SHAPING CIRCUITS FOR HIGH VOLTAGE IMPULSES

F.C. Creed and M.M.C. Collins

National Research Council of Canada
Radio and Electrical Engineering Division
Ottawa, Canada

ABSTRACT One of the circuits commonly used to generate this impulse is shown
in Fig. 2.Variations of this include having either RSi or RS2 zero and
The calculation of the resistance values necessary to generate a having a capacitive voltage divider, in which case Cs would include the
particular double exponential impulse shape or the calculation of the divider capacitance and Rd would be eliminated (Rd = oo). In the
two time constants from the resistance and capacitance values is a following calculations any of these cases may be readily taken into
difficult and time consuming problem if done rigorously. In order to account.
simplify these calculations, an approximation method has been
developed which can be carried to any degree of accuracy by an -0
Rs[
- -.
Rs2
.

iterative procedure. For normal purposes the initial calculations, done

with a slide rule, will yield results of a sufficiently high accuracy.

INTRODUCITION _ Gg Rt Rd:
The exact analysis of a particular circuit used to generate a
double exponential impulse is a process which is both lengthy and
tedious. In addition, it is generally necessary to do a complete
analysis for each different circuit used. Various attempts have been
made to produce simplified procedures for analyzing and designing
impulse shaping circuits but they usually involve special charts or Rt- tail resistor Cg-generator capacitor
computer programs1-7. Moreover, most of them do not help the Rd- voltage divider Cs - load capacitor
reader to become familiar with the operation of the circuit. The RSiI RS - front resistors
approach taken in this study has been directed towards under-
Fig. 2 Typical circuit for impulse shaping
standing the basic behaviour of the circuit and the development of
simple techniques for predicting the shape of the impulse generated.
With this approach it becomes relatively easy to determine what CALCULATION OF TIME CONSTANTS
changes should be made in a practical system if it should be necessary
In order to ascertain the shape of the impulse which will be
to correct the impulse shape.
generated using a particular generator and shaping circuit, it is
The standard impulse as specified by the International Electro- necessary to determine the two time constants of equation (1). In
technical Commission (IEC)8 and USA Standard' is shown in Fig. 1. addition, the factor K must be found if the efficiency is to be
calculated.

Impulse Tail Time Constant

The tail time constant, rt, which we will consider first is readily
seen to be dependent on the discharge of the two capacitors Cg and
go
50- -
Cs. If each of these capacitors is considered independently and the
C. 50 -2
two time constants are added, an approximation to the true impulse
tail time constant is obtained which will be called rtl.

TtI =

TCg +
s
(2)
° T30Tso Time T50
7C5 = discharge time constant of Cg ignoring C.
Virtual front time = 1.67 (Tgo - T30)
Time to half value = Tso Tc = Tc s time constant of C. ignoring
discharge 9~~~~~~~
C
(Time measured from 0')
Fig. 1 Full standard impulse (1.2/50) For the circuit of Fig. 2

This impulse is of the double exponential forn where the output Tg = Cg[Rs + Rt 11 (RS2 + Rd )] (3)
9 S
voltage, VO, is expressed by
t/Tt.
C = CS [Rd 11 (RS2 + Rt)] (4)
Vo = K e - t/ v9 (1)
The error in this approximation to rt is due to RSI and RS2 not
K - efficiency of the generator circuit being common to both discharge circuits. Since the values of these
Tt impulse tail time constant resistors govern the impulse front time constant it is logical to expect
that the error in rt, is dependent on Tf. This can be proved in a
rf - impulse front time constant complete analysis of the circuit which shows that
V - generator charging voltage multiplied by the number of
stages t +f (5)
Paper 71TP 122-PWR, recommended and approved by the Power System Instrumentation and Measurements Committee of the IEEE Power Engineering Society
for presentation at the IEEE Winter Power Meeting, New York, N.Y., January 31-February 5, 1971. Manuscript submitted September 16, 1970;made available for
printing November 25, 1970.
2239
For most applications, therefore, Tti is a sufficiently close approxi- If the ratio rf/rt is not sufficiently small to give an acceptable
mation to rt. If, however, greater accuracy is required, it can be error in rt, and rfl, a closer approximation can be obtained by an
obtained by an iterative procedure which will be discussed after the iterative procedure. From equations (5) and (9)
approximation for rf has been outlined. For the IEC Standard
1.2/50 impulse, rt = 68.5 ,s and rf = 0.405 ps so that the error in Tfl Ttl Tf tt (10)
using r;1 is only 0.6%. The standard tolerances on this impulse are
±30% on the front time and ±20% on the time to half value. This and hence
-t r
gives an approximate range for rt of 55,s to 80 As and for Tf, rf = (11)
0.3 us to 0.5 As. Thus for minimum rt and maximum rf the error rfrt -

