dielectric strength after internal failure on Table III. ASTM 495-48T Arc-Resistance ASTM Bulletin No.

ASTM Bulletin No. 185, iu«, Oct. 1952, pp.

31-38.
a 60-cycle high-potential test. This is Test
3. TRACKING IN SOLID INSULATING MATERIALS-
apparently related to the absence of in- A CRITICAL RESUME, V. E. Yarsley, W. J. Grant.
ERA Technical Report, Ref. LIT 177, The British
ternal carbonization and to the formation Average Time,
Electric and Allied Industry Research Association,
Material Seconds
of gas under pressure in the failure-path London, England, 1948.
cavity. Phenolic..... . . . . . . . . . . . . . . . . . . . . . . . . 4
4. TRACKING IN SOLID INSULATING MATERIALS-
CONSIDERATION OF REPLIES TO QUESTIONNAIRE,
An interesting experiment was con- Urea.... . . . . . . . . . 125 V. E. Yarsley. ERA Technical Report, Ref·
ducted with a hypodermic needle inserted Melamine 125 LIT 228, ius., 1950.
Vulcanized iibre.. . . . . . . . . . . . . . . . . .. . .125
between the electrodes which confirmed Phenolic utilizing internal oxidation 5. INSULATION, H. Warren. Journal, Institution
mechanism 185 of Electrical Engineers, London, England, no. 528,
the presence of gas under pressure after 1940, pp. 588-96.
the initial breakdown. As long as the 6. VULCANIZED FIBRE: NEW LOOK AT AN OLD
hypodermic needle was open to the at- PLASTIC, G. Albert. Materials and Methods,
New York, N. Y., Sept. 1954, pp. 110-14.
mosphere, the gas formed internally could
sistance to surface-creepage damage has 7. ApPLICATION OF NONTRACKING BUTYL TO
not build up pressure, and the sam ple HIGH-VOLTAGE INSTRUMENT TRANSFORMERS, R. A.
been observed in laboratory tests and Pfuntner, R. S. Norman, B. W. Wilterdink. AlEE
showed only the dielectric strength of air Transactions, vol. 74, pt. III, 1955, pp. 1173-76.
under severe outdoor exposures. The for-
at atmospheric pressure. When the 8. THE PORCELAIN INSULATOR-SILENT PARTNER
mation of carbonized tracks is eliminated
needle was closed to the atmosphere, the ON POWER TRANSMISSION, H. A. Frey. General
under these conditions. Standard ASTM Electric Review, Schenectady, N. Y., Nov. 1954.
voltage could be raised slowly to over 50 pp.52-56.
arc-tracking tests indicate that phenolics
per cent of the original breakdown value. 9. INSULATOR SURFACE CONTAMINATION, H. A.
using this mechanism can now be made
By opening the hypodermic needle gas was Frey. AlEE Transactions, vol. 67, pt. II, 1948,
superior to the best conventional arc- pp. 1420-25.
released and the dielectric strength again
resistant molding compounds. The in- 10. FLASHOVERS OF SUSPENSION INSULATORS DUE
dropped to the value for air. This cycle TO CONTAMINATION, H. A. Adler, W. H. Wickham,
ternal oxidation mechanism has also been M. S. Oldacre. Ibid., pp. 1680-85.
was repeated several times with the same
observed to prevent the formation of con- 11. PERFORMANCE OF DUST-CONTAMINATED IN-
results. A further study of this effect is
ducting carbon following an internal SULATORS IN FOG, B. Cozzens, T. M. Blakeslee.
progressing to obtain a better under- Ibid., pp. 1686-92.
dielectric failure.
standing of this phenomenon. 12. ARC-RESISTANCE TEST. ASTM 495-48T,
American Society for Testing Materials, Phila-
delphia, Pa., 1948.
Summary References
13. LETTERS TO INSULATION FORUM, W. Vassar,
C. Lell. Insulation, May 1955, p. 7.
1. ARC RESISTANCE: (I) TRACKING PROCESSES IN
It has been discovered that the appli- THERMOSETTING INSULATING MATERIALS, M. 14. PLASTICS ENGINEERING HANDBOOK, The
Olyphant, Jr. ASTM Bulletin No. 181, American Society of the Plastic Industry, Inc. Reinhold
cation of the internal oxidation mecha- Society for Testing Materials, Philadelphia, Pa., Publishing Corporation, New York, N. Y., 1954.
nism to organic materials is highly effec- Apr. 1952, pp. 60-67. 15. A NEW DRy-TYPE INSULATION FOR INSTRU-
tive in preventing carbonization under 2. ARC RESISTANCE: (II) EFFECT OF TESTING MENT TRANSFORMERS, R. A. Pfuntner, R. E.
CONDITIONS ON TRACKING PROPERTIES OF THERMO- Franck, F. R. D'Entremont. AlEE Transactions.
arcing conditions. Greatly improved re- SETTING INSULATING MATERIALS, M. Olyphant, Jr. vol. 69, pt. 1, 1950, pp. 453-58.

