TESTING OF THE INTER-TURN INSULATION OF HIGH VOLTAGE INDUCTION MOTOR COILS
Michael John Hopkins
A Dissertation Submitted to the Faculty of Engineering Uni v a r s ity o f "the Wi tw a te rs ra n d , Johannesburg f o r th e Degree o f Master o f Science
Johannesburg 1976 ABSTRACT
It is recognised that a need exists for a reliable set of tests of induction motor coil inter-turn insulation, due to the dangers presented to such insulation by fast rising surges. The occurrence and form of such surges are discussed. The behaviour of surges on coils is ana lysed both in theory and in practice to derive the maxi mum s tr e s s l i k e ly to be imposed on a c o il by a fa s t r is in g surge. Based on t h is a c r it e r io n is proposed fo r i n te r- tu rn te s ts . Two tests which do not require the coil to be cut, are proposed. The firs t consists of discharging a charged c a p a c ito r th ro u g h the c o il and a llo w in g a damped o s c i l l ation to occur which effectively stresses the inter-turn insulation. The second test consists of inducing a high frequency voltage wave into the coil using an electronic oscillator. Refinements are proposed for both tests to a llo w f o r a p ro d u c tio n te s t programme as well as t o pe rm it an unambiguous detection i_T breakdown between turns. A number of test circuits are proved in practice. Tests are derived to assess the validity and safe6\ o-' +he c a p a c ito r discharge te s t . A number o f m a te ria ls and samp les are tested to assess their behaviour in the proposed inter-turn tests. The results of such tests are compared with those obtained with the standard impulse test. The breakdown process between the turns of a coil is examined and analysed to reveal a weakness In in d u c tio n m otor c o ils which can lead to a reduced resistance to i nter-turn breakdown. I/ Michael John Hopkins do state that this dissertation is my own unaided work and has not been presented in whole or in part for any other degree or at any other university. ACKNOWLEDGEMENT
The author wishes to thank GEC Machines (Pty) Ltd. and Marth i nusen Ltd. for assistance end materials supplied. CONTENTS PAGE
1 INTRODUCTION • 1
2 AN ANALYSIS OF IMPULSES LIKELY TO STRESS THE INTER TURN INSULATION OF MOTORS ...... 3
3 THEORETICAL BEHAVIOUR OF SURGES ON MOTOR COILS .. 6
4 TRANSIENT MEASUREMENTS ON COILS USING LOW VOLTAGE SURGES ...... • .. .. 17 4.1 Mode I Coils - Results .. .. 19 4« 2 Commepci a l l y Produced M otor C o ils . - •• 26
5 THE CAPACITIVE DISCHARGE TEST ...... 32 5.1 Introduction ...... 32 5.2 A Practical Capacitive Discharge Test .. 38 5-3 Detection of Breakdown in tl.e Capacitive Discharge Test .. ., .. .. 47 5.4 An Improved Test Suitable for Factory Use ,. 55 5.5 Statistical Tests on the Capacitor Discharge Method...... 58 5.6 Various Effccts Observed Concerning the Capacitive Discharge Test .. ., 65 5.6.1 The Discharge between Turns .. .. 65 5 .6 .2 Noise on the Discharge Waveform 68
6 RADIO FREQUENCY COIL TESTING ...... 70
7 PRACTICAL BREAKDOWN TESTS PERFORMED ON A NUMBER . OF INSULATING MATERIALS .. . . 81 7.1 Introduction ., .. 81 7 .2 Tests on Rea I Co i I Samp I os : , , .. 81 7.3 Tests using Sheet Samples of Insulating Materia Is • .. ., ,. .. , 86 7.4 Analysis of the Test Results .. •., 90 ' 7«4.1 Impulse Tests ...... 90 7.4.2 Radio Frequency Tests .. .. 96 CONCLUSION
APPENDIX A 101
APPENDIX B 104
APPENDIX C 106
APPENDIX D 109
APPENDIX E 111
REFERENCES 129 1 ' INTRODUCTION
Insulation breakdown is thf most feared failure occurring in large high voltage induction motors. Damage to the ground insulation is often extensive, and it is the breakdown of this insulation which is most often cited os the initial cause o f the damage. In some cases, however, i t is suspected that the initial failure occurs between turns; heat gener a tio n a t t h is p o in t weakens the ground in s u la tio n which fails catastrophically destroying all evidence of the i n i t i a l f a :Iu re . Cl] I t has been n o tic e d by local m an ufa ctu re rs and r e p a ir ers that the incidence of insulation failure is much higher du rin g summer, [2] As the South A fric a n in d u s tria l areas are subjected to a large amount of lightning during this season it is likely that lightning surges are the root cause of this trouble, by the above process. Lightning surges can travel into the windings, which because of their distributed parameters, behave in a similar fashion to a tra n s m is s io n lin e . The steep surge fr o n t is imposed across a coil (or even a few turns) causing a high transient stress of the !> uI ati on between turns. If this insulation has o n ly been designed to w ithsta nd the normal working potential between turns oP at the most a few hundred volts, or if the coil is faulty* then i nterfcurn faljuue is I j k
2 m ANALYSIS Of IMPULSES LIKELY TO STRESS THE 1HTEK-TURH INSULATION OF MOTORS
Impulses travelling on transmission systems can be caused by a number o f events [3] . These are b a s ic a lly e ith e r switching surges, or lightning induced impulses. The classic type of switching surges found on transmission lines are usually too slow to cause trouble [4,53 ■ However,vacuum quenched circu it breakers are be I no used increasingly for motor switching and can cause extremely fast transients to occur at the motor terminals due primarily to re-ignition phenomena (6,7) L ig h tn in g surges can have very steep fro n ts and it is worthwhile considering studies undertaken on light ning in an effort to estimate the type of surge likely to cause distress to a motor's inter-turn insulation. The shape o f impulses used to te s t tra n sfo rm e rs have been s p e c ifie d by va rio u s bodies [%, 9, Id] . The r is e tim es o f such waves have been fix e d by a semi-empi r i caI process of considering the types of wave likely to be encountered in practice and deciding which is most dangerous to a trans form er [ll) . The waves chosen have r is e tim es o f the o rd er of Imi crosec, which are representative of those likely to stress the transformer due to the nature of typical trans former i nsuloti ons.Most transformers use liquid insulating mediums. These have long fo rm a tiv e tim e lags fo r high v o l t age breakdown, o f the ord er o f I mi crosec. ( l2 , 13] . Thus a faster rise time would tend to stress the oil I css and the value of rise time chosen will stress the inter-turn in sulation of on o il-fille d transformer maximally. Transformers tested on these assumptions have performed well in service. S o lid in s u la tin g m a te ria ls have much s h o rte r breakdown format Ive time lags, typically 10 to 100 nanoseconds [l] They are, th e re fo re , more s e n s itiv e to fa s t tra n s ie n ts than liquid based insulations with slower formative time lags. When considering the type of wave likely to damage a motor, oni is Forced to give consiJrrafcion to much faster fronted waves than one would norm al’ - on sid e r w ith l i q u i d - f i l l e d 4
transformers. Lig h tn in g surges on tra n sm issio n systems have been studied extensively [3,4,15,16,17,18,19/20] using various eystems and techni ques. The standard method is to use a cathode ray oscilloscope to moniter the potential across a divider chain connected to a transmission line. Times to crest vary from 0.1 to 40 microsecs, with most rise times in the region of 1 to 5 microseconds, supporting the choice of rise times used for transformer testing. Crest values were up to 6 times the working voltage of the line [ 3 ] On modern systems this would be limited by arrester sparkover values, suminari sed in Appendix A. They are ap proxim ately three times the working voltage for a 1.2 x 50 microsec wave, but may be much hig he r fo r fa s te r ris e tim e s as is indicated in the Appendix. It appears from the test waves considered in these specifications that a rise time of 0.1 microsec is not unlikely in practice, although this may be rather severe considering that there is no certainty that very fa s t waves go up to f u ll a rr e s te r sparkover. Work has a lso been done on .s tlm a tin g the maximum rate of rise of surges, using a magnetic link to measure the maximum charging current into u capacitor connected to the lin e [ 3 ] . W h ils t no in d ic a tio n o f wave shape is given, i t may be assumed th a t a s iz a b le p ro p o rtio n o f the wave fr o n t has t h is slope, The fo llo w in g re s u lts were ob ta in ed fo r a number o f lig h tn in g surges:
Li ne v o lts Maxi mum o f wave f r o n t K' r KV maxi mum average ml nimuvi
132 8 JO 220 - 33 240 100 49 12 282 6 5 25
Of part •i-.ir interest is the 12 KV line, similar to loca l 11 KV nes. fo r such lin e s values o f up to 280 KV/ microsecond could be considered. Also to he noted are the specifications for arrester sparkover voltages on wive fronts (Appendix A). A 12 KV 5
a r r e s te r used no rm a lly on an 11 KV system has a s^drkover at 50 KV when a wave rising at 100 KV/microsec is used. A wave o f t h is rate i f 1 v..>uld appear to be not unknown in practici if it is specified for arrester tests. In considering inter t.urn insulation it is the high ra te o f r is e which does damatje. T his con be o f the ord er o f 100 to 300 KV/mi ci'osec to r an 11 KV system. A useful value to adopt is» thus 20 KV/m i crosec per KV of rating# The maximum value of the crest is unimportant for arrester-pro tected systems, bufc may be taken as a maximum of five times the supply rating- In defence of such a rapid rate of rise it is interesting to study work done on cunx-nhs in d ir e c t stro ke s [17,18,19? 20]. Current rates of rise can easily be of the order of 40 KA/microsvc and these c u rre n ts a c tin g through the surge impedance of a line, or the footinn ii.ivdance of a tower nive rates of voltage rise which are extremely rapid. Also a study on 110 V systems indicated rise times of as fast os 0.0S microsvc with crests of up to 2 KV. Vaciu-". c i r c u i t breakers, because o f t h e ir advantages o f compactness and r e l i a b i l i t y , have become po p u la r For motor sw itc h in g . They do have the disadvantage o f te n d in g to gener ate Fast transients. These transients may be caused by pre s tr ik e [o) or r e s tr ik e [7] phenomena. The tr a n s ie n ts are commonly o f the o rd er o f 1 microsecond ris e tim c , but may have r is e tim es as fa s t as 0.3 microseconds. The harmful effects of these transients ma> be avoided by careful design ol the co n tro l gear or the use o f f i l t e r s . A study o f re s tr ik e s [7] has in d ic a te d ra te s o f change o f up to 5 KV per microsecond ^er KV ol" supply potential. In co n clusio n if- may bo noted th a t no allowance has buon made fo r chopped waves, which can have extre m e ly high ra te s o f chargt’ . Chopped waves are u s u a lly the r e s u lt o f poor arrester usage, being caused by I'lashowr at an unpro tected point-i It is also worth notint) that transformers are able to pass the hitih frcquvn . components of surges due to the capacitive coupling of H'<* winding. Thus transformers would appear to be not w ry effective in protecting motors from fast impulses [jilj . 6
3 THEORETICAL BEHAVIOUR OF SURGES ON MOTOR COILS
The ability of motor coils to withstand fast fronted surges requires not only a determination of the type of impulses likely to cause trouble, but also the behaviour of these impulses on e n te rin g a gi von motor c o il. Mathematical • a n a ly s is has been attempbod in refe re nce s (3,2 2,23 ,24 ,25 , 26,27,28,29,30] , the firs t three being the most useful. Q u a lita t iv e ly i t is known th a t c o lls behave in a s im ila r manner to tra n sm issio n lin e s . T h is I a due to th e d is tr ib u te d nature of their parameters. Surges are able to travel in the co 1 I and are transmitted through it, usually being ex te n s iv e ly m o d ifie d in the process. The d iffe r e n c e s between a oo'I and a transmission line lie basically in the coupling due to mutual inductance and capacitance in th e c o il. Any wave passing through Ir degraded, the higher frequency com ponents of the wave front being particularly attenuated. The wave Front tends to flatten out and thus only the first sections of a coil are fully stressed between turns. The c o il a lso pre sen ts a surge impedance which can be measured. This Is Frequency dependent, making an accurate de te rm in a tio n d i f f i c u l t . T h is surge impedance, Zc, can cause tro u b le i f s u p p lie d from a 1 Ine o f impedance Z\ . The terminal voltage then tends to change under surge conditions
h + z .
