Sparkover in the rod-plane gap under composite slow front positive impulse/alternating voltages

N.L.Allen and C.F.Huang

Abstract: Problems are discussed whch relate to the measurement of the sparkover voltage of an airgap under a composite voltage comprising an impulse superimposed on an alternating voltage. The rod-plane gap has been taken as an example. Sparkover values have been measured with the impulse generator fired at angles 0", 90", 180" and 270" in the alternating cycle. Tests have also been made with firing at random angles. Sparkover voltages do not differ significantly from those for simple impulse alone, except for the case of random firing, where a small apparent increase occurs. It is shown that the 'up and down' method of measurement is satisfactory where specific firing angles are used, but is unsatisfactory and misleading where firing is at random angles.

1 Introduction

A practical power system is subject to occasional impulse overvoltages which may be superimposed on its workmg alternating voltage. Lightning impulse overvoltages will occur at random relative to the phase of the system voltage, and switching impulses also are unlikely to occur at readily predictable times. Normal engineering practice assumes that the dielectric strength can be based on the 50% spark- over level under impulse voltage, without regard to the pre- existing alternating voltage. Nevertheless, there is a dearth of data in the literature relating to the case where a switch- ing impulse voltage is superimposed on an alternating volt- age, and it is for this reason that the following study has been undertaken.

The sparkover characteristic of the rod-plane gap forms a reference against whch the sparkover of other electrode geometries are compared, and it has been used, therefore, as the first stage of a general investigation. The general question of the applicability of the 'up-and-down' proce- dure of IEC4&l (1989) [l] is also examined.

Udo [2] reported results using a gap in whch the impulse voltage was applied at the rod and the alternating voltage at the plane. The plane was isolated from earth by a sphere gap which was designed to hold off the full alternating volt- age, but which would break down in the event of sparkover in the test gap. Th~s arrangement permits the true superpo- sition of impulse on alternating voltage, but has the disad- vantage that the impulse voltage is divided capacitatively across the test and sphere gaps, so that exact measurement of the voltage applied to the test gap is dflicult.

An arrangement corresponding more closely to engineer- ing practice is to apply both voltages to the rod. This was

OIEE, 1999 IEE Proceedngs o n h e no. 19990024 DOL 10.1W9hpmt: 19990024 Paper fmt r&vd 12th June and in revised form 23rd October 1998 N.L. Allen is with the Manchester Centre for Electrical Energy, UMIST, Man- chester M60 lQD, UK C.F. Huang was with UMIST and is now with the Department of Electronic and Electrical Engineering, University College Dublin, Dublin 4, Ireland

achieved by Hepworth et al. [3] where the alternating volt- age source was isolated from the impulse generator by a sphere gap placed between the test gap and the 'front capacitance' of the pulse-shaping circuit of the generator. There are two disadvantages to ths arrangement, however: (i) The sphere gap must be set to withstand the alternating voltage but to conduct when the impulse generator func- tions. This imposes restrictions on the combination of volt- age amplitudes used, but it also means that there is a very sharp initial rise in the impulse voltage at the test gap when the sphere gap breaks down, so that the impulse cannot be regarded as 'standard'. (ii) The maximum test gap voltage is equal to the peak of the impulse voltage only. Thus, a true superposition is not acheved; the results obtained are substantially those char- acteristic of an impulse voltage preceded by an alternating voltage.

This paper describes an alternative arrangement in which a coupling (or isolating) capacitor replaces the spark gap for separation of the two voltage sources, permitting appli- cation of both voltages to the rod. Several variations on this theme have been described elsewhere [4]. The arrange- ment avoids restrictions on the relative voltages of the two sources, though a limit to voltages of opposite polarity is set by the voltage rating of the coupling capacitor. How- ever, the voltage applied to the gap in ths case is a true superposition of the two voltages.

2 High voltage circuit

The circuit used is shown in Fig. 1. The coupling capaci- tance Cp of 550pF was placed between the front resistor Rf and front capacitor Cr. Thus, the capacitor divider Cfi C2 measured the voltage at the test gap. A capacitor CA of 487pF was placed in series with the resonant circuit trans- former as a measure of protection against the impulse volt- age; the performance has been discussed more fully in [4]. The principal limitation of this circuit is that when the impulse voltage and the half-cycle of the alternating voltage are opposite in polarity the h i t of the voltage rating of the coupling capacitor may be reached, restricting the length of the rod-plane gap to 0.5m, in this case.

