Analy >is of

Series Capacitor A | ^plication Problems

By J. W. BUTLER C. CONCORDIA
ASSOCIATE AIEE ASSOCIATE AIEE

Introduction oscillate or "hunt," under certain circuit conditions. The

tendency to hunt becomes more pronounced as the effec­

I
N T H E past several years series capacitors for the com­ tive ratio 6Ï line resistance to line reactance increases, and
pensation of line drop in power circuits have found ^putting in a series capacitor to decrease the reactance ob­
increasing use, 1 - 9 because improved and automatic' viously increases this ratio. Consequently, if the machines
voltage regulation can, in many cases, be obtained more are susceptible to hunting, the condition is further ag­
economically by this method than by any other means. gravated by the addition of capacitors. The case of hunt­
Most of the applications have been entirely successful ing of a single synchronous machine operating from an
in improving the system performance. However in a few infinite bus was investigated several years ago.11,12 In
instances unforseen difficulties of an unusual character section II of the present paper, further analytical methods
have been encountered. These difficulties were the oc­ are developed involving 2 or more machines which enable
casion of a systematic and detailed analytical study of the size and location of a capacitor to be so chosen as to
series capacitor performance. This study has been paral­ product the least tendency to hunt consistent with proper
leled, as far as possible with laboratory tests and correlated voltage regulation.
with field experience. It is believed that a satisfactory In case 3, under certain circuit conditions an induction
explanation and understanding of the abnormal system motor operating with series capacitors will be self-excited
behavior have now been obtained and that adequate and will generate other voltages of lower than normal fre­
methods for predetermining the system performance have quency and thus produce objectionable voltage fluctu­
been developed, although it is recognized that the investi­ ations. Analytical methods are presented in section III,
gation cannot be regarded as absolutely final and complete. which make it possible to determine the effectiveness of
The objects of this paper are to present a survey of some methods for eliminating this trouble.
of the troubles which may be encountered in the applica­ It has been found in this investigation that the addition
tion of series capacitors to power circuits, to present cri­ of a resistor in shunt with the series capacitor will eliminate
teria for determining the range and probability of abnormal the difficulties of cases 1 and 3, while the synchronous ma­
operation, and to describe methods for minimizing the chine hunting of case 2 is minimized or not obtained by
difficulties. proper choice and location of the capacitor. The shunt
Three types of abnormality have been investigated : resistor is of such a high value as not to interfere with the
capacitor's effective operation as a line drop compensator
1. Distorted and excessively large transformer exciting currents, and the value of the power loss in the resistor necessary to
due to saturation. ensure complete freedom from these circuit troubles will
2. Hunting of synchronous machines. in general be less than 10 per cent of the resistance loss
3. Self-excitation of induction motors. in the line tô be compensated. The proper value of re­
sistance to be used in any given case may be determined
In case 1, the application of voltage to an unloaded or from the analysis presented here. It is concluded, therefore,
very lightly loaded transformer through a series capacitor that the material offered in this paper, will enable prac­
may result in abnormally large and distorted exciting cur­ tical and economically sound series capacitor applications
rents. These large currents are not simply transients but to be made with confidence that the operating results will
persist as a steady-state condition. This type of circuit, be in accordance with predictions.
that is, a saturable inductance in series with a capacitor,
has been found to have at least 2, and probably more,
different steady states, depending on how the voltage is Section I—Distorted and Excessively Large
applied. Transformer Exciting Currents
Solutions of the equations for this type of circuit have
been found by the differential analyzer10 at the Moore A. General
School of Electrical Engineering and are discussed in sec­
tion I. In the usual case of application of voltage to an unloaded
In case 2, interconnected synchronous machines will transformer it is known that if the voltage is applied at or
near the zero point of the voltage wave a high inrush cur­
A paper recommended for publication by the AIEE committee on power trans­
mission and distribution. Manuscript submitted June 1, 1937; released for rent may result. This is because the transformer flux
publication June 28, 1937. tends to rise, in the first half cycle, to double its normal
J. W. BUTLER and C. CONCORDIA are in the engineering division, central station
department, General Electric Company, Schenectady, N. Y. value, with consequent excessive magnetizing current due
1. For all numbered references, see list at end of paper. to the very high degree of saturation. Stated in another

AUGUST 1937 Butler, Concordia—Series Capacitors 975

way, at high values of flux the effective inductance of the capacitor. However, as soon as saturation is reached, the
transformer becomes very low. The high inrush current current drawn becomes very large and a voltage soon ap­
lasts for several cycles and then the transformer draws pears across the capacitor. The voltage absorbed by the
only its usual low value of exciting current. On the other capacitor prevents the flux from building up to as high a
hand, if the transformer is supplied through a series capaci­ value as it would have without the capacitor, so the first
tor an abnormal current may persist in the steady state. half cycle inrush current is somewhat less. Moreover,
That is, the energizing of an unloaded or very lightly as the flux tends to remain offset it does not reach a high
loaded transformer through a series capacitor may result value in the negative direction in the second half cycle. To
in a continuous abnormal flow of exciting current. This the contrary, it may hardly dip below zero. Thus the
current may be equal to or even greater than full load cur­ exciting current remains very small (or, the transformer in­
rent and is of an apparently very low frequency, resulting ductance remains very large) for the interval following.
in a badly distorted current wave. Under this condition The charge is therefore left on the capacitor, which simply
of excitation the secondary voltage is also badly distorted, acts as an additional voltage source. This additional
and it likewise is largely composed of a lower frequency voltage tends continually to depress the transformer flux,
component. Oscillographic records of the magnetizing just as in the first half cycle it prevented the flux from
current, the secondary voltage and the voltage across the reaching double its normal value. Thus in the second
capacitor when a circuit is in this state of excitation are cycle the flux may not go high enough to saturate the core
shown in figure 1. so no additional charge is put on the capacitor. On the
These abnormal steady-state currents flow when the other hand, as practically no charge has leaked off, the
voltage is applied at or near the zero point of the wave capacitor voltage is still decreasing the flux. The general
just as the high transient occurs for this condition in the effect is as shown in figures 4b and 4c and is that of first
ordinary case without the series capacitor.
This phenomenon has been reported before 13-16 in the
technical press and is sometimes called "ferro-resonance."
The peculiar circuit behavior depends essentially on the
nonlinear transformer characteristics and not on any
secondary features such as hysteresis, eddy currents, etc.
The nonlinear characteristic of the transformer makes
APPLIED VOI
the mathematical analysis very difficult and to the authors'
knowledge there has been no satisfactory analytical solu­
tion made as yet.
The differential analyzer at the Moore School of Elec­
trical Engineering was used, with the able assistance of
Irven Travis and C. N. Weygandt, to obtain solutions to
the nonlinear differential equations of this circuit. Ap­ i i i/hWt Ί in i iih Ki-^M^-k^
pendix A gives the equations of the circuit and the form
into which they were put to adapt them to the analyzer.
The single-phase circuit, as shown by figure 2, was set APPLIED VOLTS* CURRE^IT™^. W^-;
up and investigated thoroughly before the 3-phase case
was studied, as it was believed that the essential features
of the phenomenon could be obtained more directly in this
manner. A schematic diagram giving the analyzer set-up
for equations 2a and Sa is shown in figure 3.
The 3-phase circuit was set up as a final check and the
results obtained indicated that the single-phase results were
',-■■ -Μ-:- ■ \i ■#:.'. Vf ■ -\r
qualitatively applicable. ï ' . : ■ "V ■■ :% - .'■:■«■% \i V
Typical curves drawn by the analyzer as solutions to SECONDARY VOLTS-^ CURRENT^
equations 2a and Sa are shown in figures 4, 5, and 6.
A short simplified explanation of the physics of the
ίι A., l i t k
phenomenon will be made here, for the sake of complete­
ness.
ITOR VOLTS-
B. Qualitative Explanation of Phenomenon
Figure 1 . Oscil-
These abnormal currents may be explained as follows : lograms of abnor­ f: Ï * I 'I ' I
Consider that the voltage is applied to the transformer mal currents and

W M WM
at the zero point of the wave. During the first part of the voltages obtained
cycle the flux in the core builds up just as though the when transformer
capacitor were not present, since the current drawn is so is excited through
small that there is no appreciable voltage drop across the series capacitor

976 Butler, Concordia—Series Capacitors ELECTRICAL ENGINEERING

Figure 2. Circuit dia­ Re
gram of transformer en­
INPUT
ergized through a series TABLE
capacitor, series impe­
dance, and shunting re­
sistance Rc ESIN ((jüt + Οθ)

n
ISi
V.Q.Q.Q.QQ&-
-TJUirLT-

tending to reduce the offset and then actually to build up

an offset in the negative direction. This process continues,
at a rate proportional to the magnitude of the voltage
trapped on the capacitor, until saturation is reached in the rJ[ESIN(url+Œ)-r,L,
negative direction. When this occurs a large negative 'f/(ii-l2)dt]clt
current is drawn, discharging the capacitor. This large
negative current may discharge the capacitor partially,
completely, or may put on a charge in the reverse direction.
Θ INTEGRATOR Θ INTEGRATOR

