Cross-connected Waveguide Lines as Standards for Millimeter- and
Submillimeter-wave Vector Network Analyzers
N M Ridler and M J Salter
National Physical Laboratory, Teddington, TW11 0LW, UK, nick.ridler@npl.co.uk
Abstract — This paper describes some investigations into
establishing primary standards of loss for waveguide Vector
Network Analyzers (VNAs) operating at millimeter- and
submillimeter-wave frequencies. The standards comprise straight
sections of waveguide, where the waveguide line is orientated
such that the waveguide aperture is at right-angles to the
waveguide apertures on the VNA test ports. This ‘crossconnected’ waveguide forms a section of waveguide that is
effectively below cut-off. The mechanical discontinuity between
the cross-connected waveguide and the VNA test ports also
generates significant reflection. The combined effect due to these
two loss mechanisms – cut-off attenuation and discontinuity
reflection loss – can be predicted from electromagnetic theory
and so can be used to establish sections of cross-connected
waveguide, of various lengths, as primary standards of loss. The
paper describes these standards in detail and compares
experimental results, obtained using a VNA operating in the
50 GHz to 75 GHz band, with values predicted by
electromagnetic modeling software.
Index Terms — Attenuation measurement, Electromagnetic
modeling,
Measurement
standards,
Millimeter
wave
measurements, Submillimeter wave measurements, Waveguide
junctions
I. INTRODUCTION
In recent years, there has been a significant increase in the
available upper operating frequency of Vector Network
Analyzers (VNAs). This has enabled VNAs to be used at
millimeter and submillimeter wavelengths. With this increase
in frequency comes a need for verifying the performance of
these VNAs at these operational frequencies. Traditionally,
this has been done in waveguide at lower frequencies using
VNA waveguide verification kits containing devices such as
sections of precision waveguide (with close to zero reflection
and transmission loss) and sections of reduced-height
waveguide (with significant reflection and transmission loss)
[1] – [3].
However, at high millimeter- and submillimeter-wave
frequencies, reduced-height waveguide becomes difficult to
manufacture reliably and to the necessary degree of accuracy.
This is because the height of conventional waveguide, with an
aperture width-to-height ratio of 2:1, is already very small.
For example, waveguide used at 1 THz has nominal
dimensions of (250 x 125) µm. To further reduce the height
of such waveguide, in a controlled manner, is mechanically
challenging in the extreme. New methods are therefore
required to realise waveguide verification devices that exhibit
significant reflection and transmission loss. It is further
desirable if the characteristics of these devices can be
predicted from first principles, or, using electromagnetic
modeling software, thus making them ‘calculable’ devices
suitable for primary standards applications.
This paper describes an investigation into one type of
device – referred to here as “cross-connected” waveguide (or
“cross-guide”, for short). This device is simply a section of
precision waveguide that is connected so that the waveguide
aperture is at right-angles to the usual connection orientation
of the waveguide. The cross-guide is aligned using the
alignment dowels found on the conventional UG-387
waveguide interfaces on the VNA test ports.
This paper presents experimental results obtained on some
cross-guide lines in the WR-15 waveguide size at frequencies
from 50 GHz to 75 GHz, this being the conventional operating
band for this waveguide size [4]. These results are compared
with electrical performance predicted using electromagnetic
modeling software – in particular, CST Microwave Studio [5].
Finally, the paper contains recommendations on how these
devices can be improved for use in all other waveguide sizes
used at millimeter and submillimeter wavelengths [6].
II. CROSS-GUIDE REALIZATION
A cross-guide device can be realized by connecting a short
section of conventional waveguide (i.e. a line) so that its
aperture is at right-angles to the apertures of two
conventionally-oriented waveguides. For example, in a
test/measurement configuration, these conventionally-oriented
waveguides could be the test port reference planes of a VNA.
A diagram illustrating this connection strategy for the
waveguide apertures is shown in Fig. 1.
For the investigations described in this paper, the crossguide line sections were aligned using the dowel pins on a
conventional MIL-DTL-3922/67D (often called UG-387) type
interface [7] – see Fig. 2. The sections of waveguide that were
used to establish the cross-guide lines were relatively short
and did not contain the dowel pins shown on the interface in
Fig. 2. (These short lengths of waveguide, when connected in
the conventional manner, can be used as Line standards during
VNA calibration techniques such as TRL [8], LRL [9], etc.)