axial ampere-turn distributions. This

TransFormer Reactance Calculations problem has been considered in the
past,I-5 but except for Roth's work"

with Digital Computers . and unpublished work done within the

General Electric Company by Lieb-
inger, Poritsky, and Tino, there has
been no consideration given the dual
L. RABINS problem of axial ampere-turn distribution
ASSOCIATE MEMBER AlEE
and the effect of curvature. I t has be-
come apparent that these two effects

S I NCE one of the important features and permits a major improvement in this must be considered together and not
of a transformer is its series or leak- phase of transformer design. separately, as in the past. Also in the
age impedance, improvements in the cal- In a cylindrical coil-core type of past, investigators looked for simplifica-
culation of this quantity are always transformer, most of the difficulties in tion of rigorous equations. In the
searched for. In power transformers reactance calculation stem from axial present work, the search has been for an
this impedance is almost all formed by ampere-turn discontinuities and un- accurate method adaptable to a digital-
its reactive component so that the balance between windings. The react- computer calculation.
accuracy of the impedance calculation ance of normal transformer designs in Roth's analysis of this dual problem
hinges on the accuracy of the leakage which both of these occur to a slight has followed in principle his double
reactance calculation. Modern methods extent is accurately calculated by simple
of calculation using digital computers design equations. However, the design Paper 56-59, recommended by the AlEE Com-
puting Devices and Transformers Committees
have been adopted so that very accurate which may occur, for example, because of and approved by the AlEE Committee on Tech-
reactance calculations can be made the request for extraordinary impedance nical Operations for presentation at the AlEE
Winter General Meeting, New York. N. Y ..
whenever they may be required in the relationships in a 3-winding transformer January 3O-February 3, 1956. Manuscript sub-
design of a particular transformer. The or as to parallel with a transformer having mitted September 29, 1955; made available for
printing November 18, 1955.
use of a digital computer makes this a peculiar impedance variation with the
L. RABINS is with the General Electric Company,
calculation quick, easy, and painless, voltage tap position, leads to unusual Pittsfield, Mass.

JULY 1956 Authorized licensedRabins-Transformer ReactanceonCalculations

use limited to: KOREATECH. Downloaded with
February 14,2024 Digital
at 04:30:33 Computers
UTC from IEEE Xplore. Restrictions apply. 26]
 YOKE

c, C

CORE
LEG WDG
ro
I 2
t REG'ION
3
REGIONJ
2-
, REGION
I
r,
r2
r3

'4

Fig. 1. Transformer with two cylindrical windings Fig. 2. TraftSformer with only one cylindrical windin,gJ shown

Fourier series method," which he earlier for the vector potential (which has only Cl, C2 = radial builds of windings; 1 and 2
applied to a reactance calculation in an angular component) is found. The Dr = effective area of leakage- field for com-·
pletely uniform flux field
which curvature was neglected. With magnetic field energy!" can now be com-
= [(r2+rl) /2] (C1/3) + [(r3:+r2) /2]g+
curvature considered, his analysis results puted and the leakage inductance found . u-,
+r3) /2] (C2/3)
in expressions difficult to evaluate not g = radial' gap between windings = r« - r»
only because Bessel functions are involved Results and Calculations Procedure n = index over which series is summed = 1,.
but because the arguments of these 2,3, ...
The final equation used to compute m = (n7r) /l
functions are themselves functions of re, r-, '1"2, ra, r, = radii defined by Fig. 1
boundary dimensions and are found by reactance is
x=m1!
solution of certain transcendental equa- 0.126 f kva ('"};!{!)"J harmonics of the pu ampere-turn.
Iln,,ff~'A =
tions. %ox= e2
--X
I
per inch distribution of windings 1
and 2
This paper is concerned with another
CI2_C22) O.063fkval Ih KI, L 1 = modified Bessel and Struve
and simpler form of solution to this (( Dr ) +--- + X functions of first order
problem, very similar in principle to 12 e2
The first term of the reactance equa-
Rogowski's" single Fourier series method,
tion results from the average ampere-
and classical methods of solution are
turns in each winding, the average being
employed." The solution presented here-
taken over the length 1. Flux lines
in is in the form of a Fourier series the
corresponding to this reactance flow in
coefficients of which are Bessel and
a completely uniform vertical pattern
Struve functions, all of which are func-
over the en tire length 1 and the reactance
tions readily programmed for computa-
derivation is quite simple, appearing in
tion by an International Business Ma-
several references, 12 so it will not be
chines (IB1I) card-programmed cal-
presented here. The remaining tenus
culator.
result from the nonuniform axial ampere-
The basic building block in the react-
turn distribution. For most designs,
ance calculation of any transformer is
symmetry exists about the axial center-
that between two windings. The re-
line so that calculations rna y be made on
actance of any combination of windings
a half-stack basis. If calculations are
can then be found by other methods,
based on 1 pu ampere-turn in a full coil
the most useful being that given by
stack and the length 1 is half the distance
Garin and Paluev. 10 Only the basic
between yokes, the same equation multi-
2-winding reactance calculation is being
plied by 2 is used.
considered herein.
The harmonic coefficients of the
Outline of Problem and Method ampere-turn distribution are given by
of Analysis (see Appendix I)

