounds.
Rates correspond to the number of incident particles per s.
the drift chamber. Figure 12-9 shows the distribution for the front face of the calorimeter,
including the polar angle distribution of photons and energy ow.
12.3.3
Hadrons
Estimates of trigger rates from photoproduced hadrons are given in Chapter 10. In this
section, we seek only to check (using a simpli ed trigger simulation) whether or not the
triggers are produced by photon interactions that could be expected to be well-modeled by
GELHAD.
Technical Design Report for the BABAR Detector
545
# of particles / µsec / .02 MeV
# of particles / µsec / .2 MeV
12.3 Background Sources
10 2
10
1
10
-1
0
2
4
6
E (MeV)
8
10
0
0.2
0.4
0.6
0.8
1
35
# of particles / µsec / 5cm
o
# of particles / µsec / 4
1
E (MeV)
10
9
30
8
25
7
6
20
5
15
4
3
10
2
1
0
10
-100
0
φ (degrees)
100
5
0
-100
0
z (cm)
100
200
Figure 12-8. Energy, azimuthal, and z distributions of photons incident on Layer 1 of the
central drift chamber for lost beam particle backgrounds. The electron rates are negligible.
Rates correspond to the number of incident particles per s.
Figure 12-10 shows the distribution of the energy of photons in electromagnetic showers that
interact hadronically; only those photons that produced triggers are modeled. The GEANT
version of GELHAD has been used to make this distribution. The trigger in this case is a
one and one-half track trigger in which at least one track reached the outermost layer of the
drift chamber and the other track reached at least the middle layer of the chamber. Note
that the spectrum is rather hard, with an average energy of 2400 MeV. This gives us some
con dence that our model, which at least conserves charge and energy, may do a reasonable
job in the relevant region.
The process of comparing GELHAD with data from the PEGASUS [Deg94] test run using
the TPC detector at PEP has just begun and will be vigorously pursued in the next few
Technical Design Report for the BABAR Detector
10 2
10
1
10
-1
0
2
4
6
8
E (MeV)
4
2
-100
0
φ (degrees)
o
8
0.2
0.4
0.6
E (MeV)
0.8
1
8
6
4
2
0
0
50
100
150
0
50
100
150
Θ (degrees)
15
Energy (MeV) / µsec / 2
o
Energy (MeV) / µsec / 4
0
100
10
10
6
4
2
0
1
o
6
0
10
# of particles / µsec / 2
# of particles / µsec / 4
o
8
10
# of particles / µsec / .02 MeV
Interaction Region and Backgrounds
# of particles / µsec / .2 MeV
546
-100
0
φ (degrees)
100
5
0
Θ (degrees)
Energy, azimuthal, and polar angle distributions of photons incident on the
front face of the calorimeter. Also included are the angular distributions of energy ow.
Rates correspond to number of incident particles and energy per s.
Figure 12-9.
Technical Design Report for the BABAR Detector
12.3 Background Sources
547
25
Number/200 MeV
20
15
10
5
0
0
2
4
6
8
10
Energy (GeV)
Figure 12-10. The energy distribution of photons in electromagnetic showers which
interacted hadronically, causing a charged track trigger.
months. We are also actively investigating alternative models as a check of the current
results.
At this stage in its development, it is dicult to evaluate the errors on GELHAD results.
Using the model in which the photon is absorbed by a nucleus (model 1) at all energies
instead of the model in which the photon is replaced by a pion (model 2) above 150 MeV
results in a factor of 3 drop in the drift chamber trigger rate, while the calorimeter trigger
rate remains virtually unchanged. The drop in the drift chamber trigger rate can be traced
to a substantially lower pion multiplicity in model 1 compared to model 2. Only a program
of comparison with data can establish the uncertainties and lead to an improved model.
12.3.4 Luminosity Backgrounds
Radiative Bhabhas
The radiative Bhabha process, e+e, ! e+e, , is a potentially large background because of
its relatively large cross section. Outgoing degraded particles from the IP in each beam are
Technical Design Report for the BABAR Detector
548
Interaction Region and Backgrounds
σ (mb), Eγ/Ebeam > Eγmin /Ebeam
Ecm = 10.58 GeV
102
region of
101
concern
100
0
0.2
0.4
0.6
0.8
1
Eγmin /Ebeam
Figure 12-11.
energy.
Radiative Bhabha integral cross section versus minimum fractional photon
swept into the beam pipe by B1 and Q1. Figure 12-11 shows the radiative Bhabha cross
section versus the minimum energy of the radiated photon that caused the degraded particle
to hit the beam pipe, normalized to the beam energy.
To study the contribution of radiative Bhabhas to occupancy and radiation damage, we
have used a Monte Carlo event generator called BBBREM, developed for LEP by Kleiss and
Burkhardt [Kle94], and modi ed for an asymmetric collider. Individual events generated at
the IP by BBBREM were input to OBJEGS and tracked until they hit beam line components
and showered. Figure 12-12 shows the energy deposited per sec in the beam pipe and other
components (the peak at 240 cm is from radiative Bhabhas bent by B1 into the Q2 septum).
For particles that hit the beam pipe between the IP and the end of Q1 (i.e., those less than
2.3 GeV), the rate from the LEB is about 5 times higher than that from the HEB due to the
di erence in cross section. However, in the HEB, there is an additional source from degraded
electrons in the energy range 5.0{6.5 GeV that impinges on the Q2 septum at 2.4 m from the
IP.
Technical Design Report for the BABAR Detector
12.3 Background Sources
549
GeV / µsec / 10cm
10 2
10
1
10
10
-1
-2
-400
Figure 12-12.
point (in cm).
-300
-200
-100
0
100
200
z position of interaction (cm from IP)
300
400
Energy deposition from radiative Bhabhas versus distance from interaction
Within 2 5 m of the IP, the power deposited by radiative Bhabhas is approximately two
orders of magnitude more than the power from lost beam particles. Fortunately, the higher
energy particles are hitting the beam pipe farther from the IP, and the resultant electromagnetic showers are mostly directed away from the vertex detector and central drift chamber.
To determine the e ect of radiative Bhabhas on the detector, they were included in GEANT
with lost particles from beam-gas bremsstrahlung and Coulomb scattering. With the addition of high- shielding around Q1 and the magnetic shielding around Q2, the rate of
photons into the detector was found to be small compared to the lost particle contribution.
:
Z
Elastic Bhabha Scattering
The cross section for elastic Bhabha scattering with angles large enough to cause the scattered
particles to hit the beam pipe within Q1 is about four orders of magnitude smaller than the
corresponding radiative Bhabha cross section. In terms of power lost within 2 5 m of the
IP, elastic Bhabhas are at least two orders of magnitude lower than both radiative Bhabhas
and lost particles. Therefore, we consider background from this process negligible.
:
Technical Design Report for the BABAR Detector
550
Interaction Region and Backgrounds
Two Photon Reactions
We neglect this process, in which the e+ and e, both radiate in the same interaction, since
its cross section is some 1000 times smaller than the radiative Bhabha process [Bai81], and
the presence of two degraded beam particles arising from the same beam crossing does not
give rise to any special background condition.
Landau-Lifshitz Pairs
In this process, an e+e, pair is produced at the IP, and a low-energy particle with fairly
large transverse momentum may hit the Be beam pipe [Che91]. The magnetic shielding
from the 1.5 T solenoid greatly reduces this background contribution to the point where it
is negligible.
12.3.5
Other Sources
Beam-Gas Interactions at the IP
Electrons and positrons in the cores of the beams scatter o molecules inside the beam pipe
near the interaction point, producing hadrons. This process does not include the more severe
backgrounds due to lost-particle hadronic interactions. Rates of triggers were estimated
using the electro-production code EPC [Lig88] interfaced to the detector GEANT simulation
BBSIM.
The EPC generator program models the double-di erential nucleon production cross section
due to electrons scattering o thin targets of nuclei. The accuracy of the program is about
20% [Lig88]. Only single-particle cross sections are produced by the program. The EPC
program was used to make tables of double-di erential cross sections for each type of nalstate hadron for beam electrons and positrons incident on a thin target of nitrogen molecules.
The single-particle nal states p, n, 0 , +, and ,, were generated and tracked by GEANT
through the detector. The simulation was lacking in that two-particle states were not
generated. The main contribution to true two-track nal states is expected to be from
! p, which is the sole source of pions in the EPC program. To obtain a high upper
limit of the two-track drift-chamber Level 1 trigger rate due to decay, it was assumed
that all protons that give rise to a track were accompanied by an A-track pion. (An A track
is de ned as a particle track which passes through all layers of the drift chamber, while a
B track passes through only the inner half of the chamber.) That is, the upper limit to
Technical Design Report for the BABAR Detector
12.3 Background Sources
551
the two-track trigger rate was assumed to be the proton-induced B-track rate (including A
tracks). A similar upper limit was estimated for the calorimetric trigger.
The estimated trigger rates due to beam-gas interactions near the interaction point [Kra94b]
are given in Chapter 10. The upper bound for the total Level 1 rate was found to be
comparable to the physics rate for a vacuum of 10 nTorr and thus is tolerable, with a large
safety margin. The trigger rate due to beam-gas events would be signi cantly lowered by a
trigger level that selects the track z position to be near the interaction point, such as a trigger
using the vertex detector. The accuracy of the beam-gas trigger-rate calculation could be
improved through the use of a program that provides nal-state particle correlations.
Cosmic Rays
Cosmic rays are caused by high-energy particles interacting with the atmosphere, producing
pions and kaons that decay into muons, and producing showers. At ground level, the majority
of penetrating particles are muons.
Cosmic rays are important to BABAR in two ways: they cause about 500 Hz of Level 1 trigger
rate when using a trigger line containing only a loose calorimeter de nition, and they can
be used to calibrate elements of the detector.
The cosmic-ray muon generator HEMICOSM [Bri94] was developed for study of trigger
rates and for future study of the detailed ux incident on various detector components.
The generator HEMICOSM was interfaced to BBSIM and used to study detector response.
The program uses an accurate theoretical ux prediction that varies as a function of both
momentum p and zenith angle [Dar84]. The ratio of + to , was taken to be 1.25 [All84,
PDG92].
There are reasonably accurate measurements of the ux incident on a unit horizontal area
at sea level for hard muons, where a hard muon is de ned to be one that penetrates 15 cm
of lead, corresponding to a momentum slightly over 0.3 GeV=c. The measured value for the
hard component of the ux, 0:013 cm,2 sec,1 [PDG92], was used to calculate the incident
muon rate.
The calculated trigger rates due to cosmic-ray muons [Kra94a] are given in Chapter 10. The
rates for the theoretical momentum spectrum were found to agree to within 5% of those
from a measured momentum spectrum [All84].
Muons from Collimators
There is a potential background from muons produced by lost beam particles on either side
of the detector near the IP. For a single beam lifetime of 1.5 hours, the HEB(LEB) beam
Technical Design Report for the BABAR Detector
552
Interaction Region and Backgrounds
loss rate is 3:5 103=s (7:7 103=s). Assuming this loss is due to beam-gas scattering and
is uniform around the ring, the fraction lost by each beam within the 60 m straight section
on either side of the IP is 60/2200 = 0.027. For muon momenta above 1 GeV=c, the number
of muons/electron [Tsa74] from full energy lost particles on copper in the HEB(LEB) is
5.010,6 (2:9 10,7), giving muon rates of 4:8 10,4 s and 0:6 10,4 s from the HEB
and LEB, respectively. Making the conservative assumption that all the produced muons
reach the detector, the combined rate is about 600 Hz.
Tunnel Shine
The detectors at PEP-I had shielding walls, ranging from one-fourth of an inch Pb to eight
inches of steel rods, separating the IR hall from the tunnels on either side, to protect against
various radiations present in the tunnels. No doubt BABAR will require similar protection.
The forward end, facing the proton alcove, is fairly well-protected by the steel plates of the
magnet endcap. The DIRC water tank on the other end may be more sensitive to radiation.
12.4
Background Rates and Detector Responses
This section describes the e ects of machine-induced backgrounds on the various components
of the BABAR detector, speci cally the vertex detector, drift chamber, particle identi cation
system, calorimeter, and muon detector. As shown below, all detector components can
function with virtually no degradation in performance at the nominal background levels for
ten years or more. It is desirable to maintain reasonably good performance and lifetime at
background levels above this nominal level for a number of reasons. First, there are inherent
uncertainties in the above predictions of background levels, both in the assumptions that
go into the generation of the source and in simulating the detector response. Second, the
accelerator running conditions may vary signi cantly, especially in the early stages when the
machine is being tuned to the optimal operating point.
For most of the signi cant sources described above, there are not large uncertainties in
the generated rates. The dominant contributor to occupancy and radiation damage e ects
is from the electromagnetic showers produced by lost beam particles hitting near the IP
(Section 12.3.2). This source also produces signi cant contributions to the background
trigger rate through the photoproduction of hadrons. As discussed in the section above
on lost particles, a major uncertainty in this process is lack of knowledge of the nal vacuum
pressure distribution in the interaction region, though preliminary designs suggest that the
uniform 1 nTorr is probably a safe assumption. The photoproduction model and its resulting
impact on trigger rates remain uncertain, and work is in progress to check them against data
and other models.
Technical Design Report for the BABAR Detector
12.4 Background Rates and Detector Responses
553
A major concern relating to machine operation is the radiation dose received during injection.
High voltage will be lowered during injection for some detector components to protect them
from harm. Components such as silicon detectors, readout electronics, and calorimeter
crystals will still be sensitive to this added dose from injection. This is currently being
studied in simulations. Experience at other e+e, machines suggests that the integrated
radiation dose received from injection can be as large as that received from normal running
conditions (i.e., the total dose should be assumed to be at least twice the nominal level).
One purpose of this section is to describe the ability of each component to handle backgrounds
above the nominal value. As the background increases, the e ects vary from reduced
eciency or high occupancy in a small region of the detector, with little impact on overall
detector performance, to actual device failures that occur at very high background levels
and result in downtime to replace components.
12.4.1
Vertex Detector
Vertex Detector Overview
The silicon vertex detector is described in detail in Chapter 4. The essential details of the
geometry and other features relevant to the background analysis will be summarized here.
The vertex detector has ve layers of sensors surrounding the beam pipe. The three inner
layers are a traditional barrel-style geometry with six detector modules per layer, each
covering 60 in azimuth. The minimum radii of the layers are located at 3.2, 4.0, and 5.2 cm,
respectively. The outer two layers are arch shaped, with tilted detectors in the forward and
backward directions to minimize the total area of silicon and to avoid small track crossing
angles. There are 16 detector modules in azimuth in Layer 4, and 18 in Layer 5. The barrel
sections of Layers 4 and 5 are located at minimum radii of 12 and 14 cm, respectively.
Double-sided silicon strip detectors are employed with strips running parallel to the beam
line and z strips oriented at 90 to the beam line. The vertex detector covers the solid angle
over the region 17:2 < < 150. The readout electronics are located outside of the active
tracking volume in a 1 cm-thick space surrounding the B1 magnets.
Vertex Detector Backgrounds
The primary source of backgrounds in the vertex detector is lost beam particle interactions
(Section 12.3.2). Synchrotron radiation is produced copiously near the IP but is carefully
shielded so that very little reaches the vertex detector.
Technical Design Report for the BABAR Detector
554
Interaction Region and Backgrounds
Beam-gas scattering produces o -energy electrons as well as photons. These lost particles
can impact the vertex detector, where they have several e ects. Radiation damage from
the ionization they produce can degrade the performance of both the silicon sensors and
the front-end readout electronics. The additional occupancy can result in ineciency for
hits from physics-related tracks and pose bandwidth problems for the DAQ system. Oine,
background hits can confuse the pattern recognition algorithms. These e ects have all been
studied, and the results are summarized here.
Radiation Damage
The BBSIM detector simulation package is used to determine the ux of ionizing radiation
through the silicon from lost beam particle showers. In units of kradper operational year, i.e.,
107 s, Layer 1 sees an average of about 33 krad/yr, Layer 2 sees 19 krad/yr and Layer 3 sees
about 6 krad/yr [Cou95]. The dose is not uniformly distributed, as shown in Figure 12-13.