wlierr U is the crest value of a surge on the line. Thus an increase in surge p o te n tia l can occur and PIaahover may occur near the machine lerminols. If the end of the winding is not grounded or terminated then Flashover may also occur here due to refection of surges travel ling through the windings. The simplest approach to surge analysis is that of con c e n tra tin g on ly on the f i r s t few tu rn s o f the c o il and assuming th a t the surge does not. m a te ria lly change on passing through them. I t is assumed to hove a con sta nt slope and 7
velocity. If this slope is M volts/sec then ui« voltage across the first turn is
Vo| = M T
= 54 volts (1)
where T is the t r a n s it tim e fo r a tu rn and the length o f each tu rn o f the c o il. The surge velocity can be estimated to be that of a tra n sm issio n lin e in the same medium as the c o il. i.e .
v = y===? mvtres/sac ...... (2)
where C is the v e lo c ity o f the l i g h t and £r is the relative permittivity (dielectric constant) of the winding insulation. For a po pu lar modern m a te ria l such as epoxy re s in and mica (or glass) the value of €r is nivinn a velocity of
-
= 1,2 * UVS metres/sec
For example take an 11 KV coil with L^. = 2 metres and usina a wave o f slope 20 x 11 - 220 KV/m i crosec, to qive the voltage across the firs t turo as
^ T i i n ?
- 3,(1 KV/U ii- i.
and T = 17 luuio&fconds When one considers that the normal working stress is perhaps 100 V then the Im p u la iv r s tr a in on t h is in s u la tio n can he appreci cited. This approach iijimiu’S l-lu- couplinf) between coil elements. Taking just Hie rdpaci t-i ve coup! i ng gives the following model. This model is suitable only for the very high fre quency components o f the sui’tje, i . e. the fr o n t o f the very steep surge. F igu re 3.1 C a p a citive model
K is capacity between turns C is capacity of turn to ground
This is the same as a string of n insulators and if the number of elements is high then the potential of any element, say m, i s
V„, U g (3)
J , the transfer function. ( See Appendix 8 )
For the first turn the potential is ...... and the potential across the Firat turn is
.... (5)
If (n-1) <| is I artier than 3 then
si nh (n -1 ) 9 =, (6) 9
4 U(i-e 9) (7)
If g .n a first order approximation results
L -i Ug (8)
A more precise analysis of surge behavior is that due to Wagner [2 2 ]. He used a ladder network as in fig u r e 3.2, taking into account oil parameters except the mutual inductance between turns.
t w utind,; i * h . . * t------— saagag;— ------L
|| M. r ... _ j ! kd” ------1 II .... II
------\
L = Inductance/unit length C = Ground capacity/unit length K = Turn capacity/unit length
F igure 3 .2 Wagnor's model
element of the coil winding the following rel ationships hold:
(9) d U + U ! - d t
(10 )
( 1 1 )
whore u and i arc rc s p c c ti vo vo lta g e s and c u rre n ts Elim inating the current terms
° ......
Assuming a solution of the Form
u - ...... ( 13)
and substituting into (12) gives
« 2 - LCw2 - LK«<2wa = 0 ...... (14)
So Ivi ng fo r w
“ - “ ■ ...... (15)
...... Where " I n F ...... = wcp , the critical frequency of the coil. This is the highest frequcnc> .1 the coil may Free I y osci I I dte. I t is the nti. frequency o f an inductance l_ dmd d cv.pacit,o»ce C. I f a t r a v e llin g wave s o lu tio n is sought then such a wave w ill have a v e Io c ity , (18) / l c ( i +)W2) i.e. i F c* tv rids to inFinity then v tends to zero. Above the c r it ic a l Frequency the v e lo c ity term is imaginary and thus the coil w ill pass only components below the critical frequency, In other words the coil octs as a low pass fiIte r, tending to pass only frequencies below a c e r ta in value*, the- c r it ic a l Frequency, Like a I I fi I vers this effect is not abrupt and higher frequencies tend to penetrate the coil to a certain extent. A sim p le r approach may be Found by con s id e rin g one element of the coil shown in figure 3-2. If the inductive and in te r - tu r n c a p a c itie s are lumped the impedance to a wave of frequency w is that of the two elements in parallel X.X, i'e' z “ xT+TT ...... (19) 1-,,-LK frequency th e impedance tends to i n f i n i t y . In practice the impedance would tend to the equivalent resistance end the component corresponding to this frequency would bo uni Formly distributed across the coil elements. Tiii s is the same as the critical frequency defined before. For freq ue ncie s above the c r i t i c a l frequecy the imped ance Z is negative, i.e. ic is capacitive. The effective capacity of the arrangement is K' and is defined by i .o. r - K - -Jj- ...... (20) 12 This forms vhv foini I i -ir tuipooitive ladder network c o n s i d o p e d o a r I i e r . Maki n« a ' “ H * ~ J V and stibsti tu ti ng into equation 3 \ ( " ) Likvwisv for frequencies betow the c ritic a l, the imped ance o f the p a r a lle l elenient is p o s itiv e and bunco forms an equivoI out inductancc 1-w-LK L ( 2 2 ) This forms an i nductor-capaci tor fodder network which behaves as a transmission line. For frequencies below the critical the coil tends to act as a Ironsm i ssion line. This has a propagation velocity of "Jtv t - r “ (23) where v ■» propayati on v e lo c ity wit-hout the i nfl neiue of i uter-tnrn capaci ty 13 S im ila r ly the c h a r a c te r is tic surge impedance o f the arrangement is where z ~ surge impedance o f th e arrangement without the i nter-turn capacity present v ' and z ' are frequency dependent end tend to v and z when th e frequency is low. This approach is useful when the frequency components o f the incoming wave are known. I t is o fte n d i f f i c u l t to estimate precisely the Frequency components of a steep surge. Wei I auer [23,3 analyses th is by considering an , impulse having the Form oF a Sine integral viz “ " 2 + T f/S'U "* d" ...... (25) This y ie ld s th e shape o f impulse shown in fig u r e 3.3 Figure 3.3 Sine integral shaped impulse 14 T h is is dssumed to opprox i niuto to a norind I impuls It contains oil the Frepuimcy components From 0 to w. The maxi muni slope ol" the impulse is at A and is du , (M) IF the r is e tim e ol" tin impulse T then the rnoxi mum frequency components in the impulse (27) This o f fe r s on approximate to s i to see i f any components Ii o above the c r it ic a l frequency, [hose above the c r it ic a l freq ue ncy, iv , have the form , M ( 29) ■I and those below the critical frequency . 129) Wellauer derives the posiHo fo r the I iiii i t i ng case of a rise time of zero which yiv s o p o te n tia l across the. f i r s t tu rn of u ref' llmax 1,11 J K (30) Compare t h is w ith eq ua tion X it is to he noted for the test wave chosen previously i.e. rising at 20 KV/m i crosecond/KV of rating to 5x rated v o lts , th a t the maximum frequency component by th is method is w. - ~2— win'in* T L- 0.2S m I croseconds i . o . w |na){ i , 2S x ^ rv u li a n s /.s t-c Choos i 119 o typiool 1110 t;oi' 1:0 1 I w ith L - 10 ^ H /tu rn ond K “• 400 p F /tu rn y ie ld s c.> ” /nc"1 1,0 \ 10° iMdi o iim/ h i' c Such o coi i onlil hovo I'ruiim'iicy tompononts obovc t h is xvliich do not tend to t.rovi I t hroiipli i t and i t can be as- siimod th a t most c o ils would tend to bo ha w in a s im ila r Fashion as thv paoamotors ohoson are rea listic ones. So Far the oFFoots oF losses and mutual inductances h a w boon ipnorod. Tho losses are complex and are frequency dependent. They tend to oFFect more the h ig he r Frequency components oV the wave Front avcvnfcuati ng the degrading of thv waw Front with passage through tho coil. Rudunberg [ 3 ] considers only the mutual inductance between adjacent turns. He states that this tend-, to give a critical Frequency as b e fo re , but m o d ifie d to cv /TT-MF"11 (31) where M is the uuitual inductance betwoen adjacent The o iT v v t o f such munuaI i ndwctance is thus to ra is e the c r it ic a l I'i'equenoy and a llo w a higher p e n e tra tio n oF high Frequency components in to th v c o i l . I t is assumed th a t the effects of mutual inductance affect motors less than trans form ers os the c o ils are screened by the s ta to r s lo ts From each o th e r. A general s o lu tio n For a system in v o lv in g mutual inductances requires thv solving of a Frcdholm type of i ntvgro-di Fl'vren t i a I equation. As th is is unso I vabl e 16 in closed form an iterative successive approximation solution must be sought involving a computer programme. In practice most coils appear to have a higher inter turn capacity than ground capacity due to the symmetry of the coil cross-section. The conductors are closer to each oth e r than to ground and expose more surfa ce area to each other than to ground. This gives a low transfer function, g. For this reason penetration of Frequency components above the c r it ic a l are lik e ly to be r e la t iv e ly deep and no s ig n if ic a n t change in the e n te rin g wave can be e .'' 2oted over the firs t turn. In these caaes a travel ing wave model is the s im p le s t and most s u ita b le to use. If the coil has a very low critical Frequency then it is w iser to use one o f the more complex methods o f a n a ly s is . I n t u i t i v e ly i t can be set'n th a t t h is would bo a rare occurrence as the travelling wave is unlikely to bo affect ed by something right on the boundary of the coil, such as the firs t turn parameters. 17 4. TRANSIENT MEASUREMENTS ON COILS USING LOW VOLTAGE SURGES . A p r a c tic a l ck'sor i pi-i on oh' surtjv behaviour on variou s c o ils is attempted. This in dom* nut only l-o aubstanfciatt* tho in fo rm a tio n conlm i m-d i it the I ,us t vh iip fvr, but u I so to p ro - v i dc vi pi'dcl-iixil ti-st app I i v.ihl v to vo I I a olrmidy constructed. As thuo i'f t i co I isiuilytsls of coil behaviour is difficu lt, tho a p p lic a tio n o f »uch .1 tx*s1 to samp I o co i I s may provide a fa r simp I or .iiisivv r to qui-stions of mhi’jii1 hvhavi our in a motor than a iiiatln-nui t i ca I study. Such a p ra c tic a l tu-nt rc iiu irc s a mi ni muni o f uqu i pinunt and c o n s is ts o f i u jc c t t uy Uni vo ltu yo wtupu in to tho c o it while monitoring the proHi*i's»h °l’ the btvps throuyh the- coil usi ng a cut>i; rov o.sc i I I ohcopr as in fi j.)in*v 4.1. A rc- pv t i t i vv sourc..' o f stepw i h n-qti I red u ml i t was due i dt-d hero to use d w r y I’a.si r i sv t i mi- puluo uenv ru to r. T his provided a onv nanosecond r i .set i me pulse ia s tin ij a number o f m icro seconds at a r e p e t it io n ra te o f approx i mate Iy 10KH„. Tho pu)sv length chosen was sul'ficienl' to allow bhv vliwplay o f tr a n s iv n t phenomena w ith in the c o i l , w h ile the re p e titio n ra te chosen yave 0 bright, stable display on the oscilloscope. The step was assumed to approximate to on ideal ste p, a reasonable aahumption consider!ny its rapid rise. To ob serve fully the rupidly•ohonyIna nature of the transients v»itin» thv coil, a biyU input impedance wideband oscillo scope is re q u ire d , p r e fe r a b ly o f 1 OOMii, b.ind-w i d t h . In ord er to a f fe c t the per|-'ormunce o f the c o il as l i t t l e as po ssib le the o s c illo s c o p e must be pro vid ed w ith M ijh impedance probes o f low c a p a c ity , so as to avo id I uadi ny the c o ll. Two types ol' c o ils w. re t nvest i yati'd; model la b o ra to ry c o ils and actual m otor c o ils . The model c o ils co n siste d o f thin wire aii-spe mled in air above a ground plane of a I u- ntinium T o il. The dimen...ionw were larye to a llo w tr a v e llin g waves a reasonable I i me to tra v e rs e the c o iI and hence in crease the effective time resolution. The motor coils con sisted of coils placed in the stator core of a iaryn