It is easily seen that, except for a fortuitous combination

Figure 3. Schematic diagram of differential analyzer connec­
of circumstances, the discharge will be either partial or
tions for solution of circuit equations
an overshoot. If it is partial the flux will reach a larger
negative value at the next cycle, and then the capacitor
will receive an opposite charge. This negative charge will, curve and the voltage is applied at or near the zero point
of course, have the effect of reducing the negative offset on the wave, the circuit will draw a sustained large dis­
and then building up a positive offset, so that the cycle of torted magnetizing current. All points below each curve
abnormal operation is complete. represent a so-called stable area where the circuit will have
Since the rate of decrease of offset is proportional to the its normal low value of magnetizing current. The various
voltage on the capacitor, the apparent frequency of oc­ curves show the effect of changing the line reactance, the
currence of these large currents (due to saturation being line resistance, and the transformer iron characteristic.
reached) is necessarily dependent on the voltage applied The 2 iron characteristics are shown in figure 8. Curve 1
and on the angle at which it is applied. For instance, the represents the magnetization curve of a modern trans­
larger the initial charge on the capacitor, the sooner it former. Curve 2 was taken arbitrarily to have the current
will build the linkages down to the saturable place on the scale of curve 1 multiplied by 3 merely to show the effect
negative side, and give a higher frequency appearance to of a different magnetizing characteristic or of a larger
the exciting current. transformer.
In order to demonstrate this property of the circuit, In figure 7 the value of capacitance, to which the left
solutions, as given by the analyzer with different values side of each curve is asymtotic, represents the value of
of voltage applied at the zero point of the wave, are shown capacitive reactance that must be used to prevent the
in figure 4. The periodicity increasing with the applied circuit from going into a state of overexcitation without a
voltage is clearly seen. If the voltage is increased suf­ damping resistor. Unfortunately these values are so low,
ficiently, the circuit can be made to saturate every half that they can seldom be useful in correcting a bad voltage
cycle and results such as those obtained by Suits 17 are then condition.
obtained. These curves were all determined for normal voltage
applied at the zero point of the wave. In certain cases,
C. Summary of Results Obtained it may be found that a resistance sufficient to insure
From the Differential Analyzer normal operation under these conditions will not be low
enough if the system is operated at a different voltage.
It is felt that the major contribution this investigation For example, figure 5 illustrates a case in which reducing
made was that of explaining and clarifying to a certain de­ the voltage caused an abnormal state of excitation.
gree the physics of the phenomenon. In order to appreci­ Curve 1 shows normal operation at normal voltage, while
ate the "goings on" in the circuit, fundamental circuit con­ curve 2 shows the overexcited condition occurring when
cepts must be kept in mind, as vector diagrams, resonance, 0.75 voltage is applied. Both voltages were applied at the
sustained harmonic solutions, and other linear circuit zero point of the wave. This indicates also that in some
concepts fail when nonlinear parameters are employed. cases the application of voltage at the zero point of the
The effect of a damping resistor shunting the capacitor wave may not be the worst condition since in a sense a
was investigated and the results are summarized in figure 7. shift of angle is partially equivalent to a reduction of
In these figures, the region above each curve represents voltage, at least as far as the initial charge given to the
an unstable or abnormally excited circuit condition; that condenser is concerned.
is, if a given circuit has a value of capacitance and shunting A circuit having a resistance shunting the saturable
resistance that fall in the area above the corresponding inductance, simulating a load being taken off the circuit

AUGUST 1937 Butler, Concordia—Series Capacitors 977

 ical equations for a general linear circuit, consisting of any
number of meshes, the characteristic determinant is in­
dependent of the magnitude or location of the applied
voltages. I t is therefore evident that the form of the tran­
sient response, or the building up or dying away of a tran­
sient current, is also independent of where the voltage is
applied.
This reasoning, showing that in a linear circuit, the loca­
tion of the resistor does not affect the periodicity of the
circuit, combined with the differential analyzer results for
the nonlinear circuit, indicate that this may well be a
general conclusion. It has been known in practice that
intervening or secondary loads prevent this overexcitation
condition, but the effect had not been previously investi­
gated quantitatively.

D. Conclusions
An unloaded transformer, operating at normal densities,
if excited through a series capacitor of reasonable reactance,
is susceptible to conditions of large distorted magnetizing
currents. These abnormal conditions can be restored to
normal by the addition of a suitable shunting resistor to
the capacitor, as indicated by figure 7, or by the locating
of a certain amount of load on the load side of the capaci­
tor.
The resistor shunting the capacitor, introduces addi­
tional losses. However, in general, this loss is less than
10 per cent of the normal line loss and this can usually be
justified by the improvement in voltage regulation. More­
over, since the presence of the abnormal condition depends
^V C =0.Q33
on the initial transient, it may be eliminated by shorting
out the capacitor only while the starting switch is being
closed, or by having the resistor connected only during the
Figure 4. Effect of applied voltage on the apparent perio­
energizing period. The shorting switch or temporary re­
dicity of exciting current
sistor may readily be arranged so as to function automati­
h = 0.049 1 / C = 0.214 (a) E - 1.0 cally whenever the abnormal condition tends to develop.
Rc = CO Magnetization curve 1 (b) E = 0.75
L = 0.197 a = 0 (c) E - 0.6
Section II—Hunting of Synchronous Machines
between the transformer and the capacitor, was also in­
vestigated on the differential analyzer. The critical points A· General
obtained with this circuit agreed very well with the points
for the same resistance when it was shunting the capacitor, An electric power system consists in general of a group
indicating that it made little difference where the resistor of interconnected synchronous machines together with
was located. impedance and induction motor loads. Any such system
From figure 2 it may be seen that the 2 locations of the is subject at practically every instant of its existence to
resistor are equivalent to 2 locations of the applied voltage, disturbances caused by changes in load, switching, pul­
the circuit remaining unchanged. Further, in the canon­ sating loads, pulsating driving torques, changes in excita-

*ψ- CURVE 1
, φ - CURVE 2
Figure 5. Effect of voltage on
excitation
Curve 1 E = 1.0. Normal excitation,
capacitor voltage decreases
Curve 2 E = 0.75. Abnormal excita­ V c -CURVET
tion, capacitor voltase builds up V c - CURVE 2
1 / C = 0.35 L = 0.394
Re = 6.3 Γι = 0.197
a =0 Masnetization curve 2