2
Fig. 1 ShOVIS a transformer with two In =-(oa-bsin ma+ob-c sin mb+
n1r
cylindrical windings. The yokes· above ... 0 l-d sin md) (2)
and below the coils are assumed of
f= frequency, cycles per second
infinite extent and are completely cir- kva = kilovolt-ampere per core leg The calculation is performed on an
cular so as to fully cover the windings. e = volts per turn IBM card-programmed calculator in a
This permits the use of a Fourier series 1= axial length over which ampere-turn lO-digit floating decimal system. The
to describe the axial ampere-tum dis- distribution is defined; will usually calculation program is divided into two
be either half or the whole distance
tribution of each winding. The general steps. First is the calculation of the
between yokes
differential equations describing a mag- 'l;NI = sum of per-unit (pu) ampere-turns harmonic coefficients. Here each wind-
netic field 11 are set up and the solution in a coil stack ing is considered separately and broken

262 Rabins-Transformer Reactance Calculations with Digital Computers

Authorized licensed use limited to: KOREATECH. Downloaded on February 14,2024 at 04:30:33 UTC from IEEE Xplore. Restrictions apply.
JULY 1956
into a minimum number of sections, each second units are used. Define a vector
section having uniform ampere-turns potential by
pu axial height. The dimensions and C and D are constants determined
VXA=B
total pu ampere-turns of each section by boundary conditions. Consider one
are punched on IBM cards and are V·A=O (4)
winding at a time and the three regions
input data to the calculator. The pro- Substituting the foregoing results in 11 defined by Fig. 2. The solutions to the
gram is arranged to handle any number
differential equations in the three regions
of sections and, by omitting parts of the (5)
are given respectively as
program deck, the length of the calcula-
The vector i defines the current density
tion is made proportional to the number A 1n = C1I1(mr)+D1K1(mr)
in the winding and since it has 'only an
of sections. After the first harmonic
angular componen t, this will also be the
is evaluated, the integer n is advanced
only component of the vector potential.
one, and an identical calculation is
Denoting this component by the subscript (14)
made, and this procedure is repeated
until all necessary harmonics have been (6)
The boundary conditions reveal in-
calculated. Harmonic coefficients are formation which permit evaluation of the
In circular co-ordinates this becomes
found for both windings in this fashion. constants C and D. These stem from
Following this, the actual reactance c)2AO 1 oAo Ao o2A .
--+- ---+-=-P,'t8 (7) continuity of the flux densities. From
calculation is made. This requires addi- or 2 r or r 2 OZ2
the definition of the~vector potential
tional input, frequency, kva per leg,
volts per turn, and all radial dimensions Describe the current density by a Fourier
oAo
series (see Appendix I) B r - -02
-
beginning at the center line of the core r:

and extending to the outer winding.

The harmonic coefficients are also neces- io=io + 2: in cos mZ (8) lorA (15)
I
Bz = - --
sary input. These coefficients will have n=l r c>r
been punched on IBM cards during their where n = 1, 2, 3,.... Assume that At the iron boundaries, the tangential
calculation. Calculation then proceeds the vector potential can be similarly flux is zero since infinite permeability
through the evaluation of the Bessel developed iron was assumed. Thus
and Struve functions and the reactance,
2:
co
Br=O'atZ=±l (16)
one harmonic at a time, each harmonic Ao=Ao(T)+ An(r) cos mZ (9)
calculation procedure again being iden- n=l This yields the equation
tical except for a change in harmonic
Substituting the foregoing, a set of
2: -mAn sin ml=O
(X)

order.
equations is obtained
n=l

Derivation of Reactance Equation d 2A

o
--+-
1 dA o A o
---=-p,'to
.
(10)
or
dr? r dr r2
n7r (17)
The following is assumed in the deri- m=T ,n=l, 2, 3,
d 2A n 1 dAn An .
vation: --+-r _._-
dr? dr r2
-m 2 A n = - P,'Z n (11)