There are sharp peaks in the bend planes of the machine at = 0; . This results in a
highly nonuniform dose per stave in the inner layers. The worst case is in Layer 1, Stave 1,
where the average yearly dose is 82 krad/yr, and the peak dose is 240 krad/yr over a region
covering approximately 6 in azimuth.
This radiation will result in increased leakage current. Experimental measurements of the
e ects of ionizing radiation on silicon detectors have shown an increase in the leakage current
of approximately 0.7 nA/cm2/krad for detectors similar in design to those planned for BABAR
[Joh09]. The vertex detector should withstand 10 times the yearly dose in order to have
a reasonable safety factor over the estimated two-year period between access for repairs.
There may be additional radiation damage caused during periods of non-optimal machine
conditions, such as during injection or machine tuning; such e ects have been known to
increase the integrated radiation dose by a factor of 2 over the nominal calculated level.
By designing the vertex detector to withstand 10 times the yearly dose, we account for this
additional dose, integrated over a period of two years, with some additional margin for safety.
For Layer 1, Stave 1, 10 times the average yearly dose results in an estimated increase in
leakage current of approximately 26 nA per strip in both and z. The corresponding noise
is 166 e, in equivalent noise charge (ENC), which must be added in quadrature to the ENC
of the input ampli er; the latter ranges from about 400 e, for the z strips to a little over
600 e, for the strips. The resulting increase in noise is less than 8% in both cases. In the
worst azimuthal region, the increase is about 72 nA per strip resulting in an ENC of 260 e,,
but this a ects only the strips. Adding this noise in quadrature with the noise of the input
device results in an increase of about 9% in the total ENC.
The front-end readout ICs are located just outside of the active tracking volume and are
mounted on hybrids supported from cones which surround the B1 magnets located on either
Technical Design Report for the BABAR Detector
12.4 Background Rates and Detector Responses
555
Dose (krad/yr/.1 radian)
250
200
150
100
50
0
0
Figure 12-13.
sec/0.1 radian.
1
2
3
φ (radians)
4
5
6
Dose as a function of in vertex detector Layer 1, measured in krad/107
side of the IP. Flexible circuits connect the silicon detector strips to the readout ICs, so they
are not constrained to be at the exact radius of the detectors. In the inner layers, the ICs
are located at a slightly larger radius than the detectors they service. The readout ICs for
Layer 1 are located at approximately the same radius as the detectors for Layer 2, so we can
use the Layer 2 radiation levels for the Layer 1 readout ICs. The worst case in Layer 2 is
Stave 1, which receives an average dose of 47 krad/yr with a peak value of 110 krad/yr.
The readout ICs must tolerate 10 times the maximum expected dose; this requirement
implies the use of radiation-hard circuits. For the planned CMOS technology, a radiationhard process is available from several vendors. For the baseline design, the Honeywell process
has been chosen because it is o ered with a 0.8 m feature size with three metal layers; these
options are desirable to optimize the IC layout and minimize the IC dimensions. The noise
characteristics of the Honeywell process have been studied before and after irradiation. A
slight increase in noise relative to the non-radiation-hard HP CMOS process is observed
before irradiation, on the order of 20% at 3 MHz, which is the frequency of interest for the
risetimes used in the inner layers. After irradiation to 500 krad, the noise increases by 15%
at 3 MHz; after 1.5 Mrad the increase is 30%.
Technical Design Report for the BABAR Detector
556
Interaction Region and Backgrounds
Stave
1
2
3
4
5
6
Table 12-3.
Layer 1
z
3.1%
1.1%
0.4%
1.3%
0.6%
1.1%
1.4%
0.6%
0.3%
0.6%
0.4%
0.6%
Layer 2
z
2.1%
0.8%
0.2%
0.8%
0.4%
0.7%
1.3%
0.5%
0.3%
0.5%
0.3%
0.5%
Occupancy in Layers 1 and 2 at nominal background.
Background Occupancies
The percent occupancy in a one s window versus stave number for Layers 1 and 2 is shown
in Table 12-3 [Cou95]. The distribution is peaked in a manner similar to the radiation
damage, with signi cantly higher occupancy in Staves 1 and 4.
The background occupancy is fairly high, and this has some important consequences for the
readout architecture. One notable consequence is that a data-push readout is not feasible
due to the large data transmission bandwidth that would be required; this in turn implies
a fast Level 1 trigger to enable the vertex detector readout [Lev94a]. The Level 1 trigger is
described in Chapter 10; it will have a latency of no more than 10 s and an uncertainty of
no more than 1 s.
The vertex detector readout system must be able to cope with up to 10 times the nominal
background occupancy. The safety factor of 10 is taken into account for occasional increases
in occupancy that may occur during non-optimal periods of running. This requirement is
met by providing sucient bandwidth in the data transmission system, together with enough
bu ers in the readout section of the front-end chip to even out the uctuations in rate.
The worst case is again Layer 1, Stave 1, where the occupancy at 10 times nominal background is 30% for the z strips. Each stave will have four data links, two on each end, to
independently transmit data from the and z strips, and each data link is expected to
have a transmission rate of at least 60 Mbits/s. At the maximum Level 1 trigger rate of
2 kHz, assuming 16 bits/hit plus a per-chip overhead of 16 bits, the total bandwidth out of
the Layer 1, Stave 1 z strips is about 6 Mbits/s per half detector module. The deadtime
fraction for 10 times nominal background, under these assumed conditions, is plotted versus
trigger rate in Figure 12-14 for one, two, three, and four back-end bu ers [Lev94b]. For three
bu ers at the expected maximum Level 1 trigger rate of 2 kHz, the deadtime is about 1%
and exhibits linear behavior.
Technical Design Report for the BABAR Detector
12.4 Background Rates and Detector Responses
557
Figure 12-14. Deadtime fraction for 10 times nominal background
one, two, three, and four back-end bu ers.
vs.
trigger rate for
Pattern Recognition
The oine pattern recognition capabilities of the vertex detector have been studied using
a detailed GEANT simulation, together with a Kalman lter track nding algorithm. The
study assumed that tracks which generated at least ten hits in the drift chamber would
be found by the drift chamber and extrapolated into the vertex detector. Tracks with a
p below 90 MeV=c which fail this criteria are found using the vertex detector for standalone pattern recognition. The overall pattern recognition eciency is found to be very
high, approaching 100%, for generic B decays with nominal and with 10-times-nominal
backgrounds superimposed (see Chapter 4). The fake rate, de ned as the number of found
tracks per event which have fewer than half of their hits from a real track, is seen to go
up as background levels increase. Most fake tracks have a p below 90 MeV=c and are due
to random combinations of background hits in the vertex detector alone. However, the ve
layers of the vertex detector provide a powerful pattern recognition capability, and the fake
rate can be kept to an acceptable level (fewer than 0.05 fake tracks/event) even at 10 times
the nominal background.
t
t
Technical Design Report for the BABAR Detector
558
Interaction Region and Backgrounds
12.4.2 Drift Chamber
Wire aging and chamber occupancy due to beam-related backgrounds are two issues of
concern for the operation of wire chambers. Aging refers to changes in gain or increases
in noise due to material deposited on the wires from the gas-avalanche process. Chamber
occupancy is the fraction of cells that contain random hits during the event resolving time,
i.e., the full drift time. If the occupancy exceeds a few percent, pattern recognition becomes
dicult.
The design of the BABAR drift chamber has been guided by the desire to minimize the amount
of material in the chamber, thus permitting the best possible measurement of the relatively
low momentum tracks necessary for the CP asymmetry studies. The drift chamber is 280 cm
in length and occupies the radial space between 22.5 cm and 80 cm, providing a polar angle
acceptance down to 300 mr. It is bounded by the support tube at its inner radius and the
particle identi cation device at its outer radius. The chamber contains approximately 40
layers of wires arranged in a small-cell geometry, with a total of about 7000 sense wires and
45,000 eld wires. The sense wires are 20 m-diameter gold-plated tungsten and the eld
wires are 55 m-diameter gold-plated aluminum. The chamber is lled with a helium-based
gas that, along with the choice of aluminum wire, represents a minimal number of radiation
lengths. The gas alone has a radiation length of 800 m. The 40 layers of individual sense
and eld wires correspond to 0.02% and 0.08% of a radiation length, respectively, for normal
incidence.
The chamber endplates and the inner and outer cylinder walls are also being designed using
low-mass materials. The endplates are foreseen to be made of carbon ber, corresponding to
less than 0:03X0. The inner cylinder wall's function is to provide a gas seal and will be made
of carbon ber or beryllium, representing <
0:002X0, while the outer wall, which supports
the wire load, will be of aluminum or carbon ber, giving 0.03X0.
Beam-related backgrounds which occur during beam injection are not a problem for the
drift chamber, for which the voltage can and will be turned o . The expected backgrounds
in the drift chamber have been calculated for a 1 nTorr beam vacuum. The drift chamber
occupancies for a resolving time of 1 s, which corresponds to the maximal drift time, are
shown for each layer in Figure 12-15. The distribution is peaked at the innermost layer, where
occupancies of 0.5% are expected. For the outer layers, the occupancies are 0.03%. These
results, based on a chamber with a radiation length of 800 m, do not include conversions in
or scattering by the wires. A simple method of estimating an upper limit on the e ect of
the wires on the occupancies would be to smear the wire material over the chamber volume,
resulting in a radiation length of 400 m, which would roughly double the occupancies. More
realistically, the e ect of the wires would increase the occupancies in the outer layers, while
leaving the results for the innermost layers unchanged.
Technical Design Report for the BABAR Detector
12.4 Background Rates and Detector Responses
559
0.005
Wire occupancy
0.004
0.003
0.002
0.001
0
0
5
10
15
20
25
30
Drift Chamber Layer
35
40
Figure 12-15. Expected background occupancies for each drift chamber layer due to
converted photons from showers produced by lost particle interactions.
The occupancies have been calculated without including the e ects of signal crosstalk,
expected to be 1{3%. Since most background events would produce low-momentum tracks
which spiral in z within a few cells (depending upon the wire orientations), producing very
large signal pulses, the number of neighboring cells which also contain pulses due to electronic
crosstalk could be sizable.
In general, chamber occupancies of a few percent are tolerable but render the problem of
pattern recognition and segment- nding more dicult. Occupancies which approach or
exceed 10% make track nding almost impossible.
For a uniform distribution along the wires, the energy deposition on a wire corresponds to an
integrated charge of 0.001 C/cm/107sec for the innermost drift chamber layer. The charge
deposition in the drift chamber is not uniform, showing factors of 2 variation in and z.
Taking these factors into account, the expected integrated charge for some regions of the
inner layers could be as high as 0.005 C/cm/107 s for the 1 nTorr beam vacuum.
Signi cant amounts of charge accumulated on the sense wires can lead over time to deposits
on the wires, resulting in electrical discharges and current draw. A conservative limit on
the integrated charge on a wire which can be safely tolerated is 0.1 C/cm. This limit can
be extended depending upon the ionizing gas mixture. For example, small amounts of
Technical Design Report for the BABAR Detector
560
Interaction Region and Backgrounds
Bars Tank
# of Charged Particles 16 0.7
# of Hits
32 8.0
Mean number of charged particles crossing the DIRC per s due to beam
lost secondaries.
Table 12-4.
H2 0 [Kad91] and CF4 [Ope91] in combination with ionizing gases have been shown to extend
>1.0 C/cm.
this limit to
12.4.3
Particle ID
The particle identi cation system described in Chapter 6 consists of a ring-imaging Cherenkov
system (DIRC) covering the barrel region and an aerogel threshold counter in the forward
direction. The DIRC consists of quartz bars read out with phototubes mounted on a water
tank in the backward (,z) direction.
The background level in the aerogel detector has not been studied in detail. The principal
source of background is likely to be the ux of low-energy photons from lost particle showers.
Backgrounds from conversions of these photons in the aerogel itself are expected to be small
due to its very low density. The contribution from the support structure and photodetectors
are currently under study.
Several background sources might a ect the DIRC's performance: event related backgrounds,
synchrotron radiation photons, beam-gas interactions, cosmic rays, phototube dark noise,
etc. The program BBSIM has been used to simulate the detector response. Each random
background source has been integrated over an assumed 50 ns readout time.
The background induced in the DIRC from lost beam
particles has been estimated as described in Section 12.3.2 with the assumption of a uniform
beam line pressure of 1 nTorr. The mean number of charged particles crossing the quartz bars
and the water tank every s and the number of hits produced at the nominal background
rate are listed in Table 12-4. These numbers are to be compared to an average of 345 hits
for a typical physics event in a 50 ns window. The deduced mean rates expected per PMT
and per readout time at 10 times nominal background are around 20 kHz for a uniform beam
line pressure of 1 nTorr.
Lost Particle Backgrounds.
Technical Design Report for the BABAR Detector
12.4 Background Rates and Detector Responses
561
The radiative Bhabha process, e+ e, ! e+e, is a potentially important source of background. A BBSIM simulation of this process predicts that about 25
charged particles per s from secondary interactions crossing the DIRC, yielding an average
of 75 hits, 40% due to Cherenkov emission in the bar, 60% in the tank. The background
from radiative Bhabhas is on the same order of magnitude as the nominal background from
lost particle interactions. The showers from radiative Bhabhas tend to originate from around
Q1 and Q2, suggesting the possibility of additional shielding to reduce this background.
Other external backgrounds such as synchrotron radiation and cosmic rays have very small
or negligible impact on the DIRC response (see Chapter 6).
Radiative Bhabha.
12.4.4 Calorimeter
The main background in the CsI calorimeter is low-energy photons from lost beam particle
showers. Energy and angular distributions for these photons are shown in Figure 12-9 for the
nominal background level. About 330 photons per s, with a total energy of 0.39 GeV, are
incident on the calorimeter. The median photon energy is 500 keV, and about one photon
per s is above 20 MeV. Major background-related issues for the calorimeter are radiation
damage, energy resolution degradation, creation of extra showers in events, and trigger rates.
The radiation dose was estimated by assuming that the photons of Figure 12-9 are absorbed
in the rst 2 cm (9 g/ cm2 ) of the calorimeter. At nominal background, the radiation dose
for 107 s/yr running is about 45 rad/yr at = 90 and 90 rad/yr at = 300 mr. Larger doses
are expected from injection and machine studies, based on the experience of existing storage
ring operation. Scaling typical CESR operations to the PEP-II design beam currents gives a
radiation dose of about 1.5 krad/yr at a radius of 45 cm, and 0.5 krad/yr at 100 cm [Blu86].
Both are well within the tolerable dose rate of 10 krad/yr.
The e ect of backgrounds on photon energy resolution was studied using lost beam particles
generated by TURTLE as input to BBSIM, where they were showered in the same events
as the photons of interest. Backgrounds corresponding to nominal and 10 times nominal
levels in 1 s windows were generated together with 100 MeV photons in the barrel and
forward endcap. Comparison of the resolutions with the zero background case, for which
= FWHM/2:36 = 1:8 MeV, shows contributions, in quadrature, of 0.55 MeV from nominal
background and 1.05 MeV from 10 times nominal background.
A small fraction of the photons from beam backgrounds has sucient energy to be detected
as extra photons in physics events. Ecient B -meson reconstruction requires measurement
of photon energies as low as 20 MeV. The rate of photons with energies above 20 MeV
from lost beam particles is 0.7 per s in the barrel and 0.4 per s in the forward endcap
at nominal background. Timing information from the shaped calorimeter signals may be
Technical Design Report for the BABAR Detector
562
Interaction Region and Backgrounds
used to discriminate between photons which belong to events and out-of-time background
photons.
Some beam background particles reaching the calorimeter have sucient energy to contribute
to the trigger rate. Work is in progress to estimate this rate reliably and to improve the
design of detector shielding if necessary. Section 10.8 contains further discussion of this issue.
12.4.5
Muons
The Instrumented Flux Return (IFR) is well shielded from potential background sources by
the calorimeter, magnet, and ux return steel. Only the endcap chambers are close to the
beam line, and these are protected by ux return and magnetic shielding for the beam line
quadrupoles. Backgrounds from the previously discussed sources and from the irreducible
cosmic ray ux are expected to be small. The outermost layer of the IFR may be sensitive
to soft electromagnetic backgrounds in the detector hall originating from the PEP-II arcs.
Shielding walls in the tunnel mouths will be necessary to isolate the detector from the arc
radiation sources.