IEE Proc.-Scr. Meas. Technol.. Vol. 146, No. 3, Muy 1999 142

(1 )

0.87kn

test gap

impulse generator

- - -

P

>

Y t

>

h time Fig. 4 RMS,atpomt (1) in Fig. I Chargrng voltage of generator = 300kV; zero voltage is indicated; time: 200pddiv

Composite voltage oscillogm for fwmg mgle 90" with Vac = NX)kV

Y

3

h time Fig.2 age qplied Charging voltage of generator = 300kV; zero voltage is indicated; time: 1001."/div

Composite voltage oscillogrm for srinple unpuke, no dtemating volt- d time

9 . 5 kVRMSutpoint (1) in Fig. 1 Charging voltage of generator = 300kV; zero voltage is indicated, time: 2OOpddiv

Cbmposite voltage oscillogram for fuing mgle 180" with VAC =

Y

>

A time

> ai P 9 -

a

I A time

Fig-3 &VIS at point ( I ) m fig. I Charging voltage of generator = 300kV; zero voltage is indicated; time: 200Hdiv

COmposite v o k e oscilbrnjorfiring Cmgre 0" with VAC = m k v Fi .6 2dkVRMSatpoint ( I ) mFig. I Charging voltage of generator = 300kV; zero voltage is indicated; time: 200pddiv

Cowposite voltuge oscillogram for frmg mgk 270" with VAC =

The pre-stress voltage modified the impulse waveshape. Impulse waveforms at the phase angles used in ths work, namely 0", 90°, 180" and 270", are shown in Figs. 3 4 . For comparison the waveform without AC is shown in Fig. 2. The superposition of the alternating voltage altered the time to peak as shown in Table 1.

These variations meant that it was not possible to main- tain a constant time to peak impulse voltage while varying the critical parameters of voltage amplitude and firing angle. It may be noted, however, that the critical time to peak for a lm rod-plane gap is - loops [5], so that for

IEE Proc -Sei Meus Technol, Yo1 146, No 3, May 1999

Table 1: Time to peak for various impulse waveforms

Alternating voltage T, p

None 250 Firing at 0" 266 90" 250

180" 220

270" 250

Charging voltage of generator = 300kV; VAC = 200kV RMS at point (1)

143

times to peak > 200p.3, as used here, breakdown would be expected before the peak is reached; this is borne out by the results described later. It must be expected that the varia- tions in rates of rise of impulse voltage, implicit in the vari- able time to crest, will contribute to the dispersions in sparkover voltage and times to breakdown, but in thls case the effects would not be large. There is also a large varia- tion in the impulse times to half value, depending strongly on firing angle, but this is unlikely to have affected the results.

Measurements were also made of sparkover voltages when the impulse generator was fired at random with respect to the phase of the alternating voltage. The 'up-and- down' method of measuring sparkover voltage was used in all cases, employmg at least 40 shots for each point. The impulse voltage was measured with a peak voltmeter which was calibrated against

divider, and the required phase angle for firing was checked oscilloscopically at the impulse generator potential divider.

The rod-plane gap was of 0.5m length. The rod diameter was 2 0 m , hemispherically ended, and the plane was an aluminium sheet 1.22m x 1.44mm. Results were corrected to standard atmospheric density but not for humidity vari- ation.

320

3101

> Y

3-

280

270 / I

230

220' I 0 50 100 150 200 250 300

VncW Fig.7 alternating voltage for four fumg angh For random firing, voltages are simple impulse peak voltages only -*- random 4- 0" -A- 90" -X- 180"

?- ::breakdown Error bars indicate f 30

50% peak composite pkover voltage m a jimctbn of pre-stress

3 Results

Composite sparkover voltages U were measured as a func- tion of pre-stress peak alternating voltage; results are shown in Fig. 7. For all four designated phase angles, there was a small decrease in sparkover voltage for alternating voltages up to lOOkV peak, with a small increase up to the maximum of 190kV used. Where random firing was under- taken, a composite sparkover voltage could not be quoted. The results given for this case in Fig. 7 are thus for the impulse voltage only.

Fig. 8 shows plots of the simple impulse sparkover volt- age U, only, against alternating pre-stress voltage. Values of three times the standard deviation about the mean, omit- ted for clarity in Fig. 7, are shown here for the random, 90" and 270" firing angles. The plots for the 0" and 180" firing angles and for random firing are unchanged from those of Fig. 7. Those for the 90" and 180" angles show an almost

144

linear variation, with respectively negative and positive slopes; in the former case, extrapolation leads, approxi- mately, to the simple alternating sparkover voltage.

Fig. 9 shows the mean times to sparkover under the four firing angles and for random firing. The scatter was large and no sigmfkant variation with firing angle, or change from the case V , , = 0, can be discerned, though there is an apparent reduction, increasing with alternating voltage, where random firing was employed.

Under simple alternating voltage, sparkover always occurred at 90°, i.e. at the peak of the positive halfcycle.

