K. Feser
E I-29
Pour la production de hautes tensions de choc de manoeuvres dans un laboratoire dessais, il faut tenir compte de la rigidite dielectrique de lair. Le probleme de cette rigidite est
examine a laide de quelques exemples. On montre que, pour
des tensions de choc de manoeuvres, les d&charges preliminaires luminescentes sous forme de Leader sont determinantes
pour le dimensionnement des electrodes, ainsi que pour les
distances entre parties sous tension et les objets mis a la terre.
Quelques exemples pratiques completent les considerations
theoriques. On examine ensuite le comportement a lirmorcage des generateurs de chocs en couplage multiplicateur de
Marx, dans ie cas de resistances intemes serie, de valeurs elev&es. Des ameliorations possibles dans la conception du couplage multiplicateur de Marx sont indiquees, pour pouvoir produire aussi bien des tensions de choc simulant la foudre, que
des tensions de choc de manoeuvres conformes aux normes.
L exposese termine par quelques considerations sur le dimensionnement des composants et sur la mesure des tensions de
choc de manoeuvres.
The strength of the air must be considered in the generation of high switching impulse voltages in test plants. The problem with the strength of the air shall be discussed on the basis
of several examples. It will be shown that preliminary leader
discharges which occur with switching impulse voltages are of
primary significance for dimensioning electrodes and clearances of live parts to earthed objects. Some practical examples
will supplement the theoretical deliberations. Then, the trigger
behaviour of impulse generators in the Marx multiplier circuit with high-impedance internal series resistors will be examined. A discussion will pertain to the possible improvements
of the previous concept of the Marx multiplier circuit for the
standardized generation of ligthning impulse voltages as well
as of switching impulse voltages. A few aspects of dimensioning some elements and of measuring techniques for switching impulse voltages will form the epilog to this paper.
1. Introduction
Ts krit = ~
- 6oo
2.08
r-
WSJ
(U in kV),
Fig. 1
kV
3000
kV
2000
I
a-
positive switching
impulse voltage
1000
Impulse shapes)
-+
500
1 0 0 0 JJS
1500
T5
Fig. 2
Fig. 3
and 1.2 ys. With a positive polarity, the lightning impulse voltages correspond to the linear dependence with a slope of
about 5 kV/cm (or 500 kV/m) which has been known for
some time. This applies also to the highest voltages ever mea.
sured (currently in the range of about 5 MV).
For gaps of up to about 7 m, corresponding to a swit,
thing impulse voltage of 1500 kV, the switching impulse voltage 250/2500 proposed in the international specifications
provides the lowest flashover voltage. For larger gaps, we obtain lower flashover values with longer front times. This is
illustrated by the dotted curve in figure 3. With a voltage of
3 Mill. V, for example, a front time of 1.2 ~.ls will yield a
flashover of only about 6 m, a front time of 250 /.LS already
provides 22 m and a front time of about 1200 ps will produ,
ce a 35 m flashover. Figure 3 will also help answer the question: Why switching impulse tests? .
Since the amplitudes of the switching impulse voltages
occurring in the network are proportional to the network voltage, the increase of the transmission line-voltage to e.g.
1500 kV will bring up problems pertaining to switching impulse voltage strength. The questions concerning test procedures of electrical apparatus with switching impulse voltages
must therefore also be considered as this is the case at present in several proposals. In the same light, it should be determined to which degree it may be possible to reduce previous
test voltages or possibly replace them [3].
In the following sections, we shall dwell specifically on
problems concerning the generation of high switching impulse
voltages, and the prerequisite according to which the test
plant must generate all existing test voltages as specified in
standards shall be observed. For completeness sake, figure 4
shows two test voltages with the associated data. The lightening impulse voltage 1.2/50 (front time 1.2 ps + 0.35 PS,
time of half value 50 ps k 10 ps) was generated to simulate
% 1
JJ,
b
_j
50
----c, h
Fig. 4
ps
while the leader pre-discharge features a highly luminous predischarge channel which pushes the potential of the electrode
with a low voltage drop into the field space. Further pre-discharge phenomena are produced at the head of the leader predischarge. We shall discuss the characteristic differences of
both types of pre-discharge later on.