in using rt, increases to approximately 1%. It should be noted that By substituting the best approximation for rf in the right hand side of
for any values of rt and rf within the above limits the impulse shape equation (I1) a better approximation is obtained. Thus the nth
will be acceptable. If, however, more precise information is required, approximation is given by
Fig. 3a shows the time to half value for a given 7r and Fig. 3b shows
the front time for a given rf. In addition, the variation of the time to rf rt (12)
half value for the usable range of Tf is shown, as is the variation of 'Ffn rt '- rf(n 1 )
the front time for the usable range of rt.
This converges very quickly; for example, even with a very low ratio
64L 2.0L of rt/rf such as 10 where rf, has an error of 10%, rf2 has an error
of only 1o. When the desired accuracy for Tf has been reached,
rt is determined using equation (5) in the form
'rt =rtl -rf
Efficiency
Once the time constants have been determined it is possible to
calculate the efficiency of the geperator. The efficiency, which is the
peak output voltage divided by the charging voltage per stage times
the number of stages, is made up of two parts. Firstly, there is the
factor K which is dependent on the generator circuit and, secondly,
a shape efficiency due to the maximum difference between the two
exponentials being less than one. For a given impulse shape the
second factor is completely independent of the generator circuit
50 60 70 80 90 0.2 0.3 0;4 0.5 0.6 and is only dependent on the ratio rt/rf as shown in Fig. 4. Knowing
TAIL TIME CONSTANT Tt-.Ls FRONT TIME CONSTANT Tf -S the time to peak,
(a) (b) p t= Tft ln rt (13)
tTf- Tf
Fig. 3 Relationship between rf, rt, front and tail times
the shape efficiency can be calculated by taking the difference of the
Front Time Constant two exponentials. For the standard impulse, however, a value of 96%
can be used since this will only vary ±1 -%6 over the allowable
If C were infinite, the front time constant, Tf, would be Cs tolerances.
multipliec by the resistance of the generator as seen from C., which
we will call Rg. Since Cg is not infinite the voltage to whlich Cs will
be charged is lower than the initial voltage on Cg so that Tf will be
reduced. Two factors cause this loss of voltage. One is the charge
transferred from C to C and the second is the charge loss from
C through the taif resistance.
By analyzing several circuits it was discovered that the charge
transfer could be taken into account by a factor rc /(Tc + rc ) so
the first approximation for rf, which we call Tf1, is g g
T
tc
Tfl :
Rg CsTg-C+r
+TC
(6)
g s- I-'
For the circuit of Fig. 2, xc and rc are as given in equations (3) and
(4) while g
Rg = [(R5 11 R) + R .2.1 Rd (7)
The discharge of Cg through the tail resistance has not yet been
taken into account but this is done quite simply since a complete
analysis of the circuit shows that