12.5 Physics Impact
The impact of backgrounds on the various detector components ultimately must be evaluated
in terms of their e ect on the overall physics goals of BABAR. Two requirements can be
compromised by high background levels: trigger selection and data acquisition for events of
interest; and oine reconstruction of the data.
As described above, a principal source of background trigger rates is the photoproduction of
hadrons in the showers initiated by lost particles, contributing to both a charged-track trigger
in the central drift chamber and to a calorimeter energy trigger. The rate is acceptable at
nominal background levels but begins to saturate the Level 1 trigger rate at around 10 times
nominal. Chapter 10 describes in detail the design of the trigger and steps taken to deal with
the possibility of a high trigger rate from this source. The conclusion of this discussion is
that some exibility is required in the trigger to allow the energy and momentum thresholds
to be raised if the rate from this source becomes too high. The loss of trigger rate for some
events of interest from the higher thresholds is also discussed in this chapter.
The e ect of backgrounds on event reconstruction within a particular detector component
has been studied in several cases, including, for example, track reconstruction in the silicon
vertex detector and the degradation of calorimeter resolution at above nominal background
levels. These are examples of two major e ects of backgrounds on oine event reconstruction:
Technical Design Report for the BABAR Detector
12.6 Summary
563
the inability to accurately reconstruct events due to ineciency and/or the confusion caused
by the high occupancy, and the degradation in resolution from the overlap of hits from
background and the physics event. The vertex detector shows a clear ability to maintain
good reconstruction eciency at 10 or more times nominal levels, due largely to its ne
sampling and the redundancy of a ve-layer detector. Similar studies with the drift chamber
are in progress. Again the large number of layers (40) and low occupancy are expected to
o er robust track reconstruction at above-nominal background levels. At 10 times nominal
background the ux of photons noticeably degrades the calorimeter resolution only for photon
energies below 50{70 MeV. This e ect will be seen as a small worsening of the reconstructed
mass resolution for generic B decays with a 0 in the nal state but will not e ect modes
with higher momentum 0 s (e.g., B ! 00).
A real assessment of the e ect of backgrounds on overall event reconstruction awaits the
development of the reconstruction algorithms, a major e ort that is in progress but not
complete.
12.6 Summary
Extensive work has been done by the BABAR collaboration to develop the tools to predict
the background rates in PEP-II and their e ects on the various detector components. Work
continues on re ning these estimates and checking the results with data where available.
There is also interaction with the PEP-II designers regarding aspects that a ect background
rates, e.g., the degree of vacuum pumping in the interaction region. The principal aim in
these studies is to ensure that the BABAR detector physics goals are not compromised by
machine-induced backgrounds. The detector design performance can be degraded due to
high occupancy causing ineciencies, deadtime, or pile-up e ects, and radiation damage,
causing impaired performance or frequent downtime to replace components. Some safety
margin (generally taken to be a factor of 10) is required to account for uncertainties in the
calculations, the assumed PEP-II running conditions, and in the case of radiation damage,
the dose accumulated during injection and machine studies. As explained in Section 12.1.4,
the design of the injection system for PEP-II suggests that injection will be cleaner than for
other machines. Nevertheless, the working assumption in the BABAR detector design is that
the dose from injection will equal that of normal running (in the case of the calorimeter, a
direct scaling of CLEO results is used).
Table 12-5 summarizes the contribution to occupancy from the dominant lost particle background for the more sensitive elements of the BABAR detector. Careful masking design and
shielding around the beam line components has reduced the contribution from synchrotron
radiation and radiative Bhabhas to a low level. Also in Table 12-5 is an estimate of the
limit on how high a value could be tolerated without compromising the performance goals of
Technical Design Report for the BABAR Detector
564
Interaction Region and Backgrounds
Silicon Layer 1
Drift Chamber Particle ID Calorimeter
Average Stave 1 Average Layer 1 (DIRC)
Photons
Occupancy 1.3%
3.0% 0.05% 0.5%
0.015%
1.1/s
Limit
20%
20%
10%
10%
2%
10/s
Safety
15
6
200
20
130
9
Occupancy in various detector components of BABAR. Calorimeter entries
are for photon energies above 20 MeV.
Table 12-5.
Silicon Layer 1
Drift Chamber
Calorimeter
Average Stave 1
Average
Layer 1
Dose/107sec 33 krad 82 krad 0.0004 C/cm 0.0045 C/cm 1.5 krad
Limit
1500 krad 1500 krad 0.1 C/cm
0.1 C/cm
100 krad
Safety
45
18
250
22
67
Table 12-6.
Radiation dose in various detector components of BABAR.
BABAR; the bottom row gives the resulting safety margin. The silicon detector has the lowest
safety factors. The 20% limit on occupancy in this detector is set by the desire to maintain
90% or better hit detection eciency. The worst stave of Layer 1 of the vertex detector has
a safety margin below 10, but a larger loss in eciency in this limited region of a single layer
is not considered a serious compromise of its overall performance.
Table 12-6 gives the radiation doses in appropriate units for the detector. The rst row gives
the yearly dose, and the second row gives an estimate of the limit at which the performance
begins to signi cantly degrade. The safety margin then determines the lifetime of the
detector at nominal background levels. Included in this safety margin, though, must be
some accounting for uncertainties in the calculated background levels and, particularly for
the silicon and the calorimeter, the additional doses likely to be delivered to these systems
during injection. The drift chamber will be ramped down in voltage during injection. The
BABAR detector is expected to be operating for 10{15 years, but it is reasonable to assume
that limited parts of the detector (such as sections of Layer 1 of the vertex detector or
the forward calorimeter) could be replaced at intervals of three to ve years. As noted in
Section 12.4.1, small regions of Layer 1 of the silicon vertex detector will see higher doses
(240 krad). These regions will have degraded signal-to-noise due to increased leakage current,
but the small extent of the damage will not seriously a ect the overall performance. The
conservative approach of scaling the CLEO results on dose from injection is used for the
calorimeter dose.
Technical Design Report for the BABAR Detector
REFERENCES
565
References
[All84]
O.C. Allkofer and P.K.F. Grieder, \Cosmic Rays on Earth," Physik Daten 25{1,
Karlsruhe, Germany, 1984.
[Blu86] E. Blucher et al.,
A249, 201 (1986).
[Bai81] V.N. Baier, V.S. Fadin, V.A. Khoze, and E.A. Kuraev, \Inelastic Processes in
Quantum Electrodynamics at High Energies,"
78, 293 (1981).
[Bri94] M. Bringle, \Cosmic Ray Muon Monte Carlo Generator using the Hemisphere
Method," A AR
# 163 (1994).
[Car91] D.C. Carey, K.L. Brown, and F.C. Iselin, \Decay TURTLE (Trace Unlimited
Rays Through Lumped Elements): A Computer Program for Simulating Charged
Particle Beam Transport Systems, Including Decay Calculations," SLAC{246
(1980).
[Che91] P. Chen, T. Tauchi, and D.V. Schroeder, \Pair Creation at Large Inherent
Angles," in
edited by J. Irwin
and D. Burke, SLAC{PUB{5652 (1991); also in Proceedings of the 1990 DPF
Summer Study (Snowmass, Colorado), pp. 751{756 (1990).
[Cou95] D. Coupal, \Lost Particle Backgrounds in A AR," A AR
# 210 (1995).
[Dar84] A. Dar, \Cosmic Ray Muons at Ground Level and Deep Underground,"
TECHNION{PHYS{84{41, Israel Institute of Technology, Haifa (1984).
[Deg94] P.V. Degtyarenko et al.,
C50, 541 (1994).
[Fes85] H.C. Fesefeldt, \The Simulation of Hadronic Showers, Physics, and Applications,"
Technical Report PITHA 85{02 (1985).
[GEA94] GEANT Detector Description Tool, version 3.21, CERN Program Library W5103,
CERN (1994).
[Hal81] K. Halbach, \Design of Permanent Multipole Magnets with Oriented Rare Earth
Cobalt Material,"
169, 1 (1980).
[Her87a] S.W. Herb, \Construction of Large Permanent Magnet Quadrupoles," CLNS{
87/61 (1987).
[Her87b] S.W. Herb and J. Kirchgessner, \Operation of CESR with Permanent Magnet
Interaction Region Quadrupoles," CLNS{87/64 (1987).
Nucl. Instr. Methods
Phys. Repts.
B B
Note
Linear Collider IR and Final Focus Introduction,
B B
B B
Note
Phys. Rev.
Nucl. Instr. Methods
Technical Design Report for the
A AR
B B
Detector
566
[Hea91]
[Joh09]
[Kad91]
REFERENCES
C. Hearty, \OBJEGS Users' Manual," A AR
# 73, (1991).
R. Johnson, \Radiation Damage Measurements on ALEPH Double-Sided Silicon
Strip Test Structures," ALEPH Note 94{009 (1994).
J. Kadyk,
A436, (1991).
J. Va'vra, SLAC{PUB{5207 (1990).
M. Danilov et al.,
A274, (1989).
R. Kleis and H. Burkhardt, NIKHEF{H/94{01 (1994).
F. Kral, \Trigger Rates due to Cosmic Ray Muons," A AR
# 164 (1994).
F. Kral, \Simulation of Trigger Rates due to Beam-Gas Backgrounds,"
A AR
# 165 (1994).
M. Levi, \Impact of Backgrounds on the Silicon Vertex Detector Architecture
and Detector Trigger," A AR
# 136 (1994).
M. Levi, \A Simulation of the Vertex Detector Readout Bu ers Using the Machine
Induced Occupancy," A AR
# 193 (1994).
J.W. Lightbody, Jr. and J.S. O'Connell, \Modeling Single Arm Electron Scattering and Nucleon Production from Nuclei by GeV Electrons," Computers in
Physics, May{June 1988, 57{64.
W. Nelson, H. Hirayama, and D. Rogers, \The EGS4 Code System," SLAC{
PUB{265 (1985).
R. Openshaw et al.,
A, 298 (1991).
Review of Particle Properties, Particle Data Group,
D45 (1992).
\PEP-II, An Asymmetric B Factory: Conceptual Design Report," LBL{PUB{
5379, SLAC{PUB{418, CALT{68{1869, UCRL{ID{114055, UC{IIRPA{93{01
(1993).
A. Snyder, \Electroproduction Triggers in A AR," A AR
# 180 (1994).
Y.S. Tsai,
46, 815 (1974).
B B
Note
Nucl. Instr. Methods
Nucl. Instr. Methods
[Kle94]
[Kra94a]
[Kra94b]
B B
B B
[Lev94a]
Note
B B
[Lev94b]
B B
[Lig88]
[Nel85]
[Ope91]
[PDG92]
[PEP93]
[Sny94]
[Tsa74]
Note
Note
Note
Nucl. Instr. Methods
Phys. Rev.
B B
Rev. Mod. Phys.
Technical Design Report for the
A AR
B B
Detector
B B
Note
13
Safety
13.1
Introduction
T
his chapter provides a description of identi ed and potential hazards, their relative signi cance, and the proposed design controls needed to eliminate or reduce the associated
risk to acceptable levels.
This chapter does not assess the risk for those identi ed hazards in terms of severity or
probability. This will be accomplished in the Safety Assessment Document (SAD), which
will further identify, evaluate, track, and resolve safety issues. The BABAR Detector will also
be reviewed by SLAC safety committees such as the Hazardous Experimental Equipment
Committee (HEEC), the Radiation Safety Committee, the Pressure Safety Committee, and
the Earthquake Safety Committee.
The purpose of performing this analysis and the committee reviews is to make sure that safety
is designed into systems, subsystems, equipment, facilities, and their interfaces, consistent
with the objective of the detector. The policy of management is to design for minimum risk.
13.2
Detector Safety Overview
BABAR will be housed in the existing PEP Research Hall at Interaction Region 2 (IR-2).
The detector weighs approximately 800 metric tonne. The oor of IR-2 is constructed of
two-foot-thick reinforced concrete. Estimated dimensions of BABAR are 6.7 m in height, 9.5 m
in width, and 8.3 m in length from the outer edge of the forward door to the outer edge of
the DIRC water tank. The design of the Research Hall incorporates a concrete shield wall,
which will contain penetrations for cableways and cryogenics. The fast electronics will be
located in a counting house just outside the wall. The existing control room area in IR-2
will be utilized. The cryogenic plant will be located on an existing pad outside the southern
wall of IR-2, and cryogenic storage tanks will be located on another pad on the hill at the
western end of the building.
568
Safety
BABAR is similar to detectors operating at SLAC and other physics laboratories, therefore
the hazards contained in BABAR are similar to previous examples.
Hazards contained in BABAR which are currently under assessment are:
Gases| ammable, asphyxiant, toxic|possibly utilized in the drift chamber and RPCs;
Electrical hazards in the various high voltage systems (the drift chamber, the RPCs,
and the DIRC photomultiplier tubes), the electronics, the power supplies, the high
current and stored energy associated with the magnet, and support equipment in the
counting house and control room area;
Radiation due to operation of the machine and possibly from a calibration system for
the calorimeter;
Cryogenics for the superconducting coil and possibly the Q1 magnets;
Magnetic forces during operation of the magnet;
Pressure/vacuum in the beam pipe, the drift chamber, the cryostat, and the associated
supply systems; and
Toxic materials such as beryllium, thallium-iodide, and cesium-iodide.
Generic hazards currently under assessment are
Earthquake,
Radiation,
Fire,
Occupational (Oxygen De ciency and Con ned Space), and
Environmental.
13.3 Beam Pipe and Support Barrel
The Q1 magnets, the B1 magnets, the IP beam pipe, and the vertex detector are assembled
into a single rigid support barrel. The beam pipe at the interaction point is a double-wall
structure made of beryllium. The inner tube is 50 mm in diameter with a separation of
Technical Design Report for the BABAR Detector
13.4 Vertex Detector
569
2 mm from the outer tube. Helium gas ows between the tubes to remove the beam-induced
heating.
The health hazard of the beryllium is minimal in the form used. The particulate created
as a result of machining, grinding, corrosion, etc., is toxic if ingested. A coating will be
applied which should control corrosion, and all machining will take place at an outside
vendor under appropriate safety conditions. If the beryllium needs to be worked at SLAC,
special procedures to control the particulate will be developed.
The thin wall beryllium tubes are easily damaged, and special handling requirements to
reduce this potential will be developed.
13.4 Vertex Detector
The vertex detector consists of double-sided silicon microstrip detectors and readout hybrid
circuits assembled into mechanical modules. These modules are glued to lightweight support
structures. At each end of the module, beryllium oxide or aluminum blocks provide the
required mechanical precision and act as a heat sink for the electronics.
The health hazard for beryllium is minimal in the form used (Section 13.3). The operating
voltage is less than 50 V, with currents less than 100 mA. Dry nitrogen or air will be used to
control the atmosphere. If nitrogen is used, the hazard of asphyxiation may exist depending
on the quantity. Water may be used for cooling. This detector system appears to have
minimal overall risk for causing injury to personnel or damage to equipment.
13.5 Drift Chamber
The drift chamber consists of a carbon- ber shell, a helium-based gas which may be ammable,
and a large number of wires with high voltage. Signi cant e ort is under way to nd an
acceptable gas mixture that is non ammable. The drift chamber may operate at constant gas
pressure resulting in a pressure di erential (as much as 30{40 mbar) between the chamber and
the atmosphere. On both ends of the chamber, rf-shields will contain an inert gas (probably
N2 ) that is circulated and will provide humidity control. The gas will be monitored for
helium content, which will allow detection of leaks from the chamber.
The drift chamber gas is an asphyxiant and may be ammable; it is probably not toxic. The
storage, transport, mixing, use, and disposal must be adequately controlled. The following
are the proposed controls:
Technical Design Report for the BABAR Detector
570
Safety
Development of a safety system to sense leaks and/or oxygen de ciency, and to automatically bring the system to a safe state. This will require the shut-o of the
ammable gas at its source, the interruption of electrical energy to potential ignition
sources, the purging of the system with inert gas, the operation of a ventilation system
to dilute any leaking gas, and the sounding of appropriate alarms to warn personnel;
Isolation of the bulk storage and the mixing operations in separate areas outside the
experimental hall;
Utilization of ow restrictors at the storage area;
Transportation and distribution of the gas in metal pipes to prevent leaks;
Comparison of chamber supply and return ow rates to aid in the detection of leaks;
and
Utilization of an appropriate re suppression system.