Based on the available test results on current, voltage, and
charge behaviour in the gas discharge gap and on the visual
image of the gas discharge, it is possible to establish a hypothesis concerning the fundamental development of flashovers
for lightning and switching impulse voltages.
The following sections will only mention the more essential exterior differences in discharge behaviour which are of
special signifiance for practical applications. The individual
physical differences are still not fully understood today.
An impulse-type of pre-discharge - the streamer pre-discharge - occurs with lightning impulse voltages after the nominal voltage is exceeded. This phenomenon makes the electrical field around the highly stressed electrode so homogenous that an impulse-type of pre-discharge cannot occur until
a very much higher voltage level is attained. If the streamer
pre-discharge reaches the opposite-electrode, a leader channel
is heated up from the positive electrode side. It is not necessary for the voltage to increase any further to heat up this leader
channel.
In contrast, continously occurring streamer pre-discharges
will be noted with switching impulse voltages as the voltage
level rises above the rated value. Due to the continous generation of current pulses, a leader will form at the anode as the
voltage level rises, even before the streamer pre-discharges
have reached the opposite electrode. Through further streamer pre-discharges, the leader can grow in steps with constant mean speed [19], whereby the streamer pre-discharges
are generated at the leader head. The leader pre-discharge
cannot grow unless the voltage in this case is increased. The
flashover occurs when - as with lightning impulse voltages the streamer pre-discharge has reached either the opposite
electrode or the opposite discharge [7].
60
I
i
IL
.
I;~b-f--lI
A
.
I
dFig. 5
~-----L
rod/plane ---/
----
d=2;
kV
double
- 10m
-L d =0,5
10 m
---\ I-+\ 1 --
-. , .-_.__. L$
\
,-----\ \ I ,/
I
I
-7
r5m -i
O,?m
iii
f
\
I---I ---in-i
;
I
I
22,! 5 m
Fig. 8
01
Fig. 7
m
I
6
I
0
I
IO
12
14 m m 16
the electrode surface would have to be reduced to below onset field strength by correspondingly shaping the electrode.
Figure 9 illustrates another example for optimal electrode
dimensioning for a voltage divider. In this case, the maximum
field strength with respect to 1 MV was calculated by a computer for various toroid electrodes and the results were com-
21412
230
II
230
350
80
80
160
230
loo
230
60
80
80
140
170
190
-
piled in table 1. We can observe that the field strength configuration at the electrode surface can only be optimtied with
two toroids of different sizes, e.g. 80/60 or lOO/SO. The onset
field strength in air depends on the electrode radius and very
substantially on the voltage shape. For switching impulse voltages and large electrode radii, we can expect a onset field
shape of about 25 kV/cm, so that the 80/60 double toroid
would be sufficient for a switching impulse voltage of approximately 2.2 MV, whereas a lOO/SO toroid would handle a
switching impulse voltage of 3.1 MV. In the meantime, these
values have been verified by measurements. Some examples of
test laboratories in which these ideas have been partially implemented should clarify the previous considerations.
Figure 10 shows a research laboratory which was specially
dimensioned for generating high switching impulse voltages.