978 Butler, Concordia—Series Capacitors ELECTRICAL ENGINEERING

tion, etc. These disturbances tend to set up rotor oscilla­
tions of the synchronous machines, with consequent volt­
age and frequency pulsations. Therefore, for satisfactory
operation, the system constants must be such as to cause
any such oscillations to disappear rapidly if due to a
momentary disturbance, or to remain of small amplitude
if due to a continuous pulsating force. A momentary dis­
turbance will, in general, produce oscillations of the syn­
chronous machines at all of the principal natural frequen­
cies of the system, while a sustained pulsating force of any
frequency will of course produce machine oscillations of
the same frequency.
Figure 6. Linkage and capacitor voltage curves with critical
values of Rc
B. Effect of Line Resistance Curve 1 — Rc = 6.1. Abnormal/ capacitor voltage builds up
and Series Capacitance Curve 2— Rc = 4.2. Normal, capacitor voltage decreases
a —0 /*i = 0.197 Magnetization curve 1
The inherent damping of the system is usually sufficient L = 0.394 1/C = 0.21
to prevent severe hunting. However, in certain cases the
damping may be either too small to be effective or may
actually be negative. That is, the rotor oscillations may be ticular case which may arise. The critical operating angle
amplified rather than damped out. There are present in of either generator in this general case is shown to be de­
all systems 2 conflicting effects. Resistances in the rotor pendent not only on the ratio of combinations of the line
circuits, some types of load, and mechanical friction will constants, but also on the operating angle of the second
produce true damping of rotor oscillations, while resistance machine and on its relative amplitude and phase of oscil­
in the lines or armature circuits tend to produce amplify­ lation. The line constants enter the equation as ratios of
ing rotor oscillations (or negative damping). the resistance to the reactance components of the several
In the case of a single synchronous generator connected transfer and driving point impedances and so are correlated
11
to an infinite bus, it has previously been shown11,12 that üo the criterion developed by Nickle and Pierce.
the ratio of line resistance to line reactance determines The equations of appendix B may be used to study any
whether or not there is negative damping, and that if all particular proposed series capacitor application to deter­
rotor circuits except the field winding are neglected, there mine whether or not it is in the danger zone and, if it is,
is a critical operating angle determined by this ratio above to determine the optimum location and maximum per­
which the machine is unstable. The critical angle is missible reactance of the series capacitor consistent with
always decreased by the presence of rotor windings, so proper line drop compensation. These equations also
the method provides a conservative criterion. enable the effectiveness of putting additional series react­
If a series capacitor is used in the line the ratio of line ance in certain machine circuits to be evaluated.
resistance to line reactance is increased, since the reactance
has been reduced. Consequently the tendency toward D. Discussion of Results
negative damping is increased.
As an example of the method of analysis, a problem in
C. System of Several Machines which the circuit may be represented as in figure 9 has
been worked out. In this case the synchronous generators
In case of a system consisting of several machines, it is at 1 and 4 ordinarily supply the load, line 3 merely acting
shown in appendix B that there is a corresponding relation as a tie to the large system at 3. However, occasionally
among the line resistances and reactances which deter­ power must be supplied to 4 from 3. Under this condition
mines the negative damping and that therefore such damp­ the voltage regulation at 4 is rather poor and it is desired
ing will be affected by the insertion of series line capacitors. to correct it by the insertion of a series capacitor between
However, it cannot be expected now to be a simple re­ points 3 and 4. As considerable line drop occurs in line 3
sistance-to-reactance ratio ; instead the criterion for stable the logical place to put the capacitor might seem to be in
operation is a ratio of certain combinations of line con­ that line so as to give loads at 2 some benefit. Unfor­
stants. For example, if the case treated in reference 11 is tunately, when this is done the hunting of machine 1 is
extended to include shunt loads taken off the line, the found to be excessive, while if the capacitor is placed in
criterion becomes the ratio of the resistance to the react­ line 4 hunting is no worse than without the capacitor. The
ance components of the transfer impedance between the desired result of correcting the bad voltage condition at
synchronous machine and the infinite bus (see appendix 4 is accomplished by putting the capacitor in line 4. How­
B, equation 8b). ever in this location the additional advantage of correcting
In appendix B a system consisting of 2 synchronous the voltage at the junction is not obtained.
generators, an infinite bus, and a shunt load (see figure 9) Table I shows the results of calculations of the minimum
has been analyzed. This is believed to be sufficiently permissible load angle of machine 1 for stability, under the
general to form a basis for the study of almost any par- 3 conditions of no capacitor, capacitor in line 4, and capaci-

AUGUST 1937 Butler, Concordia—Series Capacitors 979

tor in line 3, and also for different reactances in branch 4 Table I
and with and without shunt load. It is evident from this
table that the capacitor in 4 causes a much smaller differ­ Critical Load
Angle (δι ~ δι')
ence from the original condition than it does when in of Machine 1
Condition (Degrees)
branch 3.
Although a picture can thus be obtained of the relative No capacitor r% = <χ>, X4 0.2, n = 0.05. .. .1.38
tendencies toward hunting of several proposed arrange­ Capacitor in number 4. . . .rz = °° , Xi ■ 0.2, n = 0.05... .1.32
.4.14
Capacitor in number 3. . . . rz = <x>, xi 0.2, n = 0.05...
ments, it has not been absolutely determined by calcula­ No capacitor rz = 1, Xi 0.2, n = 0.05... .1.48
tion that one arrangement will operate satisfactorily Capacitor in number 4. . . .rz = 1, x\ 0.2, n = 0.05... .1.40
Capacitor in number 3. . . .rz = 1, x\ 0.2, n = 0.05... .3.94
and another will not. This is still a matter of judgment
No capacitor rz = °° , #4 -, 1.07, n = 0.049. .1.72
fortified by a correlation of field experience with the calcu­ Capacitor in number 4. . . .rz = «» , xi ■■ 1.07, n = 0.049. .1.70
Capacitor in number 3. . . . rz — <» , xt : 1.07, n = 0.049. .2.46
lations.
No capacitor rz = 1, X4 : 1.07, r4 = 0.049. .1.88
The critical angles recorded in table I are the load angles Capacitor in number 4. . . .rz = 1, XA : 1.07, n = 0.049. .1.88
: 1.07, n = 0.049. .2.67
δχ-δχ' where δι is found as described in appendix C and δχ' is Capacitor in number 3. . . .rz = 1, Xi
See figure 9 for circuit.
the no-load angle of machine 1 referred to the infinite bus Π = 0 r3 = 0.023 Xg ■ 0.65 0.0516
3. The oscillation of machine 4 was neglected (i.e., k = rz = » or 1 n = 0.05 X2 ■■ 0 0.20 or 1.07
xc = 0.105 = capacitive reactance
0), since it was assumed that the machine being investigated
would have the largest amplitude of oscillation. Also it Note: xg and Xi include both line and machine reactances.
was found that the necessary load angles were only slightly
affected by the operating angle <54 of machine 4. It is
therefore possible to use the simplified formula (8b .1) for modes of vibration corresponding to the 2 principal natural
nearly every case by simply substituting the value of frequencies of the system of figure 9, or to any impressed
zz (zz = rs + jxz) paralleled with s4 for the zz of equation forces, can be very simply calculated if the inertia constants
(86.2). of the machine are known. This will not usually be neces­
A more exact study of a system may be made by taking sary, however, since sufficient information can be obtained
the amortisseur windings of the machines into account from a study of the system neglecting amortisseur windings
and computing the damping coefficients by the methods and a knowledge of the general effects of such windings.
of reference 25 for every probable mode of vibration. The (See, for example, reference 19.)

■ L ' rY Figure 7. Limits

1 1 Section III—Self-Excitation of Induction Motors
0.197 0.049 of normal excita­

M
8 0.394 0.197 1 tion as a function
3. | CC=0 A. General
CURV E 2 | of Rc and 1/C.
o
y 1 /r Area below each It is well known that an induction machine will operate
1 / / curve is normal.
6 1 as a generator and that such operation depends on the
\ 1 1 Area above each
1
supplying of excitation from either a synchronous machine
\ \ // 1/

ί
curve is abnormal or a capacitor even though in case of the capacitor the
\ \ 1 f
operation has not previously been very well understood.
MAG. Now consider an induction motor supplied through a
/ ClJRVE
line containing a series capacitor. Then in addition to the

1
^ currents flowing due to normal operation of the motor, un­
Figure 8 (below).
2 / / der certain conditions the motor may act as an induction
Magnetization
\ (y curves. Curve 2 is generator of current of lower than normal frequency. This
curve 1 with cur­
low-frequency current is limited only by the impedance of
rent scale multi­
the supply circuit at the low frequency, and may reach
0
0.2 0.4 0.6 plied by 3
relatively large values. These large low-frequency cur­
CAPACITIVE REACTANCE j / C rents manifest themselves not only as current surges and
voltage swings but also as strong oscillations of the motor
I.6 M\\\ 1 ΠΤΤ7ΠΤΤ 7ΡΓ TTI illfff rotor caused by the large pulsating torques produced.
/ lifitn
W"ΊΊΊΙΙΙΙ 11 This phenomenon of self-excitation will not always take
'U M1
I.4
place and may always be entirely eliminated by sufficient
r* uL IP--· Tl ffl
4
jjlll
Ίmn
11 illll 11 line or shunt resistance, or by judiciously locating the
y
ma\
mmvC* capacitor. It may, moreover, be calculated with reason­
§0.8 ^ i™ able certainty and accuracy by equations given in ap­
Ί
I 0.6 ΊI lU-rrU
Iif / 1 if pendix C, if the circuit constants and operating conditions
ml /
IJ Jim
£o.4 are known.
K 11 Hill j11
0.2 Ml
A l fl
Um
Figures 10 to 13 show the regions in which self-excita­
0 y
\M\\ ( 1 Illll 11 tion is likely to occur for the case of a purely induction
0.OOI 0.01 0.1 i.O K
P£« UNIT CURRENT-PEAK motor load. Figures 10 and 11 show the effect of series