Control of the high voltage will be accomplished by conventional means. Design controls
will include short-circuit protection; isolation by insulating, enclosing, or covering; and
grounding. Equipment will be designed to meet appropriate codes and standards. Procedural
controls will include utilization of properly trained personnel and adherence to established
safe work practices and procedures.
Pressure is a concern in terms of damage to the chamber. The design must have features
that will prevent overpressurization which could result in subsequent damage to the chamber
and possibly other systems.
Accessibility to the chamber for assembly, maintenance or repairs (broken wire removal),
may pose a threat to personnel due to the con ned space and to the presence of hazardous
gas. Reviews of the various scenarios to assure safe access will be accomplished in the SAD.
13.6 Particle Identi cation
13.6.1
DIRC
The DIRC system will consist of quartz radiator bars, a support structure, a stando box
containing a reservoir lled with water, and conventional photomultiplier tubes.
The hazards associated with the DIRC include high voltage at the photomultiplier tubes,
water leakage from the stando box reservoir, and the possible degradation of the tubes due
to long-term exposure to helium. Control of high voltage is discussed in Section 13.5.
Technical Design Report for the BABAR Detector
13.7 Electromagnetic Calorimeter
571
A water leak could damage electronics, short circuit electrical equipment, and cause corrosion in other detector systems. The presence of water near the calorimeter may degrade
the calorimeter crystals. The interface between the stando reservoir and the window
frame/ ange must provide a positive water seal. Flexing of the door due to the magnetic and
seismic loads will need to be considered. In addition, features to control leaking water should
be fairly simple to incorporate due to the location of the reservoir outside the backward door.
13.6.2 Aerogel
The second part of the particle identi cation system is constructed of silica aerogel. Silica
aerogel is a glass-like structure which consists of amorphous SiO2.
When aerogel is unprotected, it is very fragile and pieces easily break o at edges and corners.
Cracks may develop inside during handling due to the low tensile strength. Aerogel also is
damaged by most gases and liquids, including water. For these reasons, the aerogel blocks
will be wrapped for protection. Aluminum, Te on, and paper are under consideration.
Photomultiplier tubes will be utilized and the issues addressed in the previous section
pertaining to high voltage and He gas are applicable.
In the form used in the detector, aerogel appears to present minimal risk. The very ne
dust of SiO2 generated as a result of machining requires the use of respiratory protection.
No such machining of aerogel is planned at SLAC.
The presence of helium near the forward aerogel PID photomultiplier tubes could degrade
the performance of the tubes. The drift chamber, the cryostat, and various cooling systems
will use helium. It may be appropriate to direct the air/N2 ow of the ventilation system so
that any possible helium present will be drawn away from the sensitive components.
Aerogel is an excellent insulator and the heat created by the tubes will need to be removed.
Nitrogen gas is being considered for cooling.
13.7 Electromagnetic Calorimeter
The calorimeter consists of cesium-iodide crystals wrapped in PTFE Te on or Tyvek, a support structure, photodiodes, and associated power supplies and regulators. The photodiodes
will operate at approximately 75 V. Cooling will be required and freon is currently under
consideration.
The principal hazard is the thallium iodide contained in the cesium iodide. The concentration
of thallium iodide is approximately one part per thousand. The federal OSHA permissible
Technical Design Report for the BABAR Detector
572
Safety
exposure level (PEL) is 0 1 mg m3 for airborne thallium particles [CFR93]. The thallium
iodide is contained in the crystal structure of the cesium iodide. Machining operations and
handling of the uncoated crystal will require special precautions.
The possible presence of a mutagen in the uorescent ux concentrator will also be evaluated.
This is a laser dye that will be dispersed in solid plastic.
The e ect of moisture on the crystals is detrimental, therefore the calorimeter will be
contained in a closed, shielded environment, where the humidity will be strictly controlled
to a level of approximately 3%.
One of the calibration systems under consideration contains a radiation hazard. This system
would utilize polonium beryllium (Section 7.5.3) as a source of fast neutrons to activate
freon. The freon would then be piped across the face of the crystals. The neutron source
and the pipe carrying the pressurized, activated freon liquid through the detector would be
located inside the IR concrete curtain wall. An alternative calibration scheme entails placing
low-level sources on each crystal. Both schemes will be evaluated for radiation hazard in the
SAD.
:
13.8
=
Muon and Neutral Hadron Detector
The system under consideration utilizes Resistive Plate Chambers (RPCs). These chambers
consist of large-area parallel-plate electrodes held apart at a distance of 2 mm with spacer
buttons and lled with a gas mixture. Approximately 3000 m2 will be covered by these
RPCs.
The hazards associated with the chambers are related to electrical safety, ammability of
material used in chamber construction, and the possible use of a ammable gas.
Electrical hazards are associated with the use of high voltage in the operation of the chambers. High voltage will be controlled through the use of appropriate cables and connectors,
short circuit protection, isolation by insulating or enclosing, and grounding. High voltage
supplies will not deliver currents in excess of 1 mA and will be interlocked with hazardous
gas and re alarms.
The RPCs are constructed of materials (bakelite) which are combustible only when exposed
to an intense ame or arc. As the chambers are con ned to narrow gaps between steel plates,
there is a very limited supply of oxygen available and any combustion initiated at an exposed
edge will be self-extinguishing. Smoke and heat-sensitive alarms will be located at strategic
locations around the detector.
Technical Design Report for the BABAR Detector
13.9 Magnet Coil and Flux Return
573
Proposed controls for the use of a ammable gas mixture are listed in Section 13.5. Due to
the large number of RPCs and the large surface area covered, attention must be paid to the
issue of leaks. Appropriate pressure testing before and after installation will be necessary.
Because the use of a non ammable gas mixture is highly desirable, active R&D in this
direction is currently being performed.
13.9
Magnet Coil and Flux Return
The magnet system consists of the superconducting solenoid, the ux return, and supporting
systems including cryogenics. The superconducting solenoid consists of the coil in a cryostat;
the vacuum system to provide thermal insulation; the power supply; the quench detection
and protection system; and the control system. The ux return is composed of a barrel and
endcaps, support/translation/alignment systems for each movable section and restraints to
provide structural stability. The endcap doors are designed to split on a vertical line for
access.
The principal hazards associated with the magnet are: cryogenic, electrical, magnetic, and
the control of energy during a quench.
Liquid helium (LHe) will be utilized to cool the solenoid. Issues of concern are leakage,
pressurization/vacuum, rupture, oxygen de ciency, and cryogenic burns. The helium plant
will be located outside the hall. It will contain a 4000 ` dewar. The total volume of LHe in
the experimental hall including the Q1 magnets and the solenoid is expected to be less than
250 `.
Leakage of the LHe will be controlled in the distribution system by the use of coaxial lines
with pressure switches interlocked to the supply valves. If vacuum in the line is lost (any
leak will degrade the vacuum shield), the supply of LHe will be automatically interrupted
at the source. Overpressurization of the aluminum restraining hoop cooling tubes and the
cryostat will be controlled by pressure release valves and burst discs. The worst case release
of LHe would require a catastrophic rupture of the dewar (such as a fork lift tine puncture).
Barriers will be strategically positioned to prevent this kind of accident. If the dewar were
opened in this way, the LHe would immediately evaporate and dissipate. Asphyxiation will
be a hazard in enclosed areas. One liter of LHe vaporizes to approximately 750 ` of gaseous
He at 20C and atmospheric pressure.
The solenoid current will be 7000{8000 A. Appropriate protection for the power supplies,
cabling, and connectors will be required.
Technical Design Report for the BABAR Detector
574
Safety
Whenever the magnet is energized, it poses a hazard due to the damage that may be caused
by objects, such as tools, being sucked into it. Procedural controls will be established to
control the use of equipment during this time period.
A quench detection and energy extraction system will protect the coil in the event of a
quench. The system to detect a quench will include the sensing of a voltage imbalance or
an increase of pressure. The energy extraction system will utilize external dump resistors.
Peak voltage across the magnet terminals during a quench will be less than 500 V.
13.10
Generic Hazards
The following is a brief discussion of the more general hazards:
The earthquake design requirements for the detector structure are de ned in
the SLAC Seismic Design Manual (this is more stringent than the Uniform Building Code).
The Earthquake Safety Committee will review and approve the design based on compliance
with the Seismic Design Manual [SDM91].
Earthquake.
Established radiation protections systems at SLAC include the Beam Containment System (BCS), the Personnel Protection System (PPS), the Beam Shut-O Ion
Chambers (BSOIC), shielding systems, and a radiological training program. For BABAR, the
design of the BCS and shielding, implementation of the PPS, placement of the BSOICs, and
development of operational requirements and procedures will be reviewed by the Radiation
Safety Committee.
Radiation.
An evaluation of the re prevention and re protection systems will be performed.
Life safety provisions will be provided for the facility in accordance with the NFPA Life
Safety Code. Noncombustible materials will be utilized where possible.
Fire.
The SLAC safety standards for Environmental Safety and Health are
contained in the SLAC ES&H Manual. During assembly, commissioning, operation, maintenance, and repair, the safety rules and regulations contained in this manual are applicable
to SLAC personnel, collaboration personnel, subcontractors, and users.
Hazards to personnel during maintenance and repair of components inside the detector are
under evaluation. The forward and backward end doors are designed in two halves, which
allows retraction on rails for access to the aerogel, forward end of the drift chamber, and
Occupational.
Technical Design Report for the BABAR Detector
13.10 Generic Hazards
575
calorimeter. Access to the backward components of the drift chamber will be possible through
the DIRC strong tube. A rail system allows the endcap iron plug to be retracted back to a
point where the crane could be used to set it aside. The rail system may then be used to
support a cart to carry personnel. A full-size mock-up will be used to develop and simulate
emergency evacuation procedures. The hazard of entrapment and the potential for hazardous
atmospheres, may classify the inside of the detector as a permit-required con ned space.
The hazards associated with elevated work will be considered in the design. Platforms,
railings, barriers, stairways, and ladders will be designed in accordance with the Life Safety
Code.
Environmental. Consideration of environmental issues such as the generation of haz-
ardous waste, the uses of freon, the release of gas mixtures, and environmental reporting
requirements will be addressed.
Technical Design Report for the BABAR Detector
576
REFERENCES
References
[CFR93] Code of Federal Regulations, Title 29, 1910.1000 (1993).
[SDM91] Earthquake Safety Committee, \Seismic Design Manual|Mechanical R1
Draft 2," (1991).
Technical Design Report for the BABAR Detector
14
Facilities, Assembly, Access and
Integration
T
his chapter addresses the plan for on-site assembly of the detector subsystems and their
installation in the experimental hall. It includes consideration of facilities required to
assemble the detector, and presents the philosophy [BAR93] for integration of facilities and
technical components in a smoothly owing, continuous process of assembly and installation.
14.1
Facilities
Although many of the BABAR detector components will be fabricated elsewhere, the nal
assembly and testing procedures will be done at SLAC in the IR-2 hall where the experiment
will be operated. This facility is large enough to accommodate, during the operation phase,
the detector on the beam line, the radiation shield wall, and detector services on the apron
(east side of the facility). The apron provides adequate working and storage space for the
assembly of the detector before installation of the detector on the beam line.
Figure 14-1 shows the general con guration of the hall and indicates the designated north
direction. There is a 5.96 m-wide by 8.53 m-high roll-up door at the east end of the building.
All components associated with the erection and operation of this detector are designed
to t through this door. This facility is equipped with an overhead bridge crane that will
be used during installation and maintenance. The rated capacity of this crane is 45 tonne
with an auxiliary lift rated at 9 tonne. The crane has a maximum hook height from the
concrete foundation of approximately 10.5 m and 11.1 m for the 45 tonne and 9 tonne lifts,
respectively.
The building is divided into two separate areas by a concrete shielding wall. The apron in
the east area will be used for the erection of the detector and initial testing. The west area
will house the detector during operations. After the detector has been commissioned and
experiments have begun, the apron will house the electrical, cryogenic, and data analysis
support systems as well as maintenance and upgrade facilities. Additional oces for support
personnel will be located on the second oor of the building.
578
,,,,,,
,,,,,,
,
,,
,,
,
,,
,
Facilities, Assembly, Access and Integration
Interaction Region CL
2.0 1.2
1.2 2.0
Interaction
Point
4.3
Machine
Components
Shielding Detector Out
12.4
Electronics House
Out Position
10 Ton Hook Limit
Offices
50 Ton Hook Limit
10 Ton Hook Limit
9.73
3.35
Detector Out
Position
50 Ton Hook Limit
Power Supply Racks
Designated
North
True
North
32.3 Travel
Magnet CL
+Z
2.59
Roll Up Door
5.96 Wide
8.53 High
Roll Up Door
2.74 Wide
8.53 High
2-95
7857A14
Figure 14-1.
All Dimensions in Meters
General plan of interaction hall with detector in assembly position.
Technical Design Report for the BABAR Detector
14.2 Detector Coordinate System
579
14.2 Detector Coordinate System
The BABAR detector's right-handed coordinate system has its origin 370 mm axially from the
center of the magnet. When installed on the beam line, the origin of the detector coordinate
system will be coincident with the interaction point. The positive z axis is coincident with
the detector axis and points toward the forward end of the detector. The positive y axis
points up. The positive x axis points east. During assembly, the forward end of the detector
is oriented toward the south end of IR-2. After installation on the beam line, the positive
z direction will be set at an angle of 20 mr with respect to the direction of the high-energy
beam.
14.3 Structural Support of Systems
Because SLAC is located in a seismically active area, special care must be taken to design
equipment to withstand seismic activity. Meeting this demand can be greatly eased by the
use of earthquake dampers. These dampers can be expected to reduce the peak acceleration
in the detector with respect to the ground by a factor of 3 or more. Such dampers were
used in the support of buildings in Kobe, Japan, which su ered much less damage in the
1995 earthquake than neighboring buildings supported in a conventional manner. Dampers
suited to the detector are available and were used as part of the SLD calorimeter assembly
xture.
The barrel steel rests on four feet which in turn rest on earthquake dampers resting on the
oor. These feet have mechanical jacks to adjust the height of the detector. Both the forward
and backward end doors are supported in the same manner as the barrel steel. The exception
is that during transport of the detector, the doors are rigidly attached to the barrel steel.
The coil in its cryostat is supported inside of the barrel steel. These supports are strong
enough to withstand the magnetic de-centering forces on the coil and the seismic forces, which
are substantially larger than the expected unbalanced magnetic forces. The calorimeter is
suspended from the cryostat. The DIRC is supported by the barrel steel so that the doors
at each end can be opened to service the calorimeter electronics.
It is important to protect the relative alignment of the drift chamber and the vertex detector
against temperature variations in the hall. The vertex detector is supported by the B1
magnets which are, in turn, mounted inside the support tube which holds the machine
components B1 and Q1. The support tube is attached to the DIRC strong support tube in
the backward direction and the barrel steel in the forward direction. In order to preserve
the relative alignment of the vertex detector and the drift chamber, the drift chamber is
Technical Design Report for the BABAR Detector
580
Facilities, Assembly, Access and Integration
supported at its inner radius by the support tube. Large seismically induced relative motion
of both Q2/Q4/Q5 pedestals with respect to the support tube must be minimized. The
earthquake motion problem is exacerbated by the use of the earthquake dampers beneath
the detector: the detector will probably move less than the pedestals.
14.4 Installation Overview
A preliminary installation plan, with schedule, has been formulated for the assembly of the
BABAR detector in the IR-2 hall. The detector installation requires approximately 14 months.
An additional month is needed for installation of the beam line components associated with
the interaction region.
The installation critical path is driven by the con guration of the detector systems. The
barrel assembly proceeds inward radially. The ux return is assembled. Then the solenoid
is installed within that, and the barrel calorimeter is inserted into the solenoid. The
DIRC is next installed, and the drift chamber inserted into the DIRC. The aerogel/forward
calorimeter assembly is installed. The support tube is inserted, and the doors, containing
their RPCs, are closed before a cosmic ray test. Each system's electronics is installed
simultaneously with its mechanical components.