For transport reasons, the very large electrodes needed for
this purpose can no longer be built as single units, and the
plateelectrode shown in figure 10, the so-called polycon electrode, has proven to be reliable for very high voltages. Apart
from the transportation costs, these polycon electrodes also
feature advantageous manufacturing costs (forming costs). A
damped capacitive 5.4 MV voltage divider suspended from
the ceiling is visible in the foreground. Its electrode is 4 m
sistors and in a stationary part which is given by the dimension of the parallel series resistors. With a high internal
series resistor, we can observe that the overvoltage is given almost exclusively by the stationary part of the overvoltage
which is determined by the size of the series and parallel resisters. The oscillogmms in fig. 12b clarify how the overvoltages at the second spark gap also become smaller with decreasing parallel resistance, whereby both the transient and
stationary parts are reduced. The overvoltages are also dependent of the number of stages, whereby the overvoltages at
the second spark gap become smaller as the number of stages
increases [lo]. Small parallel resistances are used with high
stage capacities for the generation of lightning impulse voltages, because the product of parallel resistance Rp times ima
Fig. 12
Natural overvoltages
time
10
5 = =
I
I
-c
--i-k
I
Fig. 13
spark erosion at the firing pin of all spark gaps calls for additional maintenance. The trigger range obtainable with this
triggering system basically depends on the dimensions of the
circuit components, whereby a smaller triggering range will have to be anticipated mainly in the parallel circuits of stages.
The second possibility according to figure 14b is a multiple
electrode spark gap designed with a triggering pulse of about
150 kV and trigger generators (TG) in each stage.
To avoid the need for additional distance or pressure adjustments, an installation was designed with a multiple spark
gap and 150 kV trigger generators in each stage which are initiated from the previous stage with 50 ns delay per stage [ 161.
This solution provides a trigger range of 15 to 100% of the
charging voltage without spark gap adjustment. Thereby, reliable firing of all spark gap is independent of the dimensions
of the parallel and series resistances, i.e. of circuit elements.
These advantages of triggering with multiple spark gaps are
contrasted by the following disadvantages:
a) In view of substantial arc cooling due to the many electrodes, the spark gaps extinguish in the tail of the
switching impulse voltages when such voltages are generated.
b) The arc voltage drop is much larger with multiple spark
gaps (e.g. obtained with 15 partial spark gaps) than
with a single spark gap. Also, this arc voltage drop depends on the breakdown voltage level, i.e. on the charging voltage. This leads to the fact that the efficiency
of the impulse generator wilI become dependent of the
charging voltage and also to a marked non-linearity in
impulse generation through the switching system [ 161.
c) Due to low partial spark gap distances (approx. 5 mm)
the service life of the spark gaps is reduced by the arc
points of the electrodes (spark erosion), since electrode
burn-off is dependent on electrode distances and has
Fig. 14
Basic schematic
u three electrode spark gap
b multiple-electrode spark gap with trigger generator TG
C lesser spark gap
11
sistance for the lightning impulse voltage [23] when a corresponding voltage drop - corresponding to the breakdown
voltage of the parallel spark gap - has been obtained over the
switching voltage resistor RP 1. Apart from good trigger behaviour, this circuit also provides the possibility of automatically switching over the tail time from e.g. 50 /.B to e.g. 2500
~(1s within a very short period of time by simply adjusting the
parallel spark gap. When generating lightning impulse voltages,
the parallel spark gap (sphere gap) is automatically adjusted
to half the distance of the main spark gap.
Fig. 15
&
c
P2
Further
designations
see
text.
Rs
CI
Pl
P2
D2
C5 e
Pl
P2
.6a
C =
b
C5 e
R
RPl
12
P2
R
R
PZ
R
R
pz
Cb
1
F
A7
/X
Fig. 16
r-
A variation of this circuit is the series circuit of both parallel resistors. Advantages mainly of economical nature are obtained by designing this parallel spark gap as a rod gap with
non-adjustable distance. In this case, the breakdown delay of
a spark gap with an inhomogeneous field with steep voltage
characteristics is utilized. Example c shows another interesting possibility in which the stationary overvoltage (instead
of the transient overvoltage) in the second spark gap becomes
equal to charging voltage UL irrespective of the number of
stages n. The disadvantage of this circuit lies in the difficult
parallel circuit.
All three possibilities have already been implemented.