1. The yoke is infinitely large, completely

The ether boundary conditions are
covers top and bottom of the windings, and for each n. This set of equations is
has infinite permeability. obtained because of the orthogonality of B z =0 at r=ro
2. Circular symmetry exists. cos mZ. The first of these equations is BIZ =B 2Z ; BIT =B2r or A 1n =A 2n at r=rl
easy to solve and corresponds to flux
3 The effect of adjacent windings and B 2z =B 3Z ; B 2r = Bs; or A 2n =A 3n at r=r2
core legs, tank walls, and current-carrying flowing in only the Z-direction. The
leads is negligible. reactance resulting from this component B 3 or As.....O as r..... co (18)
4. The field is unbounded radially. of the vector potential is the first term
The last condition requires that
in the reactance equation previously
5. The core leg is round and has infinite
permeability. given in the section entitled' 'Results and C3=O (19)
Calculation Procedure." It is the set
n Each winding is uniform radially.
since II (mr) is an exponentially increasing
of equations for general n which is mostly
7. The effect of normal ducts and insula- function.
of interest. If the right-hand side of
tion between coil sections is negligible.
the foregoing equation were zero, the For any value n, the general value of
8. The sum of the total ampere-turns in Bz n is
equation would be recognized as Bessel's
the two windings is zero.
and the solution would be 13
7rJl.in
The field equations appropriate to Bzn=m ( Clo(mr)-DKo(mr)- 2m 2 Lo(mr) )
this problem of reactance calculation are (12)
(20)
Y·B=O where II and K 1 are modified Bessel
functions of the first and second kind, Let mr=x and
VXH=i first order. A particular solution to the
complete equation may be found by the (21)
B=JLH (3)
method of quadratures" or by recogni-
where B, H, and i are flux density, tion that it is satisfied by a modified Substitute the rernammg boundary
magnetic field strength, and current Struve function L 1(mr).l3- 15 The com- conditions in the equations 14 and 20
density respectively. Meter-kilogram- plete solution for general n is then for An and Bz« to obtain

JULY 1956 Authorized licensed Rabins-Transformer ReactanceonCalculations

use limited to: KOREATECH. Downloaded with
February 14,2024 Digital
at 04:30:33 UTCComputers
from IEEE Xplore. Restrictions apply. 263
C1lo(xo) - DtKo(xoJ =0 lin/(NI) = I1n(pu) =pu ampere-turns
pu axial length (35)
C1I1(Xl)+D1K1(Xt) = C2I 1(Xl ) +
D 2KJ (XI)+in' LI(XI) 7rlp,N2lIne pu )l2n(pu)
C1lo(Xl) - D1Ko(Xl) = C2Io(xd-
REACTANCE DERIVED FROM VECTOR coef L = CC 4
t zm
(36)
POTENTIAL
D 2K o( Xl )+ i n'Lo(Xl ) (22)
With the vector potential known, the This is the coefficient of an expression
D 3K1( X2) = C211(X2)+D2KI(X2)+in'Ll(X2) for the leakage inductance; to change
leakage inductance may be found by
-DaK o(X2) = C21o(X2)-D2Ko(xl)+in'Lo(x2) to per-cent reactance multiply by
evaluating the energy stored in the
Utilize the equations magnetic field. If W is the magnetic kva
field energy, and L the inductance-! 27rf 105 - (37)
1 V2
lo(x)K1(x)+11(x)Ko(x) =- (see reference 13)
x where
(27)
x[L1(x)Ko(x)+Lo(x)K1(x)] =~ rXxKt(x)dx V=the rated voltage of a winding
'lfJo Both i and A have only angular com- kva = kva rating of a winding
f = frequency
.
x [L1(x)Io(x)-Lo(x)It(x)J = -
21:% xI1(x)dx ponents each given by a Fourier series.
dv is a differential volume = rdOdZdr. The foregoing coefficient has a length
'If 0
Thus dimension to the first power; changing

tlin
(23)

(i
from meters to inches and using the per-
The origin of the latter equations is
covered in Appendix II. Solving for
llJi·idv =111 o+ cos mZ) X
meability of free space of 41r X 10-7

