24R

ELECTRICAL BREAKDOWNIN TRANSFORMER OILIN LARGE GAPS

K. L. Stricklctt, D. M. Weidenheimer,t N. R. Pereira,t

and D. C. .Iud})

National Inst.it.ut.e of St.andards and TechnologyDept. of Commerce, Technology Administrat.ion

Gait.hershurg r-.m, 20899

tllerkeley Research Associat.es, Inc., Springfield VA, 22150

tHany Diamond Lahorat.ories, Adelphi MD, 20783-1193

ABSTRACT

The Aurora acc<'lerator uscs four parallel DlulIIlpiu puls('-forlllin~ lin('s toprovide au iut.('use flash x-ray pnlse. Prop('r tilllinl!; of tll(' pulse's I!;('npr-ated by ('a(.h lllumiein is important to the quality of the radiation. Thepulse on each Dlumlein, and the synchroni7.ation between the Dlullli<'insare affected by the c\osnre of hi!!;h-volta!!;t' tri~~('rpd oil switch('S in eachliue. The triMercd oil switch utilize's a uniform fipld p;eolllet.ry with agap spacing between 40 aud 50 CIII, a uniqne ('nvironlll('ut for ohs('["\.a-tion of an: developmcut. in t.ransfofllH'r oil. Hil!;h-spped photol!;raph~' ofswit.ch dosure shows tilllinp; to he infllH'ncpd h~. tll<' initiation and spatialdev<'loplllent of prebreakdowu stn'a 1II('rs.

INTRODUCTION

The Aurora accderator [11is a flash !!;amllla-ray sinllllat.or opendedby t.he Defense Nuclear Agency. The devin' uses four I31uml('iupulse-forming lines t.hat. may be fired separately or in eOlwert. togenerate a hurst. of hard x-ray radiat.iou. Each liue eousist.s of au oil-insulat.('d lllumlein coupl<'d t.o a.maguet ieally iusulat.ed t.ransmissionline that. is terminated in all elcdwu-healll diode. The llll\lnleins

are charg('d by an 8 M.1 l\-larx generator, awl high-voltage pulses a[('launched by closure of triggered oil switches iu each line'. fkeeutimprovements in the risctime of t.he radiation pulse generat.ed hya single lliumiein [2, 31 place new const.raint.s on the an:ept.abletiming jitter during parallel operation and mot.ivate [('-examinationof design and operating paramet('rs.

The 12 MV oil switch is one element critical to synchronizing theDlumleins. This report describes the normal operation of t.he oil

,

EAKDO\VNI IN LARGE GAPS

heimer,t N. R. Pereira,f

ldy,f

ards and Technology.1logy Administration:D, 20899

nc" Springfield VA, 22150

Adclphi MD, 20783-1193

CT

2-t9

switch, identifies the sources of variability for closure of the switch,and suggests strategies for reducing switch jitter. Other aspects ofsynchronization include: reduction of the jitter in the gas switchesand equalizing the pulse charging voltage on each Blumlein [4, 5].

FROMMARXES

l Dlumleill pulsc-fol"lnillf!; lill('s tooper timin~ of the pulses gmt('r-he quality of the radiation. The.:tization I)('tw('('n tll(' mUIIlIl'ins

0:,('triggpfI'd oil switdl('s iu (';1<"1t, uniform field g<,<)Illetry with a.niquc ell\'ironll1('nt for ohS('rva-oil. High-speed photography ofccd hy tll(' initiation and spat.ial

VMonitor

Figure 1. SclwIllatic of Aurora's FIN oil switch.

TIONOIL SWITCH OPERATION

amma.-ray simulat.or op('ra tedI<' de\'in' US('Sfour IJhuul(.ill

I separaJely or in conn'rt ton. Each lim' consists of au oil-

.t.ically insu!;lt.ed t.ransmission-heam diode. The I31umleins

II',and high-volt.age pulses arc.witches in each line. n('n'nt.

radiat.ion puls(' gene'rat.('d hyonst.raints on t.he ac("('pt.ahl<'and mot.ivat.(. [('-examinat.ion

I2m

IGas SwitchTrigger Monitor

The' oil switch is shown schematically in the' ahm'(' figure. TII<'oper-

ation of the oil switch is monit~>re'd hy a capaciti\Oe prohe mOllllte'(!wit.hin the inner IJlumlcin, the V monitor shown in t11('figure. The Vsignal is proportional to the current. from the high-\Ooltage d(,ctrod(~:I = (IQjdt = (I(CV)jdt, assuming the capacitance of the trigger

('Iedrode is constant and that st.ray CUrl"elltmay he lIeglc(y'd. Thet.rigg<'r \"(>1t.;lgeis obtained hy hardware' illtegrat.ioll of the V mOllitorsignal The oil swi tch trigger voltage is shown t.ogetll<'r with ot hermonitor signals in Figure 2.

