See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3163037

The Effect of Transmission Lines and Switching Action on the Electrical

Signals in a Powerful Capacitive Discharge

Article in IEEE Transactions on Plasma Science · September 1990

DOI: 10.1109/27.57522 · Source: IEEE Xplore

CITATIONS READS

14 278

3 authors, including:

Hector Kelly Cesar Moreno

National Scientific and Technical Research Council National Scientific and Technical Research Council
121 PUBLICATIONS 1,087 CITATIONS 81 PUBLICATIONS 1,291 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Hector Kelly on 08 May 2014.

The user has requested enhancement of the downloaded file.

IEEE TRANSACTIONS ON PLASMA SCIENCE. VOL. 18, NO. 4. AUGUST 1990 689

The Effect of Transmission Lines and Switching

Action on the Electrical Signals in a Powefil
Capacitive Discharge
HORACIO BRUZZONE, HECTOR KELLY, AND CESAR MORENO

Abstract-A numerical model for the equivalent electrical circuit of viously considered [4], [5],although without considering
a capacitive discharge that includes the effects of Rnite closure time of a realistic switch behavior.
switches and finite transit time in the transmission line is presented.
An approximate analytical method to predict the oscillations that can The core of an appropriate modeling of the transient
be excited in the circuit is developed. The numerical solutions of the behavior of fast electrical circuits is a good description of
model are compared with experimental signals measured in a plasma the switching action. The switches currently used in high-
focus device. power pulsed devices are of the gaseous type (thyratrons,
spark gaps, etc.), which operate by means of the break-
I. INTRODUCTION down of a gas. They can be conceived as a time variable
resistance r s ( r ) , changing its value from very high to a
T HE PREDICTION of the temporal behavior of volt-
ages and currents in high-power fast-pulsed electrical
discharges is important for a wide range of problems,
very low in a given time. The physical processes account-
ing for the time evolution of this resistance during the
breakdown in a given gap filled with a gas are complex,
among them plasma production. The models developed
and will not be addressed in the present work. However,
for calculating those magnitudes usually use equivalent
measurements of r, ( t ) performed by Sorensen and Ristic
circuits with lumped parameters and ideal switches which,
[6] working with an overvolted nitrogen pressurized gap,
while yielding satisfactory results for the overall behavior
yielded rs( t ) 3 t - 3 . It could seem then natural enough
of voltages and currents, are not sufficient to predict all
to model the switching action by a time variable resis-
of the experimentally observed details.
tance. However, this procedure goes through numerical
Particularly, in plasma-producing fast electrical dis-
difficulties when one tries to implement them. A better
charges (pinches, plasma focus (PF) [l], etc.), it is a
way of modeling the switching action is to represent it as
common feature to observe nonpredicted high-frequency
an element with a time-dependent voltage drop Y ( t ),
oscillations at the beginning of the discharge and also at
changing from the applied voltage uo to zero. This type
other time intervals during which some magnitudes suffer
of switch modeling was described in a previous work [7]
a strong temporal vanation (Le., pinch formation). This
and will be briefly outlined here. The voltage drop U ( ? )
last type of oscillation is sometimes associated with
was represented by means of a Fermi-Dirac-like function
plasma behavior (bouncing of the pinch [ 2 ] ; instabilities
as follows:
[3], etc.).
In this work we present a model for the equivalent cir- uO[1 + ~ X P( --t0/7)]
V(t) =
cuit of a fairly common electrical discharge, consisting of I + exp [ ( t - t0)/7]’
a capacitor bank connected to a load through a transmis-
sion line, which takes into account the effects of finite Note that v(0) = vo and that for t - to >> 7,u ( r ) =
closure times of switches and finite transit time in the line. 0. The parameter 7 is then related to the closure speed,
The model can be extended easily to other types of cir- while the parameter to defines the instant of time at which
cuits (Marx generators, Blumlein lines, etc.). Note that the switch closes. Also note that to must be chosen larger
the idea of representing some components of the electrical than 7 to have a “smooth” switching action ( t o > 37
suffices); otherwise, the initial fall of ~ ( twould ) be un-
circuit of a PF device as a transmission line has been pre-
physically sharp.
In [7] it was shown that the measured behavior of the
Manuscript received June 26, 1989; revised March 21, 1990. This work discharge current of a capacitor on series inductance and
was partially supported by a research grant from the Consejo Nacional de
Investigaciones Cientificas y Tecnicas, and a research grant from the Uni- resistance was adequately predicted by an equivalent cir-
versity of Buenos Aires. cuit using this model of switches, so we will use them
The authors are with the Laboratorio de Fisica del Plasma, Facultad de with confidence in what follows. Also, we have checked
Ciencias Exactas y Naturales, Universidad de Buenos Aires. Pabellon I ,
Ciudad Universitaria, 1428 Buenos Aires. Argentina. in simple cases to see that the equivalent impedance of
IEEE Log Number 9036728. the switch v ( r ) / i ( r ) ,where i ( r ) is the current circulating