+tl
fkva 1
C and D coef%x =0.063 (V/ N)2 I 1n(pu)I2n(pu) (38)
( Ao An cos mZ) dv (28)
The final expression is now found by
In the integration all terms not of the carrying out the integration. in equation
same order go to zero since 31 for reactance and is given as equation 1.
(29)
Bessel Function Computation
if ml=m2, or equals zero if ml~m2'
This leaves The general process by which the Bessel
and Struve functions were evaluated on
fffi . Adv = tff ioAord()dr + the card-programmed calculator con-
co
sisted of a computation using either a
L1/2ff inAnrdOdr (30)
power or asymptotic series.
n=l
The card program is set up so that for
Carrying the integration over only low arguments a power series is used
The solution for the vector potential is the windings since it is only here that and for high arguments an asymptotic
now complete. i ~ 0 and omitting the first term in the series. The changeover point is 6 for
Consider now that two windings are foregoing 10 , J' xlIdx, and J' xLvlx; and 3.5 for
present, as shown in Fig. 1. Two solu-
K o and fxK1dx. In general, 12 terms
tions of the same form as that just found
were used in the power series expansion,
may be superimposed by properly identi-
7 or less in the asymptotic. To obtain
fying regions 1, 2, and 3. The space
a smooth transition of the integrals at
harmonics of the ampere-turn distribu-
the changeover point, it was necessary
tion of windings 1 and 2 will be repre-
to adjust the last coefficient of the
sented by ii« and i 2n respectively. In Except for the integral of the BesseI
asymptotic series.
the region occupied by winding 1 functions, a typical term of this expression
The general principle involved in the
has the coefficient
computation of the series was built on
adding terms of the form
(32)
xC+D
Refer both windings to ones with an each such operation being performed in
equal number of turns N. Then, with one card cycle. A power series with
the current density being uniform over seven terms
the radial build, it may be written as
P =Ao+AlX+A2X2+A3X3+A4X4+
. lIn A ax5+A 6x6
'hn= C (33)
1
may be factored starting at the end to
where lIn is the ampere-turns pu axial give
length. Then
P =A +x[A 1+x(A 2+x{A 3+x[A 4+
x(A 5+xA 6 ) ] } ) ]
(34)
so that it is seen that one card operation
Multiplying numerator and denominator advances one term in the evaluation of
by N2 and recognizing the series. Based on a speed of 150

264 Rabins-Transformer Reactance Calculations with Digital Computers

Authorized licensed use limited to: KOREATECH. Downloaded on February 14,2024 at 04:30:33 UTC from IEEE Xplore. Restrictions apply.
JULY 1956
 0 0
0
0 0 0 T T
2.55 5' 5"
9.16 .50
I 4.59
.50

~--
12.25

r-
4 .7~ 1.27 6.98 1.27 4 .75 1.27 _~~8 1.27

•!
10
0) ~ It)
0
III
CD
f-o:
t::
.50 0
./
.50 0 22.9 18.36 ~
I 22.95 10.70 UJ
a:
.50 ~

~
0
o
33.8 37.09

B C
0
9.9

A
t

0 .209
2.07 .50 17.60
29.16
0 .048
12.83 6.00

4 .75 l27 4.23 1.27

12,5 3.776 3.026 2.216 4.133

.037 .398 0
.50 35.69 33.59 80.03
22.93 Fig. 3. Ampere-turn distribu-
.50 tions of tested transformers LU
a::
iTI7 0
o
A-Case 1 .048
B-Case 2 6.00
C-Case 3
D-Cases 4 and 5; case 5 has the ~ .&.Q.L
windings interchanged 26.98 17.60
o E-Case 6

cards per minute, a 12-tenn series is

evaluated in 5 seconds.
The coefficients of the series were
evaluated by a desk calculator using
general expressions.w-! The series for
the integrals were found by direct or if the solution had been left in the each block on the figures contain the
integration of the power series and by integral. form given by the method of pu ampere-turns (the top number) and
repeated integration by parts of the quadratures, the result for the reactance the axial height of the block (the bottom
asymptotic series. would have contained various combina- number). Distances to the yoke and
The functions were calculated over tions of terms such as holes in a winding are shown as blocks
the range 0 to 50 and compared to existing with zero ampere-turns. All values of
tables,17-20 or to combinations of existing pu ampere-turns are based on a total of
tables when direct comparison could which is not in a form in which the 1 pu ampere-turn in a stack. Table I
not be done. Better than 5-place ac- coefficients of the power or asymptotic summarizes the test results. In all
curacy was obtained throughout. series are readily found. cases the calculation was made through
It should be noted that considerable the tenth harmonic.
simplicity results in the determination Test Results Table I. Test Results
of the series of the functions to be
evaluated from the use of Struve func- Several comparisons between tests
Per-Cent X
tions and the recognition of the integrals and calculated reactances on production
which are equal to certain grou ps of transformers and models which were Case Kva Test Calculated
Bessel and Struve functions. If the built to get an exaggerated ampere-turn
1 1 . 63. . . 14 . 5. . . . . . . . . 13 . 75
Struve functions were not used as a distribution are listed in the following. 2 3.60 58.7 58.29
solution of the differential equation, but These examples are all normally difficult 3 3. 60 97. 6 95 . 53
4 3. 60 40.5 42. 10
a solution built up by superposition and to calculate. Fig. 3 shows the general 5 3. 60 43. 5 43. 66
summation of the effect of current sheets, winding configurations. The data within 6 2.53 34.3 35.99