980 Butler, Concordat—Series Capacitors ELECTRICAL ENGINEERING

 Figure 9. Circuit at normal load, a sustained overload may produce self-
t J (^^KSlûfiflÛfiflÛOr—i INFINITE diagram of 2 syn- excitation.
BUS
E3 c h r o n o u s ma­
r
F A 4 *4
chines connected C. Calculations and Test Results
LOAD to infinite bus and
impedance load In order to check the theory offered for this so-called
"self-excitation phenomenon,'' tests were conducted on a
miniature set-up, in order to get good control of the vari­
line resistance in preventing self-excitation, while figures ables. A 3-phase 3/4-horsepower wound-rotor induction
12 and 13 show how a resistance in shunt with the capaci­ motor was connected to a power source through a series
tor can eliminate the phenomenon. For any practical capacitor and 0.39 per unit external reactance. The
case the reactance of the series capacitor will of course be values of series capacitance were varied over the complete
much less than x', probably about half of x', since x' is range of self-excitation, and the corresponding value of
the short-circuit reactance of the motor including line shunting resistance to insure normal operation was de­
reactance and since the capacitor reactance will usually termined. These tests results, and corresponding calcu­
be no greater than line reactance. Therefore figure 14 lated points on the curve are shown plotted in figure 16.
is given, showing the shunt resistance necessary to pre­ The agreement between test and calculated values was
vent self-excitation in the useful range of capacitance, for very gratifying. I t is felt that the equations offered can
2 sets of system constants. be used with confidence in predetermining these regions
The values of shunt resistance given in the curves will of abnormal operation.
usually be conservative because the effect of series line The motor was tested running light, hence the speed was
resistances and machine losses, which themselves tend to taken as unity to simplify the calculations. The armature
damp out self-excitation, has been neglected. Equations resistance of the motor and external circuit was 0.10 per
are given in appendix C which take both shunt and series unit. This high circuit resistance is the reason for the
resistance into account and which should be used to^a^a- curve in figure 16 intersecting the a = 0 point at such a
lyze any specific problem. These are based on the an­ kigh value of xc. The rotor resistance was likewise very
alysis presented in reference 20. Figure 15 shows an ex­ high, hence the motor had a correspondingly small rotor
ample of how the value of a. necessary for stability de­ time constant. The small rotor time constant gives the
creases with increasing line resistance. motor a greater tendency to self-excite as shown in figures
10 and 12.
B. Discussion of Results In view of the type of motor tested, that is, one having
high resistances, these test results should not be generalized
Figure 10 shows that as long as the capacitive reactance as significant of normal motor performance.
is less than half of the system short-circuit reactance, The line of demarkation between normal operation and
self-excitation will not occur for a normally loaded group of
induction motors. This is because the line resistance is
usually great enough to damp out any tendency to self-
excite. However, as soon as a certain critical value of
short-circuit reactance is exceeded the amount of series
resistance required begins to increase rapidly. This crit­
ical value of reactance is, of course, a function of the motor
characteristics and the impedance of the connected circuit.
If the line is relatively short (i.e., if its reactance is less
than the motor reactance) it may be compensated more
than if it is long, unless a shunt resistor is added to pre­
vent self-excitation.
Figure 11 shows that as the motor speed decreases the
region of self-excitation moves down along the xc ordinate.
That is, the tendency to self-excite occurs at a lower value
of capacitive reactance at lower motor speeds. Such low
speeds occur during starting and during periods of over­
load. We should thus expect some oscillations as the motor
comes up to speed. This is especially true if a starting 10.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
resistance is used in the rotor, as figures 10 and 12 indicate 2 M
1
X
SERIES LINE CAPACITIVE REACTANCE = Xc
I2x
that an increase of rotor resistance increases the tendency
to self-excite. In general it will not be necessary to elimi­ Figure 10. Self-excitation of induction motors with series
nate these transient oscillations. Actually, in service capacitors
when the system behaves correctly under normal operating Showing series resistance required For stable operation as a function of
conditions no trouble during starting has been experienced. the line capacitance and showing effect of varying rotor resistance
If, however, the system does not have very much margin- 0.4 x = 3.24 T = 171.4 and 17.14 ω = 0.975
0

AUGUST 1937 Butlery Concordia—Series Capacitors 981

self-excitation was very sharp up to xc = 0.6. Calcula­ oL a found for each value of xc. The current wave was*
tions and tests both showed this. Beyond xc = 0.6 the "watched through an oscilloscope and self-excitation could
calculated curve became more difficult to determine be­ be readily recognized by a change in the wave. In the
cause it seemed to be very broad in the upper regions, lower regions below xc = 0.6, the phenomenon was more
that is, it was difficult by ordinary slide rule methods to violent and critical. A change of a few per cent in the
get a smooth curve, indicating the critical region was not value of shunting resistance R in this region, would be
very sharply defined. Tests bore this point out in the up­ sufficient to change the circuit from normal to a definite
per regions. The test points plotted are the largest values state of self-excitation. The region of self-excitation in
the practical and useful range of xc can be readily recog­
nized by calculation since this range is below xc = 0.6.
1.4
Figure 17 is an oscillogram of the current flowing in one
phase during the motor starting period, and of the subse­
quent flow of current when the circuit is in the state of
1.2
self-excitation. It will be noted that in this case the motor
\ came up to speed in the normal manner, with normal
\
c 1.0 inrush and sustained current. However in several cycles
1 \ the current built up to several times normal with obvious
o
/ \ superposed frequencies, which are identified as the natural
Λ
<0.8
currents of the system having positive decrement factors.
ί \
tu
"θ.δ
m
A < \ \
These currents are ultimately limited in magnitude by
saturation and the current builds up until the circuit con­
z \ \ \ \
-I stants change to a value that will not permit further self-
/ \ \

ift(i
0.4 \ excitation.

0.2 1 D. Self-Excitation of Synchronous Machines

A B c P
0 VL 0.4 0.6 1.2 1.6
L_
2.0 2.4 2.8
I t has been stated above that in circuits containing
series capacitors and induction motors, there may be a
3.2
SERIES LINE CAPACITIVE REACTANCE—Xc
tendency to self-excite and that in circuits containing series
Figure 1 1 . Self-excitation of induction motors with series capacitors and synchronous machines there may be a
capacitors tendency toward hunting. It is evident, however, that
the tendency to self-excite (that is, to generate amplifying
Showing effect of varying motor speed
A~ ω = 0.25 Β—ω = 0.5 C—ω = 0.707 D—ω = 1.0
current oscillations even though the rotor inertia prevents
x' = 0.4 x = 3.24 T0 = 1 7 1 . 4 hunting) exists also in synchronous machines. It has been
possible to obtain a satisfactory explanation of operating
I.4 experience with synchronous machines on the basis of
hunting alone but it is conceivable that under certain
1.2
conditions purely electrical self-excitation may be en­
""*« countered.
V*
^ While this matter has not been thoroughly studied, the
I.0 ^
// V phenomenon is familiar in relation to the determination of

7 0.6 To-
/> f
V the
S. B.
line charging
Crary, 21
M.
capacity of synchronous generators.
Takahashi, 22
and Y. H. Ku 23 have made
I7.I4 fl
■ \

cc
a study of this subject and have shown the tendency of
/ UN 5TA 3LE
II
8 0.6
/ ' /// \ line resistance to reduce the self-excitation. By means of
\ the theory given in reference 20, it may further be shown
0.4 1 'h
.D=I7 .4
\
that a resistance in shunt with the capacitor may also be
used to eliminate the self-excitation (see appendix D).
/ 1 \
Figure 18 shows, for the same generator as in figure 1 of
0.2 /

L1
reference 21, the effect of shunt resistance.
UN ST/û BLE

k
Since the effect of rotor circuits in addition to the main
1

0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2

L
field has not yet been determined, no definite conclusions
SERIES LINE CAPACITIVE REACTANCE — Xc can be drawn from this curve, but it is believed worthwhile
to point out the existence of such self-excitation and to
Figure 12. Self-excitation of induction motors with series indicate that it may be corrected exactly as in the case of
capacitors induction motors. The phenomena of self-excitation and
hunting may of course exist simultaneously and are
Showing resistance (in shunt w i t h the capacitor) required for stable
operation as a function of the line capacitance and showing effect of
mutually dependent to a limited extent. Their mutual
varying rotor resistance effects have not yet been studied exactly.
x ' = 0.4 x = 3.24 ω = 0.975 r = 0 The ideal procedure, from a mathematical point of view,

982 Butler, Concordia—Series Capacitors ELECTRICAL ENGINEERING

 Figure 14. Self-
ΓΤ\
1.2
excitation of in­
1.0 duction motors
with series capaci­ \
/f
0.8 20
tors
\
Λ y
s
=- Showing resistance \
π \ \
8 (in shunt w i t h the
0.4 yy / / \ capacitor) required \ \ v
/ ' \
r7 1 \