0093-3813/90/0800-0689$01.OO O 1990 IEEE

690 IEEE TRANSACTIONS ON PLASMA SCIENCE. VOL. 18, NO. 4, AUGUST 1990

through the switch, can be adjusted by a t law with n --I’

between 3 or 4 during the time interval in which the switch

impedance is larger than the equivalent impedance of the
rest of the circuit, in agreement with [6].
The set of equations for describing the electrical dis-
charge are presented in Section 11; whereas its solutions
for some simple but interesting cases are given and dis-
cussed in Section 111. In Section IV, an approximate an- Fig. I . Equivalent circuit for discharge.
alytical method to predict the oscillations that can be ex-
cited in the circuit is developed. Finally, in Section V the These equations are coupled by the transmission line
numerical solutions of the model are compared with ex- relationships:
perimental signals measured in a PF device.

11. THE MODEL

The equivalent circuit adopted to represent the dis-
charge is shown in Fig. 1. It is composed of two networks
( 1 and 2 ) connected through a transmission line ( a - b in
the figure) with characteristic impedance z o and transit
time 7/.Network 1 contains the capacitor bank (with ca-
Vb

12 =
= A(1; -
z; + 1;
I;)
1 at the extreme b (6)

Zl(T)= Ii(T - TI) Z,(T) = Z,(T - TI) ( 7 )

pacity e,), a stray induction C , , a switch SIwhich con-
trols the beginning of the discharge current i , circulating where If are the right and left current propagating solu-
in this network, and eventually a resistance r , . Network tions along the line, respectively (subscripts a and b in-
2 can contain several elements: a time variable inductance dicate that the I’ are evaluated at the corresponding ex-
C a ( t ) ,a resistance r2, and a switch S2 that controls the tremes of the line), and the line is assumed to be lossless.
beginning of the current circulation in this network ( i 2 ) The corresponding initial conditions are
and allows the introduction of an arbitrary time delay for
Zl(0) = 0 dZl/dT(O) = 0 Z,(T) = 0, T 5 T2.
the beginning of the current circulation in network 2. To
generalize the solutions, we found convenient to express (8)
the equations and the initial conditions in a dimensionless
form, by defining the following dimensionless variables: 111. RESULTS
To get a gradual insight on the combined action of the
line and switches, we will start presenting the results ob-
T/ = ( C ,e , ) - I ’ * T / tained in very simple cases. As the number of parameters
involved in the calculations is rather large, we will limit
V = ./U0 L2 = & / C l our analysis to fixed values for A and TI:A = 9 and TI =
0.017 (chosen to match our PF device, see Section V).
11.2 = (Cl/e0)l/2i,.2/u0 Anyhow, we will show in Section IV that, in most prac-
= (eo/Cl)”2rl,2 A = (eo/Cl)I/2zo. tical cases, these results can be easily extended to other
numerical values of those parameters. We have chosen
Using them, the dimensionless equations can be written Vb( T ) and dl, / d T as representatives of the circuit behav-
as ior, because these are the quantities usually measured in
d2Zl/dT2+ d ( R , I , ) / d T + I , + dV,/dT + dVu/dT = 0 the experiments.