While the detector is being assembled and tested in the east end of the IR-2 hall, a concrete
shielding wall is placed between the detector and the interaction point so that studies of the
machine components may occur during detector assembly. The eastern edge of this wall is
located approximately 4.5 m east of the interaction point.
After testing of the assembled detector is complete, the concrete shielding wall is disassembled, and a transport mechanism is used to move the detector to its nal position in the
PEP-II colliding beams. Figure 14-2 shows the detector in its operational location.
14.5 Detector Component Installations
14.5.1
IR-2 Detector Hall Preparation
Base plates of 50 mm nominal thickness are provided on the oor of the experimental hall.
The plates are leveled using hex bolts and secured to the concrete foundation with adhesive
anchors. The tops of the plates are positioned 60 mm above the oor of the experimental hall.
After erection and leveling of the base plates, the space beneath the plates is grouted solid
with a nonshrinking, cementitious, owable grout. The purpose of the grout is to provide
Technical Design Report for the BABAR Detector
,,,,,,
,,,,,,
,
,,
,,
,
,,
,
,
14.5 Detector Component Installations
581
Interaction Region CL
Magnet CL
9.73
7.19
3.5
1.1
+Z
Maximum Curtain
Forwarded Position
Designated
North
True
North
Shielding Detector In
3.35
Power Supply Racks
Electronics House
Stair to
Upper Housing
Roll Up Door
5.96 Wide
8.53 High
2-95
7857A13
Figure 14-2.
Offices
Roll Up Door
2.74 Wide
8.53 High
All Dimensions in Meters
General plan of interaction hall with detector on beam line.
Technical Design Report for the BABAR Detector
582
Facilities, Assembly, Access and Integration
the base plates with a level supporting surface and to provide uniform bearing between the
plate and the concrete foundation. The grout is placed as a continuous ow without air
pockets and allowed to cure 28 days prior to commencement of the erection of the barrel
ux return [KRE93].
14.5.2
Coil and IFR Installation
The assembly of the barrel ux return assembly is begun at a point approximately 12.4 m
east and 0.5 m north of the interaction point. The barrel axis is parallel to the bridge of the
overhead crane. During assembly, it will be necessary to move the detector barrel toward
the south end of the building. The magnet barrel is a self-supporting structure and houses
its own transport mechanism. The transport mechanism will be tested before installation
commences. Figure 14-1 shows the location of the detector in IR-2 during the assembly
process.
The barrel ux return is composed of six inner blocks and six outer blocks. The barrel RPCs
are installed in these blocks while the blocks are horizontal. The muon system is serviced by
a variety of signal cables, high voltage cables, and gas piping. All of the connections to the
individual detectors are made during the assembly process. During the assembly process,
the 45 tonne capacity overhead crane is used to lift the blocks into place. Figure 14-3 shows
the assembly procedure.
The electronics platforms and stairways will be partially preassembled into large components
and installed on the outer surfaces of the barrel ux return at this time. The manifold
connections for the barrel RPCs are now made. These connections link the barrel detectors
to the hall utilities [TRI92]. These manifold connections are located on the upper service
platform. The manifold gas utility feeds run along the west wall of IR-2 and connect with
the detector at both ends of the hall. The signal cables and high voltage cables are routed to
crates located externally to the detector on the electronics platforms. The detailed routing
of services from the RPC detectors to the crates and manifolds is still being investigated.
All connections are inspected and tested to ensure proper operation of the system.
The solenoid junction box and cryo-chimney are mounted to the cryostat prior to its installation. The cryostat is installed into the barrel ux return from the backward direction
as shown in Figure 14-4, using a beam and carriage system that will transport the coil's
gravitational load along the axis of the beam. To install the coil into the magnet steel, a
structural steel beam, referred to as the transfer beam, is inserted through the bore of the
coil from the backward side. There are three tripod supports for this beam. One is located at
the north end of the assembly area and a second at the south end. The third tripod support,
the oater, is positioned on either the north or south side of the detector, depending on
the device to be installed. For insertion of the coil, the oater is positioned on the north
Technical Design Report for the BABAR Detector
14.5 Detector Component Installations
583
Step 1
Position the bottom outer block on
spacers. Attach the two diagonal
outer blocks with their support legs.
Step 2
Remove the spacers. Install the
bottom inner block. Attach the
upper diagonal outer blocks.
Step 3
Install the lower diagonal inner
blocks. Attach the top outer block
with the far corner braces.
Step 4
Install the upper diagonal inner
blocks. Erect a raised platform to
install the top inner block.
Step 5
Roll the top inner block in and attach
the corner braces. Remove the
platform. Add the gap spacers.
Figure 14-3.
Magnet barrel assembly.
Technical Design Report for the BABAR Detector
584
Facilities, Assembly, Access and Integration
Figure 14-4.
Installation of the Solenoid using a beam and carriage system.
side. The transfer beam is temporarily supported on the two tripod supports on the north
side and the strongback carrying the barrel calorimeter is positioned on the forward side of
the barrel ux return. The transfer beam is then threaded through the coil and secured to
the tripod located at the south end of the assembly area. The tripods are provided with
x and y translation capability to enable accurate alignment of the coil relative to magnet
steel. The coil will be cooled and the cryogenic safety systems tested. The cryogenic services
are supplied via exible transfer lines from the helium lique er/refrigerator which is located
outside the IR-2 experimental hall, near its southeast corner. The exible transfer lines are
routed along the building periphery from the refrigerator to the west wall. These lines pass
through the wall to the cryo-chimney. This chimney is accessible from the upper platform.
The location of the magnet dump resistor and the magnet power supply has not been decided.
It is preferred that the power supply be installed on the apron outside the radiation wall to
permit access for maintenance.
The DIRC support structure is mated to the backward end of the barrel ux return assembly
using the overhead crane. This assembly consists of the DIRC strong support tube and the
structural bracing that connects the assembly to the barrel steel. The structure is properly
aligned and bolted into place.
Technical Design Report for the BABAR Detector
14.5 Detector Component Installations
585
Each endcap ux return is assembled as two doors. Each end door has a transport mechanism
built into its base. Using the overhead crane, an assembly of 20 vertical plates, containing
RPCs which were installed along with cables and services while the assemblies were horizontal, is mounted on the base along with a 30 tonne counterweight. The counterweight
helps minimize the potential for overturning during movement of the doors. The forward
end doors also contain the plug assembly that is used to divert the magnetic ux away from
the beam line components. The doors are attached to the barrel for magnet testing. The
backward steel plug is temporarily inserted into the backward end doors and removed after
eld mapping is completed. Services for the door RPCs are connected.
The coil is powered at low current and the electrical safety systems tested. While the current
is ramped up, the cooling system and decentering forces are monitored. It may be necessary
to de-energize the magnet and adjust the position of the solenoid at this time to minimize the
forces on the coil due to nonuniformities in the ux return. After initial testing is complete,
the doors are opened, and the eld mapping hardware is installed. NMR probes are installed
to monitor the eld during eld mapping. These probes will be removed after eld mapping,
and reinstalled after drift chamber installation. After the eld is mapped, the doors are
removed and stored against the wall of the IR-2 experimental hall, and the eld mapping
hardware is removed. The location at which the end doors are to be stored has not yet been
decided.
14.5.3 Barrel Calorimeter Installation
After magnetic mapping is complete, the installation of the barrel calorimeter commences.
The fully aligned, cabled, and tested barrel calorimeter arrives at IR-2 from the calorimeter
assembly building, B109. A system test is performed to ensure that no damage occurred
during transport to the hall. A nal check of the module alignment may also be done at this
stage if necessary.
The installation of the calorimeter into the coil bore is very similar to the installation of
the coil into the magnet steel. Once the transfer beam has been threaded through the coil
a structured beam is bridged between two pins on each side of the calorimeter and tied to
the end- ange pick up points. To maintain internal alignment of modules the pickup points
and support points must be the same. Figure 14-5 shows the installation method and the
transfer of loads. The ange on the forward end of the calorimeter support structure is
bolted directly to the end ange of the solenoid cryostat. Eccentric pins on the end ange of
the cryostat allow the horizontal and vertical positioning of the calorimeter. The backward
end of the calorimeter support structure is secured to the cryostat using adjustable radial
support arms. Shimming may be required between the anges of the calorimeter and the
cryostat.
Technical Design Report for the BABAR Detector
586
Facilities, Assembly, Access and Integration
Figure 14-5.
Sequence of barrel calorimeter installation into the coil.
The distribution system for the freon cooling for the barrel calorimeter is installed at this
time. It consists of multiple manifolds which are routed to the calorimeter along the
cableways between the barrel ux return and the forward and backward doors. The beroptic cables and the dry nitrogen supply and return lines are also installed in this space.
14.5.4 DIRC Tube Assembly Installation
The central support tube containing the DIRC quartz sectors and the stando box assembly
are transported to IR-2 while the barrel calorimeter is being installed.
The transfer beam is removed from the forward end of the detector barrel, which is then
moved 0.87 m to the south. No additional movement of the barrel will be necessary prior to
installation into the beam line. The oating tripod support is then moved to the forward
end of the barrel.
Using the overhead crane, the DIRC tube assembly is positioned at the backward end of the
barrel. The transfer beam is then threaded through the DIRC tube assembly and the bore of
the barrel calorimeter, and secured to the tripod support stands. The gravitational load of
the DIRC tube assembly is transferred to the transfer beam and the DIRC tube assembly is
rolled into the bore of the barrel calorimeter. The DIRC is aligned and secured to the strong
support tube of the DIRC support structure. The central support tube is then suspended
at the forward end from the barrel calorimeter end ange using temporary supports. The
transfer beam and the tripod support stands are removed from the barrel and moved to
storage.
Technical Design Report for the BABAR Detector
14.5 Detector Component Installations
587
Using the overhead crane, the DIRC stando region is now mated to the backward ange
of the DIRC strong support tube and securely fastened. A structural steel pedestal is then
inserted and installed on the lower surface of the strong support tube. The pedestal is
provided with a set of rails that extend beyond the backward end of the DIRC stando box.
14.5.5 Drift Chamber Installation
While the DIRC is being installed, the drift chamber and its assembly xture are mated
to an insertion frame. The base of the insertion frame is provided with rollers and also
with vertical height adjustment capability. The drift chamber temporary backward support
frame is inserted through the DIRC assembly from the backward end and secured. The drift
chamber is then inserted into the bore of the DIRC central support tube from the forward
direction. The drift chamber is properly positioned and temporarily cantilevered from the
its backward support frame. The drift chamber insertion frame is then removed and taken to
storage. The temporary supports that hold the DIRC central support tube are then removed
from the forward end of the barrel.
The signal cables, high voltage cables, and gas lines exit the drift chamber from the backward
end where they are routed outward radially to the inner surface of the DIRC strong support
tube. They are then routed in z along this surface until they exit the detector at the backward
end of the DIRC stando box. The detailed routing of these cables and gas lines is under
evaluation.
14.5.6 Aerogel/Forward Calorimeter Installation
The assembly frame, carrying the loaded and tested endcap segment, will be transported
to the experimental area where the endcap segment is transferred to the installation jig.
The aerogel segment is attached to the front of the calorimeter. The half-endcap is brought
into its mating position with the barrel, de ned by conical pins locating into holes drilled
into the barrel end ange, and xed in place, as shown in Figure 14-6. At this stage in
the installation, there is 5{10 mm between the adjoining barrel and endcap conical surfaces.
The endcap is aligned in the vertical plane to ensure uniform clearance, and nal xing holes
drilled. The endcap is now in its park position.
The second endcap is similarly installed in its park position and the two segments bolted
together to form the complete endcap, making a conic section. Provision is made for the
endcap to be wound into the barrel, using adjustable bolts at three points on the barrel
ange, and contact switches used to enable precision mating. Removal of the endcap involves
winding it back to its park position, unbolting the two segments, attaching each in turn to
Technical Design Report for the BABAR Detector
588
Facilities, Assembly, Access and Integration
Schematic view of the installation of an endcap segment including mating
to the barrel forward ange. (The aerogel is not shown.)
Figure 14-6.
the installation frame and removing them one after the other. This procedure decouples the
endcap from the barrel and ensures that the alignment of the latter is not disturbed.
The distribution system for the freon cooling for the forward calorimeter is installed at this
time. These cooling tubes are routed along the cableways between the barrel ux return
and the forward end door. The high voltage and signal cables for the aerogel are routed out
and around the forward calorimeter. They also exit the detector radially through the space
provided between the barrel assembly and the forward end door.
14.5.7 Vertex Assembly Installation
The vertex assembly that contains the beam pipe, vertex detector and the two B1 and two
Q1 beam line magnets is installed from the forward direction. This is done by cantilevering
the assembly from a temporary external support xture and carefully inserting the assembly
through the bore of the drift chamber using the overhead crane. When the vertex assembly
is properly positioned, the permanent drift chamber supports are installed on both ends.
The vertical pole structures that support the vertex assembly from the barrel ux return are
installed in the backward and forward ends of the detector. The temporary supports on the
drift chamber are then removed.
Technical Design Report for the BABAR Detector
14.5 Detector Component Installations
589
14.5.8 Backward End Plug Installation
Using the overhead crane, the backward end plug is mated with the DIRC rail system. The
backward end plug is rolled into position and securely fastened to the DIRC strong support
tube.
14.5.9 Electronics House
The initial processing of data occurs in the detector-mounted electronics. This equipment is
not accessible while beams are circulating. The closest accessible electronics are located in
the electronics house. The quantity and layout of the electronics located there are being evaluated. The electronics house is mounted to a structural steel frame. The frame is provided
with heavy-duty rollers similar to those used for the detector barrel. Prior to movement of
the detector into the beam line, the frame of the electronics house is mechanically connected
to the barrel support structure so that they move as a single mechanical unit. The concrete
radiation shielding wall is re-erected between the detector and the electronics house after
the detector is moved onto the beam line.
14.5.10 Detector Transport System
The detector transport system, which is similar in concept to those used on previous detectors, is used to move the detector in the IR-2 hall. It consists of a structural steel frame with
two hydraulic cylinders. The frame is bolted to the oor base plates, and the cylinder rams
are pinned to the detector assembly. Pressurizing the hydraulic cylinders extends the rams,
thus pushing the detector. When the ram reaches its maximum travel, the transport system
is removed and the rams retracted. The system is then reinstalled closer to the detector and
the process is repeated until the detector is located in the desired position.
14.5.11 Final Position of Detector in Beam Line
After the detector has been moved to the interaction point, it is rotated 20 mr about its
vertical axis using the detector transport system. The support legs of the detector rest
on seismic isolation pads; the pads will absorb energy caused by ground motion during an
earthquake. Jacks provided in the barrel assembly allow the vertex detector to be aligned
vertically with the interaction point. The detector legs are shimmed appropriately at the
seismic isolation pads and secured to the concrete foundation.
Technical Design Report for the BABAR Detector
590
14.5.12
Facilities, Assembly, Access and Integration
Service Space
Services for the various systems enter the detector via three paths. The rst path is radially
out between the barrel steel and the door. A gap of 100 mm is provided between the coil
cryostat and the door for calorimeter services. The blocks between the barrel steel and the
door occupy about 40% of the azimuth. These are placed at 60 intervals at the corners of
the hexagon. Beyond the coil, a gap of 150 mm is provided between the barrel steel and the
door. This allows space for the IFR services. All services passing this way will be in closed
conduit to assure that they t in the space available and to protect them when the door is
open and when the door is being closed.
The services for the drift chamber exit the detector at the backward end along the inside of
the DIRC tube. They proceed through the DIRC and around the water tank.
Routing of vertex detector services has not been resolved.
14.6 Detector Maintenance Access
Ready access to detector components eases the reliability requirements on those components,
thus reducing cost without compromising performance. The ability to make accesses of
short duration when needed to ensure data quality reduces loss of beam time. Parasitic use
of accelerator downtime allows repair of less pressing detector problems if access times are
commensurate with typical planned and random accelerator outages [BAR93]. The design
timescale for access for minor repairs is a few hours to a day. Annual accelerator shut-downs
will be used for more extensive repairs and for upgrades. Access to the electronics house is
not limited. Access to electronics, power supplies, and services located inside the radiation
wall but outside of the detector is possible by turning o the beams. All DIRC readout
components can be repaired in this way. The timescale for such a controlled access is less
than an hour.