1
for stable operation XURVEi \
02 \ / \ as a function of the
A k \ line capacitance for > | x'=0.4,To='7l
LÛLL
0.1
_L 0.2 0.3
Γ
0.4 0.5 0.6 0.7 0.8 0.9 1.0
various systems
x = 3 r = 0
X'=0.I6JO=I7I
SERIES LINE CAPACITIVE REACTANCE-Xc χ'=0.4,Το=Ι7.Ι
ω = 0.975
Figure 13. Self-excitation of induction motors with series ! I x'=0.4,To=m
0.8|
capacitors H CURVES X'=0.4,TO=I7.I
0.6
Showing effect of varying motor speed
-ω = 0.25 Β—ω = 0.5 C—ω = 0.707 D—ω = 1.0 0.4
X'=0.I6,TO=I7I
0.4 3.24 To = 1 7 1 . 4 r = 0
0.2
0.2 0.4 0.6 0.8 I.O
RATIO χ ς / χ '
would be to set up the equations of motion of the circuit
including both electrical and mechanical motions, and
then to find out whether the system as a whole was stable R — resistance in shunt with the capacitor
or unstable. This method brings out very clearly the basic k
identity of the phenomena of hunting and self-excitation
m, n = subscripts given to currents, voltages, impedances, and
but is, from an engineering point of view, at present im­
angles in relating them to certain machines and circuits
practicable because of the excessive labor involved in the as used in section I I
calculations. It is believed that the results obtained by
the approximate studies are practically the same as would
be found by the unified method. Appendix A—Derivation of Equations
Set Up on the Differential Analyzer
Nomenclature
The single-phase circuit analyzed is represented in figure 2. T h e
Section I magnetization curves shown in figure 8 are the 2 iron characteristics
t h a t were investigated and discussed in the body of the paper.
E — applied voltage, maximum value The differential equations applying to this circuit are:
/ = time in radians
a — phase angle of applied voltage dii 1 Ç,.
E sin (t + a) = nil + L i2)dt + (la)
fi = line resistance dt
i\ = line current
ii = current through Rc HR> i<ù dt (2a)
Rc = resistance shunting the capacitor
L = line inductance Rearranging and integrating ( l a ) into a form applicable to the
C = capacitance differential analyzer, there is:

=S[Esi
ψ — flux linkage
Vc = voltage across capacitor Lix + ψ sin (t + ex.) — riii I (ii — it)dt «lit (3a)
C
Sections II and III The schematic hook-up of the analyzer, giving the solution t o
E = internal voltage equations 2a and 3a is shown in figure 3. The input voltage was
ea, eq — direct and quadrature axis terminal voltages actually not cranked in by an operator. Instead an auxiliary set-up
Ψα> Ψα — direct and quadrature axis flux linkages was made t h a t gave the solution to the differential equation
id, iq = direct and quadrature axis currents <Px
xd, xq = direct and quadrature axis synchronous reactances (4a)
dt2
(section II)
r — stator line resistance per phase which is a sinusoid, and it was impressed automatically into the
Ô = angle of voltage E from some reference point hook-up. An operator was needed to crank t h e current into t h e
0 = angular displacement of rotor system by keeping a pointer representing ψ, t h e linkage, on t h e
ω — ρθ — rotor speed linkage-current characteristic. Values of applied voltage, linkage,
xc — line series capacitive reactance capacitor voltage, and total current were plotted by the output
T0 — rotor time constant with open-circuited stator tables.
x = direct or quadrature axis synchronous reactance of induc­ The procedure used to determine the points on the résistance-
tion motor including line reactance capacitive reactance curves of figure 7 was as follows. The differen­
x' — direct or quadrature axis transient reactance of induction tial analyzer was started at a point corresponding to application of
motor including line reactance the voltage at the zero point of the wave and t h e several dependent
a = xc/R variables of the circuit plotted automatically on an output table.

AUGUST 1937 Butler, Concordia—Series Capacitors 983

 0.12 Figure 15. Relation 2 different values of shunt resistance, one just sufficient to produce
between line resistance normal operation, the other slightly higher.

0.10 and a required to pre­

vent self-excitation
Appendix B—Criteria for Negative Damping of
0.08 = 3.24 = 0.4
= 0.2 Synchronous Generators Neglecting Amortisseur
er Winding
^0.06 Figure 16 (below).
II
ö Self-excitation of in­ Consider the circuit of figure 9 with 2 generators connected to an
0Ό4 duction motors with infinite bus with a shunt load a t the branch point and with resistances
series capacitors and reactances as indicated.
The effects of salient poles will be considered only in machine
0.02 Comparison of calculated
number 1, the stability of which is being investigated, while machine
and test results
number 4 is considered to be round rotor. Moreover, the reactance
X—Test points
x4 will usually be taken as the transient reactance or some modified
0.0I 0.02 0.03 0.04 Curve—Calculated
value depending on the frequency of the oscillation being studied.
SERIES RESISTANCE — r x' = 1.03 x = 2.32
The steady-state equations for this circuit are:
« = 1.0 r = 0.10

0.30 ~~ iql + % + iqZ — iqi = 0

« Xâidi +#2^2 = Ei
0.25 5C ^ ^
-Xzidz + r 2 % ~niqz — Ez cos Bi
— idi + id2 + Hz — id* = 0
0.20 —nidi — nidi +# f f % +*&% = 0

2-0.15
V V nid2
Xiidi
—nidZ
+#4%4
~X2iq2+Xzig3
-\~niq2
= E3 sin δι
+ 7-4% = E4COS(5i — δ4)
nidi +nidi —xdqi — £ 4 sin(5i—δ 4 )
0.10 (lb)
/
0.05 If the effect of the amortisseur windings is neglected it has been
shown 12 t h a t the load angle below which negative damping of
i J
0.2 0.4 0.6 0.fi I.O I.2 I.4 I.6 I.8 2.0
\ j i
2.2 2.4
generator number 1 may occur is accurately given by the condition
PER UNIT x c (when the frequency of the mechanical oscillation is small compared
to normal-voltage frequency),

As indicated by the equation above there are several variables which (2b)
can be used in the interpretation of the analyzer results. The
quantities which would ordinarily be measured in test of an actual Solving equations lb for idl, we find
circuit are t h e transformer voltage and current. However, it was
0 0 0 0 -1
found by experience t h a t the most significant quantities, as far as
the efficient determination of a criterion for proper circuit operation Γ 1 #2 0 0 n n 0 0
Es cos δι x2 -Xz 0 0 n —n 0
was concerned, were the capacitor voltage and transformer flux.
0 1 1 -1 0 0 0 0
Since, as shown in the oscillograms, the current only appears as an àidi (3b)
0 -n 0 0 xq x2 0 0
appreciable magnitude at widely separated intervals of time (e.g.,
Es sin δι n —n 0 0 — %2 xz 0
figures la and 4c) and, when it does appear, it is likely to bear no
obvious relation to its previous value, thus it is very difficult to use
Ei cos ( δι - δ 4 ) X2 0 XA 0 n 0 n
E4sin(5i - δ 4 ) n 0 n 0 — %2 0 — Xi
current as a continuous indication of the behavior of the circuit.
On the other hand, the transformer flux, which bears a direct func­
where Δ is t h e determinant of the coefficients of the currents in
tional relation to the current as given by the saturation curve, is equations lb.
always of a reasonably large magnitude and of a nearly sinusoidal
Differentiating with respect to δι, there is
wave shape. I t appears, as shown in figures 4 to 6, as a sinusoidal
wave having a slowly varying displacement, and the rate of change 0 0 0 0 -1 1 I -il
and magnitude of this displacement were found to be useful factors 0 x2 0 0 n n 0 0
in determining as quickly as possible from a given analyzer run, —Ez sin δι
whether or not the circuit would approach its proper steady state.
Χ2 — Xi 0 0 n -n 0
d 0 1 1 -1 0 0 0 0
Δ
The capacitor voltage was found to be still more useful in this τ ~ idi = 0
respect; t h a t is, if one had to judge from the behavior of only one
ddi -n 0 0 x< x2 0 0
Ez cos δι n -n 0 0 ~~x2 Xz 0
quantity whether the circuit was to approach a normal or abnormal - £ 4 ( l - £ ) s i n ( o i - -δ 4 ) X2 0 XA 0 n 0 n\
condition, one would choose t h e capacitor voltage. This voltage Eiil-k)cos(ôi- -δ 4 ) n 0 TA 0 -Xi 0 —Xi\
determined by its magnitude the rate of change of flux wave dis­ (4b)
placement, b y its rate of discharge the effectiveness of the shunt
resistor, by its duration the continuance of the shifting of displace­ 5δ 4
ment of the flux wave. I n general, it was discovered t h a t if ever where k = — , i.e., k is the ratio of the motion of machine number
Oui
the capacitor voltage increased in magnitude from cycle to cycle
4 to t h a t of machine number 1. For sinusoidal oscillations k may
(it will be understood here t h a t by "cycle" is meant the apparent
be taken as the ratio of the amplitude of the machine oscillations.
cycle of the low frequency current pulsations and not that of the Equation 4δ, together with equation 2b, may be written
applied voltage), for example, if it ever went higher the second time
than it did the first, the system behavior was abnormal. On the
0= Δ -Ez(Mzi sin δι + M6i cos δι) —
other hand, if it continually decreased in magnitude from cycle to ;*Γ-
cycle the circuit operation was sure to be normal. This is illus­ E4[Mn sin (0! - δ4) + Mm cos (δχ - δ4)](1 - k)
trated by figure 6 which shows curves of capacitor voltage taken a t (5b.l)