(1) Case A : Open Line Circuit

Vb - V2 - R2Z2 - d(LJ2)/dT = 0 We will first consider the case in which the line is open
at its end ( d 1 2 / d T = 0). In Fig. 2(a) and (b) we show
(2) Vb( T ) corresponding to a l= 0.1 and 0.02, respectively,
where for the cases R I = 0 and 10. As can be seen, Vh( T ) os-
cillates around 1 ( U , , = u o ) , with amplitudes depending
on the value of a,.Also, we note that the period of the
(3) oscillation in Fig. 2(a) is not equal to an integer number
of TI, as could be guessed beforehand. Furthermore, Fig.
2(b) shows the presence of higher frequency components
superimposed to the same frequency of Fig. 2(a). In what
follows, we will call fundamental frequency to the lowest
frequency appearing in the results and harmonics to the
for describing the closure action of switches SIand Sz. higher frequency components (the values of the harmon-
 BRUZZONE et al. : EFFECT OF TRANSMISSION LINES ON ELECTRICAL SIGNALS 69 1

-1 --
0 2
1
T
4 6

Fig. 3. Curves of d1,ldTobtained in short-circuited line case for T ,= 0.1

and R , = 0, 1, and2.

V dl /dT
-1 I !

0
0 0.5 1
T

(b)
Fig. 2 . Curves of Vh( T ) obtained in open circuit case for R , = 0 and 10. -1 4 I
, o 2 4 6
(a) r I = 0.1. (b) r l = 0.02. T

Fig. 4. V h ( T ) for: 1) R , = 0, 2) R , = 1, 3) R , = 2 and of d12/dTfor: 4)

R , = 0, 5) R, = 1 , and 6) R2 = 2. In all cases L, = 0.1.
ics should not necessarily be an integer number of the fun-
damental frequency). From Fig. 2 we find a dimension-
less value of a23 for the fundamental frequency and a Case C: Line Loaded with Constants R, L
value of a 190 for the first harmonic. It can be seen that The results obtained from the model using R I = 0 and
when R I = 10 the fundamental frequency is fairly L 2 = 0.1 will be presented. In Fig. 4, V,( T ) and d12/dT
damped, but not so the first harmonic. are given for three values of R,: R2 = 0, 1, and 2. For R2
= 0, V, = L 2 d12/dT and both curves differ only in a
Case B: Short-circuited Line scaling factor. A fundamental frequency of g 0 . 9 3 and a
We will now consider the case in which the impedance first harmonic with frequency = 65.3 can be seen, this last
of network 2 is null ( V , = 0). Fig. 3 shows plots of distorted by higher harmonics. The effect of R2 # 0 is
d12/dT for al = 0.1 and R I = 0, 1 and 2. The funda- illustrated by the other curves, and strong dampings of the
mental frequency is 2 0 . 9 3 (very close to l , the fre- fundamental frequency and harmonics are observed. On
quency corresponding to the C Ie,, oscillation) and is the other hand, the frequency values of the first harmonics
strongly damped for values of R I much smaller than those seems to be unaffected by the presence of R 2 . In these
of the previous case. Note also that the amplitude of the cases, V,( T ) is essentially proportional to Z2, because the
first maximum in the case R I = 0 is = 0.85 and not 1 (that resistive drop is much higher than the inductive one (note
is, u o / C l ) .This is due to the fact that the line inductance that the maximum of Vb coincides roughly with the cor-
C I contributes significantly to the total inductance in the responding zero of d 1 2 / d T ) . It can be seen that a value
circuit ($/ICI = ATI G 0.15). A first harmonic is also of R2 = 2 seems to be sufficient to damp critically the
present and is clearly less damped than the fundamental fundamental oscillation.
oscillation. The results obtained using a l = 0.02 (not
shown) are very similar, but the amplitudes of the first Case D: The Role of a Switch at the Output of the Line
harmonic are .larger and higher harmonics are also dis- The addition of a switch S2 in series with other load
cernible. impedances should not change the general picture that has
From the foregoing results we can draw the conclusion been drawn up to now. The closure of S2 has two main
that the main effect of diminishing the value of a lconsists effects on the circuit: on the one hand, it changes the open
in introducing in the oscillations larger components of circuit condition at the extreme b of the line to a closed
higher frequencies. For reasons of brevity, we will use in one; on the other hand, its closure speed a;' (discussed -
what follows al = 0.1. later) will act as a new excitation source. A complete de-