One of these lines is shown in Fig. 3 – where Fig. 3a shows
the waveguide line with its aperture orientated in the usual
way for connection to the interface in Fig. 2.
Dowel pins
Fig. 1
Diagram to illustrate a section of cross-guide inserted
between two conventionally-oriented waveguides
A key feature with the UG-387 interface is that the dowel
pins and dowel holes are equally spaced and positioned
symmetrically around the connecting face of the interface.
This means that the waveguide line shown in Fig. 3a can also
be connected to the interface in Fig. 2 after the line has been
rotated through 90°, so that the aperture of the line is at rightangles with respect to the aperture shown in Fig. 2. This
orientation of the waveguide line section is shown in Fig. 3b,
which shows the same waveguide line from Fig. 3a except
with the aperture orientated at right-angles to the interface in
Fig. 2.
The alignment accuracy of waveguides connected using
UG-387 interfaces is governed solely by the accuracy of the
diameter and position of the dowel pins and dowel holes found
on UG-387 interfaces used to make the connection. Since
both the conventional connection and the cross-guide
connection of the line shown in Fig. 3 uses essentially the
same dowel pin/hole alignment process (i.e. two dowel pins
on each VNA test port fitting into the four dowel holes on the
waveguide line), the expected alignment accuracy for both the
conventional and cross-guide connections will be the same.
This indicates that the quality of connection (in terms of
alignment and repeatability) for a cross-guide line will be
comparable with that of a conventionally-connected line
(e.g. when used during a VNA calibration process).
Fig. 2
Conventional MIL-DTL-3922/67D (UG-387) waveguide
interface, showing the two dowel pins
Outer dowel holes
Fig. 3a A short waveguide line, showing the four outer dowel holes
that are used for aligning the waveguide during connection to UG387 interfaces. Note that the conventional waveguide aperture widthto-height ratio is 2:1
Aperture width-toheight ratio is now 1:2
III. MEASUREMENTS AND MODELING
A. Measurements
A series of measurements were made on some candidate
cross-guide lines using NPL’s Primary Impedance
Measurement System (PIMMS) [10]. A millimeter-wave
VNA was configured with WR-15 waveguide test ports.
These test ports were established as reference planes by
performing a ‘Thru-Reflect-Line’ (TRL) calibration [8] using:
a “¼-wave” line (as the Line standard); a flush short-circuit
connected, in turn, to both test ports (as the Reflect standard);
and, joining the test ports together (as the Thru standard).
Fig. 3b The same waveguide line shown in Fig. 3a, except rotated
through 90° so that the ratio of the waveguide aperture width-toheight is now 1:2. The outer dowel holes remain in the same location,
due to their symmetrical positioning around the face of the interface
Three relatively short waveguide lines, with approximate
lengths 1.6 mm, 2.0 mm and 3.6 mm, were connected as
cross-guide lines to the VNA test port reference planes. In
addition, a relatively long cross-guide line (of approximately
50 mm) was also measured by the VNA in order to investigate
if signals in a section of cross-guide line could propagate
along relatively long lengths of line.
B. Modeling
The S-parameters of three different lengths of WR-15 crossguide line were simulated using the time domain solver in
CST Microwave Studio [5]. The cross-guide line lengths
simulated were 1.6 mm, 2.0 mm, and 3.6 mm corresponding
to the three shortest measured cross-guide lines. The longest
measured cross-guide line of length 50 mm was not simulated.
In the simulations, the background material in the
computational domain was set to perfectly electrical conductor
(PEC) and vacuum bricks were defined to represent two
aligned sections of WR-15 waveguide, corresponding to the
two VNA test ports with an orthogonally oriented section of
WR-15 waveguide connected between them, corresponding to
the cross-guide. The frequency range of the simulation was set
as 50 GHz to 75 GHz. Energy was fed into and extracted from
the computational domain by means of waveguide ports at the
inputs to the two aligned waveguide sections. The
computational domain was discretized using a hexahedral
mesh. The S-parameters of the cross-guide alone were deembedded by shifting the reference planes from the waveguide
ports to the cross-guide ports. Since all the waveguides in the
simulation were assumed to be lossless, the de-embedding
only affected the phase of the S-parameters. Because of
symmetry and reciprocity the simulated S-parameters S11, S22,
S21 and S12 satisfied the following conditions: S11 = S22 and
S21 = S12.