984 Butler, Concordia—Series Capacitors ELECTRICAL ENGINEERING

 1.O1

" ■ ^

EA THIS AREA APPLIES FOR F:

IFID
■Λ
MM + ( 1 — k) ■— {M%i cos 6A — M71 s i n δ4) ! 0.8

J
CIRCUIT OPENED OR CLOS ED
Ez
t a n δι = — (5b.2) ^"
E4
Mzi + ( 1 — k) — (Mn c o s h + Λίβι s i n δ4) 3 0.6
■Ë3

The relation — idl = 0 thus leads to a transcendental equation

αδι
£0.4 pi
!
which may be solved numerically for δι as a function of 64. DOTTED LINE APPLIES FOR T 0 S30
i 0.2
In equations 56, there is (APPROXIMATE SOLUTION)

*2 0 f\ — r2 0
-r2 0 Xq — x2 0 0 0.04 0.08 0.I2 0.I6 0.20 0.24 0.28 0.32 0.36
Mzi = ^2 + ^3 —r% xz x2 + x 3 #8 (6b. 1) œ=WH
x2 xA 0 — r2 7-4
r Figure 18. Regions of self-excitation of synchronous
* TA 0 x2 — *4
machine with armature capacitance as affected by shunt
x2 0 n r2 0 resistance R
x2 4 - xz - x 3 —n r2 + r 3 —rs
M,i = —r2 0 *« X2 0 (6b.2) xd = 1.0 xQ = 0.6 xd' = 0.30 T0 = 1,000 r = 0
X2 XA 0 r2 r4
r
2 TA 0 — X2 — XA

X2 0 *Ί — ?-2 0
x2 #3 0 —r2 — r3 Z2\ri + rz) + xq(nx2 — Xzr2) + ^1(^2^3 + r2rz)
tan 6i = (8b.2)
M 7i —r2 0 *« —x2 0 (6b.3) Z22(xg + Xz) + xq(x2Xz -f r2rz) + ri(x3r2 — x2rz)
r2 n 0 x2 Xz
where
r2 - f U —rA x4 x2 - f x 4 — XA

Z22 = r22 + *22

x2 0 n ^2 0
x2 -x3 0 r2 —r 3 The right-hand side of equation 86.2 is recognized as the ratio of
Msi = —r2 0 *β
x2 0 (6b.4) the real to t h e imaginary part of the transfer impedance from ma­
r2 —n 0 —x 2 Xz chine number 1 to the infinite bus if xq is used for the machine re­
X2 + X4 XA —r 4 r2 + ^ TA actance. Further, if the shunt load is removed (Z2 = œ) it checks
the criterion of reference 11.
In the special case for which the motions of the 2 machines are
If the shunt load is removed b u t machine number 4 is kept, t h e
the same (k = 1), equations 56 reduce to
form of equation 5b is unchanged b u t equations 6b are reduced t o :
Mu
tan δι = (7b) —rz Xq + *3 Xz
Mu
-Mzi = XA —n rA (9b.l)
but this is not likely to be a useful case. If k = 0, there is no sim­ rA Xq —XA

plification. Xz ri + n rz
If machine number 4 is eliminated by letting xA and/or YA be -Mn — XA r\ -r4 (9b.2)
infinite, Mn, Mu become negligible compared to M3U M*i, and x
YA a — XA
equations 5b reduce t o :
-Xz fi rz
x2 — r\ r2 -Mn = rz Xq Xz (9b.3)
X2 + Xz rz r2 + r z —rA Xq + Xi — XA

—r2 ~xQ x2
tan δι = (8b.l) xz n rz
x2 r\ —r2 — M&1 -rz Xq Xz (9b.4)
— r2 xg —x2 —XA f\ + ^A — rA

r2 + r 3 Xz x2 + #3 ons 5b EA = 0, or if x 00 t h e r e is,

_, Mn
δι = —tan (10b)
Mzi
which is equation 7b if EA = 0, and is equation 8b if XA or r4 = œ.
If EA = 0 b u t XA and r4 are finite, equation 10b m a y of course be put
in the form Sb by paralleling Z 2 and Z 4 , i.e., in equation 8b replace
r2 and x2 by the corresponding real and imaginary parts of the
parallel values of Z 2 and Z 4 . This reasoning shows that since tan δ
in (86.2) equals the ratio of t h e real and imaginary components of
the transfer impedance from machine number 1 to the infinite bus
(number 3) therefore —Mu/Mzi, where Mu and Mzi are given by
(66.1) and (66.2), must also be the ratio of the real and imaginary
parts of the transfer impedance of machine number 1 to the infinite
bus with ZA finite. In fact, directly from equation 36 it may be
seen that since :

Figure 17. Starting and self-excited current of induction Δίάι = —Μ2ϊΕι -f MziEz c o s δι — Μ6ιΕζ s i n δι +
motor with series capacitance in primary circuit M7IEA COS (δι - 6A) - MSIEA s i n (fc - δ4) (lib)

AUGUST 1937 Butler, Concordia—Series Capacitors 985

Then: tions (see reference 20) :

— Μ21/Δ is the imaginary part of the driving point admittance t e + a)[ed — ρψα + rid + ωψϋ] + xcid = u[eQ — ρψϋ + riq - ωψά]
at machine number 1 (p + a)[eq — ρφα + riq - ωψα] + Xch = —<a[ed — ρψα + ηα+ωψ9]
Mn/ Δ is the imaginary part of the transfer admittance from 1 to 3 (le)
Mn/ Δ is the corresponding real part where, for an induction motor with completely symmetric rotor and
Mn/ A is the imaginary part of the transfer admittance from 1 to 4 stator,
Msi/ Δ is the corresponding real part Ψα = -x(p)id (2c)

It follows that since

and
/ real \ / real \
(12b) x'T0p + x
\ imaginary/admittance \imaginary/i m p edance x{p) = (3c)
ToP + 1
we have
Substituting (2c) in (lc) and rearranging; there is:
MM _ / real \
(13b) lip + *)[px(p) + r] + xc - ωΗ{ρ))π - « { ( / > + a)x(p) + px(p) +
Mn \ imaginary/impedance 1-3 AH = —(p + <*)ed + o>eq
(4c)
and ω{(ρ + a)x(p) + px(p) + r} id + {(p + a) [px{p) + r] + xc -
real œ2x(p)}iQ = — œed — (p 4- oL)eq
_Msi = (
(14b)
Mn \ imaginary/impedance 1-4 and for constant rotor speed the characteristic determinant (the
denominator of the expression for current as a function of voltage),
All of these transfer impedances and admittances are computed for
a direct-axis current in machine number 1. Also, since it is only the Δ = {(p + ct)[px(p) + r] + xc - c A ( £ ) } 2 + ω*{(ρ + a)x(p) +
relative magnitude of the components of impedance which enter into px(p)+r\* (5c)
the final criteria, the impedances calculated entirely by means of
Substituting (3c) in (5c) and rearranging in descending powers of
xq can be used. T h a t is, even though xd occurs in Δ and affects
p, there is:
the true values of the impedances, Δ does not appear in the results.
The method of extension to more than 2 machines is evident from (Γ 0 £ + 1) 2 Δ = Δ' = p\x'*T0*)
equations 2b, 3b, and 4b. + ρ5[(2χ'Τ0)(χ + αχ'Το + rTo)]
+ ρ*[(χ + αχ'Το + rToy + (2χ'Τ0) X
Composite Loads (αχ + ω2χ'Τ0 + r + arT0 + xcT0) ]
+ p>[2x'T0(Xc + ω2χ) + 2(x + ax' T0 + rT0) X
If the shunt load of figure 9 is a composite load 18 of any kind all {ax 4- ω*χ'Τ0 + r + arT0 + xcT0) ]
the criteria given here may be used unchanged if r2, #2 are defined + p*[(ax - ω*χ'Τ0 4- r 4" *rT0 + xcT0Y +
by the relation: 2(x 4- ax'To + rT0)(xc — ω2χ) +
de2 , . (ax'To 4- 2x 4- rToyœ* + 4coVr o (oa; 4" r)]
—- = r2 4-7*2 (15b) + p[2(ccx - ω2χ'Τ0 + r 4- arT0 4- XcTo) X
(xc - cA) 4- 2ω\αχ'T0 + 2x + rT0) X
at the load voltage e2. (ax -\-r)]+(xc - ω 2 χ) 2 + (ax + r)2a>2
(6c)
Equation 5c is of the form ( Δ' = a2 -f- b2) and so may be factored
Appendix C—Self-Excitation of Induction Motors as [ Δ ' = (a 4- jb)(a — jb)] thus at once reducing the order of the
equations to be solved by half. However, this is accomplished at
the price of introducing complex coefficients. The real parts of the
The object of this appendix is to present criteria to determine the
roots of the 2 factors are identical in pairs; the imaginary parts
limits of stable operation of induction motors with series line capaci­
equal and opposite. Thus only one factor need be considered, but
tors. Under certain conditions of load it is found in practice t h a t
it is found t h a t all methods for determining the character of the
there will exist undamped current pulsations of apparently low
roots or solving the equation require about the same amount of
frequency and large magnitude. These pulsations are here con­
work as doubling the order and in effect obtaining equation 6c.
sidered to be caused by undamped or negatively damped (ampli­
In complex form the equation may be written as :
fying) free electrical oscillations. Thus, to determine whether a
given (electrical) system is stable or unstable the natural currents a + jb = x'ToP* + [(x + ax'To + rT0) +ja>2x'T0]p2 +
are examined. The time variation of these natural currents is l(xcT0 - ω2χ'Τ0 + ax + r + arT0) - H « ( 2 s + ax'T0.+
specified by the roots of the characteristic determinant of the system. rT0)]p + l(xe - c A ) +ja>(ax + r)] (7c)
If the real part of any one or more of these roots is positive the
system is unstable since then the corresponding component of the with a similar conjugate expression for (a — jb).
transient current is amplifying rather than decaying and will tend I t will be observed that equations 6c and 7c are almost symmetrical
to increase indefinitely until limited by changes in the circuit caused in r and a except that a has sometimes the coefficient xf and some­
by saturation, slowing down or oscillating of the motor rotor, etc. times x, which destroys the symmetry. It is therefore expected
Thus, it is not necessary actually to solve for the roots but only t h a t all the conclusions and curves obtained will be very similar if
to find the signs of their real parts. To find the actual frequency the effect of shunt and series resistances are investigated separately,
of the unstable oscillation it is necessary to solve the equation, using a as the variable in case of the shunt resistance. Here a
which is of the sixth degree (or of the third degree if expressed with may be thought of as the reciprocal of the time constant of the RC
complex coefficients) and has in general 3 conjugate pair of complex circuit considered by itself. I t is the negative of the coefficient of
roots Or 3 natural frequencies. time in the exponential solution of the RC circuit equations.