H,V. motor, fixed in position, but not yet taped toyether. This allowed access to the ends ol the coils for connection, full deto ili» o f co 1 I dimensions ami param eters are to be found 18 CRO PULSE GENERATOR .10 ., F igu re 4.1 Surge measurement on mode 1 c o il MAIN SURGE 0. ------ INDUCED SURGE INDUCED SURGE Suroe e n te r ! nn coi I INDUCED SURGE ------3------T T H U I N SURGE INDUCEd SURGE 1------1 INDUCED SURGE L—^ INDUCED SURGE Surge after almost a turn Figure 4.2 Propagation of induced surges on coil 1 9 i n Append!x C» 4 i t Mode I Coils - Results ( i ) Mode I c o iI 1 Mode I coil No. 1 represents a motor coil in layout having 10 turns equidistant From ooch other and set a fixed distance from e a rth . The spaei ntj chosvn gdvu j redsonabl e ratio between turn capacity to earth and capacity turn to earth. The C/K ratio was 0,7 5 giving u transfer function g of 0,86. This a I I owed u neasui'ab I e depa rture from i ded I trans.mi ss ion I i up behaviour due to inter-t-urn cup-icitive coup I i ng. A number o f fa c ts wvre rvvoa ivd. (a) An in itia l o I ectrostat i c d i sti* i init i on of 0.4 of the applied potential was established i imiiedi ate I y on the f i r s t tu rn . O s c illo g ra p h 1 shows fclut p o te n tia l one turn from the start compared to the input of the coi If for an unturmi natod (open circuited end) coil. The case for a short circuited coil is little different (Oscillograph 2), as would be expected from th e o ry. (Appendix C). The case fo r the t h ir d tu rn (O scillo g ra p h 3) i s however d if f e r e n t and emphasises that the original mathematical m >deI was only an approximation for a more complex fie ld arrangement. The initial potential «jt uhe third turn should have forcen 0,2V. This distribution is 0,3V due to the effects of shared fields between capacitive elements of the original model. This is n o t' lik e ly to be so in an a ct la I motor c o i l , where • tin- conductors are large and fiit, tending to screen each othe r and th ereby sep a ra tin g th e variou s ca p a c itiv e elements and the i r f ie ld s . It does i l l u strate that c 0,4-1 o f fu ll v o lta g e ). (b) A travelling wave is observed on I:ho coil. This x= 50 ns/div x =-50 ns/div 4,1 C oil 1, % , End o/c 4 .2 Coil 1, V, , End s/c x = 50 ns/div x=100 ns/div 4,3 Coil 1, V3, End s/c 4.4 Coil 1, V10,End o/c x = 5 jjs/d iv x- 5 ps/div 4.5 Coil 1, V10 , End o/c 4.6 Coil 1, v5, End s/c 21 takes 0,1 microsecond to transverse one turn of 30 metvos { 1 and 2) and 0 ,3 microseconds For 3 tu rn s (90 m litres) ("3). This yields having a low impedance (SOjt.), causes a negative wave to be returned. This explains the abrupt drop in the main wove at point I after 0,2 microseconds, (“ l ) which is due to th e re tu rn e d m inor wave a r r iv in g a t t h is p o in t. Those m inor waves can cause small transient overvoltages at various points in the coil. The main effect is that of reflecting energy back towards the so u rce , O sc illo g ra p h 4 shows th e v o l tage at the end of the coil (turn No. 1.0) and it can bo seen th a t about 0,1) o I" the surge volta ge arrives at this point, taking into account the p o te n tia l d o ub ling o f the open c ir c u it e d end. In Oscillograph 5 taken at the same point over an ex tended time scale, it can bo seen that the electro statically induced waves eventually begin to pre dominate. The coil is now resonating with a Fre quency fix e d by it-s to ta l induetanou and in te r - tu rn capac i nonce. The main wave can s t i l l be ob served as^'iving every 2 microseconds until it dis appear* a f te r approxi mate I y S in i croseconds (^5 ). The waveform becomes smoother u n t il i t becomes a damped o s c illa t o r y s in u s o id . The e q u iv a le n t c i r c u i t e x p la in in g t h is behaviour is shown in fig u re 4.3. f“ / L = total inductaitcc of coil Cj = i ntenidI capacity of the coil turns Cg = end capacity of coiI to earth C|, in this cose, is ignored os it is considered to be ten capacitors (inter-turn) in series and hence ra th e r a low v alu e. On t h is ba sis the f r e quency o f o s c illa t io n is found to be 184 KH which yields Cg = 250 pf (see Appendix C), a reasonable" fig u r e For the end c a p a c ity o f the c o il tu rn s . Oscillograph 6 shows the potential at the midpoint of the coil with the end short circuited. A similar e f fe c t occurs here, w ith a smooth damped o s c i l l a t io n emerging as the reflected components are eliminated. The oscillation exists between the coil inductance and the in te r - tu r n c a p a c itie s and is much more ra p id . This is duo to the reduced capacities involved. (d) As the coil bohaves as a transmission line it ex hibits curtain wo 11-known transmission line character ih tio u . To o b ta in th e surge impedance o f th e c o il it s end was te rm in a te d by a v a ria b le re s is ta n c e and th e effects oF varying this termination observed. Oscillo graph 7 shows the e f f e c t o f a 1000 ohms te rm in a tio n . X. 200 ns/div x= 200 ns/div 4.7 Coil 1, V10,1000 ohms 4.8 Coil 1,V9,300 ohms x 200 nsldiv x 100 ns/div 4.9 Coil 1,V10, 500 ohms 4.10 Coil 2.V,, End o/c x 100 ns/div x- 100 ns/div 4.11 Coil 3.V., End s/c 4.12 Coil 3, V5, End s/c 24 This is too high os a positive surge is retransmitted from the termination and arrives back at the source. This causes a small positive rise in the source wave form after 2 microseconds. This is reflected back as a negative surge arriving back at the termination a f te r 3 microseconds. A valuo o f 300 ohms tvouid appear to cause the complete absorption of the surge (Oscillograph 8). This is deceptive and does not indicate correct terminot Icn. The minor wove behind the main surge, which is negative, tends to inter act with the main surge in such a way that the main surge, its slightly negative reflection, and the m inor wave and r e f le c t io n a ll ca n cel. The t e r mination is actually too low. A termination of 500 ohms would appear to be very nearly correct. (Oscillo graph 9 ). Compare t h is w ith the open c ir c u ite d t e r mination (m4). The main surge is now almost half the open circuited valuo indicating correct termin ation. This agrees with the theoretical value derived from th e eq uation where L ” inductance per tu rn C ™ turn capacitance to ground For c o il 1 L = 55 x 1 0 '° Henry C = 220 pF /is. * lo r 6 Z, o 0 x 1 0 "" = 50 0 jx An interesting point is that although the termination effects the main surge, it appears to have little effect on the minor surge travelling behind it. Also observe that while the main surge is attenuated to 0,6 of its initial value, the minor surge appears to bo unattenuatod by its passage down the coil. The only explanation of this lies in the additive in d u c tio n o f a ll m inor waves as they tra v e l down 25 the c o i I . (e) Little Ioss of steepness occurs in the front of the main surge i t s e l f . The surge is a tte n u a te d by the reflection of energy due to mutual capacitance (or inductance) as discussed above and s w if t ly ceases to be important, it is still very significant after o n ly one tu rn o f tr a v e l in to th e c o il and fo r pra c tical purposes may be considered to be unaffected in th i s pegi on. Thi s a I Iows the c a I c u la ti on o f maxi - mum stresses at this point to be simplified ,t° that proposed in the theoretical treatment of the pre vious chapter, namely a simple travelling wave. (i i ) Model c o iI 2 The configuration of this coil is different from the first one in that although the dimensions are the same the con ductors are arranged vertically instead of horizontally (see Appendix C). As is seen in Oscillograph 10, the beh a vio u r is alm ost id e n tic a l t o Model C oil 1, te n d in g t o con firm that electrostatic behaviour is better explained by a f i e l d c o n fig u r a tio n model than a c a p a c itiv e model. This coil allows shielding of the upper conductors by the lower, yet behaves in exactly the same manner as the previous coil even though a c a p a c itiv e model would be made up o f c a p a c ito rs whose ground capacitance C varied with position. In analysing a motor coil, care must be taken to see that littie common dielectric is shared between the coil turns before a capa citive analysis is used. In most motor coils capacitive analysis is justified. (ill) Model coi I .1 This coil simulates the behaviour of a double layer coil compared to the previous models which simulated single layer coils. The dimensions are similar to Mode I Coils 1 and 2 (see Appendix C). Although the potential at point 1 is sim ilar to that obtained at the same point in the other coils (OsciIlograph It) the potential across the layers is markedly different (Oscillograph 12). from this it would appear that half the surge potential is imposed across the insulation layer between coil Iaycrs for a considerable period. This This is likely to heavily stress the insulation at this 26 p o in t and a c e r ta in amount o f o v e r-i n s u Ia ti on is in d ic a te d . Designers are aware of tliis and most coils are well insulated at thi s poi nt. ( iv ) Mode I co iI 4 Thi s coil demonstrates the effect of high ground capacity and low in t e r - t u r n c a p a city (see Append!x C). I t behaves more as a trunsmi ss i on I I no than a c o i l . Observe in O s c illo - graph 13 the lack of overshoot or initial electrostatically induced potentials, indicating perhaps a link between the two as suggested. Note, also,the decreased velocity of p ro pa ga tion . This is due t o th e d ie le c t r ic enamel co a tin g o f the. w ire (Formex) having a Par g re a te r share o f the ground field than before, due to the proximity of the conductors to ground. The effective permitivity is thus greater than 1. 8. Commercially Produced Motor Coils The same technique was used on a set of commercial induction motor c o ils . The c o ils , wound and in s u la te d , were placed in the s ta to r s lo ts o f t h o ir motor and wedged in p o s itio n . The te s t was performed ju s t p r io r to the fin a l ta p in g up, allowing access to the bare coil ends. No access was possible" to the individual coil turns without damaging the coil in sulation so that measurements wore conducted purely on inter coil voltages. (Set- Appendix C for coil details) Four coils wore connected together in series and measurements taken a f t e r th e f i r s t c o i l . O sci11ographs 14 and 15 show the pe r formance at this point with the end of the chain open cir cuited. Oscillographs Iti and 17 show the performance with a short circuit at the end of the chain. The initial behaviour is v i r t u a ll y the same For both cases (*14 compared t o ” 16). A propagation delay occurs of 0,15 miorosecond equivalent to a velocity of propagation of l,0bxl0^ metres/second. This is e q u iv a le n t to a wave tro v e I I in g in a medium o f d i electric constant = 7.7. The actual value for the epoxy re s in mica ground in s u la tio n I a between h and 9, depending on the m a te ria l used. A number o f in te rn a l r e fle c tio n s are di scerneble at regular Intervals. These are reflected versions of the initial electrostatic waves. The driving waveform I ticks sharpness due. to the increased capacitance of the coil to ground compared to th e pre vio us model c o i l , i.e . 