JULY 1956 Authorized licensedRabins-Transformer

use limited to: KOREATECH. Reactance Calculations
Downloaded on with
February 14,2024 DigitalUTCComputers
at 04:30:33 from IEEE Xplore. Restrictions apply. 265
 Fig 4. Typical co

AMPERE
YOKE
axial ampere-turn 1=10 +L..J
"""" In cos mZ; n7r
m=--
TURNS PER n=l 1
UNIT-LENGTH
I .....- - - - - - ~
__1 -------11....
_
density plot
where

In =- 2fl I cos mZdZ

I 0
a b
Let
AXIAL LENGTH - r

All calculations are based on a value Concerning the assumptions used in

of unity volts per tum and arbitrary the derivation, only two are of signif-
Equation 2 for the harmonic coefficients
kva. Frequency was 60 cycles-per second. icance. The first concerns the yoke
of the ampere-turn density immediately
Case 6 involves a calculation between the being considered to cover the windings follows.
inside winding and the other two in and extending to infinity; whereas it
series. On a full-winding basis with the actually covers only a portion of the
middle and outer windings sharing 1 pu windings. It has been found by com-
ampere-turn, the turns are so divided parison to test results that the position Appendix II
that the total pu ampere-turns in the of the yoke has little effect, of the order
middle winding is 0.896 and in the outer ef 3 per cent, if it is about 4 or more
0.104. 1\ complete reactance calculation inches distant from the winding when the To prove the several relations in equation
23 certain equations are required for the
requires three independent 2-winding winding stack height is about 90 inches. derivatives of the Bessel and Struve func-
calculations, each based on 1 pu ampere- Since this separation usually corresponds tions H
tum in each winding. If the inner, to the normal insulation distances, the
d
middle, and outside windings are des- actual distance to the yoke is used in - l o(x ) =I1(x )
ignated by the subscripts 1, 2, and 3 dx
the calculation as being the distance to
respectively, then the calculated im- the assumed infinite yoke. For any d
pedance values are - Ko(x) = -K1 ( x )
specific case, if it appears that it may dx
"Yl_2 =40.39 cause error, calculations are made by
arbitrarily changing the yoke distance d 2
X 1 _ 3 = 112.42 - Loex) =- +L1(x)
over a large range and noting its effect dx 7r

X 2 _ S = 127.76 on the reactance. A compensating effect d

which tends to counterbalance the re- ~ xI1(x)=xlo(x)
Following the technique of Garin and
Paluev." the reactance actance increased caused by the infinite
yoke is the use of only ten harmonics in d
- xK1(x) = -xKo(x)
Xl -~:) = O.896X I _ 2+O.1041 _ 3 - the reactance calculation. While nor- dx
(O.896)(O.104)X2 _ 3 =35.99 mally only ten harmonics are required for d
accurate calculations, any number may - XLI (x) =xLo(x)
I t was not possible to test the values of dx
these reactances individually. be used. The second important as-
sumption is the neglect of adjacent wind- It may now also be verified that the solution,
Considering the extreme nature of equation 13, to the differential equation is
the cases listed, accuracy is good with ings and tank walls. These corrections correct. The integral relations given in
the maximum error of about 5 per cent. which are usually negligible are probably equation 23 were first obtained by noting
Some of this error may be due to manu- best handled emperically, though some a general equation given by McLachlan
qualitative effect may be found by and Meyers!' for
facturing tolerances or slight errors in
test. locating a highly permeable or highly
conducting cylinder outside the windings
an d by determining the reactances for and by comparing the equation thereby
Discussion and Conclusions obtained for
these cases. The reactance equations
Equations and computational tech- for these conditions may be found by the ! Io(x )dx and! Ko(x )dx
niques for the calculation of the leakage same methods as described herein.
to an integration by parts of
reactance of core-type cylindrical trans- It should be emphasized that the one
former windings have been presented. major item which makes the use of these !xI1(x)nxand!xK1dx
While emphasis has been placed on the accurate methods of reactance equations
use in transformer design, the results practical for regular design application This procedure does not have to be
followed here since the integrals may be
are applicable to air-core reactors also. is the modern high-speed digital con- verified by using the derivative equations
The interpretation of the reactance puter. in the foregoing and noting that
equation is straightforward. Each har-
monic con tribution is seen to consist of
two self-reactances terms and a mutual
reactance term. The reactances are
Appendix I
d 2
only that in a leakage field sense, i.e., - (xL1ex)I1ex)-xLo(x)I1(x» = - - xl1«x)
dx 7r
they do not include the contribution of A plot of the ampere-turns per unit axial
the core. This avoids the difficulty of height versus height is a series of straight- By making use of the facts that for small
line sections, as shown in Fig. 4. Over values of x
finding a small leakage reactance by the length l the ampere-turn density may
taking differences of very large reactances. be defined by a Fourier series Io(x) =1