APPROXIMATE C R I T E R I O N FOR INSTABILITY

Analysis
I t will be found that a very simple and sufficient condition for
For an induction or synchronous machine with capacitance and instability is that the coefficient of p in equation 6c be negative.
resistance in shunt connected to the terminals, there are the equa­ T h a t is, if this coefficient is negative the system is surely unstable,

986 Butler, Concordia—Series Capacitors ELECTRICAL ENGINEERING

but if it is positive the system is not necessarily stable. Putting Since at low speeds the capacitive reactance a t which the maxi­
the coefficient equal to zero there is: mum series resistance or shunt conductance is required is shifted
toward the actual capacitive reactance likely to be used in the
(ax + r) circuit, equation 10c or 8c then becomes a useful criterion. Having
(x + χ')ω2 — ar —
To the boundary curve of instability at any one speed, those at any
(x - χ') 2 ω 4 l|~ , (*x+r)-] other speed may then be quickly estimated by the relations l i e .
(x + Sxf)ar + — \ ar ^ —
Figures 11 and 13 show the effect of varying the rotor speed as
determined by equation 10c. Figure 11 shows the effect on the
x)
— ω2(χχ'α2 + r2) ■ (3x +
-fr) ° value of series resistance requirod if there is no shunt resistance,
while figure 13 shows a similar curve of a versus xc if r = 0.
Equation 10c also provides a means of determining quickly an
The 2 values of x c obtained from (8c) for a given system determine
approximate expression for the value of a required to eliminate
the limits of a region which is theoretically surely unstable.
instability completely at any speed. This expression is found by
If a set of circuit constants is selected, leaving r and xc free, then equating the radical in equation 10c to zero, which locates a point
r is varied and xc computed, a region is outlined as in figure 10 very close to the maximum of the a versus xc or r versus xc boundary
(dashed curve). This figure shows r a s a function of xc if there is curve. Then
no shunt resistor (a = 0). If the effect of shunt resistance alone is
determined, letting r = 0, the dashed curve of figure 12 is obtained. (x + 3*')r'
Equation 8c outlines the unstable region very close to the true
boundary over most of the range of instability. However it indi­
4xx' — r'2 + 4xx' -
X

cates that instability lies wholly in the range, V(x + 3 x ' ) V 2 + (4xx' - r'2)[(x x') 2 - 4r' 2 ] (12c)
where r' — r/ω.
A ' < Xc < ω2Χ (9c)
If r = 0, there is
so that if xc < ω2χ' the system is stable when in fact there is required
a certain finite series or shunt resistance to assure stability. Equa­ M = (*-_*2 (13c.l)
tion 8c is therefore not a good criterion when xc is small as in the \ω/ιηα Χ 2Λ/ΧΧ'
case of series capacitors for line reactance compensation except
when the motor is running at low speeds. which occurs a t
Xç X ~j* X
(13c.2)
EXACT CRITERION
Z2 " ~~2
To find the true boundary in the region xc < ω2χ' the criterion of Similarly if a = 0, there is
Routh 24 is used. This is carried out numerically b y assigning values
to all of the system constants and testing equations 6c or 7c for fr\ = (x - x')
(13c.3)
stability with various values of line or shunt resistance. The boun­ \w/max 2
daries found in this manner are shown in figures 10 and 12 (solid
lines). These figures are calculated for a motor of 0.16 per unit which occurs a t the same xc as in equation 13c.2. These last rela­
transient reactance and an external reactance of 0.24. In figures tions determine equations,
10 and 12 there are also plotted boundaries corresponding to opera­
tion at the same speed but with some external rotor resistance, the (x — x')2xc
for < (14c.l)
total rotor resistance being 10 times the rotor winding resistance.* 2(x + x')
The effects of series and shunt resistors have been determined sepa­
and
rately in order to show as simply as possible the orders of magnitude
of the various quantities involved. However, since there is always (x — x')2xc
some line resistance present, the value of a required need not be as f or r = 0 (14C.2)
2(x + x')
large as that given by the curves of figure 12. Figure 15 shows an
example of how the a necessary for stability can be decreased with which are the equations of the loci of the maxima of the r versus xc
increasing line resistance, and has been computed from equation Qc and a versus xc curves as the speed ω is varied. Then, if the capaci­
by Routh's 24 method. tive reactance xc in a given circuit is known equations 14c determine
the value of r or a needed to eliminate self-excitât ion at all speeds.
EFFECT OF ROTOR S P E E D However, as mentioned in the body of the paper, it will not in general
be necessary to do this.
Equations 6c or 7c may be used to determine exactly the regions
of unstable operation at any rotor speed. However, the effect of FREQUENCY AND T I M E CONSTANT
speed may be more clearly and easily seen from the approximate OF THE AMPLIFYING OSCILLATION
expression (8c). Since the rotor open-circuit time constant is
usually rather large, equation 8c may be further simplified by neg­ The frequencies and time constant of all these components of the
lecting terms containing T0, whence transient currents may of course be found directly by solving equa­
tions 6c or 7c. However, it has been found that within the normal

ï[-+-fe)Q]
range of circuit constants only one of these currents is amplifying
and so likely to be observed as a large current pulsation. Moreover,

+ this current is t h a t corresponding usually to a small root of 7c,

V^H^K)1-^)(;) K-:)"(;)" (10c)

over most of the range of xc. I t should therefore be possible to
find the root approximately by neglecting higher powers of p in
equation 7c whence the solution is

Equation 10c shows that as the speed ω is varied the scales of the (xc — ω2χ) -{-jœ(ax + r)
r versus xc or a versus xc diagram are simply changed so that Pi 2
(15c)
[(xc — ω χ' + ar)T0 -\- ax -\- r] -\-jœ [(αχ' -f- r)T0 + 2#]
xc is proportional toco 2
(lie) A further obvious simplification is obtained by neglecting the
r and a are proportional to ω
terms in the denominator which do not contain T0. If pi — pr + jq
* If rotor resistance is zero (To — °») there is no instability. Also if T0 then pr is the decrement factor and (ω + ç) the frequency of the
or ω = 0 there is obviously no instability. possibly amplifying oscillation. The frequency is ω + 2 rather than

AUGUST 1937 Butlery Concordia—Series Capacitors 987

q since the solution is for id, iq. The phase currents are obtained b y manner similar to that of reference 21 there is found an approximate
Park's 2 5 equations formula for t h e lower region:

ia — id cos Θ — iq sin Θ, etc. , , (Xq - Xc)(Xd - Xc)