1 0 0 n s l d i v 2 0 0 n s / d i v 4.13 Coil 4, V,, End s/c 4.14 M otor Coil, End o/c 5 j j s l d i v 2 0 0 n s ld iv 4.15 M otor Coil, End o/c 4.16 M otor Coil, End s/c 2 p s / d iv 1 p s i d i v 4.17 M otor Coil End s/c 4.18 Motor Coll,410ohms CR=120 nanoseconds, which is in agreement with the measured beha viou r. The tr u e behaviour is not t h is sim ple as the cap acitance is distributed. The capacity presented the first p a r t o f the wave is th u s e ffe c i v c Iy low and allo w s on i n i t i a l rapid rise to take place before the effects of the entire capacity are fe lt. The wave is further complicated by minor travelling waves arriving from Inside the coil. A transient of rise-time 250 nanoseconds, in practice, corrvsponds fa irly wo I t to this waveform and is likely to stress the first coil of a winding by an opprec- iable amount. Considering a 20 KV per microsecond per KV of r a tin g wave Pom give s approxim ate i y 20 KV across th e c o i l . This coi ! should be tested at 3,3 KV per turn if troublefrce operation is desired in on a p p lic a tio n where sharp surges are likely to bo encountered. It is of interest to note that the attenuation, when the coiIs are in the motor, is much higher then of air cored coi Is. This coil, in particular, if provided with an earth of alum inium foil attenuates surges to a much lessor extent than when such an earth is provided by the iron of a stator core. This attenuation is more thai, iikuly increased by hysteresis and eddy current losses within the iron op the stator. This is not designed to cope with the high Frequencies present in th e wave and a high a tte n u a tio n r e s u lts . An anomalous situation exists with a chain, of four coils. An open circuited termination lends to an under-damped circuit which o s c illa t e s a t 73 kHz and lias an a tte n u a tio n fa c to r or 9x10 4 sec""*. Terminating in u short circuit also yields an under-damped circuit, tills time of frequency 2 1 0 kHz and a tte n uation factor 3,3x10"' sec”* . Terminating the end of the coil chain In a resistance tends to increase the apparent damping w ith a value o f 410 ohms g iv in g nn ap p a re n tly c r it ic a l damping ( see o s c illo g ra p h 18 ), Assuming a lumped param eter mode I y ie ld s an e q u iv a le n t c i r c u i t as shown in fig u r e 4-4. w vV w -^OOOOO/ F igu re 4.4 L — Equivalent total series inductance of coils Cj = Equivalent total inter-turn capacitance of col I a = Equivalent total ground capacitance of co i I >; Rj = Equivalent total series resistance of coils 1*2 = Equivalent total shunt resistance of coils Rx = External resistance at end of coi l.s Since the coils are oscillatory at a very much higher frequency when R = 0, (compared to the case when R is i n f i n i t e ), i t may be concluded t h a t is very much sm a lle r than Cg ( by about lOx ). Furthermore, even at this high freq ue ncy, th e value o f Rj is low. In th e open c ir c u it e d case and can be ignored leading to a sim p le r e q u iva le n t circuit. Analysing this equivalent circuit yields:- a ) Also considering the damping factor <% Now in the case o f c r i t i c a l damping ie when R = 410 ohms (3) LC2 ~(2C2I?)2 where .R Knowing'R^, foc and c< , and substituting these valuer, into eq ua tion s 1, 2 and 3 y ie ld s C2 = 3340 pF AI so R2 may bo found R% = 1660 ohms And the inductance L L - 1.44 x 10~ 3 H These values are consistent with the coils acting together, the inductance being that of all four coils plus their mutual inductances. The capacitance is of the same order as that of a single coil to ground. This reveals on obvious fact, that the coils behave in it ially like a transmission line which then tends to a highly damped resonant system later on. The later oscillations, being h ig h ly damped, c o n s titu te l i t t l e danger to the c o iI tu rn in s u l a tio n . Those o s c illa t io n s ore assumed to bo e ve nly d is tr ib u te d across the coils and result in very little stress between actual turns. The initial situation is altogether different. A steep incoming waveform w ill impose a voltage differential across only the firs t coil or part thereof resulting in a high stress in this region. The motor tested in this last section may bo considered fa irly typical;, it is significant that even a half microsecond rise-time surge is capable of f u l l y ste ssin g the f i r s t coi L As has been seen e a lie r a c e r t- a i n amount o f s tre s s r e l i e f is p.",si bIe by v ir t u e o f the 31 eIectostat ic distribution within the coi I. Thi s mer i ts further stu dy. Such a stu dy is p o s s ib le o n ly when th e c o ils used are expendable a nd can th us be pe n e tra te d to o b ta in in t e r - t u r n potent! al s, a s in th e study on model c o ils . 32 5 m CAPACITIVE DISCHARGE TEST 5,1 INTRODUCTION One method of test i nq coil i nter-turn insnlation consists o f di schargi nq a h i gh vo lta ge c a p a c ito r through the c o il to form an o s c illa t o r y underdamped RLC c ir c u it . T h is leads to an exponenti a I Iy decaying sin u s o id a l wave which is imposed across the coil and stresses the i nter-turn insulation. The firs t form of this test was proposed in 1926 by Rylander [2 9 ] . The c i r c u i t used is th a t shown in fig u re 5.1 Hyllimli'r ummecUon for Irallng I lie Iiisiilullon of turns, C <'em mi luir nr Mi'mhiii Iiik npiirk gov l| if[ Tlmi'ii i-livillt "s Vi,lingo on ironiforini'r ur Voltiigc on loot cull FIgure 5.1 A high vo lta ge AC wave from the tra n s fo rm e r was imposed across the capacitor, coil, spark gap combination. When the gap broke down, u s u a lly near th e c re s t o f th e AC wave, the capacitor was di scharged through the coil producing a damped o s c illa t io n . The vo lta g e across the c o il was measured using a sphere gap o r a peak fo llo w e r c i r c u i t fe eding an electrostatic voltmeter. Detection of breakdown was achieved by magnetically coupling another coil and capacitor combin ation to the test coil and tuning for resonance with a healthy test coil and HV capacitor combination. When break down occurred, the change in inductance of the test coil was sufficient to alter hlu- characteristic frenuency so that resonance with the sensing coil circuit was lost and a much reduced reading on the meter resulted., The Rylander method s u ffe re d from a number o f disa d vantages, principally in the detection of breakdown within 33 th e c o il. As AC was s u p p lie d to the system approxi mate Iy one hundred impulses a second were imposed across the c o il. This could lead to e Ffects not usually found with sinqle impulses, such as damaqe due to heating or partial discharge phenomena. Breakdown o f the spark gap tended to occur at different points on the AC half wave leading to a misleading pulsation of the meter pointer. This could mask an inter mittent breakdown within the coil. 7he most serious defect was no absolute method o f d e te c tin g a c o il f a ilu r e as the method of detection relied on a sudden relative drop in the meter reading when breakdown occurred. This method would tend to fail to detect a direct short circuit between turns. This event, although unlikely, is possible due to damage in the fo rm ing process. Breakdown a t very low vo lta g e s would also be hard to detect and would allow a coil which was not only p o t e n t ia lly f a u lty , but a c tu a lly damaged by the te s t, to be used in a motor winding. A further disadvantage of this detection method was that the reading on the meter was pro portional to the coupling between coils and the test voltage, leading to further uncertainty in the meter reading. Vi iui»fiinw'i' Ihftitlrilimt Tnih.InMnvv for HrilllllJT 111.' IITllriiT lli'WilJ'tliw mrtxinr llluw I'nlluuiv rrrllflrr lihvtiln- voimvllor nl 0.1 /I F igure .‘5.2 We I I aue-r [32] has attem pted to improve the method by usin g a DC fed capacitor discharged through the test coil This circuit is illustrated in figure S.2 and probably arises from the availability of high voltage rectifiers I 5 years after RyIander's original circuit was proposed. It does solve the problem of inconsistant voltages being 34 a p p lie d to the c o il. The de te ct i on method used is ob vio u sly an attem pt to a si mpli Pi ed go/no go method, w ith a neon lamp being used as a vo lta ge d e te c tio n element. I t s u ffe rs From all the limitations of Myladder's method with the added disadvantage o f not having a c o n tin u o u sly graded ana i ogue reud out on the detection circuit. An additional Feature of We I Iauer's method is the inclusion of a damping resistor which causes the c i r c u i t to be more damped and hence pro duces a vo lta g e wave form o f th e Form shown in fig u re 5.3 VllildKM VulliW Figure *>.3 Although this waveform is better for coil testing than th e underdamped one, personal experience has shown th a t a large proportion oF l:he capacitor voltage is dropped across th e r e s is to r making such a p ra c tic e i ne FFi ci e n t, '/’] .. Luw-vuilime Imdifunncr 7', Hleli-voliuiiii Ininiliinncr - Illbh'Wiiluiiii cupucilof L, - Tcil siimilinen luoil) ...... rnin.ir.l «UIII|llmu cull f, Htisiimnitc-tlining cujiucllof #i HmiMuiiro ufioi niKclmeii A, SltIuji ru»iinir for 2 1 - - Triwivrei! Mxirk nun 2 liMriiinum lu liiiiiu.iic liileMurri nharl clicuili f i tnilH- 5.4 35 KrankeI and Schuler [33] have proposed a method which is virtually the same as Rylander's (figure 5.4). The in c o n s is te n t volta ge s have been overcome by usin g a tr ig g e re d spark gap which is operated at the same point on the cycle every time. Their method retains all the other disadvantages of Rylander's method. They test the coils in position in the motor stator which may be rather rash as a failed coil w ill r e s u lt in the motor having to be s trip p e d down. Vhey give suitable test voltages (figure S. <), a Feature which other workers have ignored in their articles. - Coil lot voHiibbs 0, In relmlon in imictiino voliiieo ratings I/. fig u r e 5.5 They arrive at their values by considering that in practice not more than 50% of surge voltage is likely to be imposed across the entrance c o il. They th e re fo re con sider a surge voltage to be Us - 4tln + 5KV where U = normal phase to phase vo lta ge o f the motor, Taking 50% of this gives curve 3 in figure 5.5. It can be 36 seen t h a t the values used are hi qh and c o ils te s te d on t h is basis are I ike Iy to be sound in practice. It is noticed that these values are based on the motor voltage only. A more accurate value would have been based on the actual coil character!sti cs. Coils having a small nurnbe- of turns would be u n nccce ssariIy stresse d usin g the values shown. A number o f workers have used in d ire c t means o f te s tin g c o ils which have already been in s ta lle d in machine w indings. Such methods are useful fo r te s t!n g machines which are a I - ready in service. The difficulty here lies in obtaining an adequate voltage across the coil without imposing a high voltage across the entire wi ndi ng, a situation which can damage the ground in s u la tio n . The usual way o f a ch ievin g this is to induce u high voltage into the coi! using an external coil magnetically linked, via an iron armature piece, to the coil concerned. O liv e r e t a I [l] have used t h is method to te s t the in t e r turn insulation of large hydro-generators. They detect break down by d ir e c t exam ination o f th e induced waveform w ith an oscilloscope. Due to the high inductances, frequencies tended t o be low, ty p ic a lly 10 000Hg. The apparatus is shown in Figure 5.6. They gi vis no in d ic a tio n o f how they o b ta in ed s u ita b le te s t vo lta g e valu es, bu t the volta ge s used were sufficient to breakdown some of the windings tested. They presumed that the waveforms obtained at low surge voltages were in d ic a tiv e o f soundness and o b ta in in g a d if f e r e n t wave Form at higher voltages indicated that a breakdo .n had occurred, ('itnllHilhilliiti fiir illilni'iiiK Inil. voIIiiri’ III mil, Figu re 5.6 37 A somewhat s im ila r te s t was used by Sexton and Alke to detect faulty turbo-generator windings [34] . They also used an induced method to obtain high coll voltages, but refined the breakdown detection by using a surge comparison method. This allows the coil under test to be compared with another coil in the winding. An unsound coil will give a different disploy to the rest of the coils in the winding. A weak point in their apparatus was the use of thyratrons to control the discharging of the capacitors through various c o ils . As th y ra tro n s are e a s ily damaged by c u rre n ts higher than the design lim its, a resistance was incorporated in series with the thyratron which tended to lim it the voltage i nduced In to the col I . A number o f workers have used surge comparison methods to test motors as well. Catlin and Rohats [35] used an Ignitron controlled circuit, but found a large volt drop in the leads to the commutator. Comparison was done by tapping across the commutator bars and comparing voltage wave forms ob ta in ed between ad jace nt bars. Moses and H arte r (36) have a method whereby the con nections to th e winding are alternatively reversed by a mechanical switch. An os cilloscope is connected to the midpoint of the winding and th e a lte r n a tiv e surges appear d is p la y e d as p o te n tia ls above ground. Reynold's method [ 3 ?] is similar but employs in due! ng coi i s, The methods describ ed have a lso been used on low powered, utilita ria n type motors, where automatic testing methods are in d ic a te d , Weed [38] has developed an e le c tr o n ic te s te r with an ingenious method of comparing surges in different" coils. Capacitors are discharged simiItaneousIy through d if f e r e n t c o ils in the w inding and dis p la y e d on X and Y plates of an oscilloscope. A Lissouja's figure results. If both coils are sound then a regular figure results, the form usually being a spiral. Should they be different then an irregular Figure appears. Strain [,39] also describes a surge comparison tester For fractional horsepower motors. His article is hclpFul in establishing a statistical basis for such tests assuming a 0,1% rejection rate. 38 5.2 A PRACTICAL CAPACITIVE DISCHARGE TEST A fact disclosed by careful study of the above references is that few workers have chosen to test coils before they are installed in the motor. Most of them test only the completed motor winding, The only worker obviously testing before installation is We I I auer [32] « Testing before in stallation has the advantage that any defective coils are rejected before they are installed in the motor. This elim inates having to strip the motor when defective coils are discovered later i it the manufacturing process. The disad vantage of testing before installation is that if a coil should be damaged when being in s ta lle d in th e m otor, then this damage is unlikely to be detected unless further tests are done la te r on. To assess the right stage at which to test the coils requires a knowledge of what faults are likely to occur in the coil inter-turn insulation. The most common faults are those likely to occur during manufacture of the coil. These i no Iude 1. Voids in the insulating material due to air bubbles,tears or discontinuities in the insulat- i ng mater I a I . 2. Foreign matter occluded in the insulation. This includes water, dirt or metallic particles. 3. Sharp di scon ti nui t i es on the conductors which pierce the insulation. These are usually burrs ra is e d by one o f th e mechanical processes invo lved in manufacturing and winding the coils. 4. Cracks in the in s u la tio n caused by th e p u llin g of the coils to form the final coil shape. Damage to tho coll during installation in the stator is most l i k e l y to he due to excessive physical deformation of the coil. Any such damage is likely to be accomoanied by damage to the ground in s u la tio n as w e ll. I f t h is is the case then the motor w ill fall any subsequent corona or pressure test, such tests being normally done before the m otor is deI ivored, 39 This justifies testing the i nter-turn insulation of coils just before they are installed in the stator of the machine. T h is means th a t the in d iv id u o l c o ils have been wound, pul led and finally cured (usually by heat) before being tested. Such c o ils have low inductances, ol: the o rd er o f 50 to 200 mi croHenri es and re q u ire sp e cial te s tin g c ir c u it s to stress the turns udequatuIy. Duo to the i nconsi stand es of Rylander's method it was decided to adopt a method similar to We I Iauer's method, A variable AC supply is rectified and fed to a capacitor. The capacitor is discharged through the te s t c o il v ia an a i r yap I'orming a damped RLC c ir c u it . The o r ii|in a l c a p a c ito rs employed fo r t h is to s t were those c o n stru cte d by J . J „ Kr i tz i ngcr fo r liis Impulse genera tor (40) . TlifSf consisted ol" TO two mi croFarad paper capa citors connvcf-vd in svri 0 8 tind scaled in a co n ta in e r o f insulating groaso (Poneti'ol ,). They gave a nominal capacity of 0,05 microfarads and a total working voltage of 100 KV. These same ca p a c ito rs had been used by Jones fo r h is te s ta on motor c o ils usin g th e c a p a c ito r disch arge method (41) . These ca p a c ito rs nave poor r e s u lt s and i t was found subse quently that- they presented a very high internal inductance du& to the comHned effects of the capacitors in series. This i ndvt'itxsnce woa of the order of 10 to 20 microhenries and caused a large proportion of the capaci tator voltefle to be dropped in te r n a lly when khv c a p a c ito r was discharged. Coupled with this was the large physical si zo of the capacitor which necessitated iong leads to be used for connection. I t was found th a t the inductance o f these loads added app reciably to the stray inductance present, It was also Found that Kritzinger had installed a low resistance in the tube joining the capacitor to its attached discharge sphere. T h is a lso caused an a p pre cia ble v o lt drop and had to be removed. As Jones had conducted iiis experim ents w ith c e rta in of these impediments present, it is doubtful whether hi s test voltages were anywhere near os hi tjh as expected. This in d ic a te d th a t a moans o f measuri n$) the actual vo lta ge on the c o il is re q u ire d . Such vol tayos are e.asi I y measured by connecting a non-condncti ve resistive divider across the 40 To avoid the volt drop due to the high internal induc tance of the copacitor, it was decided to suerch for more s u ita b le ca p a c ito rs . These were e v e n tu a lly found in the form of ceramic encased capacitors of Admiralty pattern, probably scrapped from a radio traiismi ttrr. Although of lower capacity. Q.012Syuf, and lower ra te d vo lta g e . 30 KV, they had a very low in te rn a l impedance. T h e ir p h ysical dim ensions were very much smaller, allowing test coils to he connected directly to the capacitor terminals. This allowed the entire test arrangement, coil, capacitor, shunt resistor and sphere gap to bo placed on a tablrtop. The close proximity of the components cut down stra y ind ucto ncrs to the minimum and ensured that full vnI rage appeared across the coil. 1 he sphere gap used had 7, S cm copper spheres mounted on an adjustment screw which could be turned, via an insulating rod, to change the gap, and henee th e breakdown vo lta g e . This feature allowed the test voltage to he changed with out de-energi si ng the circuit. It was found unnecessary to use a tr ig g e r e d spark gap as the n a tu ra l breakdown o f a p la in sphere gap was c o n s is te n t and pro vid ed an a c cu ra te ly control!obie test voltage. Experiments with triggered gaps revealed that they tended to interfere with the CRO used to study the oscillation, Jones complained of high capacitor leakage currents lim iting the value of voltage to which lie could charge the capacitor, it. was found that the leakage of the capacitor was not causing this effect. The trouble was traced to poor rectifier diodes with excessive leakage. These rectifiers, o f the selenium p e ncil type, were replaced by modern s ilic o n avalanche type. These were capable o f passing O .F A and avalanched at 22.5 KV. four of these rectifiers each intended fo r norma I working a t 10 KV, were connected in se rie s . This allowed voltages of up to 40 KV to he generated. It was found in practice that the capacitors used were capable of with standing this voltage despite their nominal 30 KV rating. 41 Figure 5.7 The f i r a l best c i r c u i t is shown in fig u r e 5,7, Note th a t measurements are taken From a voltage sensing point, Jones took samples from a current sensing series resistance. It was found in practice that this tended to suffer severely from in te rfe re n c e . The arrangement shown tended to place Pull capacitor voltage across the coil and the advantage was that the veltage displayed on the electrostatic voltmeter was the penk value of that placed on the coil. Thus it was unnecPasarv to keep referring to the C.R.O.trace to obtaim the coll test voltage. Great care must be taken to ensure that all connectors have as g re a t a cross s e c tio n as p o s s ib le to keep down t h e ir inductance and reduce the inductive drops in the circuit so as to avoid p la c in g undue vo lta ge s tre s s between the measuring instru m e nts and e a rth . The e a rth in g p o in t must be p la ced r ig h t ol: the ju n c tio n o f the c o il end and the lower end of the divider. Early tests did not do this rigorously and allowed some alarming fIashovrrs to occur In -the C.R^O. connected to the divider. It is obvious that such a situa tion in dangerous to operators as we I I as to equipment. 4 2 Equi vii I I’nr I'ii'c u it u f i mpul bp genvrutor Fi gun- S. 8 Antil ysi ng th v v iv c u it involvt-tl in the discharge y ie ld s the complvx u r u i t shewn in fig u re fi, This c i r c u i t is e a s ily s im p lif ie d p ro vid ed due oqre Inis been taken to m in i mi so inUm 'tanvvs end resi stanci>s, Choice o f a s u ita b le capacitor elimimit-'s the capuci tor stray inductance and re s is ta n c e , Lc v c . The a ir gap c o n s titu te s a n o n -lin e a r resistor, R . This may he oIimiuatvd as it is cnty effective before breakdown and at low currents, Inductance and resis tance o f the leads. £.f and Rt , are assumed negli lib le . The in te r - tu r n c a p a city o f rlu- c o il, Cmc, is assumed i n s i g n i f i cant compared with the main capacitor, with which it is effectively in parallel. The inductance of the divider, is m inim ised by the use o f n o n -in d u c tiv e asbestos based resistive tape. Ry lumping the divider resistance. with the effective m il rcsis unco, ii|]K. , an effective resistance o f R i s obtained in 1 he c ir c u it shown in fig u r e S.9. An aiui 1 > >. i * of this simple circ u it i a presented In Appendix C. 43 S i mpl i f i cvl i-tin i v .i! tin t Fi yurv S.O TI»p .nut i \ s I s is utic l"ii I lo r t.ht> i n i t i a l de-si nn oI" thv ti°st i"I i •u i r s, d I I owi ivi u roiKih clio iu r oF ca p a city and d iv id e r I’t--: sl-dni:»'. Thr I'i no I choicp o f such compocipnts rest s in di i'vv expvr i tucnl-a h i on ds the c o il i nductancp and resji stdiici' ilvpviHlid on on sk skin I ; p HYct s wi tlv n the coil dur to tiiP h r t)ll r i ’ eCT.U-IK-\ i'll Ild illlK -n t, vomponvnl. ol; th r te s t wave form., li.is skin p n't-vt de-pvnds m cvndm -tor slz e and con Fi y u ra t i on ,.i.d is di ITi fii 11 fo prod t th o o re t i cn ihv ii'sl- results T i,;, s .