266 Rabins-Transformer
Authorized licensed use limited Reactance
to: KOREATECH. Downloaded Calculations
on February with Digital
14,2024 at 04:30:33 UTC fromComputers
IEEE Xplore. Restrictions apply.
JULY 1956
 10. TRANSFORMER CIRCUIT IMPEDANCE CAL-
References CULATIONS, A. N. Garin, K. K. Paluev. AlEE
.KoCx)= -.l.n~ Transactions (Electrical Engineering), vol. 55, June
2
1936, pp. 717-30.
1. TRANSflORMER REACTANCE AND LOSSES WITH
NONUNIFORM WINDINGS, H. O. Stephens. AIEE 11. STATIC AND DYNAMIC ELECTRICITY (book),
Transactions (Electrical Engineering), vol. 53, W. R. Smythe. McGraw-Hill Book Company,
1934, pp. 346-49. Inc., New York, N. Y., first edition, 1939, pp.
256-322.
2. A FGRMtULA FOR THE REACT ANCE OF THE
INTERLEAVEL> COMPONENT OF TRANSFORMERS, 12. MAGNETIC CIRCUITS AND TRANSFORMERS
H. B. Dwight" L. S. Dzung. Ibid., vol. 56, 1937, (book), John Wiley and Sons, Inc., New York,
pp. 1368-71. N. v. 1943, pp. 358-62.
3. INFLUENCE ,OF VARIOUS FACTORS ON LEAKAGE 13. BESSEL FUNCTIONS FOR ENGINEERS (book),
REACTANCE 0'1' TRANSFORMERS, A. L. Morris. N. W. McLachlan. Oxford University Press,
Journal, Institution of Electrical Engineers, Oxford, England, first edition, 1934.
London, England, vol. 86, 1940, pp. 485-95.
2x 14. THE STER AND STEI FUNCTIONS, N. W.
L 1( x )= - 4- CALCULATION OF THE MAGNETIC FIELD OF
McLachlan, A. 1. Meyers. Philsophical M aga-
37r R~T ANGULAR CONDUCTORS IN A CLOSED SLOT
A:!llD ITS ApPLIeATION TO THE REACTANCE OF zine, Taylor and Francis, Ltd., London, England"
vol. 21, 1936, pp. 425-36.
it may be verified that the proper lower TRANSFORMER WI.DINGS, E. Billig. Report Q/
limit on .the integral is zero since the left- T 1Of. British Electrical and Allied Industries Re- 15. THEORY OF BESSEL FUNCTIONS (book), G ..
seareh Association, London, England, 1945. N. Watson. Cambridge University Press, Cam-
hand side of equation 23 reduce to zero bridge, England, 1922, pp. 328-38.
5. OX THE NEW METHOD OF TRANSFORMER
atx=C). LEAKAGE REACTANCE CALCULATION, K. Ozawa. 16. DIFFERENTIAL EQUATIONS (book), L. R. Ford..
This is shown as J apa1lt Science Review, Tokyo Institute of Tech- McGraw-Hill Book Company, Inc., New York.;
nology, Tokyo, Japan, vol. 1, Mar., 1949, pp. N. Y., p. 76.
{x.[Lt{x)Ko(x) +Lo(x)K1 (x )] } 1:=0 69-75.

-+-l)J
17. BRITISH ASSOCIATION MATHEMATICAL TABLES ..
6. MAGNETIC LEAKAGE INDUCTANCE IN TRANS- Cambridge University Press, Cambridge, England ,
- 2X x 2X(X-1ft x
=limx.[ --In-+- =0 FORMERS WITH CYLINDRICAL WINDINGS AND FORCES vol. 6, BESSEL· FUNCTIONS.
X-'O . 37r 2 7r, 2 2 x ACTING ON THE WINDINGS, E. Roth. Revue
Generale de L'Electriciti, Paris, France, vol. 40, 18. T ABLE OF THE FUNCTIONS, f Kodx f 10d»,,
1936, pp. 291-303, 323-36. V. R. Bursian, V. Fock. Travaux, Institute of
-:{ x:[L 1 (x) Io(JC) - Lo{x) 1f1 (x) ] } r =·0
Physics and Mathematics, Leningrad, USSR, vol.
7. ELECTROMAGNETIC PROBLEMS IN ELECTRICAL

= lim x(
2X X2)
. - - - =0
ENGINEERING (book), B. Hague. Oxford Uni-
versity Press, Oxford, England, 1929, pp. 298-325.
2, 1931, pp. 1-10.