-\- a i = (5d)
and it may be shown t h a t the component of phase current of fre­ — XqXd
quency ω — q vanishes identically.
If the real p a r t pr of pi is equated to zero (corresponding t o t h e Xq i Xd , ^{\^Q ^d )
XqXd i («d)
point a t which t h e oscillation is just neutral) equation 8c is arrived 2 Λ 4
at by a different route.
The approximate equation 15c holds only for xc > ω2χ' and so in Similarly, an exact formula for the upper region is obtained b y
the useful range of capacitance a t normal motor speeds, equation 7 c replacing xd b y xd in equation 5d, whence
must be solved directly. I t is still possible to avoid the solution of (xq - xc)(xd - xc )
a cubic equation if it is desired t o find t h e frequency only a t points + a2 (7d)
XqXd
on the boundaries between t h e stable and unstable regions, since
at such points t h e root is obviously pure imaginary and if p — jq,
equation 7c may be split up as Xq+Xd . J{Xq -Xd)2 2
Xc = * ^ - XqXda2 (8d)
a + jb = - [(* + *x'T0 + rT0)q2 + ω(2χ + ax'T0 + rT0)q -
(xc - ω2χ)] -j[x'T0qz + ω2χ''T0q2 - (xcT0 - ω2χ'Τ0 +
ax + f + <xrT0)q — ω(αχ -f r) ] (16c) References
and the real a n d imaginary components of (16c) equated separately
1. SERIES CAPACITOR INSTALLATION AT BALLSTON, N . Y., E. K. Shelton.
to zero. Sets of system constants on t h e boundary may then be General Electric Review, volume 31, August 1928, pages 432-4. SERIES
first found by Routh's 2 4 method and q then obtained by solving t h e CAPACITORS FOR TRANSMISSION L I N E S (BALLSTON, N. Y.), E. K. Shelton. Elec.
quadratic equation a = 0 in equation 16c. Of t h e 2 real roots thus Engr. of Australia and N. Z., volume 5, October 15, 1928, pages 245-7.
found only one will in general satisfy t h e cubic equation b = 0 of 2. APPLICATION AND PERFORMANCE OF SERIES CAPACITORS, M. I. Alimansky.
General Electric Review, volume 33, November 1930, pages 616-25.
(16c) and this will of course be the proper frequency. For example,
3. SERIES CAPACITOR IMPROVES 4,150 VOLT L I N E OPERATION, W. S. Hill.
taking circuit constants corresponding to t h e points xc — 0.2, General Electric Review, volume 36, October 1933, page 461.
r = 0.035 in figure 10, the frequency is found t o be very nearly 4. W H Y N O T SERIES CAPACITORS FOR DISTRIBUTION F E E D E R S ? C. L. Dudley
0.7 times normal (rotor speed). At values of xc > x' the frequency and E. H. Snyder. Electrical World, volume 103, June 30, 1934, pages 942-5.
is always very close to, and slightly less than t h a t corresponding t o 5. SERIES CAPACITORS, R. E. Marbury and W. H. Cuttino. Electric Journal
the rotor speed. volume 33, March 1936, pages 145-9.
M. Takahashi 2 2 has an approximate solution of equation 7c for 6. LARGEST CAPACITOR BY INSTANTANEOUS RESPONSE ELIMINATES STUBBORN
FLICKER, ORIGINATING IN INDUSTRIAL LOAD O F F RESIDENTIAL FEEDER, J. J.
the case of no shunt resistance and has determined regions of self- Lewis. Electrical World, volume 106, November 7, 1936, pages 3488-90.
excitation for this case. His results are in general agreement with
7. SERIES CAPACITOR PROVES ECONOMICAL, A. E. Perreten. Electrical West,
those given here. volume 77, December 1936, page 48.
8. IMPROVED VOLTAGE REGULATION W I T H SERIES CAPACITORS, R. C. Buell.
General Electric Review, volume 40, April 1937, pages 174-9.
9. EFFECT OF SERIES CAPACITORS U P O N STEADY STATE STABILITY OF POWER
Appendix D—Self-Excitation SYSTEMS, E. S. Allen and J. L. Cantwell. General Electric Review, volume 33,
May 1930, pages 279-82.
of Synchronous Machines
10. DIFFERENTIAL ANALYZER ELIMINATES BRAIN FAG, Irven Travis. Machine
Design, July 1935, pages 15-18.
From equations 15 of reference 20, the equations of a synchronous 11. STABILITY OF SYNCHRONOUS MACHINES AS AFFECTED BY ARMATURE
machine operating a t synchronous speed, with only one rotor wind­ RESISTANCE, C. A. Nickle and C. A. Pierce. AIEE TRANSACTIONS, volume 49,
ing, with series capacitance and resistance in t h e armature circuit, June 1930, pages 338-350. Discussion, pages 350-1.
and with additional resistors shunting t h e capacitors, are: 12. EFFECT OF ARMATURE RESISTANCE UPON HUNTING OF SYNCHRONOUS
MACHINES, C. F . Wagner. AIEE TRANSACTIONS, volume 49, 1930, pages
1011-26. Also refer to discussion of reference 9.
(P + oi)(ed - ΡΦά + rid + tq) + XcU = eq - ρψα -f riq — ψά , ^
13. L E S RÉSEAUX DE TRANSMISSION D'ENERGIE (a book), Jean Fallou.
(P + °ö(eq - Ptq + riQ - ψά) + xciq = -ed + ρψά - rid - φί
Pages 390-408.
where 14. P. Boucherot, Comptes Rendus des Séances, Académie des Sciences, volume
170. Paris.
, /^\· Xd'T0p -f- xd .
Ψά = -Xd{P)U = ^ L , , Π (2d) 15. Jean Fallou, Revue Générale de L'Electricité, volume 19, June 1926, page
Top + l 987.
16. THEORY OF ABNORMAL LINB-TO-NEUTRAL TRANSFORMER VOLTAGES,
~Xq(P)h = -Xqiq (3d) C. W. LaPierre. AIEE TRANSACTIONS, volume 50, March 1931, pages 328-42.
If (Id) are solved for t h e currents id, iq in terms of the applied 17. STUDIES I N NON-LINEAR CIRCUITS, C. G. Suits. A I E E TRANSACTIONS,
volume 50, June 1931, pages 724-32.
voltages ed, eq, it is found t h a t the operational common denominator
18. STEADY STATE STABILITY OF COMPOSITE LOADS, S. B. Crary. ELECTRICAL
of the 2 formulas for current is : ENGINEERING, volume 52, November 1933, pages 987-92.
Δ' = pHqXd'To 19. FIELD T E S T S TO DETERMINE THE DAMPING CHARACTERISTICS OF SYN­
CHRONOUS GENERATORS, F. A. Hamilton, Jr. AIEE TRANSACTIONS, volume
+ p*[xqXd + rT0(xq + xd') + 2 Oil oXqXd J 51, September 1932, pages 775-9.
+ p*[(2xqxd' + xc\xq + xd'))T0 + r(xd + xq + rT0) +
20. Two REACTION THEORY OF SYNCHRONOUS MACHINES W I T H ANY BALANCED
a(2xdXq + aXqXd'To) + 2arT0(xq + xd') ] TERMINAL IMPEDANCE, C. Concordia. Submitted for publication.
+ p2[2xdXq -1- Xc(Xd + Xq) + K T0 {2xc + Xq + Xd } + 0 + 21. T w o REACTION THEORY OF SYNCHRONOUS MACHINES, S. B. Crary. ELEC­
a(T0xc{xq -{-Xd} + 2T0xqXd + oiXdXq) + ατ(2χά + TRICAL ENGINEERING (AIEE TRANSACTIONS), January 1937.
2xq + 2Ts + T0a{xç+xd'\)] 22. TRANSIENT PHENOMENA OF AN ALTERNATOR UPON CONDBNSIVB LOAD,
+ p[T0(xc - xq)(xc - Xd) + r(xd +Xq + 2xc -f rT0) + M. Takahashi. Researches of the Electrotechnical Laboratory, Tokyo, Japan,
No. 350, August 1933.
ct(xc{xd + xg\ + 2xdxq + aXqXd'To) + ar(2xcT0 + 2r +
23. TRANSIENT ANALYSIS OF ALTERNATING CURRENT MACHINERY, Y. H. K U .
a{xd +xQ} + arT0)] Journal of Electrical Engineering, China, volume 3, number 4, August 1932,
+ l(Xd - Xc)(xQ ~ xc) + r2 + a2xdxq + 2arxe + (ar)2] (4d) pages 179-206.
24. ADVANCED RIGID DYNAMICS (a book), E. J. Routh. Page 168.
If the same generator as in reference 21 is taken, with r = 0, and
the stability of the system as affected by a is studied by means of 25. T w o REACTION THEORY OF SYNCHRONOUS MACHINES, R. H. Park. Part I,
AIEE TRANSACTIONS, volume 48, 1929, pages 716-27; part II, AIEE TRANS­
Routh's 2 4 criterion, t h e results of figure 18 are obtained. Also in a ACTIONS, volume 52, 1933, pages 352-5.

988 Butler, Concordia—Series Capacitors ELECTRICAL ENGINEERING