Repairs that require access to the inside of the detector will be discussed in the balance
of this section. Before the doors are opened, the solenoid is de-energized. However, the
solenoid will remain cold. Only when speci cally required will the solenoid be warmed to
room temperature.
Forward Detector Internal Access
Two schemes for access into the detector at the forward end are under consideration. In
the rst, the doors are moved outward at an angle of 30 relative to the beam line. This
Technical Design Report for the BABAR Detector
14.7 Detector Integration
591
option provides 4.6 m clearance between the barrel and the forward doors. In the second, the
doors are moved outward at an angle of 85 relative to the beam line. This scheme provides
minimal but acceptable access to the RPCs and minimizes potential con icts with machine
components Q2, Q4 and Q5. In both schemes, the barrel and endcap calorimeter electronics
are readily accessible. The aerogel inner ring of readout devices can also be repaired. The
time needed to open the doors is approximately one hour. In order to access the end of the
drift chamber or the outer ring of aerogel readout, it will be necessary to remove one or both
endcap segments. This requires use of special xtures during a longer access.
Backward Detector Internal Access
Inner detector access in the backward region is complicated by the extension of the DIRC
assembly through the backward end doors. The barrel RPC detectors and the electronics
on the backward end of the barrel calorimeter are accessed by opening the backward ux
return end doors. These doors open perpendicular to the magnet axis. Access to the vertical
DIRC support plate and the cryo-chimney is also available. The time needed to open the
backward doors is approximately one hour.
The backward end of the drift chamber and some of the vertex detector electronics can be
accessed by removing the backward end plug. This requires installation of a special jig and
rails. This operation will take about one shift. The backward end plug is rolled back until
it is positioned beyond the DIRC stando box. It is then removed using the overhead crane.
Once the plug is removed, it is possible to enter the interior of the DIRC along the beam
line to access the face of the drift chamber. Quarters will be tight, since the inside diameter
of the DIRC is only about 1.6 m. Special jigs will provide a temporary oor for access.
Repair of drift chamber wires requires that the forward doors be opened, the forward
calorimeter and aerogel segments be removed, and the backward end plug removed.
Vertex Detector Access
Access to the vertex detector will require removing the support tube from BABAR. This
requires dismantling numerous beam line components and will entail substantial downtime.
14.7
Detector Integration
The Detector Integration Group is responsible for ensuring integration of the detector subsystems and of the detector with the collider and the experimental facility. The detector
Technical Design Report for the BABAR Detector
592
Facilities, Assembly, Access and Integration
Figure 14-7.
Overall detector dimensions( mm).
integration group maintains control of the physical interfaces including envelope dimensions,
mechanical attachments, obscurations, cable and utility routing, and alignment [KIR93].
Figure 14-7 shows a cross section of the central section of the BABAR detector indicating the
subsystem envelopes.
The detector integration group has the responsibility for installation of the detector elements in the experimental hall as well as for preparation of the experimental hall prior to
commencement of the detector assembly process.
14.7.1 Assembly Clearances
In order to avoid damage to the detector systems during the assembly process, an intersystem clearance of 20 mm is generally required. The 20 mm criterion is based on experience
Technical Design Report for the BABAR Detector
14.7 Detector Integration
593
with SLD. The 20 mm clearance is intended to provide for inaccuracies in fabrication of the
large parts, for tolerance build-up, and for assembly without extremely elaborate guidance
xtures. In addition, the clearance must be adequate to accommodate any relative motion
of systems during an earthquake. As a typical example, the clearance between the cryostat
and the inner at RPC is 20 mm.
An exception to the 20 mm rule has been made for the clearance between the drift chamber
and the DIRC central support tube because it is expected that both systems can be made
with close circularity and straightness tolerances and that a simple but e ective installation
xture can guide the drift chamber inside the DIRC for assembly. Another exception is the
gap between the aerogel and the forward calorimeter, where only 3 mm are provided because
the aerogel will be mounted directly on the calorimeter. Only 10 mm is provided between
the aerogel and the drift chamber because these objects are not large and because there is
good visibility available during the assembly process. The clearance between the forward
calorimeter and the endcap steel is 36 mm because the steel is not expected to be very at
and is expected to de ect by about 8 mm due to magnetic forces.
14.7.2 Good Neighbor Policy
Each system must obey a Good Neighbor Policy, i.e., the system must not pass its problems
to its neighbor without the neighbor's agreement. For example, each system must dispose
of any heat it generates without simply passing it on to its neighbor. This is particularly
important in cases in which temperature stability is required for a system for mechanical
alignment or accuracy of response. Each system must return its temperature to ambient by
whatever means is required, and do so within its own geometrical envelope.
Heat is not the only operating problem. Electrical and mechanical noise must also be
considered. Each system must take care to prevent or contain such interference with its
neighbors.
It is expected that each system will take care to minimize the risk of leakage of uids or
gases which can harm other systems. For example, a leakage of helium could damage the
phototubes of the DIRC, and water could damage CsI crystals.
14.7.3 Common Services
Use of common xtures in the assembly process will be optimized. For example, the xture
used to install the cryostat may be suitable for installing the barrel calorimeter.
Technical Design Report for the BABAR Detector
594
Facilities, Assembly, Access and Integration
Easily erected, modular sca olds will be used during the assembly phase and during the
operation stage. Such sca olds were employed to great advantage in SLD.
A centralized system will provide HVAC (heating, ventilation, and air conditioning), clean
AC power, emergency power, uninterruptible power, and oxygen de ciency and re alarms.
Technical Design Report for the BABAR Detector
REFERENCES
595
References
[BAR93] B. Barish et al., \GEM Technical Design Report," GEM{TN{93{262 (1993).
[KIR93] T. Kirk et al., \Solenoidal Detector Collaboration|Project Management Plan,"
SGT{000006 (1993).
[KRE93] H.J. Krebs et al., \Solenoidal Detector Collaboration|Preliminary Design
Report|Muon Barrel Toroid Support System," SDT{000168 (1993).
[TRI92] G. Trilling et al., \Solenoidal Detector Collaboration|Technical Design Report,"
SDC{92{201 (1992).
Technical Design Report for the BABAR Detector
596
Technical Design Report for the BABAR Detector
REFERENCES
15
Collaboration Issues and Project
Management
T
he BABAR Collaboration had its inaugural meeting at SLAC in December of 1993, soon
after the approval of the PEP-II accelerator project in October. An Interim Steering Committee was appointed to organize and direct the Collaboration until a permanent
organizational structure could be set up.
The Steering Committee, in consultation with the collaboration membership and laboratory
management, drafted a governance document for the Collaboration. The governance document was approved by the newly formed Collaboration Council in May of 1994. Following
the procedures described in the governance document, the ocers and management of the
collaboration were selected. These structures are now fully in place.
In this chapter we summarize the collaboration organization as detailed in that document [BAB94] and describe the organization of the project management aspects of the
construction and operation of BABAR. Figure 15-1 shows a schematic organization chart of
the BABAR Collaboration.
15.1 Membership
Ph.D. physicists, engineers, and Ph.D. thesis students who contribute signi cantly to the
BABAR detector, as well as those who contribute signi cantly to the accelerator and plan to
participate in the physics program, are eligible to be members of the collaboration. Through
the submission of the Technical Design Report, membership has been open to all individuals
who meet these criteria, and their institutions. Henceforth, new groups may apply for
membership in the collaboration by submitting an application to the Spokesperson, which
will be voted on by the Collaboration Council.
The Technical Design Report has been signed by 77 institutions from 10 countries.
598
Collaboration Issues and Project Management
Collaboration Council
Chairman - L. Piemontese
Vice-Chairman - R. Wilson
Institution Representatives
Project Management
Spokesman - D. Hitlin
Deputy Spokesman - R. Aleksan
Technical Coordinator - V. Luth
Project Engineer - R. Bell
Technical Board
Executive Board
Canada - D. MacFarlane
France - G. Wormser
Germany - K. Schubert
Italy - M. Giorgi
UK - J. Fry
US - K. McDonald
US - A. Seiden
US - M. Witherell
SLAC/US - [V. Luth]
LBL/US - M. Pripstein
Spokesman - D. Hitlin
Deputy Spokesman - R. Aleksan
Technical Coordinator - V . Luth
PEP-II - J. Dorfan
Figure 15-1.
Technical Coordinator (Chairman) - V. Luth
Project Engineer - R. Bell
Chief Electronic Engineer - G. Haller
Chief Software Engineer - D. Quarrie
Integration Physicist - H. Lynch
PEP-II Representative - J. Dorfan
Safety Officer - F. O'Neill
Spokesman - D. Hitlin
Deputy Spokesman - R. Aleksan
System Managers:
PEP-II/BABAR Interface - H. DeStaebler
Vertex Detector - F. Forti/N. Roe
Drift Chamber - D. MacFarlane
DIRC PID - G. London/B. Ratcliff
Aerogel PID CsI Calorimeter - R. Schindler
IFR - C. Sciacca
Magnet - R. Bell/[T. O'Conner]
Electronics - A. Lankford
Computing - N. Geddes/F. Porter
A AR Collaboration.
Organization of the B B
Technical Design Report for the BABAR Detector
15.2 Collaboration Council
15.2
599
Collaboration Council
Institutions with three or more collaborating members who are Ph.D. physicists are represented directly on the Collaboration Council. Members of the collaboration from institutions
with fewer than three Ph.D. physicists may aliate with another institution for the purpose
of representation on the Collaboration Council. Large institutions have one vote for every
ten collaboration members. The Collaboration Council has an elected Chairperson and
Vice-Chairperson.
The Collaboration Council deals with issues related to the overall framework of the collaboration. It is responsible for membership policy, publication policy, and the selection of
speakers for conferences.
The Council also appoints a Nominating Committee to nominate the Spokesperson every
three years, in consultation with SLAC management and with the Collaboration at large.
This nominee must be rati ed by a two-thirds majority of the Collaboration Council. The
Council may vote to remove the Executive Board, also by a two-thirds majority.
15.3
Spokesperson
The Spokesperson is the scienti c representative of the collaboration, and is responsible
for all scienti c, technical, organizational, and nancial a airs. On nancial matters, the
Spokesperson's authority is consistent with the requirements of the various funding agencies.
The Spokesperson consults with the Technical Board on technical matters. In addition, the
Spokesperson is responsible for keeping SLAC management and the Collaboration Council
informed about collaboration a airs.
The Spokesperson, after broad consultation with the collaboration and SLAC management,
nominates a Deputy Spokesperson, who must be rati ed by the Collaboration Council. The
term of appointment for both oces is three years, renewable.
15.4
Executive Board
The Executive Board advises the Spokesperson on all scienti c, nancial, and organizational
matters. It consists of members distinguished by their scienti c judgment, technical expertise, and commitment to the experiment. The Executive Board may remove the Spokesperson
with a two-thirds majority vote. The membership will re ect the national composition of the
Technical Design Report for the BABAR Detector
600
Collaboration Issues and Project Management
collaboration: initially, the Executive Board consists of one representative each from Canada,
France, Germany, Italy, and the U.K., and ve representatives from the US. This composition
may change as the collaboration grows. The Spokesperson, Deputy Spokesperson, Technical
Coordinator, and a PEP-II representative serve as non-voting ex-ocio members.
The Executive Board meets eight times per year. Several of these meetings are held jointly
with the Technical Board.
15.5 Technical Board
The Technical Board advises the Spokesperson on technical and nancial matters. It consists
of the Technical Coordinator as chairperson, the Project Engineer, a PEP-II representative,
the Collaboration Safety Ocer, the Spokesperson ex-ocio, and the managers of the
detector subsystems. In cases in which there are co-system managers, one of the comanagers sits on the Technical Board. The Spokesperson nominates the subsystem managers
in consultation with the Executive Board and may nominate additional members of the
Technical Board.
The Technical Board meets monthly to review the progress of the project. Several of these
meetings are held jointly with the Executive Board.
15.6 Finance Review Committee
The Finance Review Committee monitors the nancial aspects of the experiment as set
forth in the management plan for the detector and agreed upon between SLAC and the
collaboration. This will be detailed in Memoranda of Understanding between SLAC and
each participating institution. The committee will be chaired by the SLAC Director of
Research, and will include representatives of each of the funding agencies. The Spokesperson
will normally also attend meetings of the committee.
15.7 Communications
Collaboration Meetings are held four times per year, typically in March, June, September
and December. At least one meeting each year will be held outside the United States.
Technical Design Report for the BABAR Detector
15.8 Construction Responsibilities
601
The international nature of the BABAR collaboration requires the use of up-to-date communications techniques to facilitate communication. Frequent use of video and telephone
conferences promotes communication with the collaboration.
The BABAR World Wide Web Home Page, located at
http://www.slac.stanford.edu/BFROOT/
and reachable from the SLAC home page and from the home pages of many collaborating
institutions, is a major center of communication. Weekly event calendars, netnews groups,
and BABAR Notes are posted there, as are announcements of special interest.
15.8
Construction Responsibilities
The design of the BABAR detector described herein has been arrived at through an extended
process of optimizing the design of individual subsystems and the detector as a whole against
the required physics performance, the interests and technical capabilities of the collaborating
institutions, and available nancial resources.
In cases in which there were competing technologies, the process of arriving at a single
susbsystem design has been managed by a series of ad hoc task forces charged with evaluating
alternatives and recommending choices. In making these technology choices, the task forces
have considered physics performance, cost, maturity of technology, schedule implications,
and the integration of a given system into the overall detector context. Task forces have
been employed in arriving at nal designs for the vertex detector, particle identi cation,
electromagnetic calorimeter and instrumented ux return systems. The reports of these task
forces are available as BABAR notes.
Responsibilities for design and construction of the various detector subsystems have been
assigned through the traditional process of matching interests, capabilities, and resources.
Final responsibilities will be detailed in the Memoranda of Understanding. The following tables provide a brief synopsis of the expressed interest of many of the participating institutions
in detector construction responsibilities. In the case of computing, the institutions listed are
those that have been involved in the ongoing e ort to plan the computing infrastructure. It
is expected that responsibilities for the generation of simulation, reconstruction, and analysis
code will be distributed more broadly throughout the Collaboration.
It has been agreed by the Collaboration that certain detector items will be nanced through
a Common Fund. These include, but are not limited to, the magnet and its supporting
hardware, certain utilities, online computing hardware, and certain aspects of the software
development project. Details are under discussion.
Technical Design Report for the BABAR Detector
602
Collaboration Issues and Project Management
Subsystem
Vertex Detector
Institutions
Italy:
Milano, Pavia, Pisa, Torino, Trieste
US:
UC Santa Barbara, UC Santa Cruz, LBL,
Stanford
Main Tracking Chamber Canada: UBC, Carleton, CRPP, McGill, Montreal,
TRIUMF, Victoria, York
US:
Colorado, Colorado State, MIT, SLAC
DIRC PID
France:
Ecole Polytechnique, Orsay, Paris 6/7, Saclay
US:
UC Santa Barbara, Cincinnati, Colorado State,
LBL, Rutgers, SLAC
Aerogel PID
France:
LAPP Annecy
Italy:
Ferrara, Milano, Padova, Roma
Russia:
BINP
US:
UCLA, Caltech, Maryland
CsI Calorimeter
China:
Beijing, BGRI, SIC
Germany: Dresden
Russia:
BINP
UK:
Bristol, Brunel, Edinburgh, Lancaster,
Liverpool, Manchester, ICSTM, QMW,
RHBNC, RAL
US:
UCIIRPA, UC Irvine, Caltech, Iowa, LLNL,
UMass Amherst, Mississippi, Mt. Holyoke,
Notre Dame, SLAC
Instrumented Flux Return China:
Beijing
Italy:
Bari, Frascati, Genova, Napoli
US:
LLNL, Vanderbilt, Wisconsin
Table 15-1.
Institutions interested in construction of BABAR detector systems.