Fig. 4
Transmission for the 1.6 mm cross-guide line: measured
(PIMMS); modeled (CST MWS)
Fig. 5
Transmission for the 2.0 mm cross-guide line: measured
(PIMMS); modeled (CST MWS)
IV. RESULTS
Figs. 4, 5 and 6 show plots of the measured transmission
magnitude, as a function of frequency, for the three short
lengths of cross-guide line (i.e. 1.6 mm, 2.0 mm and 3.6 mm).
These plots show that, for all three cross-guide lines, the
transmission magnitude varies smoothly with frequency. The
plots also show that each cross-guide line produces a different
amount of attenuation – the longer the cross-guide line, the
greater the amount of attenuation. This suggests that short
lengths of cross-guide line can be used to realize standards of
attenuation – the value of attenuation being related to the
length of the cross-guide line.
Fig. 6
Transmission for the 3.6 mm cross-guide line: measured
(PIMMS); modeled (CST MWS)
Figs. 4, 5 and 6 also show the modeled transmission
magnitude, for the same three cross-guide lines. The modeled
transmission values show the same type of variation with
frequency and with length of cross-guide line. Any
discrepancy between the measured and modeled electrical
performance should be accountable by the uncertainties in
both the measured and modeled values.
Fig. 7 shows a plot of the measured transmission magnitude
for the 50 mm long cross-guide line. This Fig. also shows a
plot of the measured transmission between the two VNA test
port reference planes when both ports are terminated with
short-circuits. The measurement of the two short-circuits
provides an indication of the level of isolation between the
two VNA test port reference planes, since effectively no signal
should be transmitted under these conditions. Since the
measured transmission for the 50 mm cross-guide line and the
two short-circuits is very similar (at around –80 dB), this
indicates that no detectable amount of signal propagates along
the full length of the 50 mm cross-guide line. This suggests
that the electromagnetic wave in the cross-guide line is not
able to propagate in the usual fundamental TE10 mode but
instead decays evanescently as the wave enters the cross-guide
section of the line – i.e. the cross-guide section of line acts as
a waveguide below cut-off [11].
band1 and upper frequencies for this waveguide band
(i.e. 50 GHz, 62 GHz and 75 GHz).
TABLE I
MEASURED ATTENUATION FOR THE THREE CROSS-GUIDE LINES
AT 50 GHZ, 62 GHZ AND 75 GHZ.
Attenuation (dB)
Cross-guide
length (mm)
50 GHz
62 GHz
75 GHz
1.6
24.29
17.63
11.39
2.0
28.96
21.36
13.55
3.6
47.40
36.05
21.70
These attenuation values are also plotted in Fig. 8, where it
is clear that the attenuation values for the three cross-guide
lines, at each frequency, lie on a straight line. This indicates
that it is feasible to design cross-guide attenuation standards,
choosing the length of the cross-guide line to give the required
value of attenuation. A series of such devices could then be
used as verification standards for VNA transmission
measurements by providing a wide range of known values of
attenuation.
Fig. 8: Attenuation versus cross-guide line length at 50 GHz, 62 GHz
and 75 GHz
Fig. 7
Transmission for the 50 mm cross-guide line and the
isolation measurement
A straight-line fit to the values at 62 GHz has the following
equation:
α = 9.335 L + 2.662
V. DISCUSSION
In the previous section, it was suggested that short lengths
of cross-guide line could be used to realize standards of
attenuation. To investigate this further, selected attenuation
values for the three short cross-guide lines measured
previously are shown in Table I, below – at the lower, mid-
(1)
where L is the length of the cross-guide line and α is the
associated attenuation (dB). This equation can be re-arranged
to show what length of cross-guide line needs to be used to
achieve a required value of attenuation:
L = (α - 2.662)/9.335
1
(2)
The mid-band frequency value used here is the
approximate geometric mean frequency of the recommended
lower and upper frequencies for this waveguide size.