•snrci d I I y to th r d'uount C>1" 6it r ‘ iss i tnposi d u turns A lowpr frrourncy Hu- use o I' a l.ir iu r cdpfici to r wliirh mrons th a t uf- 1) I' sto re d oni-rq> must hr di ssi pai rd whrn t M iisv l"u I to fivnporr th is t r s t with t-1 . s p io ific d l>y I.E.C. -spi-c i Fi cat i on L ie ] h ,1 stuiuld imI iiiipu I sr wdvr Form with I tdiji- to 1 hi crosh volur in l .2 ‘r iki in •"0 111 i i-rost-conds as sliown ir •i' <. IT' 1,1 I. Sp< c i I 1 vdt i ona surh ot. th is om’ , allow I unh i nip ul >»(* t vh! 1 in) oF v I vrt r i ro I cnuipmrnt. I 43 n Si mpl i I'i ud i-pui vu I vnt c i r c u i t Fi siuro i,') The c ir v u i t .n u ils s is i s use Till Tor the i n i t i a l rirsinn of the test circuits, dllowitvi d rouuh choice of capacity and d iv id e r re sista n ce . The filia l choice o f such components r e s ts in d ir e c t exper imenttit i un as the c o il inductance and re sista n ce depend on shin e f fe c t s w ith in the co i i due to the high frvouem 'v fundaments I component o f the te s t waveform.. This skin effect depends on conductor size and configuration and is difficult to predict theoretically, justify ins such a practical approach. Previous workers have not indicates! any standard waveform for this test. I hi s is o st'r i ous omissiot i different’ fundamental frcuuvnci es ansi decay ra te s cou ld have an e f f e c t on the te s t r e s u lts . This a p p lie s e s p e c ia lly to the amount of stress impost-I across initial turns. A lower frrourncy also im p lie s the use ol a l.iru r r c a p a c ito r which means th a t a larcier amount of stored rm r-p must he dissipated when t h is is di schiirnid. 11 is useful to compare t h is te s t w ith the slumlord impulse test , spicifiiil hy l.t.C. specification iiO [lO] . This spi c i lies a standard impulse wave form with on impulse nsriHi in % u I tape to t hi crest value in ^.1 C m I crosi-t onds and deca\ i ik | in r 0 microseconds as shown in fig u re A, 10 I ,i I , Speci f i cat i ons such as t h is one. a llo w constant impulse tislin I EC sl.mdui'.l i mj>iils I i iii'i’i- S. IV A ht.md.ii'd wd\vl\irin lo r v.ii'-ivi t i vi iii scIuii' cif ti’stiivi voulvl vunx i-ni iMit I \ hv on rh is stdmJdrd wdv T w hit-h iinisi In- I . ' " m i v ro s i'i. om ls . ! h i », 11 > -1 ds f ns 1<>7 KHz. The ii'ipulhi sp, r, I i i\ii i on flive a wave which is s u f f ic ie n t ly underdamped to have an envelope decay slower than the standard one. Adjustment of the value of the shunt resistor used to monitor voltage allows the exact decay to be achieved. Such a procedure resulted in the waveform shown in photograph 1. Selection of the correct capacitor combination with a coil of a certain inductance may he ascertained from figure 5.H. fC/C //V aU {.7ANCl£ yu// -l-C combinations for allowable impulse frequencies fig u re 5.11 X 10 ps/div Y 5 KV/div X-10fjsidiv Y” 5 KV/div 5,1 Sound coil waveform 5.2 Breakdown waveform 5.3 Discharge track IBSS X 10 ps/div X 5|JS/div Y -500V/div 5,5 Current waveform 5 . 6 E n d o f d i s c h a r g e s . 3 DETECTION OF BREAKDOWN IN THE CAPAC H IV E DISCHARGE TEST The most difficu lt aspect of the test is the detection of any breakdowns between turns of the test coil. As stated before, previous workers have mostly used a tuned secondary coil. This is designed to resonate at the same Freouency as the healthy coil capacitor combination. Potentials in duced magnetically in it are rectified and fed to a meter. If the coil should break down,the freouency of oscillation changes, reducing the induced voltage.and hence the reading on the meter. The method has a number o f disadvantages. The resonant circuit must be adjusted for each new set of coils if these are o f d i f f e r in g inductance. I t may even be necessary to adjust for each separate coil. Voltage must be applied in gradually increasing steps; breakdown being indicated by a sudden drop in the meter reading. Thus th e meter reading is relative and is tied to the test voltage. Obviously with this method it is not possible to determine absolutely whether or not breakdown has occurred without some sort of reference. In most cases this reference is the test coil itself at a tower voltage. Another standard coil which is known to be sound may also be used as a re fe re n c e . In t h is case both coils must be identical. A number o f methods have been t r i e d in an -attem pt to fin d a r e lia b le means o f in d ic a tin g breakdown. The most commonly used method in the tests done was to monitor the voltage across the coil, via the potential divider, using a cathode ray o sc illo s c o p e . I t was Found t h a t the waveforms given by sound and unsound c o ils d iff e r e d so markedly th a t p o s itiv e Identification by this method is assured. This may be observed by comparing photograph 1, (a sound coil), with photograph 2 (an unsound c o il) . Photograph 2 shows a wave Form which is not only of higher Frequency, but also of greater attenuation. I t a lso appears fa r more complex w ith noise and high freouency components superimposed on th e fundamental wave, T h is wave appears different from the sound one because the breakdown appears to absorb a large amount of energy whilst at the same time effectively reducing the inductance of the coil. The short circuited elements of the coil also tend to os c i l la t e , producinq complex h i oh freouency waves The break down di scharqe causes noise to be generated in the circuit. All this leads to a highly distinctive waveform. Previous workers have noted th a t breakdown is d i f f i c u l t to d e te ct usin g t h is method. In those cases the machines te s te d were often large generators with the coils already installed in the stator. It can only be surmised that the effects of the s ta to r masked the breakdown e ffe c ts . As can be seen in Chapter 4- this leads to considerable attenuation due to the iron of the stator damping out high frenuenci es This iron is designed to have a low loss a t low frenuenci es o n ly, Coil geometries are also important. It was found that most discharges within the motor coils did not tend to take place between adjacent tu rn s , but r a th e r between a number o t turns (see later}. This accentuated the drop in inductance as well as a llo w in g more vo lta ge and energy to be absorbed by the fault. Obviously the lower the stray inductance of the c i r c u i t , the more n o tic a b le t h is e f fe c t becomes. I f one is Forced to di“charge through a series of coils in a winding then the e f fe c t o f one tu rn on a c o il going s h o rt could e a s ily be missed. I t is not known whether more than one tu rn would tend to be involved when a large generator, with its differing coil geometries, is tested. Despite the findings of other workers, it i • ‘‘••It that although this is not en tirely an absolute fault detection, it is still the most useful. 11 •mnended for any laboratory tests as more in fo rm a tio n th .. ,at de te rm inin g breakdowns is gained from i t . I f one wishes to avo id us inn mi o s c illo s c o p e then o the r methods ot hriok down detection may be used. This may be re - oui red when testing in a Factory environment using semi skilled labour, Jones [41^ discovered that an electronic counter connected to the current sensing ou-put q s v p an in d ic a tio n o f breakdown by the number o f counts r e n ls ta re d on the counter. The counter was operated in the manual mode i.e. it remained activated by a pushbutton during the test and counted all crossings of a preset trigger level. No in dication of the actual trigger level is given. The drawback 49 with this method is thi-vfc thr coun H var i r s with trincifr I f vr I and w ith te s t voltaeiPs, joiif-s Found th a t he ob ta in ed court s o'!5 about 2 ^ wit!' sound n>i Is and Q or 9 with failed coils ^typically). He surmised correct Iy that this was due to the much hi ghei rate o f dpctiy o f tin1 fa ile d c o il wave „ T his allows less oscillatory wavi-1 crests Vo exeede the triqqer I eve I and bp counted. It was decided to improve on this iiieflns of detection usin g eon i pmetit th flt was simple t'o operate and economical to purchase o r co n s tru c t. It wrfs Found th a t none o f the counters available in the El rc tri cd I Encii neeri nci Deoartment were s u itd b le fo r t h is purpose. I t was decided th a t a custom built counter uni k offered the best solution to this problem. Such d counter pttiploytnq TTL i ntpqrrfted circuitry, was de signed and c on structed . The c i r c u i t o f t h is counter is shown in figure 5,1 1, The si qua I from the volta ge d iv id e r , in t h is case 1:1000, Is fed to atte n u a to rs . The s ttr m ia to rs have zener diodes to l i m i t the voItan e and to remove neoeti ve half wave sicinals. Roth attemuitors feed 7A,I tl monostftH» multivibrators via their Schmitt Trionrr' inputs. Thrsr are desi tii.f-d to o o e ra tr ai: 1 . c v o lts p o s itiv e no i >1 n le v p ls . The less sensitive input (hi oh attenuation) mu 11ivibrator con - tr o I s the o p e ra tio n o f coui\feii\« and {fttc h im i c i r c u i t On receivintj the impulse, this monostab I c starts a timina cycle which is de‘si F ig. 5.12 Sptu' I a I i sod cou nter For breakdown dv booti on 51 IateS preserving the BCD output of the counters brfore they are reset, This method avoids fjavirg to reset the counters manuaI Iy each tim e. The refinements in this circuit lie in the correct adjustment of the monostable circuits. The First monostable is designed to be fcri ggered on ly once by the impulse wsve- Fonn, it s i input be i nn below the t r ig g e r vo lta g e a t the end o f one tim in g cycle. The timinc) pe rio d can be ad juste d to be longer than the impulse decay period or shorter. If the counters are s till beincj fed at the end of the timimg cycle, then a consistent number of counts are obtained regardless of test voltage. Allowing a fairly unlimited period for this allows more counts to be recorded at higher tee voltages. The period of the second monostable has an imp .’bant effect on the d e te c tio n o f breakdown. The p e rio d o f the* pulses p ro duced are adjusted so that they allow the monostable to reset in time'to catch t>v following positive half cycle in the impulse. Should breakdown occur then the freouency increases and the monostable. beiiuj activated, itisses those h a lf cycles arriving wi titi n the pulse This reduces drastically th e number o f pulses l ik e ly to be counted when breakdown occurs. In addition, the faster rate of decay of the impulse allows less half cycles to excode the trigger level, The detector was tested with the attenuators both set to their most sensitive setting, with the time cycle set to 110 microseconds and the pulse I enqth to .1 microseconds. The wave form shown in photograph 1 was a p p lie d and f o r a ll test voltages gave a consistent count of UG. When the coil was fa u lte d to produce the wave form o f photograph 2, the count dropped to I, occasionally 2. Testa on other coils confirmed these fi quves and at lowed a powerful discrimination o f sound and unsound c o ils . The d e te c to r is simple to set up. With th e te s t c o il in position, a low voltage impulse is opplied. (A value of S K V i s suggested). The timing attenuator,.A, is adjus te d so th a t the numosl-tihIe is ju st a c tiv a te d a llo w in g a count to be regi Kteved. This can be seen when the d is p la y operates aftor* having been se t to zero by the manual rese t button. The timing monostable is then adjusted to obtain a 52 con ven ien t count, u s u a lly 20. The o th e r rnonostable ( B) is then adjusted to cjive pulses just narrower than the impulse o s c illa t io n p e rio d . T his is done by in cre a sin g the pulse period until the count suddenly drops to ha If its previous value. The adjustm ent is then hacked o f f by a convenient amount, u s u a lly ju s t enough to give • t " | | count. T est v o l tages may now be a p plie d. This detection method still has t disadvantage that a comparison must be made between an assumed sound c o il and the tested coil, in this case the sound coil being the test coil at a lower voltage. Care must also be taken with this detector that it is adequate Iy screened and supplied from an interFere nee-free DC supply, otherwise interfering signals From the H.V. circuitry will cause trouble. A Fur ther refinement to the circuit was contem 'ated but never attempted. This was to allow a delay of a Few microseconds in the enabling of the counters to avoid recording any of th e pulses due to a broken down c o i l . A sound c o il would give a Full count as usual. It was also thought possible to eliminate the decoders and readouts by having a simple go/no go system whereby a count oF more than 10, say, would cause a light to opperate. This would allow the use oF only one counter and a b is ta b le m u ltiv ib r a t o r , inste ad oF the d ig it a l elements shown, (see Figure S. 13)' Figure 5 .U tio/No go counter 53 F I g u i'v S. 14 S i mp I c Jc l e c t o r Anotlu-r simplt' mt-f-iivil o f Hi ven of F by f hr trst setup A louiirr sharper crack usually r e s u lts . However, t h is means o f d e te c tio n is fa r too em pirical and unreliable. It is also possible to observe the discharge directly in some cases. A red flash can be observed on the ends of the coils if they are wrapped with glass tape insulation. 