Tf1
1_ 1
Tf
1+7t - (8)
or Tf (9) 30 40 50 60 70 80 90 1100
f
=

I + Tf /Tt SHAPE EFFICIENCY- %

This gives the same percentage error as the approximation for the tail
time constant. Fig. 4 Shape efficiency for a double exponential
2240
 The factor K is dependent on two things. Firstly, the ratio of Since TC and rc are not known, a first approximation for Rg must
the capacitors since any charge used to charge Ca results in a voltage be made from
drop on C3 and, secondly, the resistive voltage division caused by
the front resistors and the tail and voltage divider resistors. tf C + C
(18)
Rg -C ag
Thus K can be approximated by
For the circuit of Fig. 2,
K%:~, Cg 'DR (14)
C +Cs (DR Rg9 -(Rsa ll Rt) + Ra2 (18a)
where DR is the resistive voltage division between Cg and CS. For so it is necessary to pick the value of Rai (this may be a fixed internal
the circuit of Fig. 2 generator resistance) and then to calculate Rs2. Using this first
approximation for Ral and Ra2, rc and rc can be calculated and a
Rd Rt 11 (Rd + Rs2 ) check made of the value of rta afd hence att using equations (2)
Rd+sR2 (14a)
-
DR =
Rd [Rt ll (Rd + R a)] + Raa and (5). If Tt is within the desired limits, Rg can be recalculated using
equation (17) and any modification to Rs2 can be made using the
This approximation gives a result which is about 1% low for the expression
standard 1.2/50 impulse. The error increases as the ratio Tt/Tf decreases
and is approximately 200 %. Rg = [R2 + (Ra Rt)] ll Rd (19)
ft /f
A rigorous analysis of the circuit shows that the efficiency is If rt is not within the desired limits, Rt will have to be increased or
given by decreased approximately as the ratio (rt desired)/(T, calculated). Rg
is then calculated and a new value of Ra2 determined as above.
I1= I
F H.
T

CS (Rs, R32 + RS2 Rt + RtRsj)i NUMERICAL EXAMPLE

I )Ta-tT
(15)
For this example a 12-stage generator will be used which has a
TT capacitance of 0.25 Mt-F per stage. This makes Cg = 20.8 nF. The
generator has 315 Q2 built into tih_.sphere gap arms so that Rai will be
The first term is the factor K while the second term is the shape 315 Q2. The load capacitor, Ca, is 700 pF and the resistive voltage
efficiency. It might appear at first that the approximation is more divider, Rd. is 30 kQ2.
complicated than the true expression for K but the advantage of the
approximation lies in the fact that the underlying principles are The problem is to find Rt and RS2 to give a 1.2/50 impulse for
easy to remember. Thus it is readily applicable to any circuit used which t = 68.5 ,s and Tf = 0.405 gs. For this impulse shape the
to generate the double exponential. difference between Tf and Tf1 and rt and rcan be neglected so that
only rf and wt will be used. The first approximation for Rt will be
obtained from equation (16) which gives R 3,180 Q. Since Cg is
DESIGN OF IMPULSE SHAPING CIRCUITS the dominant capacitance a closer approximation could be obtained
by subtracting RSi from this value and then taking account of the
The second type of problem confronting the imnpulse generator paralleling effect of Rd. These two corrections tend to cancel each
user is that of determining the approximate front and taiL resistances other, however, and since R1 can be readily corrected later, the value-
to give the desired impulse shape with a particular value of load of 3,180 Q2 will be used.
capacitance. Here again the exact solution is tedious and time
consuming so an approximation method has been developed. Equation (18) gives us the first approximation to Rg of 598 Q2.
Rs2 is then calculated from equation (18a) in the form
The starting point is Cg which is a function of the number of
stages used and is thus fixed once the number of stages has been set. Ra2 = Rg- (Rai 1 Rt)
C. will also be known since it is dependent on the test object
capacitance plus any parallel loading capacitor and capacitive voltage This gives a value of 312 Q2 for Rs2* The next step is to calculate rc
divider which are generally of fixed value. If the tail resistance is and rc from equations (3) and (4). 5
fixed, which it frequently is, then the only values to be determined
will be Rsa and R82 since the voltage divider resistance will be fixed.
If the tail resistor is not fixed, a very rough initial approximation Tc = 66.5 gs
can be obtained from
Tt 2.2 pls
R
CG+Cs
(6
(16) TCa =

The first step in calculating Rsa and Ra2 is to calculate Rg which Checking rt, from equation (2) gives 68.7 As. Rt is therefore close
from equation (6) is enough and will not need to be corrected. Recalculating Rg from
equation (17) gives 598 Q2 which is the same as the first approxi-
TC59+ TCa
R =
I
(17) mation. Now the final value of Ra2 is calculated from equation (19)
C
Cs 7Cg giving RS2 =324 2.