Technical Design Report for the BABAR Detector
15.8 Construction Responsibilities
Subsystem
Magnet
China:
Italy:
Russia:
UK:
US:
Electronics Canada:
France:
Germany:
Italy:
603
Institutions
Beijing
Genova
Dubna, BINP
RAL
LLNL, ORNL/Y12, SLAC
Montreal
Ecole Polytechnique, Orsay, Paris 6/7
Dresden
Genova, Napoli, Milano, Pavia,
Pisa, Torino
Taiwan: Academia Sinica
UK:
Bristol, Edinburgh, Lancaster, ICSTM,
QMW, RHBNC, RAL
US:
UCIIRPA, UC Irvine, UC Santa Cruz,
Caltech, Colorado, Iowa, Iowa State, LBL,
Penn, SLAC
Computing Canada: McGill
France:
LAPP, Orsay, Saclay
Germany: Dresden
Italy:
Padova
UK:
Manchester, RAL,
US:
Caltech, UC Davis, UC Irvine, UC Santa Cruz,
LBL, LLNL, Mississippi, Pennsylvania, SLAC,
Prairie View
Table 15-2.
Institutions interested in construction of BABARdetector systems (continued).
Technical Design Report for the BABAR Detector
604
References
[BAB94] PEP-II Detector Collaboration Governance (1994).
Technical Design Report for the BABAR Detector
REFERENCES
16
Cost and Schedule
16.1
Introduction
the
, the design of the B B detector has evolved as a result
Sbothofinceindividual
a continuous performance analysis and cost optimization. These studies involved
systems as well as the detector as a whole. Much scrutiny was given
Letter of Intent
A AR
to the requirements for acceptance, resolution, and granularity. Furthermore, the designs
were updated following revisions of the fabrication, assembly and installation methods, and
reassessments of the need for redundancy and reliability.
In cases where there were competing technologies, ad-hoc task forces were formed to evaluate
the options taking into account the projected performance with regard to the CP measurements, the technical feasibility based on prototype measurements and comparisons with
systems of similar design, the integration into the overall detector, and the projected cost
and schedule implications. In this way the Collaboration selected the CsI calorimeter, the
resistive plate chambers for the IFR, and the DIRC and aerogel for particle identi cation in
the barrel section and forward sections, respectively.
The BABAR detector design is based on well established technologies and takes advantage of
state-of-the art electronics, data acquisition and computing techniques. Furthermore, the
design of each subsystem builds on extensive R&D work that was started, in some instances,
well before the Collaboration was founded in December of 1993. The current focus is on
detailed design, prototyping and testing; in some cases preproduction activities have been
initiated. One of the highest priorities is the building of engineering teams to develop detailed
designs and prepare the procurement of long lead-time items.
16.2
Project Cost
A cost estimate has been developed for the detector described in this Technical Design
Report. This estimate includes all activities and materials associated with the project,
except for R&D and the design, construction and tests of prototypes that are currently
606
Cost and Schedule
underway and are funded from other sources. The current estimate is mostly based on
grounds-up analyses, though in some areas engineering estimates remain to be replaced. The
estimates are based on speci c designs and where applicable, comparable work performed
by the same group. In many cases, in particular for large items, multiple vendor quotes have
been obtained and samples have been purchased and tested.
All costs are stated in 1995 US$. The estimates cover the design, engineering, procurement
and fabrication, the assembly, tests and installation. The duration of the project is assumed
to be 42 months, beginning in April 1995 and ending in September 1998.
All activities necessary to complete the project have been organized into a detailed Work
Breakdown Structure (WBS) which at present contains more than 8000 elements. Each
WBS element is identi ed by a number and a short title. A description eld is used for
detailed de nitions, key parameters, components, as well as information on the basis of
estimate. Documentation related to the basis of estimate, vendor quotes, etc. will be collected
separately in workbooks. The WBS is available in a uniform database format, and it is
network accessible. At Level 2, the project is partitioned into nine systems, each headed by
a system manager who is responsible for the design, construction and installation, as well as
the cost and schedule management.
Because of the international nature of the collaboration, local conventions for cost accounting
are being used. This means:
The engineering and labor are recorded for each task in terms of the time required.
All materials, fabrication and procurement costs are recorded on the basis of local
rates.
For work performed at US or Canadian institutions, local rates for engineering and
labor are applied; the assigned rates are not burdened with operating overhead charges.
For activities that are supported by non-US/non-Canadian institutions, expenses for
engineering and labor are not recorded, they are included in operating budgets of those
institutions.
For items funded from US sources, contingency is added to the base cost to cover additional costs, above and beyond the base, that are necessary to ensure the completion
of the task.
For items funded from non-US sources, no contingency is added, since the base cost
includes a certain allowance for uncertainties in the cost projections. Signi cant cost
overruns can be dealt with by additional grants.
In all cases support from operating or infrastructure funds are recognized and not
included in the estimate. Also, work performed physicists is not charged to the project.
Technical Design Report for the BABAR Detector
16.2 Project Cost
607
US Cost in US Accounting (1995 k$)
Description
Funding
Agency
EDI&A
(mm)
EDI&A
($)
Labor
(mm)
Labor
($)
Base
Cost
M&S
Contin
gency
Total
1.1
Vertex
US DoE
112
821
114
527
733
2,081
25%
1.2
Drift Chamber
US DoE
3
24
16
103
255
382
22%
2,599
467
1.3.1
DIRC
US DoE
122
1,049
183
897
1,862
3,808
26%
4,799
1.3.2
Aerogel
US DoE
1
7
12
63
198
268
25%
334
1.4
Calorimeter
US DoE
136
1,219
436
1,507
11,257
13,983
16%
16,258
1.5
IFR
US DoE
7
58
63
219
196
473
16%
551
1.7
Electronics
1.7.1
Vertex
US DoE
59
506
0
0
132
638
27%
813
1.7.2
Drift Chamber
US DoE
84
296
2
6
493
795
32%
1,051
1.7.3
DIRC
US DoE
79
529
9
56
341
926
20%
1,111
1.7.4
Calorimeter
1.7.5
IFR
1.7.6
Aerogel
US DoE
6
37
0
2
8
47
32%
62
1.7.7
Trigger Level I
US DoE
44
409
0
154
563
28%
718
1.7.8
Trigger Level II/DAQ
US DoE
60
523
0
187
710
32%
938
1.7.9
Controls
US DoE
5
44
0
47
91
37%
125
1.7.D
System Engineering
US DoE
71
533
0
533
42%
757
US DoE
156
1,289
137
446
243
1,978
31%
2,600
US DoE
US DoE
504
84
3,097
696
126
529
1,027
10
4,653
706
12%
8%
5,227
764
1,533
11,137
1,098
4,355
17,143
32,635
20%
39,174
1.8
Computing
1.9
Management/Integration
1.9.1
1.9.2
SUM
Tech. Coordination
QA/ES&H Oversight
Table 16-1. Estimated Detector Cost in FY1995 US dollars, for items supported by US
Institutions. Not included in these tables are items that might be nanced through the
Common Fund. These include the magnet and some of the infrastructure, some of the
expenses for computing equipment, the electronics common to all systems, and expenses
for detector installation and project coordination.
Following a procedure developed at other HEP laboratories, the contingency was derived for
each WBS elements from a weighted sum of four factors that assess the potential risk or
uncertainty in the design, engineering solution, cost and schedule. This method introduces
a certain objectivity into the assessment, however, it can produce distortions. These have
been corrected.
In Tables 16-1 and 16-2, the cost estimate is summarized at Level 2/3 of the WBS. The
data are \rolled-up" from much more detailed information. Table 16-1 refers to the cost
of items that are to be funded by US institutions, thus EDI&A and labor costs are fully
accounted. Table 16-2 refers to items that are to be funded by non-US institutions. Not
included in either table are items that could potentially be nanced through a Common
Fund. While there is consensus that such a fund should be established, the size of the fund,
Technical Design Report for the BABAR Detector
608
Cost and Schedule
Non-US Cost in Local Accounting (1995 k$)
Funding
Agency
Description
EDI&A
(mm)
1.1
Vertex
INFN Italy
60
1.2
Drift Chamber
NSERC Canada
95
1.3.1
DIRC
France
111
1.3.2a
Aerogel
France
EDI&A
($)
Labor
(mm)
Labor
($)
Base
Cost
Contin
gency
Total
2,864
2,864
0%
2,864
888
1,806
0%
1,806
106
2,641
2,641
0%
2,641
12
12
173
173
0%
173
INFN Italy
0
4
146
146
0%
146
Calorimeter
BMFT Germany
0
89
2,768
3,054
0%
3,054
72
151
2,268
2,268
0%
2,268
1.5
IFR
INFN Italy
8
21
1,060
1,060
0%
1,060
1.7
Electronics
1.3.2b
1.4a
PPARC U.K.
1.4b
1.7.1
Vertex
INFN Italy
1.7.2
Drift Chamber
NSERC Canada
1.7.3
DIRC
France
1.7.4
Calorimeter
PPARC UK
1.7.5
IFR
1.7.6a
Aerogel
1.7.6b
1.7.7
1.8
Trigger Level I
Computing
162
M&S
316
0
55
364
149
602
286
0
120
120
0%
120
0
1,945
2,309
0%
2,309
1,734
1,734
0%
1,734
185
18
1700
1,700
0%
1,700
INFN Italy
19
29
409
409
0%
409
INFN Italy
6
0
126
126
0%
126
France
2
0
33
33
0%
33
23
53
0
0
78
0
78
0
0%
0%
78
0
18,953
20,521
0%
20,521
PPARC U.K.
France
84
785
SUM
680
741
888
Estimated Detector Cost in FY1995 US dollars, for items supported by nonUS Institutions. Not included in these tables are items that might be nanced through the
Common Fund.
Table 16-2.
the obligations of various national funding agencies to this fund, and the list of items that
should be supported from this fund remain to be negotiated.
The total cost of materials and services (in Europe referred to as investment costs) to be
nanced from the US and non-US resources, including the Common Fund, amounts to
45 million US$.
16.3
Schedule
Having a detector ready to confront CP violation physics of PEP-II on a time scale commensurate with the PEP II completion date of October 1998 is a formidable task.
Detailed schedule information will be incorporated into the Integrated Project Schedule.
The summary level schedule presented in Table 16-3 serves mainly to identify interactions
Technical Design Report for the BABAR Detector
16.4 Detector Systems
609
between systems and critical path items. This schedule is technology-limited and allows
installation and commissioning of BABAR at a date compatible with the PEP-II schedule.
16.4 Detector Systems
A brief discussion of the cost and schedule projections for the individual detector systems
follows.
16.4.1 Vertex Detector
The ve-layer double-sided silicon strip vertex detector chosen for BABAR is an extrapolation
of existing designs. The emphasis is on precision in the z direction, basic to the main physics
thrust of the experiment. Reduction of material to decrease multiple scattering is a primary
design requirement. The performance of 300 m double-sided silicon strip detectors has
been satisfactorily tested. For the read-out electronics, radiation-hard fabrication techniques
developed for SSC and LHC are more than sucient for the PEP-II environment.
Members of the vertex detector group have been active in the design and construction of
similar devices for CLEO-II, SDC, ALEPH, D0, CDF and Mark II. There is substantial
expertise within the group in detector and readout chip design, in trigger design and in
mechanical structures. The cost estimate is based on this recent experience and actual
vendor quotes. Full size prototype detectors are on order. The design and prototyping of the
VLSI read-out chip is underway, it is on the critical path for the assembly and tests of the
prototype detector modules and the beam tests scheduled for later this year. The assembly
of the detector modules is a delicate, time consuming task that will be performed in parallel
in the US and Italy.
16.4.2 Tracking Chamber
Several technical innovations in the main tracking chamber are motivated by the requirement
of minimal multiple Coulomb scattering. Low-mass gases have been developed within the
group over the past several years. Several chambers of comparable design and dimensions
have been constructed for HEP experiments. Low-mass eld wires can be made of aluminum,
although thin gold plating of the Al wires remains to be developed.
Construction of the main tracking chamber on the needed timescale requires the early resolution of mechanical engineering issues involving endplate and feedthrough design, and choice
Technical Design Report for the BABAR Detector
610
Cost and Schedule
of eld wire material. The electronics and data acquisition are reasonably straightforward. It
is planned to use a full-length prototype test chamber as a test bench for chamber operation
and monitoring, electronics and data acquisition.
The stringing of the wires is the most time consuming task and it is judged to be a critical
path item. The task will be performed at a laboratory where chambers of similar dimensions
have been assembled recently.
16.4.3 Particle Identi cation
While the DIRC is a novel concept for a Cherenkov detector, its construction is conceptually
simple. The production and assembly of the quartz Cherenkov radiators present a relatively
well-characterized problem that should have no schedule impact. Tests with small samples
of quartz bars have produced excellent results. A test with a full length prototype detector
is in preparation.
The readout system uses large numbers of conventional photomultiplier tubes; system engineering is thus the principal concern. A rst-article procurement of 500 phototubes is the
basis of the cost estimate, a rm o er for the whole photomultiplier purchase is in hand; it
has a signi cant savings for large quantities. Concerns about the detailed design of the water
tank and the modi cations to the ux return have been addressed. Access to inner detector
components has become easier with the elimination of the backward calorimeter endcap.
The aerogel threshold counter extends the coverage for particle identi cation in the forward
direction to 300 mr. This system is by far the smallest in BABAR, but requires sizable support
for R&D, engineering and photon detectors that can operate in the 1.5 T magnetic eld.
The present choice, ne-mesh photomultipliers, will become commercially available in the
near future. The cost estimate is based on a preliminary vendor's quote. Several other
experiments are in need of such tubes, and it is expected that the price will drop over the
next few months.
In the past, the production of high quality aerogel has been a serious concern. Facilities exist
for the production of the required quality and modest quantities. Prototype pieces have been
tested in a beam and have shown very satisfactory light yield and absorption length.
The mechanical mounting structure, and the calibration and readout systems are relatively
simple and of limited scale, requiring a modest engineering e ort.
Technical Design Report for the BABAR Detector
16.4 Detector Systems
611
16.4.4 CsI(Tl) Calorimeter
Large CsI calorimeters of similar design and comparable volume have been built in the
past and have proven to work well at e+ e, storage rings. Nevertheless, the production of
6,780 crystals of high quality is the principal challenge in this system, it is also the principal
cost driver. The procurement of the crystals is clearly on the critical path, and there remain
concerns about the projected rate of delivery of the crystals.
The BABAR calorimeter is divided into a barrel and a forward endcap section, with a total
volume of 6 m3 . The procurement plan assumes that a production rate of 250 crystals per
month can be established by the end of 1996 and can be sustained for more than a year.
The crystal procurement relies on three vendors in three di erent countries, each with prior
experience in growing non-alkaline halide crystals and with interest in increasing the capacity
of their production facilities. Extensive contacts with all three crystal manufacturers have
been established and it it expected that new, more advanced growing techniques can be
applied (similar to those being developed at SLAC) which will produce excellent quality at
signi cantly lower prices than in the past. The choice of a single supplier for the raw material
will assure uniform quality and will bring the bene t of volume and delivery discounts. It
also allows for a bene cial arrangement for the recycling of crystal material.
The engineering and fabrication of the low mass structure that will support the crystals
is the second largest item. The proposed concept is rather simple. The design has been
tested using nite element analysis and involves the use of Kevlar-epoxy composites, while
allowing fabrication by low-technology and relatively inexpensive methods. The present
endcap design is based on more conventional support structure, but an alternate design,
similar to the barrel is under study.
Based on experience in other experiments the readout system based on photodiodes and
wavelength-shifters can be expected to be reliable and perform as designed. Prototypes for
the silicon PIN photodiodes are presently being tested. The price for the 13,560 photodiodes
is being negotiated. Since the production capacity is more than adequate, the delivery
schedule is not expected to be a problem.
The modular design of mechanical support structure is simple and inexpensive and it provides
great exibility in scheduling the assembly and installation. The schedule allows for about
one year of nal assembly and tests. The production and assembly of the wavelength
shifter/photodiode/preampli er readout will be widely distributed and will be performed
in parallel to the crystal production.
Technical Design Report for the BABAR Detector
612
Cost and Schedule
16.4.5 Flux Return Instrumentation
The construction of large area chambers for muon detection is a task of a scale that has
been achieved by nearly every major 4 detector built in the last decade. The design and
operation of large numbers of resistive plate chambers (RPC) is well established. Facilities
with adequate capacity for the fabrication of the RPCs exist, as do experienced teams of
scientists at collaborating institutions where these chambers will be tested.
The schedule foresees the installation of the chambers prior to the assembly of the steel
structure, and thus places constraints on the schedule for the chamber fabrication, tests, and
nal assembly at SLAC.