For example, Table II gives calculated values of cross-guide
line length, using (2), to give 10 dB steps of attenuation in
WR-15 at 62 GHz. Table II shows that changing the length of
the cross-guide line section by approximately 1.07 mm
produces a change in attenuation of 10 dB at 62 GHz.
TABLE II
CALCULATED VALUES OF CROSS-GUIDE LINE LENGTH TO
PROVIDE KNOWN STEPS OF ATTENUATION AT 62 GHZ
Attenuation (dB)
10
20
30
40
50
60
Required crossguide length (mm)
0.79
1.86
2.93
4.00
5.07
6.14
the phase varies smoothly across the full waveguide band.
However, the phase at any given frequency does not vary
appreciably with the length of the line. For the transmission
phase, this further suggests that the wave in the cross-guide
line section is not propagating using a conventional waveguide
mode but is instead decaying evanescently with no appreciable
phase change associated with the wave. It is also interesting to
note that there is an almost constant 90° phase difference,
across the full band, between the reflection and the
transmission coefficient phases.
VI. OTHER S-PARAMETERS
Although the main focus of this paper has been the use of
cross-guide lines to develop primary standards of attenuation,
it is also informative to examine the reflection properties of
these lines, and also to look at the phase behavior of both the
reflection and transmission coefficients.
Fig. 9 shows a plot of the measured linear reflection
coefficient magnitude of the three short lengths of cross-guide
line measured in the previous sections. This Fig. shows that all
lines exhibit very high reflection across the full waveguide
band. The amount of reflection is only mildly dependent on
the length of the line – the longer the line, the higher the
reflection coefficient. This can be seen clearly at 75 GHz –
i.e. the maximum recommended operating frequency for this
waveguide size.
Fig. 10
lines
Reflection coefficient phase for the three short cross-guide
Fig. 11 Transmission coefficient phase for the three short crossguide lines
VII. IMPROVING CROSS-GUIDE ALIGNMENT
Fig 9
Linear reflection coefficient magnitude for the three short
cross-guide lines
Figs. 10 and 11 show plots of the phase of the reflection and
transmission coefficients, respectively. Both Figs. show that
The mechanism used for aligning the cross-guide lines
investigated in this paper has been the dowel pins and dowel
holes as specified for the conventional UG-387 interface [7].
However, it is well known that this interface does not provide
good alignment accuracy for waveguides used for precision
measurements at higher millimeter- and submillimeter-wave
frequencies. This has led some manufacturers to introduce a
‘precision’ version of the UG-387 interface (see, for
example [12]) which contains two additional alignment holes
above and below the waveguide aperture (at ‘north’ and
‘south’ positions). Fig. 12 shows a photo of a ‘precision’
UG-387 interface that contains these two additional inner
dowel holes.
Inner dowel holes
decay (i.e. attenuation) and path length (i.e. cross-guide line
length). In this respect, the cross-guide line acts like a
waveguide below cut-off (WBCO) [11]. Such a line can
therefore be considered as a primary standard of attenuation in
any waveguide size that can be realized to a suitable degree of
accuracy. However, cross-guide lines are considered to be
most applicable in small waveguide sizes where other devices
(such as reduced-height waveguide) become difficult to
realize reliably.
For any given waveguide size, a series of cross-guide lines
of different but known lengths will provide a series of
‘known’ values of attenuation. These devices could be used to
construct waveguide VNA verification kits in any waveguide
size. Similarly, such devices could also be used to verify the
linearity of a VNA – for example, during a measurement
uncertainty evaluation process.
Addendum
Fig. 12 ‘Precision’ UG-387 waveguide interface, showing the two
inner dowel holes at ‘north’ and ‘south’ of the waveguide aperture
During connection, two precision dowel pins are inserted
into these holes, and into similar holes on the device under test
(assuming the device is also fitted with the precision version
of the UG-387 interface). In fact, the waveguide line shown in
Fig. 3 already contains these inner dowel holes. However,
when this line is rotated through 90° to form a cross-guide
line, these inner dowel holes are now positioned at ‘east’ and
‘west’ positions (see Fig. 3b), rather than ‘north’ and ‘south’
positions (as in Fig. 3a), and so these holes no longer align
with the inner dowel holes found on the ‘precision’ UG-387
interfaces. Therefore, a method of improving the alignment of
cross-guide lines used with ‘precision’ UG-387 interfaces
would be to include an extra pair of precision inner dowel
holes (situated at ‘east’ and ‘west’ with respect to the
conventionally-aligned aperture) so that these additional holes
can be used to align the waveguide, with respect to ‘precision’
UG-387 VNA test ports, when the waveguide is configured as
a cross-guide line. This alignment process will enable
cross-guide lines to be aligned with the same degree of
accuracy as conventionally-orientated waveguides connected
using ‘precision’ UG-387 interfaces. This will facilitate the
use of cross-guide lines as accurate primary standards at all
millimeter- and submillimeter-wave frequencies.