550 - 500--

400-

loo--

VAC, k v Fig.8 5&6 simple m p h e volitzge re@edfor kover against pre- stress alternathg voltages v,, forfarafvbtg mgh a m i j Y r d r n f & g A = sparkover voltage of gap under simple impulse. B = sparkover voltage of ga under alternating voltage. C = impulse voltage required to break down gap at p$: negative alternating voltage -*- random 4- 0" -A- 90" -x- 1809

?- ::breakdown Error bars indicate f 3 0 Dashed lines show ideal linear relationships U, + VAC (peak) = U

I 0 50 100 150 200 250 300

VAC, kv Fip.9 50A CO r d r n z g 3- random 4- 0" -A- 90" -x- 1800

&r bars lndicate f 10

Mean tbne to sparlcow, m a*tim of VAC and f e g mg&, for ate sparkover voltugo and for simpk @e voltago m the use of

2709

4 Discussion

4.1 impulses The result shown in Fig. 7, that the composite sparkover voltage showed only small changes with VAC, regardless of firing angle, is in contrast to the results obtained with a direct voltage pre-stress condition, where si&icant effects on sparkover voltage were obtained [q. However, the result is in general agreement with that of Hepworth et al. [3],

Sparkovers under point on wave firing of

IEE Proc.-Sci. Meas. Technol., Vol. 146, No. 3. May I999

who were not using superposition of voltages. In both cases, corona discharges were set up at the rod, at peak voltages 2 lOOkV, prior to impulse application; they are evidently less important in affecting sparkover in the alter- nating voltage case.

The plot of Fig. 8 shows that with firing at 90" the extrapolated line cuts the VAC axis at approximately the simple alternating sparkover voltage; at all points below the line the gap is safe against sparkover and the line represents the minimum sparkover voltage of the gap under all condi- tions of firing angle. As the impulse voltage is raised, at a given alternating voltage, the probability of sparkover increases, over all angles, from zero to 100Yo at the line obtained for the 270" firing angle, representing the maxi- mum impulse voltage required to break down the gap.

The impulse sparkover voltages U. at 0" and 180" where, of course, the instantaneous alternating voltage is zero, dif- fer insignificantly from the case of sparkover under simple impulse voltage alone. However, the curve shown in Fig. 8 obtained with random firing is not significantly different from those at 0" and 180".

4.2 Results with random firing The sparkover voltages under simple impulse and under simple alternating voltages differ, for the impulse waveform used here, by < 15kV. Also, under simple alternating volt- age, the threshold condition for sparkover always occurred at the peak of the positive half-cycle, i.e. at 90".

3 . 1 0 Altematmg voltage waveform showmg ranges ofjrmg mgk. €4, 8, 8,. e,, withm which critical Lomposite voltages OLLW m rrmdom $rwg of

m@e

The IEC procedure Class 2: Up-and-down tests [l] for sparkover level requires that trial voltage levels U , at around the anticipated sparkover level, shall be increased by an amount AU until a disruptive discharge occurs, upon whch the voltage is reduced by an equal amount in steps until a withstand is obtained. In the present experiments, with random firing of the impulse generator, the upper limit of composite voltage using this procedure must repre- sent the application of positive impulses just before, or just after, the 0" or 180" angular positions. The notional range within which these voltages were required for sparkover is shown in Fig. IO by the zones e,, & and &, e,, for when firing occurred between & and e,, sparkover must always occur and a reduction of voltage of at least one step, AU, would then be made. Since the probability of occurrence of the impulse within the positive half-cycle was 0.5, it is inherent in the procedure, as applied in this case, that an upper limit to U, would be sharply defined since the deter- mination of the upper lurnt itself implies a 50% chance of sparkover at the two levels separated by a voltage AU. It follows that the possibility of applying higher voltages in the negative half-cycle is negated by the up-and-down method, which is therefore not applicable to this situation; the true spread in sparkover voltages is not apparent.

Confirmation that the sparkover values under random firing are characteristic of the zones of 8 defined in Fig. 9 is derived from Fig. 8, where it is seen that the curve of S i - ple impulse voltage U. required with random firing lies slightly above those for the 0" and 180" point-on-wave positions. This reflects the fact that, in the random case, the angles 8, and 84 lie in the negative half-cycles, thus requir- ing a slightly larger impulse for composite voltage sparko- ver. Sparkover at other angles in the negative half-cycles, that is at < 8, and > e,, would require higher random impulse voltages over a decreasing angular range; the prob- ability of ths occumng is obviously less than that of spark- over in the positive half-cycle at a smaller voltage.