-
692 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. la, NO. 4. AUGUST 1990

scription of all possible situations exceeds the scope of

this work and only the results of two calculations will be
presented here, both with r I = 7r2 = 0.1 and T2 = 1. The
effect of changing T2 is trivial if Vb( T ) is nearly constant
‘i
around T2; however, if Vb( T ) strongly changes in this
time interval, the closure of S2 can introduce higher har-
monics than those expected from the value of r 2 ,because
of the dependence of V2(T ) on Vb (see (2)).A time-de-
pendent inductance C2(r) was set in network 2, chosen
to represent the inductance of the current sheath of a PF 4 , , , . , , , , , . .1 ,, , , , , , 1
device and was calculated self-consistently from a simple 0 1 2
T
snowplough model similar to the one described by Elt-
(a)
groth [7],coupled with the circuit equations. Anyhow, we
2
will restrict the presentation of our results (for reasons
that will be discussed in Section V) to the initial phase of ”b

the phenomenon ( T I2)and in this time interval, L2( T )

does not change significantly from its initial value L2(0 )
= 0.15. In one of the calculations we set RI = 10 and R2
= 0; while in the other one, R I = 0 and R2 = 1 was used.
Fig. 5(a) and (b) show, respectively, the obtained behav-
ior of Vb( T ) and dZ2/dT for these cases. It can be seen
that, before S2 starts to close, Vb(T ) evolves identically
as in the open-circuit cases shown in Fig. 2, and once it I / I I
0 1 2
is closed, the circuit behavior is similar to the one found T
for lines ended with LR elements. (b)
Fig. 5 . Effect of switching action of Sz on V,,( T ) and d l , / d T for L , =
IV. ANALYSIS OF THE RESULTS 0.15. (a) R , = 10, Rz = 0. (b) R , = 0,Rz = I .
In the previous section it was shown that a complex set
of oscillations in the current and voltage are predicted by This equation can be numerically solved for W iand W,.
the model. An interpretation of these oscillations can be The values of the fundamental frequencies and the first
made as follows. The equivalent dimensionless input two harmonics obtained for A = 9,T, = 0.017and RI =
impedance Zeqof a lossless transmission line connected
to a load of impedance Z, at its end can be written as
0 and 10 are WO = 22.46and WO = 21.95 j4.75;W 1 +
= 187.62and W ’ = 187.61 +jO.15;W 2 = 371.03 and
+
W 2 = 371.02 j0.04,respectively. As can be seen, a
good agreement with Fig. 2 is found. Besides, if the con-
dition W T , << is satisfied (“short” line), (10)can be
when acted upon by an excitation source with a dimen- solved analytically, yielding in this limit:
sionless frequency W ( j is the imaginary unit). As V, =
ZeqZI,an alternative equation for network 1 can be written
e E (1 - R:/4 + A/T/)”’
as z R1/2
Q = dZl/dT + RIZl+ Ze,Zl + VI and the condition A / T , +
1 > R:/4 must hold in order
to have oscillations. This approximate solution has an im-
where Q is the dimensionless charge of the capacitor. By mediate physical interpretation when written in dimen-
considering T large enough to make VI = 0, and looking sional terms:
for a solution of the type II oc exp ( j WT ), we arrive at a
+
complex algebraic equation for W E W, j W j :