2241
Comparison With Actual Results Comparison of the calculations using the rigorous analysis with
the approximate.calculation and measured results is shown in the
In order to check the usefulness of the above calculation the following table.
generator was set up using the calculated values. Since the exact
values were not available the closest existing resistances were used. Exact Approximate Measured
For Rt this was 3100 Q2 and for R.2 305 Q2. Tt 66.8,us 67.1 ps
Tf 0.395,us 0.393,us
To ensure that the actual resistances used would give an impulse
K 0.872 0.863
shape within the tolerances, rf and Tt were calculated using only the
first approximation. From equations (2), (3) and (4) T, rt = 67.1 ps Shape efficiency 0.964 0.96
and from equations (6) and (7) -f rf = 0.393,us. These are both Overall efficiency 84.1% 83%
well within the standard tolerances. The factor K from equation (14) IEC impulse shape 1.17/48.9,s 1.19/48.5 Ps
is 86.3% so that the overall efficiency is 0.863 times the standard Output voltage 114kV 1121 kV 114 kV
shape efficiency of 96%o. This gives an efficiency of 83%.
Effect of Inductance
A brief investigation of the effect of inductance in this circuit
has been carried out and the calculations show that if the inductance
is considered to be distributed, as is the case in practice, the front
time of the impulse is unaffected'°. In an actual impulse the shape
will not be a true double exponential. Rather it will have rounding
at the start of the impulse as is illustrated in Fig. 1. This rounding is
caused by reflections within the generator which are associated with
the travel times and surge impedances of the whole system. The net
result of this is to cause a shortening of the front time by some few
percent.

CONCLUSIONS
The approximation method which has been described can be
used to calculate the values of the generator components needed to
(a) generate the standard 1.2/50 impulse or to check the impulse shape
which will be obtained from a given set of components. The cal-
culations can be readily done with a slide rule and a minimum
amount of information need be remembered along with the simple
theory which enables the user to develop the specific equations
needed for any given circuit.