The oil trigger electrode is a circular, steel plate 95 ('In in diameter.The electrode is 1.25 cm thick and the circumference is machined

to an edge having a 125 ItIII radins of cun'ature. Thc trigger is heldapproximately 5 cm into the gap and is electrically connected to the

critical to synchronizing the~ normal operation of t.he oil

250

5.0

0.0

MV -5.0-10.0

100

KV 0.0

-1001.0

0.0MV

-1.0

-2.010.0

0.0MV

-10.0

-20.00.0

Gas Switch Trigger

V/N TriggerElectrode Voltage

Output Voltage

0.5 1.0 1.5

TIME (Ils)

2.0 2.5 3.0

lOil Switch ClosesGas Switch Closes

Figure 2. mumlpin voltage monitor signals.

high-potential side of the gas switch. Tlw gas swit.ch's ground ('Iec-trode is electrically connected to the inner I3lumlein through a solidaluminum frame. The intermediate electrode is charged to approx-

imately -10 MV by erection of the Marx generat.or. During pulsecharging, the oil switch trigger reaches a ,'oltage FIN, wh('re t.hecapacitance of the trigger to the inner I3lumlcin is N - 1 times thecapacitance to the intermediate Dluml('in electrode. Closure of thegas switch allows the charge on t.he trigger el(.ct.rode to flow to theinner Dlumlein, and dramatically increases the electrical stress atthe t.rigger's edge. Prebreakdown streamers originate at tll(~sharpedge and propagate across the gap toward the intermediate elec-trode. Conducting arcs develop along the path of the streamers,connecting first between the oil switch trigger and the intenlwdiateDlumlein and then to the inner I3lumlein, thus closing the switch.The negative going spike appearing in the trigger signal indicates

.- --

~

251

that electrical contact has been made between the high-voltage elec-trode and the trigger.

The switch run time, that is, the delay between closure of the gasswitch and closure of the oil switch, is one measure of switch per-formance. In general, switch run time, tr, depends on the averageelectric field strength, Eau, in the gap [61:

tr ~ 22 E-1.33au .

Here Eau is in units of MV fcm and tr is in ns. The intermediate andinner Dlumlein electrodes are separated by approximately 50 cm,and, for the conditions described, the average electric field strengthis on the order of 200 kV fcm, which yields a mean run time of180 ns.

OPTICAL DIAGNOSTICS

JS)

2.0 2.5 3.0

lOil Switch Closes

Two high-speed photographs, obtained during switch closure underidentical conditions, are shown in Figure 3. Each photograph con-sists of a sequence of eight exposures: The time separations betweenframes and the frame exposure times are approximately 50 ns and10 ns, respectively. ~he field of view js repre:;ented by tl~e brokenline in Figure 1.

These data arc consistent with the initiation of multiple prebreak-down st.reamers that arc uniformly distributed on the circumference

of the trigger. The luminous regions along the edge of the trig-ger that appear in the second frame of each sequence correspondto streamer initiation sites. The streamers appear to grow rela-tively uniformly at initiation. However, at switch closing only a fewstreamers have bridged the gap. Indeed, a single channel appears toclose the switch in many of the photographs. Open-shutter picturestaken under similar conditions also confirm that the gap is bridgcdby a fcw large arcs.

These photographs verify that the spatial dc,.elopment of the stream-ers contributcs to the shot-to-shot variability in the switch run time.Photographs A and H corrcspond to switch closurcs having runtirilcs of 160 and 140 ns, rcspectively. Note that the luminous regionin photograph H, the switch closing with the shorter run time, is

~llalS.

The gas switch's ground dec-nncr Hlumlein through a solidlectrode is charged to approx-\{arx generator. During pulseles a volt.age Ff N, where ther Dlumlein is N - 1 times the

llein electrode. Closure of the

rigger electrode to flow to the.reases the electrical stress at

eamers originate at the sharptoward the intermcdiate elec-

Ig the pat.h of the streamers,:1trigger and the intermediate"llein, thus closing thc switch.in the trigger signal indicates

252

2 4 6 8

A

1 3 5 7

B

Figure 3. High-speed photographs of switrh closure. Individual frameexposures are 10 ns and the frames are separated by approximately 50 115.The field of view is shown by the broken line in Figure 1. The second

frame is approximately 75 ns after closure of the gas switch. Closureof the oil switch occurs between the third and fourth frames in A andbetween the second and third frames in D.