0 = j ( W - 1/W)+ RI + Zeq. (9)

which is the frequency corresponding to the discharge of
For given values of T,, R I , Z,, and A , the real parts of the capacitor eoon the line capacity e,
through C I and
the complex roots of (9)give the natural frequencies of r l . Also, as A / T , >> 1 holds in most practical situations

imaginary parts (
damping constants.
e, (e,
oscillation of the system W:, * ), whereas the

W l , * ) give the corresponding

( A / T , = e0/e,), these results indicate that only rela-
tively large RIvalues can damp the open-circuit oscilla-
tion.
For the open-circuit case, Z, = 0 0 , (9)becomes In the short-circuit case, (9)becomes
0 = j ( W - 1/W)+ R I - j A cotan ( W T , ) . (10) 0 = j ( W - 1/W) + RI + j A tan (WT,).
BRUZZONE (’I c l / . : EFFECT OF ‘TRANSMISSION L I N E S O N ELECTRICAL. S I G N A L S 693

As before, an analytic solution for the first frequency is

obtained in the limit W T , << 1 in the form:

Wu = R , / 2 ( 1 + AT/).
Using the values of the parameters corresponding to the
calculations shown in Fig. 3, we obtain: W: G 0.93,
0.82, and 0.34 for the cases RI = 0, 1 , and 2. respec-
tively. The first two values are in good agreement with
the corresponding ones obtained from the numerical
Fig. 6. Oscillographic traces of voltage (upper) and current time deriva-
curves, whereas the third value cannot be matched into tive (lower) measured in plasma focus device operating with filling pres-
that graph because the circuit is very near to its critical sure o f O . 3 mbar. HoriLonlal division is equivalent to g 0 . 2 . in dimen-
damping for RI = 2. sionless units.

V . COMPARISON W I T H A N EXPERIMENT oscillations predicted by the calculations (mainly those

For comparison purposes, the voltage between elec- that we called fundamental) admit simple physical inter-
trodes and the time derivative of the current were mea- pretations in limiting cases (and are then predictable with-
sured in our PF device, operated at a relatively low filling out the need of complex calculations); however, the rest
pressure value ( 0 . 3 mbar of deuterium ) to have a sizable of the observed features are not amenable to easy inter-
time delay of the breakdown in the gun. The values of the pretations.
relevant parameters of the device are uo =
20 kV; e,, = We would like to emphasize that it is a common expe-
14 pF; C , 21 nH; zo =
0.35 R; t , = 9 ns ( 2 0 parallel rience for people working with fast-pulsed electrical cir-
1.8-m 7-R coaxial cables); and C 2 ( 0 )3 2.4 nH. In Fig. cuits (plasma-producing devices, trigger generators,
6 we show typical oscillograms of the voltage and the cur- pulsed lasers, etc.) to observe “strange” components in
rent time derivative, which correspond to uh and di,/dt the measured signals that are frequently discarded as
with the notation of the model. From the rise time of ZJ/>, “noise.” thus implying that these features are not ac-
we obtain T~ = 50 f 10 ns, then 7rl G 0.1. In the same tually present in the observed phenomenon but rather are
way, from the rise time of d i 2 / d t (or the corresponding spuriously introduced by the measuring system. Indeed,
fall time of t i b ) , we obtain T~ = 50 f 10 ns (7r. G 0. I ) . it is true that electromagnetic noise and other perturba-
On the other hand, the experimental time delay between tions (i.e., resonances of the measuring device) are mat-
the starting of both signals is t? = 540 & 20 ns ( T , 3 1 ). ters of concern in many experimental situations but, in
With this set of values for r l , 7 r 2 , and T z , a matching view of the results shown in Section 111, the interpretation
between the experimental results and the predictions of of measured signals should be made with caution.
the model can be attempted. By comparing the signals of The attempt to reproduce with this model the voltage
Fig. 6 with the theoretical curves of Fig. 5(a) or (b), good and current measurements in an actual plasma discharge
agreement is found in all the frequencies predicted by the as discussed in Section V is encouraging because it yields
model but not in the dampings. In fact, we have found good agreement with many of the observed features, but
that before the breakdown in the gun occurs ( 0 I T I it also makes evident the need for a deeper understanding
l ) , a value of R, 3 I O is needed to reproduce the damp- of the underlying phenomena (an investigation in this di-
ing of q,(see Fig. 5(a)), but once breakdown of the gas rection is already under way). The results presented in
takes place, this value of RI drastically damps the main this work enable us to predict that, if an electrical element
circuit oscillation, which is not observed. On the other in the circuit undergoes a relatively fast variation of its
hand, we cannot find a suitable constant value for R2 that magnitude (like the plasma inductance during pinch for-
allows us to reproduce in detail the behavior of the ex- mation), oscillations in currents and voltages should be
perimental signals after the closure of SI.We are then led excited in much the same way as during the closure of
to conclude that an adequate simulation of the initial stage switches.
of PF discharges (and, likely, of any similar discharge)
requires the introduction of damping elements of a more REFERENCES
complex nature than constant resistances. J . W . Mather. in Mcrhod.\ of E.rpcrirfierirrr/ Phy.\ic,.c, vol. 9B. R . H .
Lovberg and H . R . Griem. Eds. New York: Academic Press. 1971.
VI. FINALREMARKS ch. IS.
Y . Kondoh and K . Hirano. P/iy.\. Niricls. vol. 21. p. 1617, 1978.
The model presented in Section I1 for calculating the T. Oppenlandcr. G . Pross. G . Decker. and M . Trunk. Pltrsrfitr P I ~ J s . .
time evolution of currents and voltages in pulsed circuits vol. 19. p. 1075. 1977.
M. J . Rhee and R . F. Schneider, f E E E Trci1i.c. N u d . S u . , vol. NS-30.
yields complex behaviors for these magnitudes in the p. 319-1. 1983.
rather simple circuit configurations studied. Some of the R . F. Schneider. “An experimental study of the plasma focus device
 694 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 18, NO. 4, AUGUST 1990