REFERENCES
(1) P.H. McAulay and F.R. Benedick, 'Calculating Surge Generator
Waves', Electrical Journal, vol. 30, pp. 326-328, August 1933.
(2) J.L. Thomason, 'Impulse Generator Circuit Formulas', Electrical
Engineering, pp. 169-176, January 1934.
(3) A.B. Lewis, 'Calculation of Electrical Surge Generator Circuits',
Journal of Research for the Bureau of Standards, vol. 17,
pp. 585-603, October 1936.
(b) (4) J.L. Thomason, 'Impulse Generator Voltage Charts for Selecting
Circuit Constants', Electrical Engineering, pp. 183-189, January
1937.
Fig. S Impulse shape oscillograms
(5) Arnaldo M. Angelini, 'New Formulae for Practical Calculations
of Impulse Waves', CIGRE paper No. 335, 1950.
The generator output is shown in Fig. 5. The impulse shape
mneasured according to IEC specifications is 1. 19/48.5 As. For these (6) Arnaldo M. Angeliii, 'Impulse Generator Circuit Formulas',
mneasurements the time resolution is about 5 ns and the voltage University of Rome, November 1966.
resolution about 0.4% of peak. The actual voltage generated will (7) H. Anis and R.P. Comsa, 'Identification and Generation of Non-
have a shorter front time than that measured because of the response Standard Switching Surges', paper 70-TP-603-PWR, presented
characteristics of the measuring system. It is estimated that the at the 1970 IEEE Summer Power Meeting.
measuring system will lengthen the front by about 50-100 ns. (8) High Voltage Test Techniques, IEC Publication No. 60, 1962.
The measured output voltage is 114 kV. The charging voltage was
1 L.~ kV so that the above efficiency gives a calculated output (9) Techniques for Dielectric Tests, USA Standard No. C68-1, 1968.
voltage of 1 12 -'2 kV. (10) Discussion on IEEE paper No. 70-TP-603-PWR - F.C. Creed.
2242
 Discussion A. Bizzul, S. T. Mak, and S. D. Northrup (Joslyn Manufacturing and
Supply Company, Chicago, Ill. 60609): The authors outline a good
F. W. Heilbronner (The Technical University, Munich, Germany): The method for quickly determining the shaping circuits for standard high
authors are to be congratulated for a lucidly written paper which com- voltage impulse testing. They neglect the effect of circuit inductance,
bines understanding of the circuit and its basic electrical operation. and the results they obtain from their calculations are in good agree-
The results measured, however, show in the oscillogram of fig. 5b ment with the waveshapes they obtain from actual circuits. The
the whole problem of technical impulse generation: whilst the equiva- authors' method is well suited for impulse circuits of small size,
lent one-stage diagram is suitable for determining the time constants of which is usually the case for voltages to about a megavolt. In larger
the generated impulse from an ideal double exponential function, the circuits for several megavolts, larger clearances are necessary. Hence
actual generator delivers a much more complicated voltage, especially longer leads with larger inductances are required, and the effect of in-
in the impulse front. ductance on the output waveshape is no longer negligible. Figure 1 is a
It can happen, for instance, that the 30-T-line intersects three suggested modification of the authors' circuit, to incorporate these in-
times the oscillation near the start of the impulse; hence an experi- ductances.
mental proof of the front time calculated would be rather unreliable. We studied such a circuit when designing large impulse test circuits.
In particular, when one considers testing with chopped impulses, it Computer studies were made and later verified by actual test set-up.
would be advantageous to the practical impulse circuit design, if an In order to simplify analysis of the front of the output waveshape,
equally simple procedure for estimation or suppression of these oscil- (refer to Figure 1), the divider resistance Rd can be neglected, since it
lations could be found, for until now only complicated circuits involv- was large compared to R3. For the values Co = 0.033 ,F, Cl = 1800 pF,
ing computer programs 1,2 offer a solution. neglecting Ll and L2 and assuming an impulse tail time constant of
rt = 68.5 ,us, and an impulse front time constant of rf= 0.405 Ps, the
REFERENCES authors' method of analysis yields the results R2 = 164Q and R3 =
[1 ] I. Gallimberti, M. Rea: Identificazione ed analisi di circuiti 19002. The corresponding output waveshape is shown by curve I of
equivalenti per generatori di alte tensioni impulsive. Energia Figure 2. Next, two computer studies were made by assuming values of
Ellettrica XLVI (1969), pp. 772-782 (Italian). LI = 30 pH (the approximate inductance of our 1.5 Mv impulse
[21 F. W. Heilbronner: Firing and Voltage Shape of Multistage Impulse generator), L2 = 60 pH (L2 corresponds to the inductance of the shap-
Generators. Paper No. 71 TP 121-PWR, Winter Power Meeting ing circuit) and the values of Co, C1, Rl, R2 as above. The first com-
1971. puter study was made for R2 = 21Sf2 and the second for R2 = 95Q2.
The results are shown by curves II and III respectively. Curve III shows
Manuscript received February 12, 1971. a large, superimposed oscillation with a calculated natural frequency of
376 KHz. For the larger value of R2, curve II shows a smaller oscillation
which is damped out sooner than for the smaller value of R2. The cal-
culated natural frequency for the waveshape shown is curve II is
309 KHz.
S. W. Zimmennan (Cornell University, Ithaca, N. Y.): Solutions of the To verify the computer studies above, tests were performed using
basic equations for waveshape control have been simplified greatly by our impulse generator with the circuit parameters above. Oscillograms
the authors. I cannot resist making the suggestion that the voltage for R2 = 95Q2 and R3 = 215Q2 are shown in Figures 3 and 4, respec-
efficiency of about 80% attained practically in this circuit can be tively. There is good agreement between curve III and the trace in
boosted by the addition of inductance. The authors are indeed aware Figure 3, and curve II and Figure 4. We note similarity in oscillations
of this. Although beyond the scope of their paper I wonder if they in the theoretical results and test data.
would care to suggest what approximate voltage efficiency they might We conclude that for impulse circuits with lead travel time of a
expect to realize by the addition of reasonably well chosen properly few tenths of microseconds, as in our case, that lumped-parameter
placed values of inductance to obtain acceptable (a) front of wave or analysis is indeed applicable. We also conclude that the effect of the
(b) acceptable full wave test data. inductance of this size circuit should be taken into consideration for
The advantage suggested here is to be able to use a given generator impulse wave-front analysis.
to produce a higher test voltage at the expense of slight aberrations
of a "legal" exponential shape, with an overshoot provided by circuit Manuscript received February 17, 1971.
inductance.
The attached references might be of interest to the authors.
REFERENCES R. P. Comsa and H. Anis (McGill University, Montreal, Canada): The
[1 ] J. R. Eaton and J. P. Gebelein, "Circuit Constants for the Produc- authors present a very interesting analysis of the double exponential
tion of Impulse Test Waves." General Electric Review, 1940. impulse, establishing relationships between two groups of parameters,
[21 J. L. Thomason, "Impulse Generator Circuit Formulas," AIEE the network components and the time constants. They give useful
Transaction, 1934, pp. 169-176. methods to determine one group from the other. Particularly valuable is
[3] J. L. Thomason, "Impulse Generator Voltage Charts, for Selecting the use of a single, more general, network with the front resistance
Circuit Constants," AIEE Transactions, 1937, pp. 183-188 and spread on both sides of the tail resistance. The usual practice employs
p. 882. two distinct circuits, each representing a particular case.
Manuscript received February 16, 1971. Manuscript received February 18, 1971.