The results thus show that use of the up-and-down method for the practical case where a switchmg impulse is superimposed at random timing upon an alternating volt- age yields a mean value of composite sparkover voltage which is close to that given by the point-on-wave results, but gives a totally misleading impression of the spread in impulse voltages needed for sparkover. Indeed, the limits of real spread in the impulse voltage required for composite voltage sparkover are given by the curves for 90" and 270" firing angles. Thus, the IEC Class 2 procedure is not appli- cable to the practical case in power systems, where switch- ing impulses are unlikely to occur at readily predicted angles. The voltage level method should be used as an alternative, but this would be a very time-consuming proce- dure for a composite impulse/alternating voltage. However, this work has shown that a better estimate of the sparkover voltages to be expected with random timing is obtained by use of the up-and-down method with point-on-wave firing at the extremes of polarity.

4.3 Times to sparkover Fig. 9 shows that the mean time to sparkover, under ran- dom firing of the impulses, decreases with increasing VAC Ths is in contrast to the mean times under all point-on- wave firing, where there is no significant change, within a large scatter, with VAC.

It has been noted that, under random firing, and using the 'up-and-down' technique, almost all recorded sparko- vers will occur in the positive half-cycle of the alternations. A large proportion recorded occur between the angles 82 and & (Fig. lo), but the impulse voltage levels have been set (Section 4.2) at values characteristic of the zones 81 to 82 and 6, to 8,. By contrast with sparkover voltages, for example, at the 90" angle, the actual composite voltages occurring between & and & in the up-and-down procedure would be larger. Thus, a large proportion of the recorded sparkovers, under random firing, occur under overvoltage conditions. Therefore, times to sparkover tend to be less than at specific firing angles.

4.4 Effects of corona It has been shown in the case of direct voltage activation of the rod that a negative corona markedly reduces the spark- over voltage under a positive switching impulse and that a positive corona tends to suppress the positive leader, during the impulse with a resulting increase in sparkover voltage [6]. These effects do not occur in the present case with RMS values of alternating voltage up to 200kV. It must be assumed, therefore, that residual ionisation remaining after corona at each voltage peak sufficiently neutralises the charges in the succeeding corona to remove effects due either to space charge fields or to ionisation of either polar- ity.

TEE Proc -Scr Meas Technol, Vol 146, No 3, May 1999 I45

5 Conclusions

The sparkover voltage of a rod-plane gap, under a com- posite voltage comprising a switching impulse superim- posed on an alternating voltage, is not significantly different from that for a switching impulse alone. The mag- nitude of the switchng impulse voltage which can be with- stood before breakdown occurs varies between 0 and the peak alternating voltage applied, depending on the angle of firing of the impulse within the cycle.

The IEC ‘up-and-down’ method of measuring sparkover voltages is satisfactory when the impulse is initiated at spe- cific firing angles by a ‘point-on-wave’ technique. It is not satisfactory where the impulse occurs at random, where it is capable only of yielding a mean sparkover voltage close to that obtained by firing at 0” and 180”.

The ‘up-and-down’ method therefore cannot be employed where initiation of the impulse is at random. The true spread of composite sparkover voltages in this case would be revealed only by lengthy voltage level measure- ments. The ‘up-and-down’ method defines only the lower

and upper limits of the dielectric strength by using the point on wave technique at 90” and 270”, respectively.

6 Acknowledgment

This work has been supported by a research grant from the UK Engineering and Physical Sciences Research Council.

7 References

1 IEC 60-1, 1989 2 UDO, T.: ‘Switching surge sparkover characteristics of air gaps and

insulator strings under practical conditions’, IEEE Trans., 1966, PAS-

HEPWORTH, J.K., KLEWE, R.C., LOBLEY, E.H., and TOZER, B.A.: ‘The effect of AC bias fields on the impulse strength of point- plane and sphere-plane gaps’, ZEEE Trans., 1973, PAS-92, pp. 1893- 1903

4 HUANG, C.F., ALLEN, N.L., and GREAVES, D.A.: ‘High voltage circuits for application of composite voltages to test gaps’, IEE Proc., Sci. Meas. Technol., 1999, 146, (2), pp. 6469

5 WATERS, R.T.: ‘Electrical breakdown of Gases’, in MEEK, J.M., and CRAGGS, J.D. (Eds.): ‘Spark breakdown in non-uniform fields’ (Wiley, London, 1978), Chap. 5

6 ALLEN, N.L., HUANG, C.F., CORNICK, K.J., and GREAVES, D.A.: ‘Sparkover in the rod-plane gap under combined direct and impulse voltages’, ZEE Proc., Sci. Meas, Technol., 1998, 145, pp. 207- 214

85, pp. 859-864 3

146 IEE Proc.-Sci. Meas. Technol., Vol. 146, No. 3, May 1999