as a charged particle accelerator,” Ph.D. dissertation, Univ. of Mary- Hector Kelly was born in Medoza, Argentina, on
land, College Park, 1985. February 14, 1948 He received the M.S. and
[6] T. P. Sorensen and V. M. Ristic, J . Appl. Phys., vol. 48, p. 114, Ph D. degrees in physics from Buenos Aires Uni-
1977. versity in 1972 and 1979, respectively.
[7] H. Bruzzone, H. Kelly, and C. Moreno, Amer. J . Phys., vol. 57, p. Since 1973 he has been a Researcher at the
63, 1989. Plasma Physics Laboratory of the Science Faculty
[8] B. Carnahan, H. A. Luther, and J. 0. Wilkes, in Applied Numerical of Buenos Aires University. In 1980 he joined the
Merhods. New York: Wiley, 1969, ch. 6. National Research Council of Sciences of Argen-
[9] P. G. Eltgroth, Phys. Fluids, vol. 25, p. 2408, 1982. tine. His current interests are in powerful electri-
cal discharges, MHD structure of current sheaths,
* and strong collisional shocks.

Horacio-Bruzzone was born in Bahia Blanca,

Buenos Aires, Argentina, on March 3, 1943. He
*
received the M.S. and Ph.D. degrees in physics
from Buenos Aires University in 1967 and 1976, Cesar Moreno was born in Buenos Aires, Argen-
respectively. tina, on July 1 1 , 1959. He received the M.S. de-
In 1969 he joined the Plasma Physics Labora- gree in plasma physics from Buenos Aires Uni-
tory of the Science Faculty of Buenos Aires Uni- versity in 1989.
versity. During 1979 he was at Frascati (Italy), Since 1989 he has been a Fellow of the Na-
involved in plasmas focus research with the I-MJ tional Research Council of Sciences of Argentina.
device. Since 1978 he has been a member of the His current interests are in powerful electrical dis-
National Research Council of Sciences of Argen- charges and computational physics.
tine. His current interests are in powerful electrical discharges, breakdown
gaseous phenomena, and MHD structure of current sheaths.

View publication stats