Ll ~~~R, Lt Re
S _ __,v

Co

_R.

Fig. 1. Equivalent Circuit of Surge Generator & Impulse Wave Shaping

Circuit.
2243
 S. 0.4 4l
. I__ _{-

I I

:~~~~~~~Fg
f:ICSadr
2/ mus

0.
L~~~~~~~I 1.

Fig. 2.
Uni Ste
2.
Repos Fo
1: IEC Standard Impulse
=21hi
3. 4.

II: Unit Step Response For R2 = 215 ohms

111: Unit Step Response For R2 = 95 ohms

While the paper is limited to the study of the 1.2/50 impulse, the
methods suggested by the authors are general and could be applied to
other wave shapes. A slight complication arises when the network com-
ponents are to be determined for a specified impulse. This is due to the
fact that, generally, switching impulses are given by the actual time to
peak tp and the actual time to the 50% value t2 and not by the time
constants. In this case it is suggested that first the time constants rt and
rf be determined from tp and t2 by using the following two simple
equations (author's reference 7)
Tt = (t2- tm)/0.7
=rt/rf = exp [1.35 + 1.20 In (t2/tm)]
The first equation is obtained analytically from the double exponential
equation. Its very simple form is due to an approximation based on
*tlvfi;0S____S
0000000
_t _ rrt >rf.ForThechecking
second equation is obtained by a curve fitting process.
purposes the standard 1.2/50 impulse, for which the
Fig. 3. R2 = 95S2, SWEEP: I /lsec./div. time constants are known to be rt = 68.5 ps and rf = 0.405 pAs, could be
considered. Using the actual time durations measured from the true
origin (not virtual origin) tp = 2.09 ps and t2 = 49.74 us the values ob-
tained for the time constants, by applying the two suggested equations,
are rt = 68.10 ps and rf = 0.395,us.

E. H. Povey (Doble Engineering Company, Belmont, Mass. 02178): This

paper will be most helpful to those engaged in impulse testing over a
wide range of test voltages. We have found it useful in connection with
our 12-stage generator, where only sufficient stages are activated to
meet the voltage requirements of a specific test.
Using the method given in the paper, a table of front and tail re-
sistors has been calculated for 3 to 12 active stages. Some of the values
have been checked on the generator itself, and,the wave shapes were
found to be very close to standard.
All calculated values were checked with similar success, using a
model of the generator made up of standard capacitance and resistance

Fig. 4. R2 = 215Q2, SWEEP: I usec./div. Manuscript received February 18, 1971.