However, a mutual feature of circuits b and c is the fact that
the overvoltages are damped too much with high internal series resistances and that the trigger range will become limited
with high per stage series resistance values. This can be avoided in multi-stage impulse generators in that the first 3 to 5
stages are equipped with partially low-impedance series resistors, so that the transient overvoltages in the first stages lead
to firing: from the fourth stage on, the high stationary value of
the overvoltages in a total of e.g. 3 stages will lead to the firing of the following spark gaps. For resistance values of 1 to
5 kOhm per stage, this solution is practical and simultaneously economical. The number of stages which must actually be
equipped with low-impedance series resistors depends basically on the series resistance values and the number of stages.
In an 8-stage impulse generator with 10 kJ per 200 kV stage,
it is sufficient to choose a low-impedance series resistor in
the second stage for, say, internal series resistances of 5 kOhm
per stage, in order to obtain a trigger range of more than 20%
for the series connection of all stages.
For very high-impedance series resistances of several 10
kOhm (for example to generate front times of 1000 ps), it is
necessary to use the circuit illustrated in fig. 16. In this case,
the high-impedance series resistor is shunted with a firing capacitor Cz as the impulse generator fires, thus increasing the
transient, high-frequency overvoltages in the following spark
kn
\
300200-
0.2
Fig. 11
6 nF
10
Series resistance Rs plotted against load capacity Cb for the generation of a standardized
front time for switching impulse voltages
(example: 2Gstage 4 MV impulse voltage
generator with a per-stage-energy of 20 kJ)
13
6kn
543
-______
- - - - - - - - - - - - -
T
c200 IlOOO)Js
1
-I'
0,2
\
30011000jls
0,s
1
c-
34nF
10
Fig. 18
gene-
rator with a
%I
IO04
c 30
se
R =lO.R undR
m
, ,,,
4 5 nF
IO
cb
I
I
Fig. 19
c-lo---I
;P WRp
C;
RP
charging resistor
c,
external series resistor Rsi
impulse capacitor
4
parallel resistor
firing capacitor
internal series resistor
charging voltage
Literature
H. C. Barnes and D. E. Winters: UHV transmission design requirements - switching surge flashover characteristiw of extra long air gaps.
Tram, IEEE PAS 90(1971)4, p. 1579...1589.
[?] Phtnombnes damorgages dam Iair des intervalles pointe-plan de 5 m
et 10 m soumis & de5 chocs de manceuvre de polarit positive. Rapport
CigrC no 33...15, 1972.
[3] fi;l;eeser: Gedanken zur StoBspannungserzeugung
und Priiftechnik fiir
Entwcklung
YOII U H V - S y s t e m e n . B u l l . S E V 64(1973)15,
s. 911...917.
[4] Parafoudres. Premi&re partie: Parafoudres ?+ rCsistance variable pour
rCseaux a courant altematif. Publication de la CEI 99-l. Deuxi&me
Cdition. 1970.
[5] Essais ?I haute tension. Publication de la CEI 60. Deuxibme Cdition.
1962.
[6] E. Lemkr: Der Durchschlagmechanismus van Luftfunkenstrecken be!
Schaltspannungen. Wiss. Zeitschr. der Techn. Unlversitlt
Dresden
17(1968), S. 105...115.
[7] K. Feser: Mechanismus zur ErklPrung des Schaltspannungsphlnomen~.
STZ 46(1971)46, S. 937v.946.
[8] M. F. Simon and G. L. Leroy: Contribution to a better understanding
of impulse voltage measuring systems. Trans. IEEE PAS 91(197?!?,
p. 478...484.
[9] D. Kind: Einfiihrung in die Hochspannungs-Versuchstechnik. UniText. Braunschweig, Vieweg-Verlag, 1972.
[lo] A. Rodewald: Ausgleichsvorg~nge in der Marxschen Vervielfachungsschaltung nach der Ziindung der ersten Schaltfunkenstrecke. Bull. SEV
60(1969)2, S. 37...44.