16.4.6 Magnet
The speci ed superconducting solenoid is similar in dimensions and speci cations to several
existing magnets. The preliminary design is being prepared by a collaboration of scientists
and experienced engineers in the UK and Italy and will be completed in April 1995. It forms
the basis for the cost estimate and will also be used to initiate the procurement process.
The segmented ux return is also similar in design to other detectors. It must be completely
speci ed in the next few months, so procurement can be initiated in FY1995. The superconducting solenoid and the steel structure forming the ux return are both on the critical
path since they must be completely assembled and the magnetic eld must be measured
prior to the installation of all other detector systems.
16.4.7 Electronics
Great emphasis has been placed on the adoption of a common architecture and the use of
common solutions wherever possible. This approach is not only attractive from a design
and operational point of view, it is also most cost e ective, since it minimizes duplication of
e ort and allows for economies of scale.
The electronics system described in the TDR serves as a model based on present day
technology and experience. The WBS divides the electronics system into three principal
segments:
the detector speci c electronics, i.e., front-end electronics,
Technical Design Report for the BABAR Detector
16.4 Detector Systems
613
the trigger and data acquisition combined with electronics that are shared among
detector systems, and
the system engineering and the infrastructure.
For most detector systems, the design and prototyping is a very signi cant expense. The
costs for the front-end electronics for the drift chamber, calorimeter and DIRC are largely
driven by the channel count. For the drift chamber and calorimeter, the requirement for
analog information and the extraction of trigger information adds signi cant costs. The
fabrication costs include expenses for QC and tests. The front-end electronics for the vertex
detector are very closely integrated with the detector design and assembly and they are
therefore included in the detector costs.
The scale of the trigger and data acquisition system is conventional. Technical innovations
relate mostly to the asynchronous nature of the system and its fully pipelined architecture.
Trigger, data, acquisition and other electronics contribute about 20% to the total cost, shared
about equally between engineering and M&S.
The schedule for the detector mounted electronics is given by the assembly and installation
schedule for a given detector system, i.e., the vertex detector, the drift chamber, PID,
calorimeter, and the IFR. This means that there are 12 to 18 months for design and
prototyping, and another 12 to 18 months for fabrication and testing. For all these systems
it is crucial that prototype front-end circuits be available for the beam tests later this year.
The bulk of the o -detector electronics will not be needed until the assembly and installation
is completed in early 1998, thus allowing a total of three years for design, prototyping and
fabrication. However, it is planned to have a production model of the data acquisition system
available for testing more than a year before the completion date, so that design problems
that might not be apparent on a single circuit test can be detected and corrected prior to
the large production. This would allow for complete system tests, i.e., from a prototype
detector to the DAQ for a substantial number of channels.
16.4.8 Computing
The quantity of data that will need to be transferred from the BABAR detector to mass storage
and then analyzed pose signi cant challenges for the computing and network systems. For
costing purposes the system is divided into two major segments:
the online infrastructure, consisting of Level 3 data acquisition, control and monitoring
hardware, as well as the networking, and
Technical Design Report for the BABAR Detector
614
Cost and Schedule
the software development, including both online-speci c items such as VxWorks, EPICS,
expert system tools; and general items like IDL, methodology software, OODBMS, GUI
builder, documentation and development aids.
Not costed are items that are conventionally part of laboratory or university infrastructure
(e.g., the large clusters of workstations that perform the oine analysis and simulation tasks),
and activities that are generally supported from operating funds (e.g., maintenance of hardware and software, software engineering that is available on a continuing basis throughout
construction and operations, and data-aide activities). Also not included in the computing
system cost estimate are computers for administration and workstations for engineering
and design (covered in WBS 1.9), and workstations for software development (covered in
WBS 1.7), as well as embedded CPUs in the DAQ/trigger systems (covered in WBS 1.7).
The largest cost items for the computing system are the required software engineering and
the hardware for the Level 3 compute farm. The engineering e ort is derived from three
sources:
professional software engineers, at a level of approximately three to four FTEs for the
duration of the construction (not including the project computing engineer, covered in
WBS 1.9);
undergraduate students, who are frequently recruited from computer science departments. Past experience has proven that many of the programming tasks (such as
development of speci c computing utilities) can be performed by this kind of low-cost
part-time student labor; and
physicists, who will form a large part of the team for developing application code, but
are not included in the construction cost.
For the purpose of this estimate, it assumed that the Level 3 farm will consist of a networked
system of UNIX workstations. The principle data ow will be via FDDI connections, with
a DEC Gigaswitch to provide routing.
Future cost savings that can be expected from advances in technology over the next few
years are presently being examined. While there is incentive to delay the acquisition of the
computing hardware in order to take advantage of rapid developments and lower costs, a
large portion of the computing system needs to be in operation well before the detector is
ready to take data. Prototype framework, data model, and code management capabilities
need to be in place as soon as possible, because so many other development activities depend
on them. Hence, these items represent relatively early expenditures in the BABAR project.
The software for detector simulation, data reconstruction and analysis will be developed
and used by a large number of members of the collaboration, involving very few computer
Technical Design Report for the BABAR Detector
16.4 Detector Systems
615
specialists. Because of the geographic diversity of the collaboration, coordination is a
complex task and organizational and technical systems are being developed to facilitate
communication and remote operation.
16.4.9 Management, Installation and Integration
This WBS item includes the support of the BABAR management and detector integration,
the QA and ES&H management as well as the detector installation.
Substantial emphasis is being placed on system integration from the start of the project.
This involves coordination of the service and utility requirements of the various subsystems,
as well as detailed planning for the assembly of the detector and its commissioning. This
e ort will be coordinated by an experienced sta of engineers and physicists.
The overall plan, the cost estimate and schedule projections are based on recent experience
with SLD at SLAC. In particular, the estimate for the installation relies on the availability
of a sizable crew of laborers and electricians to prepare the hall, assemble the electronics
building, install utilities, and install the detector components. Special jigs and tooling for
the installation of the individual detector systems are considered the responsibility of that
system. The same is true for any detector speci c utilities like cables, cooling and gas
supplies.
The development of the detailed installation procedure has resulted in a top down schedule,
determining the date at which each of the detector systems will have to be delivered to the
experimental hall, together with any special tooling and jigs required for installation.
Technical Design Report for the BABAR Detector
616
Cost and Schedule
ID
1
Task Name
INSTALL BaBaR
2
1.1 Vertex Detector
3
Design & Prototype
4
Procure & Fabricate
5
Assemble & Test
6
7
Install
1.2 Drift Chamber
8
Design & Prototype
9
Procure & Fabricate
10
Assemble & Test
11
12
Install
1.3 Particle Identification
13
1.3.1 Barrel PID (DIRC)
14
Design & Prototype
15
Procure & Fabricate
16
Assemble & Test
17
Install
18
1.3.2 Forward PID (Aerogel)
19
Design & Prototype
20
Procure & Fabricate
21
Assemble & Test
22
Install
23
24
1.4 CsI Calorimeter
1.4.2 Barrel Calorimeter
25
Design & Prototype
26
Procure & Fabricate
27
Assemble & Test
28
Install
29
1.4.3 Forward Calorimeter
30
Design & Prototype
31
Procure & Fabricate
32
Assemble & Test
33
34
Install
1.5 Muon System
35
Design & Prototype
36
Procure & Fabricate
Table 16-3.
1995
1996
1997
1998
Qtr 2AQtr 3 Qtr 4AQtr 1 Qtr 2 Qtr 3 Qtr 4AQtr 1 Qtr 2 Qtr 3 Qtr 4AAQtr 1 Qtr 2 Qtr 3 Qtr 4
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
4/3 AA Design &AA Prototype
12/23
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
10/30
7/7AA
A
A Procure & Fabricate
A
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A Test
AA
3/4
Assemble
&
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
9/4
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
4/3 AA Design &AA Prototyp 3/8
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A Procure & Fabricate
A
AA
11/6
12/11
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
12/11
Assemble 7/2
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/16 7/
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
7/
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A9/5
AA
4/3 AA Design &AA Prototype
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
4/1 Procure &AA Fabricate
2/20
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A 10/1
AA
10/1
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA Install
11/17
7/1
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
/3 A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
AA
4/3 AA Design &AA Prototype
5/23 AA
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
6/30AA
A
A 1/2 Procure & Fabricate
A
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/1 Assemble
& Test
7/1
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/31
8/
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
4/24
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
4/3 AA Design &AAPr 11/28
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A 8/2 Procure
A
A
AA
& Fabricate
3/28
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
12/26
Assemble & Test
3/2
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
3/16 I 4/24
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
4/3 AA Design &AAPr 11/28
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A Procure
A & Fabricate
A
AA
5/3
3/28
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A & Test
AA
5/6
Assemble
5/29
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
8/7
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
12/23
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
4/3 A Design &A Prototype
12/23
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
10/2 AA Procure & FabricateAA
5/21 AA
A
A
A
A
AA
Projected project schedule, aimed at completion in October 1998, part 1.
Technical Design Report for the BABAR Detector
16.4 Detector Systems
ID
37
38
39
Task Name
Assemble & Test
Install
1.6 Solenoid Magnet
40
Design & Prototype
41
Procure & Fabricate
42
Assemble & Test
43
Install
44
1.7 Electronics
45
1.7.1 Vertex Detector Electronics
46
Design & Prototype
47
Procure & Fabricate
48
49
Install
1.7.2 Drift Chamber Electronics
50
Design & Prototype
51
Procure & Fabricate
52
617
Install
53
1.7.3 Particle ID Electronics
54
Design & Prototype
55
Procure & Fabricate
56
57
Install
1.7.4 CsI Calorimeter Electronics
58
Design & Prototype
59
Procure & Fabricate
60
61
Install
1.5 Muon Electronics
62
63
64
65
Procure & Fabricate
Install
1.7.7 Level 1 Trigger
66
Design & Prototype
67
Procure & Fabricate
68
Install
69
1.7.8 Level 2 Trigger & DAQ
70
Design & Prototype
71
Procure & Fabricate
72
Install
Table 16-4.
1995
1996
1997
1998
Qtr 2AQtr 3 Qtr 4AQtr 1 Qtr 2 Qtr 3 Qtr 4AQtr 1 Qtr 2 Qtr 3 Qtr 4AAQtr 1 Qtr 2 Qtr 3 Qtr 4
A
A
A
AA
A
A
A
AA
A
A
10/1 AA Assemble & T 6/30AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/2 Install
A
A
A
AA 11/12
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
3/13
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
4/3 AA Design &AA Prototype
1/1
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA9/10
3/14 Procure &AAFabricate
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA9/9
3/13
Assembl
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/11
Install
3/13
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
4/3 A Design &A Prototype
12/31
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
12/30
A
A
A 1/1 Procure & Fabricate
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/1 Install
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
4/3 AA Design &AA Prototype
11/29
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A Fabricate
AA
4/1
Procure
&
5/28
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
9/1
Install
7/
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
4/3 AA Design &AA Prototype
11/29
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
12/2
Procure & Fabricate
5/29
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
6/1 Instal
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A 9/30
AA
4/3 AA Design &AA Prototype
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A 1/1 Procure & Fabricate
A
AA
3/30
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
9/2
Install
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
10/29
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
4/3 A
A
A 9/30
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
4/1 Procure &AA Fabricate
A
A
AA 10/29
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/1 Instal
A
A
A
AA 10/29
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
6/
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA 9/30
4/3 AA Design &AA Prototype
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/1 Procure
& Fab 3/30
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
4/1 Inst 6/29
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
AA
/3 AA
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
AA
4/3 AA Design &AA Prototype
7/31
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/1 Procure
& Fabricate
6/29
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
A
A
A
AA
7/1 Install
A
A
A
AA
A
A
A
AA
Pro jected project schedule, part 2.
Technical Design Report for the BABAR Detector
618
Cost and Schedule
ID
73
Task Name
1.7.9 Controls & PEP II Interface
74
Design & Prototype
75
Procure & Fabricate
76
Install
77
1.7.B Common Electronics
78
Design & Prototype
79
Procure & Fabricate
80
81
82
Install
1.8 Computing
1.8.1 On-line Infrastructure
83
Interactive Stations Procurement & Installation
84
Level 3 Farm Prototype
85
Production Level 3 Farm
86
On-line File Server
87
Networking R&D
88
89
90
Network Procurement & Installation
1.8.2 Software Development
On-line Software
91
On-line Prototype
92
Initial HER Operation
93
HER Background Studies
94
BaBar Installation
95
BaBar On-line
96
Infrastructure Prototype
97
Reconstruction
98
Design
99
Implementation
100
Simulation Design & Implementation
101
Analysis Tools Design & Implementation
Table 16-5.
1995
1996
1997
1998
Qtr 2AQtr 3 Qtr 4AQtr 1 Qtr 2 Qtr 3 Qtr 4AQtr 1 Qtr 2 Qtr 3 Qtr 4AAQtr 1 Qtr 2 Qtr 3 Qtr 4
AA
AA
AA
AA
AA
AA
AA
AA
/3 AA
AA
AA
AAA
AAA
AAA
AA
AA
A
AA
4/3 AA Design &AA Prototype
7/31
A
AA
AA
AAA
AAA
AAA
AA
AA
AA
7/1 Procure
& Fabricate
6/29
AAA
AAA
AAA
AA
AA
AA
AAA
AAA
AAA
AA
AA
7/1 Install
AA
AA
AA
AA
AA
AA
AA
AAA
AAA
AA
AA
/3 AA
AA
AA
AAA
AAA
AA
AA
AA
AA
AA
AA
AA
4/3 A Design &A Prototype
7/31
AA
AA
AA
AAA
AA
AA
AA
AAA
AAA
AAA
AA
7/1 Procure
& Fabricate
6/29
AA
AA
AA
AA
AA
AA
AA
AAA
AAA
AAA
AA
7/1 Install
AA
AA
AA
AA
AA
AA
A
AA
AAA
AAA
AA
/3 AA
AA
AA
AAA
AAA
AAA
AA
AA
A
A
A
AA
/3 A
A
A
AA
AA
AAA
AAA
AAA
AA
AA
A Stations Procurement
A & Installation
AA
4/3 AA Interactive
5/29
A
A
AA
AA
AAA
AAA
AAA
AA
AA
A
A
AA 9/30
4/3 AA Level 3 Farm
Prototype
A
A
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA Production Level 3
10/1AA
AA
AA
AA
AA
AA
AAA
AAA
AAA
AA
AA
10/1
On-line
File
5/30
AA
AAA
AAA
AAA
AA
AA
AA
AA
AA
AA
AA
4/3 A Networking
R&D
4/30
AA
AA
AAA
AAA
AA
AA
AA
AAA
AAA
AAA
AA
5/1 Network
AAPro 12/31
AA
AA
AA
AA
AA
A
AA
AAA
AAA
AA
/3 AA
AA
AA
AAA
AAA
AAA
AA
AA
A
A
A
AA
/3 A
A
A
AA
AA
AAA
AAA
AAA
AA
AA
A
AA
4/3 AA On-line PAA 9/29
A
AA
AA
AAA
AAA
AAA
AA
AA
A
AA
10/2 AA Initial HER Operation
3/31
AAA
A
AA
A
A
AA
AA
AAA
AAA
AAA
AA
AA
10/2 A HER Background Studies
12/31
AA
A
AA
A
A
AA
AA
AAA
AAA
AAA
AA
AA
AA 1/1 BaBa 4/30
AAA
AAA
AAA
AA
AA
AA
AA
AA
AA
AA
5/1 BaBar
AA
AAA
AAA
AAA
AA
AA
A
A
AA
AAA
AA
4/3 AA Infr 6/30AA
AA
AA
AA
AA
AA
AA
AA
AAA7/3
AAA
AAA
AA
9/30
AA
AA
AAA
AAA
AAA
AA
AA
AA 7/3 Design
AA
A
AA
3/29
AA
AA
AA
AAA
AAA
AA
AA
AA
A
AA 9/30
10/2 AA Implementation
AAA
A
AA
AA
AA
AAA
AAA
AA
AA
A Design & Implementation
A
AA
4/3 AA Simulation
A
A
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
AA
4/3 A AnalysisATools Design & Implementation
A
AA
A
A
A
AA
Pro jected project schedule, part 3.
Technical Design Report for the BABAR Detector