2:'iJ6 8

significantly less extensive than for A. The transition from denselypacked streamers to the propagation of a few dominant streamersis one characteristic of switch closure. An early transition, as in thecase of runaway growth of a single streamer, reduces the run time.Thus, modification of the switch to favor early transition to a fewdominant streamers would reduce switch jitter. For example, re-stricting the number and location of streamer initiation sit~s wouldfavor -all early transition.

The st.reamer initiation time is another source of jitter. The im-pulse breakdown voltage of insulating fluids depends on the rate of

voltage rise: Tests using a small gap and nonuniform fields [71showthat. t.he variance in the time-to-breakdown decreases wit.h increased

rate-of-rise. Thus increasing t.he rate of rise of the trigg('r volt.ag('will reduce swit.ch jitter. In a recent s('ries of t('sts. tll<' stalldard

de,'iation of the swit.ch run time' was reduced fr"m 9 ns t.o 3 ns bya modest. reduction in t.he capacitance of t.h(' trigger electrode audhy providing a low induct.anc(' cUlTent path hdw('('n tll<'gas swit.chawl t.h(' inner I31umlein [81.

5 7

CONCLUSIONS

Tillling jit.t.('r for closure of the oil swit.ch has two COIllIH)JI('nts: ,'ari-

ahility in tll<' streamer init.iation t.ime and in the t(,llIporal d('vel.oplllent. of the st.reauwrs t.hat. hridgc the gap. The \"arian("(' in t.he

st.realll<'r init.iat.ion tillle lIlay he reduced by increasing the rate' of

voltage. rise' ilt. t.he trigg('r d('drod(.. Th(' "ilriability in t.h(' sp;ltialalld, I1<'1In" temporal development of t.he sh'('alll('l"S is likely t.o 1)('n'duced hy rest.ricting streamer init.i,ttion t.o a f(.w sites.

ACKNO\VLEDGEMENTS

switch closure. Individual frame

'parated by approximately 50 ns.en line in Figure 1. The second-mre of the gas switch. Closureird and fourth frames in A andn.

This work is support.ed ill part by t.he Defens(' 1\ uc!('ar Ag('ncy, the'

US Depart.ment of COllml<'rc<~, an(1 the US Depart 11\('11t. of Energy.The iluthors thallk Aurora's t.edmical stalf for th('ir capabl(. assis-t a lice"

REFERENCES

[IJ Smith. I. and Dernst('in, D., "Aurora, an EI('ctron Acc<'l('rator".

254

IEEE Trans. Nucl. Sci., Vol. NS-20, pp 294, 1973; Agee, J., "NewCapabilities of tbe Aurora Flash X-ray Machine" , Nucl. Instr. Meth.,1991.

[2] "Bushell, M., Fleetwood, R., Judy, D. C., Merkel; G., Smith~ M.,and \Veidenheimer, D. M., "Bremsstrahlung Risetime Shorteningby Diode Geometry Reconfiguration", this conference, paper PC-35.

[3] Sanford, T. W. L., Halbleib, J. A.,McAtee, W. H., Mikkelson, K. A.,Mock, R. C., and Poukey, J. W., "Improved Fla.<;hX-ray Uniformityat 19 MeV Using a Compound-lens Diode", J. Appl. Phys., Vol. 69,PI' 7283, 1991.

[4] \Veidenbeimer, D. M., Pereira, N. R., and Judy, D. C., "AuroraSynchronization Improvement", Proc. 8th International Pulse PowerConference, San Diego, CA, pp 924, 1991.

[5] Weidenbeim~r, D. M., Judy, D. C.,.Lindsay, D., Salvan, L. E.,Golden, J.,and Pereira, N. R., "Jitter Reduction in Aurora", DNAAdvanced Pulse Power Conference, Albuquerque, K~L 1990.

[6] \Varren, T. F., Hammon, H. G., Turman. B. N., and Pr<,stwich,K. R., "Jitter Improvement on the 12 MV Oil Switch<'son Aurora",Proc. 5th IEEE Pulse Power Conference, Arlington, VA. 1985,editedby Turchi, P. J., and Rose, M. F.

[7] Kelley, E. F., and McKnigbt, R. H., "Str<,aIIlN Propagation inTransformer Oil under the Influence of Submicros<,cond Rise-time

Pulses", in press.

[8] Weidenheimer, D. M., Pereira., N. R., Judy, D. C., and Stricklett, K.L., "The Aurora Accelerator's Triggered Oil Switch". Proc Cone.High Energy Beams, Washington DC, 1992.~