VIII. SUMMARY
The cross-guide lines investigated in this paper have
exhibited some interesting properties. Firstly, it appears that
the wave in the cross-guide line does not propagate in a
conventional manner (e.g. using the TE10 mode) – instead, the
wave becomes evanescent and decays exponentially. This
suggests that there is a direct relationship between signal
Since having this paper accepted for publication, the
authors have become aware of another published paper
that has already described the use of cross-connected
waveguide for verifying scattering parameter
measurements [13]. In [13], the term “90° shim” is used
to describe the type of device that is called “crossconnected” waveguide in this paper.
It is clear that [13] was responsible for introducing this
type of device for the purpose of verifying scattering
parameter measurements. The authors of this current
paper apologize to the authors of [13] for not referring to
their work in the version of this paper that was
distributed at the ARFTG conference.
ACKNOWLEDGEMENT
The work described in this paper was funded by the
National Measurement Office of the UK government’s
Department for Business, Innovation and Skills, UK.
REFERENCES
[1] R W Beatty, "Calculated and Measured S11, S21 and Group
Delay for Simple Types of Coaxial and Rectangular Waveguide
2-port Standards" NBS Technical Note 657, December 1974.
[2] Maury Microwave, “Two-port Standard Sets”, Technical Data
Sheet 3A-322A, April 1998.
http://maurymw.com/pdf/datasheets/3A-322A.pdf.
[3] Agilent Technologies, “11645A R, Q, U, V, W Waveguide
Verification Kits User's and Service Guide”, Part Number
11645-90013, November 2004.
[4] MIL-DTL-85/3C, “Waveguides, Rigid, Rectangular (Millimeter
Wavelength)”, October 2005.
[5] www.cst.com. (Computer Simulation Technology company
web-site.)
[6] N M Ridler and R A Ginley, “IEEE P1785: A New Standard for
Waveguide Above 110 GHz”, Microwave Journal, Vol 54,
No 3, Cables & Connectors Supplement, pp 20-24, March 2011.
[7] MIL-DTL-3922/67D, “Flanges, Waveguide (Contact) (Round,
4 hole) (Millimeter)”, December 2009.
[8] G F Engen and C A Hoer, “Thru-Reflect-Line: An improved
technique for calibrating the dual six-port automatic network
analyser”, IEEE Trans, Vol MTT-27, No 12, pp 987-993,
December 1979.
[9] C A Hoer and G F Engen, “On-line accuracy assessment for the
dual six-port ANA: extension to nonmating connectors, IEEE
Trans, Vol IM-36, No 2, pp 524-529, June 1987.
[10] N M Ridler, “News in RF impedance measurement”, XXVIIth
General Assembly of the International Union of Radio Science
(URSI), paper no 437, session A1, Maastricht Exhibition and
Congress Centre (MECC), The Netherlands, August 2002.
[11] R N Griesheimer, “Cut-off attenuators” in C G Montgomery
(Ed), “Techniques of microwave measurements”, McGraw Hill,
1947.
[12] C Oleson and A Denning, “Millimeter-wave Vector Analysis
Calibration and Measurement Problems Caused by Common
Waveguide Irregularities”, Proc. 56th ARFTG Microwave
Measurement Conference, Boulder, CO, December 2000.
[13] T Schrader, K Kuhlmann, R Dickhoff, J Dittmer and M Hiebel,
“Verification of scattering parameter measurements in
waveguides up to 325 GHz including highly-reflective devices”,
Adv. Radio Sci., Vol 9, pp 9-17, 2011.
View publication stats