55 5.4 AN IMPROVER TEST SUITABLE FOR FACTORY USE. The c a p a c itiv e discharge te s t describ ed e a r lie r has a number of drawbacks M iitinq its use as a routine testing procedure in a factory environment. These .ire 1. Lack o f any d e fin ite absolute method o f d e te c tin g breakdown 12. Potential physical danger to test personae, T his is unavoidable hie to the use o f a danger ous charged c a p a c ito r as well as the d ire c t handling of metallic contacts. 3. Changes in c o il types lead to changes in impulse freout ncy and a tte n u a tio n . This leads to s e ttin g up delays and expenses every time a new type o f coiI is tented. To overcome tin se problems an induced impulse method is proposed. This method uses a caoaci tor which is discharged i hrough a fixed coil which acts as the primary of an air core d f cans former, the secondary being the test. coil. In t h is way •‘•he I'reouenCv and a tte n u a tio n o f th e impulse is controlled ma i nl> H tin capacitor and primary coil and not to any large e xte n t h> th< te n ' c o il. Thus a large v a r ie ty of test coils may be impulsed witl-ou1 ll." indu ed impulse exceeding the tolerances specified for basic freouency and a tte m u if i on. The test c o il ine\ bt tui I v ani ca M y separated from the circuit containing the capacitor. Such a c i r c u i t has t|.« drawback th a t induced volta ge s arc* highly deoemlvnt on the coupling between coils A voltage measuring aysti m is reouired. which can be attached to the test coil, fo 1 1 s.sen the e fleet of this measurement on the primary circuii, it is preferable that this instrument have a high impedance, As the sv stern could be designed for go/no go •o n d iIi ons, a spark nap was decided upon The spark gap, set for the test value, brinks down when the test vol tage reaches this cornet value. The 1 sI vo lta ge is achieved by raising I.he capacitor discharge voltage in the primary c ir c u it by steps tint i I the spark gap breaks over on the 56 ocHco! production tester The c ir c u it to rnnduvt t li i ^ t r s t is shown in I'i cmrr 1 tt. I t is v v r\ si mi Im* to I In- or c. c STATISTICAL TESF^ OH FHL CAPACITOR DISCHARGE METHOD BpPoi'p the ceprtci for di fjolmi'iic iicHiod c I oti ili i h i to o hisih on i I foi I tiro rnto durim i the 2. I s I I I- i n l r r - f ui'ii i h-.ii I o I i m i m ol o r i u I w o o k e n rd in <3in w«i> In I ho I r w I ? Should the t o.sl' Criuso doiiuiuv I o r ho i nsu lot i on fln-n it is of little use for production to s tiu i. Such dumuuo may be due t o r h o m io u l c h u n n r s . m o c h .jii i ca I e I' f o r t s o r p a r t i a l i l i s c h .tru e .s , o r a com bi n A t t o n o f th e to d. How dors I hi’ test re I -iti to f hr s t .iildard mrtluid ol i mini 1st tosHim ? This mi thod employs the stan dard i -iioul yt- i \ . 'lc.' 1 on di sc cl i’d c o ils as dr - 1 ho answers t o t hose uursl i ons nin tin I s hr >|i vt n by a fill I st a t is t ic a l si ink o f I ho to s t usi nt) a lar’uo number o f m otor co ils , lln lo rt uuat v I s such an rx o rc i so would prove exfromelv cast-v ilue In I ho hi oh prodm'l i on oosl of hiiih voI tape motor c o ils . Such c o ils , besides ro u t ain in u a f a ir amount o f copper and expensive n is u la tiiu i m a te ria l, also use a I time amouul o f l.ihour, as I hr tapi nt\ is usual I v do no by hiiiul. Such a tent proeedum is o n li w iih in t hr ri sources o f .i I tir.ic iiiiihul act ure r and m u then i’i a \ he considered too expensive in just i t \ , i'u rl hrriiiort , tii f f i cui t \ can ho ex pect dl in produe i mi a I ,ii"ie numbvr o f eompl r I e I \ iile n tic a l c o ils , le s ts iisiiHi iic I mi I mol or t o i l s lirethus impossible in a low Inul-iri pro |r., I such as th is . A I i m i i i'll number ol coils are available I ruin manufac t u r e r s an d cumuli r c i a I m o to r r r w i n d e r x , I h e se a r e s p a r e c o i l s 59 uvpp dl- flip end o f ,1 motor w indino joh. When a I arn*3 motor is wound, more f o i l s 1 lion iivci'ssnr> ore produced. Should a coil he ddiiici cevjiuc.r«« -•^==M -4 • • • / '/ teu»u A s tu rd y , ivrl I insulutc-d m ulor u u il wds se le cte d from those aviii I dhl e (see figure S.Id), Tlie first, turn was split in the p o s itio n shown and two copper p la te s in se rte d . These plates acted as c-> I ectrodes with the turn potential across them. They wore care fu lly p o lis h e d and m aintaine d in a smooth condition to ensure consistent contact with any sam ples hvtwven them. The coil was used in a capacitive dis charge te s t con I i cpirat i on au in fi pure S. 7. I t pavr a healthy d ischdrqe as shown in phot opraph I . I lie c o il was placed on a wooden t.uhle one met re a hove tiround level in a h o riz o n ta l position, with the capacitor and sphere wap next to i t.„ Tlie split turn has eonveni entI\ exposed for insertion nf samples. Hie samples used should In ri >|lit s he actual samples o f i nte r-t.u rn in s u la tio n m a te ria l, iln fo rt unal e I y such samples arc d i f f i c u l t to oh fu iii and use. I he m a te ria l used u su a lly c o n s is ts a I an enamel laye r on t lie conductor fo llo w e d by a layer (or layers} ol pi ass fibres improuiioted with an a r t i f i c i a l re s in , A 1 rue sample o f such a m a te ria l must he s trip p e d from a conduct or and fla tte n e d before he i mi tested. T his would severe I \ duiu.uje the sample. The ideal m ate ria l foi these tests must bo readi I\ ol fainahle in flat uniform Sheets and .hou I d p re fe ra b ly la e l e c t r ic a l ly fra s iilo . I t was e v e n tu a li, decided to use paper, w ith due care he i no taken to ensure a co n sisten t nun Ii I v o f samples. The paper used was a phot ocopv i no paper p. oduced by the Company. I t is a hi uli ciua l i l y paper w ith \i ro iira p h i« coa ti n»i Such paper has a h i uh risistance to m ail" the e I ec.l rostat i c i ma \ ! 61 cool dry ploci- timl i-ovori'd w ith ,i dry c lo th to exclude d u st. The mixi nt| prove dun- wos ri-prutvd v i qorousl y once a day fo r a wrrk. Al the end o f t h is p e rio d I he samples were considered to have no rna Ii sed the wrothc e havinn hern c o n s is te n tly dry dur i up th is pi r i oil. I'nr i ruj the te n ts sump I vs were removed a t random I'rom the covered howl. T his was done one at a time usi nit tw erso rs t'o nrip the paper at the corner. This was placed betwien the e le ctro d e s which were under a si i i|ht vomprrssi on due t .> t h e ir shape and the na tu ra l spri ii'b ness ol the c o il tu rn s. I t was esl ahl i shed a I the s ta rt th a t -sanphs had breakdown values ol" about >' kV or hiuher measured over the e n tir e coi /, fhe te s ts th e re fo re hep.in a t 0 KV. Two d is - vharues at approximat e I\ 10 second in te r v a ls were imposed on the sample. The w i t a-ie was ra is e d hv ‘iOO \a I t s and the p r o cedure repealed, ,?0 serond.s I at er. The p o te n tia l o f the di s- ciianio was ra is e d in th is manner u n til breakdown occurred. The breakdown was di reel I\ v is ib le between the ele ctro d e s. f i v e different test conl'i M u ra t i ons w e re used, fach day f o r five days, ten s a m p le s were te s te d on each c o n fig u ra tio n and t lie te a ts were done in a d if f e r e n t sequence on d if f e r e n t days, The temperature re m a in e d between IS°C an d !iO°C du rin q all tin tests. After each of the croup of ten breakdowns pe r ro n f i tiurut i on. the electrodes were repo I i shed to esta b l i s h the same conditions for the n< V croup of t e s t s A number o f M a n ip les with deliberately introduced h o le s were also tested to establish I he breakdown of an air nap of the same separation as the thickness of the paper. It was found that flits ndK' breakdowns at a max t mum of .3 KV, ha I f of the lowest value w ith th e paper in position. The leafs done were:* f e a I I The c o il was a rr.line d so thal i I oresi nl rd the f i r s t turn to I he sample 1 in cunaciI\ to earth was mini mised l-s Inn i in as few 'e a rl h\ ' objeclw as p o ssib le f a p .ii i f \ I o e a r l h T ’.p I I iiduct ae >• of eo i I n) m i rrollenr i e s I esl I Similar in (.•<• i i-nt wilh coil arrauovd I o p re sent I aw 1 urn to sa.upl e , 62 Ti'Bt .! I-1 i’sh iii’<'ri<‘nrvU I o Siiipl**, wt Hi rin Ti’sl 4 Si m i I ,ir hn ii's f l>tit >• i t-li c o ils iii-rtiiKu H to p n • sent I ,ist t urn t o s.vipl i Pffi-I S 7lu stvu- ,i.s I vst I. Init M't- samp I r v.,is ili .suh.ii-urd t iini'a *ir n KV ut 10 svvond i nU’t’vdl » dv i op to In-iii'i tt-s lfil i ii I In* ustKil iiutniu'i*. Tliv rc s u I t ,ii'f suiiiiiitii' i sv il hi'iN '. A iiioi’i- ilv i N iim h vr o f s.ii'in I v s v!i 1 f s I s 1 u m l ’! . !-c i,u I r s w hi c li d i’f wi t.K i n 'I, r V. i n d i - ctiV i ii.| f.init tin i’c is I i I 1 I v Test 1 No e a rth i) No e a rth screen - g o , : 0,4 Test 3 First turn S c re e n e d 0,4- Test 4 Screened 12KV Breakdown p o te n tia l Figure 5.17 Histograms of initia l turn breakdown tests Liddi. 1 Test 1 S in g le te »t : F a « 0 ,4 Test 5 i . 25 repeated discharges; 1 0 2 4 6 8 10 Breakdown p o te n tia l Figure ,5.18 Histogram o f repeated disch arge te s t d wn voItage* docs not appear to Iuwrr the breakdown vol tage o f th e samples, The r e s u Its o f th e te s ts show very l i t t l e spread in the standard deviation of the readings. This can be inter preted as evidence of the consistency of the samples due to the careful preparation and testing of the samples. Although the material used in these tests, paper, is not a material used in motor insulation i b is nevertheless fr a g ile enouflh to sim u late the worst extremes o f motor in te r - tu r n in s u la tio n . Ii the te s ts done i t may be co n clu ded th a t weak poi ribs in in t e r - t u r n in s u la tio n have been sim u late d e f f e c t iv e ly . The te s t s have shown th a t the break down level for inter-turn insulation is more or less con stant regardless of the turn considered, provided the earth capacity of the coi I is minimised. It also appears that the di seharoc test causes little or no damage to the insu la t io n p ro vid e d s reasonable number o f disch arge s are im posed on i t . The t e s t can th us be s a fe !y used to te s t i n t e r turn insulation without causing the coil breakdown character istics to change. 