[Ill A. Rodewald; Die Ziindwahrscheinlichkeit der Schaltfunkenstrecken in
B u l l . S E V 60(1969)18,
d e r M a r x s c h e n Vervielfachungsschaltung.
s.
175 Rfil
-. _._...
__.
[12] F. W. Heilbronner: Firing and voltage shape of multistage impulse
generator% Tram IEEE PAS 90(1971)5, p. 2233...2238.
[13] G . Carrara e L . Dellera: Sullinnesco d e i g e n e r a t o r i d i impulsi d i
tensione. Energia Elettrica 42(1965)1, p. 15...20.
[14] T. E. Broadbent: New high-voltage multistage impulse generator circuit. J. Scient. Instruments 37(1960)7, p. 231...236.
[15] M. J. Bishop und R. Feinberg: Grundsltzliche Verhesserung des Hochspannungs-Stossgenerators.
Anwendung des Polytrigatrons als Schaltg&t. E und M 88(1971)2, S. 62...67.
[16] M. J. Bishop and M. F. Simon: The impulse generators at les Renard&es. Trans. IEEE PAS 91(1972)6. D. 2366...2376.
1171 K . Rongs: N e u e Grosstra&f&a&n-Montagehalle
d e r Transformatoren Union. ETZA 94(1973)2, S. 125...127.
[18] K. Feser: Erweiterung des Auslijsebereiches
van vielstufigen Stossgeneratoren fiir die Erzeugung ~0x1 Schaltspannungen. ETZ-A 94(1973)3,
s. 171...174.
[19] B. E. Giinger and E. G. Maim: Sudies of spark formation at high
switching voltage of positive polarity. Trans. IEEE PAS 91(1972)6,
p. 2427...2436.
1201 K. H. Schmitler: Funkenstrecken mit optischer Triggerung. Laser
2(1970)3, S. 9...10.
[21] W. Lampe: Triggerung vcm Stossgeneratoren
mit weitem Arbeitsbereich. ETZ-A 83(1962)18, S. 591...596.
[22] A. Rodewald: Zur prazisen AuslBsung van Stossgeneratoren bei einer
vorgew%hlten Ladespannung. Bull. SEV 59(1968)20, S. 947...952.
[231 A. Rodewald: Ober eine Marxsche Vervielfachungsschaltung
mit Hilfsfunkenstrecken
zur V e r b r e i t e r u n g d e s AuslGsebereiches. ETZA
92(1971)1, S. 56...57.
1241 K. Feser: Ein neuer Spannungsteiler fiir die Messung hoher Stoss- und
Wechselspannungen. Bull. SEV 62(1971)19, S. 929...935.
[l]
Fig. 20
Ir
0, c2
1.8 ,uF)
7. Conclusions
When generating switching impulse voltages in test plants,
the strength of the air must be incorporated in the design concept of the test plants. For optimum utilization of available
space, it must be observed that the shape of the electrode does
not have an influence on the breakdown voltage after the leader pre-discharge has occurred. A technical solution in the
UHV range must be to prevent leader pre-discharges by correspondingly dimensioning the electrodes. To use flashover
distances as an additional dimensioning criterion is only purposeful for cost reasons where lower switching impulse voltages (up to about 2 MV) are involved, especially if load capacities are high. Special circuits have been developed for
triggering Marx impulse generators for the generation of
switching impulse voltages and also for the generation of
lightning impulse voltages, so that it is possible today, to generate unipolar impulses in the range of ps to ms with one
and the same test installation. Apart from generating the various voltages according to valid standards, the flexibility of
a test installation is one of the most important tasks of an
economical impulse generating system.
Authors address :
Dr. K. Feser. Engineer, Emil Haefely Rr Cie AG, Postfach. CH-4028 Basel/Switzerland
15