%-+{I 3r 7r
19. TABLES OF THE STRUVE FUNCTIONS, MATHE-
8. ON THE LEAKAGE FIELD AND THE LEAKAGE MATICAL TABLES PROJECT, National Bureau of
INDUCTION COEFFICIENT OF A TRANSFORMER WITH Standards. Journal of Mathematics and Physics,
I:t i£ necessary to determine proper SANDWICH WINDINGS AND DIVIDED END COILS, Massachusetts Institute of Technology, Cambridge,
limits on the integrals because the term W. Rogowski. Mitteilungen uber Forschungs- Mass., vol. 25, 1946, pp. 252-59.
to be evaluated is the left-hand side of arbeiten, Zeitschrift des Vereins Deutscher In-
genieure, Dusseldorf, Germany, vol. 71, 1909, 20. TABLES OF INTEGRALS OF STRUVE FUNCTIONS,
equation 23, not the integral. The integral pp. 1-36. M. Abramowitz. Ibid., vol. 29, 1950, pp. 49-51.
representation of this group of functions
il. STATIC AND DYNAMIC ELECTRICITY (book), 21. INTEGRALS INVOLVING BESSEL AND STRUVB
'was used to obtain a faster calculation FUNCTIONS, N. W. McLachlan, A. L. Meyers.
W. R. Smythe. McGraw-Hill Book Company,
procedure since only two functions need Jnc., New York, N. Y., first edition, 1939, pp. Philosophical Magazine, Taylor and Francis, Ltd.,
be calculated rather than six. 290-93. London, England, vol. 21, 1936, pp. 437-48.

..
---_ ---------------

No doubt Mr. Rabins considered the The amount of data available for making
Discussion existing expressions and their derivations. a comparison between calculated and test
But knowing the capabilities of digital values is much larger now than it was at
B. A. Cogbill (General Electric Company, computers, he was confident that they the time the paper was written. Therefore
Pittsfield, Mass.): The title of Mr. Rabins' could handle with comparative ease the a more realistic appraisal can be made by
paper might imply that he has done a more complicated calculations involved using only reactances of transformers which
more or less routine job of adapting a in the use of more rigorous methods. So, have been built and are in service. Based
method of computing transformer reactance he chose the more difficult, but more on nine reactances which have been cal-
to digital computer calculations. However, rewarding, course of analyzing the field culated by the method under consideration
it is obvious that this is a minor part of his problem to obtain a solution which met and tested after final assembly of the
accomplishment; more important' is his his needs. This was accomplished in an transformers, the average of the ratio of
mathematical analysis of the field problem. admirable manner. The mathematical calculated value to test value is 0.978;
I t may be correctly argued that the analysis of the field problem alone is an and one standard deviation from this
field problem has been analzyed many outstanding contribution and should be of average is only 0.0101. This is quite
times in the past by very capable in vestiga- interest to many engineers and practical impressive, especially when it is considered
tors. But a thorough study of published mathematicians although they might not that some of the error may be due to manu-
material reveals that, in practically all have any particular interest in digital facturing tolerances and slight inaccuracies
cases, the earlier investigators were looking computers. in test. Moreover, none of the transformer
for simplification, Digital computers were The paper contains a number of compari- designs on which the method has been used
unknown. So earlier efforts were directed sons between test values and values cal- is of the simplest type where standard equa-
toward obtaining expressions which were culated on the digital computer using the tions provide us with sufficiently accurate
practical for use in calculating transformer author's method. As stated, these cases results.
reactance by methods available at the time. are of extreme nature. The author cited
Hence, a number of simplifying assumptions them to demonstrate the accuracy of the
were usually injected at an early stage of method under conditions for which other L. Rabins: I wish to thank Mr. Cogbill
the analysis; the almost universal one methods are known to be quite unreliable. for his discussion and for the presentation
being that curvature would be ignored. But he had no intention of conveying the of the data on production transformers.
As a result, the use of the expressions for impression that the method is not practical As he has stated, existing rigorous methods
reasonably accurate calculations is quite for use in the more usual cases. Actually, for reactance calculation were in vestigated.
restricted; and, even in the simpler cases, a large number of reactances for power Calculations on the card-programmed cal-
results are somewhat less accurate than transformer designs have been calculated culator showed these to yield insufficiently
those which could be obtained from use of by the method; and it has proved to be a accurate results so that the need for a less
more rigorous expression. practical design tool. restrictive approach became apparent.

JULY 1956 Rabins-Transformer

Authorized licensed use ReactanceonCalculations
limited to: KOREATECH. Downloaded with
February 14,2024 Digital
at 04:30:33 UTCComputers
from IEEE Xplore. Restrictions apply. 267