2244
boxes. The box representing the generator capacitance was charged to Frank C. Creed and Malcolm M.C. Collns: We are very grateful to
a few volts through a high-ohmage resistor, and discharged into the re- the discussers for their interest in the paper and for their comments
mainder of the circuit through a no-bounce switch, the closing of which which add much to the value of the paper. The major area of concern
represented the firing of the generator. The resulting wave was viewed
on an oscilloscope, using a low-capacitance input probe connected is the inductance of the circuit which we have chosen to neglect
directly across the box representing the load capacitance. since our work has indicated that distributed inductance is not a
Previously, the model had been adjusted on a "cut-and-try" basis significant factor when the circuit has sufficient resistance to
to assist in unusual test circuit configurations. The method described in
the paper offers a faster and more logical way of arriving at approximate prevent oscillations. To illustrate this, Fig. 11 shows the effect of
resistance values, and reduces trimming of the actual circuit to a changing from a distributed inductance (46 meters of lead, Fig. I la)
minimum. to a lumped inductance (110 pH, Fig. I lb). With all the series
resistance short-circuited these two circuits give the same resonant
frequency. When approximately 500 12 is added, the difference
between the lumped and distributed inductance cases becomes quite
clear showing that the circuit cannot be analyzed solely on the basis
D. Train (Hydro Quebec Institute of Research, Varennes, Quebec, of a lumped inductance. If neither the lead nor the lumped inductance
Canada): The authors must be congratulated for presenting this is used, the impulse shape is the same as that with the 46 meter lead
interesting new solution to an old problem. The method presented will
be useful to all engineers working with high voltage impulse generators. shown in Fig. Ila.
Circuit inductance has been ignored in the analysis because the
front time is unaffected if the inductance is considered to be dis-
tributed. However, in practice the circuit inductance is not uniformly
distributed and the case may arise where the inductance of the load
capacitor is very low compared to the rest of the circuit. This condition
can then be represented by the circuit shown in Figure 1. (ignoring the
effect of tail resistance). If
R =
2,>/C
where
CS X CG
C=
CS + CG
the circuit is critically damped and the output voltage is given by:

v = E X C [-I( I + 2 .t) exp - ]

These equations and the I.E.C. definition of Virtual front time

= 1.67 (Tgo - T30) were used to derive the curves shown in Figure 1.

2.5 3Jv
LC
L Rs / ^

2.0-

10 pH
1.5

CoC
1.0
0
Figure 11 Comparison of lumped and distributed in-
ductance. (a) distributed inductance, (b) lumped inductance
0.5
As Dr. Heilbronner points out there will in general be an
oscillation at the foot of the impulse. In the practical case, provided
0100 1000 10,~000 pF the oscillation is not larger than allowed by the Standard, the best fit
CS Co
X line through the oscillation must be used to determine the impulse
SERIES CAPACITANCE CS Ca shape.
Professor Zimmerman raises a good point about the use of
Fig. 1. Front time for critically damped circuit. inductance to increase the voltage efficiency of the circuit. Where
only the wave front is of interest our generator can produce a peak of
about 150%. When considering full wave data, however, the limit
These curves indicate that for a given circuit inductance it will only be would depend on the overshoot acceptable for a specific test. To
possible to obtain the standard wavefront up to a certain value-of series date we have not looked into the possibility of deriving an approxi-
capacitance and that under these conditions the wavefront duration is
certainly affected by the circuit inductance. mation technique which would include inductance in the circuit.
Manuscript received February 19, 1971. Manuscript received March 25, 1971.

2245
 The results of Messrs. Brizzul, Mak and Northrup illustrate the actual time of crest and of half value on the tail tkt the time
one of the limitations of our simplified technique, which is the case constants should prove to be a useful extension of the approximation
where there is insufficient resistance in the circuit to prevent an over- formulae.
shoot. If they were using a smaller load capacitor so that the generator The verification of the technique by Mr. Povey is very gratifying
resistance was sufficient to eliminate the overshoot then their to us.
calculated resistance value would have been sufficiently accurate. We would agree with Mr. Train that there is a limit to the series
Considering the time involved in determining the required resistance capacitance if one is to obtain the standard impulse shape. What this
by our simplified calculations, it is probably faster to do this and limitation is cannot, however, be accurately determined from the
make corrections when the overshoot is observed rather than to go lumped constant equivalent circuit. Rather the location of resistors
through a complete calculation including the circuit inductances. and how much mismatch occurs at the ends of the leads will have an
This is particularly borne out in our example above where the lumped important bearing on the limit. As illustrated above, an increase in
circuit approach would give the wrong answer. inductance by extending a lead cannot be equated to adding a lumped
The equations developed by Messrs. Comsa and Anis for relating inductance.

2246