6 4 down vo lta g e does not appear to lower the breakdown v o l ta ge o f th e samples. The results of the tests show very little spread in the standard deviation of the readings. This can be inter preted as evidence of the consistency of the samples due to the careful preparation and testing of the samples. Although the materia! used in these tests, paper. is not a material used i n motor insulation it is nevertheless fragile enough to simulate the worst extremes of motor in t e r - t u r n in s u la tio n . In the te s ts done i t may be c o n c lu ded th a t weak p o in ts in in te r - tu r n in s u la tio n have been sim u late d e f f e c t iv e ly . The te s ts have shown th a t the bre ak down level f o r in t e r - t u r n in s u la tio n is more o r less con stant regardless of the turn considered, provided the earth capacity . of the coil is minimised. It also appears that the discharge test causes little or no damage to the insu lation provided a reasonable number of discharges are im posed on i t . The te s t can th us be s a fe ly used to t e s t i n t e r turn insulation without causing the coil breakdown character is t i c s to change. I I 65 5,6 VARIOUS EFFECTS OBSERVED CONCERNING THE CAPACITIVE DISCHARGE TEST 5. h. 1 The d isch am e between tu rn s To observe the effects of the inter-turn discharge on the coil insulation a mini her of deliberately overstressed coils were di sected and examined. The ty p ic a l r e s u lts o f an i n t e r turn fIashover are shown in photograph 3. which is of a coil with the main ground insulation removed. Extensive blackening of the conductor covering has occurred. Thio is not due to carbonising of the insulation, but is merely a loose covering of fine particles as is found after an arc has occurred between metalic electrodes. Damage t. the con ductor Insulation has occurred only at the small regions where the arc has l e f t o r r e io in r d the m e ta llic conductors. It would appear that the arc has travelled between the ground insulation and the conductor wrapping, jumping across the thickness of several conductors. This appeared to be the normal procedure and no case was de te cted where breakdown occurred directly in the space of linear field confi gurati on between the conductors. A ll breakdo ms tended to e n te r and leave from the region of high stress at the radius at the edge of the conductor. A high proportion of breakdowns occurred in the overhangs of the coil. Not enough coils were examined to constitute a reasonable study. It would be o f in te re s t i f such a study were to be done as i t is certain that two facts would be revealed, These a r e :- 1. The overhang inter-turn insulation is weaker than the main insulation where the coil passes through the stator. This is possibly because the stator rcoion is compressed and heat cured causing voids between the edges of conductors to be minimised an-' removing th e lin e o f weakness between ground in s u la tio n and conductors. 'I. The i ntv r -tu rn in s u la tio n is damaged by th e c o il pul Iin g o p rra fI on. Having the ovvrlnimi region slightly weaker than the slot region is actually desirable when one considers that a break down w ith in the s lo t cou ld cause e xten sive damage to the u6 m otor iron. Because o f th is , accurate co n tro l o f overhang strength would hp an aid to efficient motor design. As men'1-' -nr d he t ore , i t appears th a t the in te r -tu r n di schargp •? usually between a nunber of conductors, A ll c o ils . di sp I ay i’d t h is cha ra c te r i a t i c. When te s ts were f ir e nun an a r t i f i c i a l c o il o I" s ole no id Form was used For fin d itiH tho most e f fe c t i vidt i mi th is situation Quantitatively yields an explanation. F i iiui'c S. ICoil so e tio n 67 Consider in I'inuri- ^. 1 ‘I, a coil having n turns and which breaks down over x burns when a voltage V is applied t o the c o il. Presume th a t the conductor In s u la tio n has a breakdown value o I' u v o lts and the weakness alo ng side the conductors, a breakdown strength oF b volts. No account is taken of' Field distortions decreasing breakdown fevt-ls. The voltage per turn is V ^ —— There For the vo lta ge across the tu rn is VX V, * “ V—— v o lts When breakdown occurs thio is eoual to the breakdown voltage a+(x-1 Ihtg i.e . Vx V -5 - - -’u H x - l )b thi re I'or-r V — + nb - ^ I'irKling max i nw and minima bv di ITerent i ati ng and equating to 0 i.e . when '2a -• b T his means t ha i the e o il has maximum stre n g ih when the v o l tage between turns is e c, l). 2 Noist' on the- d is c h ^ r tit- wevp Tor n Jom-s ohst-rv i-d in hi s studv t !ui o I iiriiv amount o f noi s<* ccuri’Vk! on h i » us-, i I I l r.HM*. His measurements were tikvn on th i’ fiii’i’i-nt wu vv lorni whicii vxh i I'i ts t'xe<*ssi ve oi s f of this n«i t him- , dt, is svi-n in pluitoijraph t-. I t is eeii th a t t liis noihC, vvhii-h is o f o h i >ih frc-out-noy n a tu re , i;cur>. a I Mk - < nr-i-cni m.ixi mo. I I i » i nlv r v s t i mi t.o co'ipare hi s u ith H h- volt.iui- wax t- fvnii us in photoijraph I. T his also xhi hi ts noi st* ti I I hotnih not rib qrv.it I > us thv c u rre n t wave- orm. This noi -.v is u I so voi ni-i di-nt w ith the peaks o f the ,i\ c Vorni. A llhouuh it .inin-.u's th .it th is noist* is provided > t lit- siiiiii- soui-t v, I hi s i h nut p o s s i! I v os i-ur-rcnt «md ol tu i|f -ire out o I' ulhisv I \ DO1’. 11 is ixi'vt-umi’xl I hoi I Ik- no i sv is dm- to cui-i-e nt or vo I - Hoc i't-\ i i'-. 11 in i 'u v i rcii i t . I In1 \ o I I U'iv noi no is Ciniard ' \ t ho f u i- i- i'iii n ,u h i m t i r n uh< n i in -.p o rk 6 RADIO FREQUENCY C O IL TESTING To achieve a reasonable test potential between the turns of a coil without drawing excessive current implies having a_ high impedance c o i l . An uni nscrte d m otor c o il has to o low an inductance to achieve t h is a t power freouenci es and a test waveform having high freouency components is reoui red. For example, a 100 mIcroHenry c o il would have an impedance o f 0.03 ohms a t SO ','z, whereas at 1 MHz the same coil would have an impedance o f oOOohms. To achieve a t o t a l c o il v o l tage of 1 KV would therefore respectively reoui re 33 KA or 1.7 A of current. The simplest test waveform to achieve this is a con tinuous sine wave of the correct freouency. This may con veniently be produced by an electronic circuit. Little information is available concerning this method. An extensive search of the literature failed to reveal references to any continuous radio freouency tests of solid insulating mat erials, A local motor winding firm, L.H.Marthi nusen Ltd. has a piece of eouipment of uncertain vintage which is de signed to perform such a test on induction motor coils. The values o f vo lta ge induced and freouenci es needed by t h is tester could not be ascertained from the user. The circuit appeared to be a single valve oscillator which coupled in d u c tiv e ly w ith the t e s t c o il. As i t had to be tuned fo r a mir, mum g r id vo lta ge a t resonance, i t is presumed th a t the circuit is very similar to that of the familiar grid dip oscillator used by radio workers to ascertain the reso-' riant freouency of circuits (sre figure h. 1 ). This test was applied only in special cases and even then appeared to be used more to discover inter-turn short circuits than to scientifically stress the insuIating mat e r ia l. The te s t was conducted by p la c in g th e te s t c o il w ith in the magnetic influence of the test circuit and then tuning the circuit to the self-resonant freouency of the coil by ob serving the dip on an ammeter connected to the g rid . Proper class C oscillations caused the minimum current to flow in the grid circu it. The supply voltage was raised until the desired potentials were achieved. If the coil Fig. 6.1 Grid dip oscillator Fig. 6.2 Mei ssener oscillator 220 VAC Fig. 6.3 Practical Me i ssener oscillator 72 flashed over internally then the resonant freouency and Q of the coil changed causing the grid dip meter to rise as the class C condition was lost. A te s t method such as t h is one has a number o f disa d vantages, There is a lack of preciseness in the induced voltage levels which make the test levels d ifficu lt to de termine, leadinq to the possibility of overstressino the insulation. If one induces very low voiteqe levels then this may not occur, but it negates the usefulness of the test in not stveaainq the insulation close to lim its which would reveal pote-; ally weak coils. As the voltage level is de pendent on tin vuiplina bet-ween o s c illa t o r and the te s t coils it can o-i.ously vary over a large range depending on the p ro x im ity v f the c o ils und th e ir o r ie n ta tio n to each o th e r ds well as to the number o f tu rn s on the te s t c o il, Another disadvantage is that the circuit must be re- trimmed for each coil so as to obtain the optimum qrid dip. I f the circu it were s Ma ( /M u - Ma))> - Mo Is the mutual inductance between the anode c o ll and the te s t c o il, is the mutual inductance between the gi'M 0 anti the te s t c o il. fj is the voltage amplification factor of the Rd is th<» plate resistancp of the valve R is the equivalent series resistancr of thf te s t c o il (anH o th e r coupled components) W is the annular velocity of the oscillations 73 W = 2 Tf f where f is th e frequency. Obviously a pr i ini' requisite fur oscillation is that - & r < / " Thus a largo // is an advantage; also a largo Mg. It would appear that a smalt Ma would be an advantage, how ever this conflicts with the other appearance of Ma in the equation and Ma is best maintained in terms of Mg. A low valu e o f , 11 ate resistance Ra is also indicated. Most im p o r ta n t, a itrgh frequency and low e f fe c t iv e s e rie s re s is tance in the test coil may be indicated. This is important under breakdown conditions as the effective parallel re sistance decreases suddenly i.e. effective series resistance increases. Oscillations w ill then cease when a breakdown occurs. A practical circuit employing this principle is shown in fig u r e (1.3. I t has the advantage o f a u to m a tic a lly reson ating the coil as we I I as certain safety features. The coil could be placed ; n the live circuit and oscillations would commence. I f the o p pe retu r was in m e ta llic co n ta ct w ith the coil conductor then sufficient effective, series resistance was intro du ced to damp th e o s c illa t io n s . The c i r c u i t had the disadvantage that oscillations wore not always produced, probably due to some or other infringement of the maintain- anco conditions above. Under those circumstances a certain amount of juggling of the test coil was required to obtain the required Mg. The tost also suffered from the disadvantage of not having a standard test frequency, independent of the coil being tested. As the breakdown potentials of solid insulations are very frequency-dependent, this could be a grave disadvantage where consistent testing is required. For t h is reason th e te s t c i r c u i t is on ly recommended for non-rigorous , fast checking of coils by unskilled per sonnel where a certain safety element is required. For rigorous testing a different approach is required. Although the principle of induced high frequency voltages is still required, to obtain the correct frequency stability a circuit must be designed which w ill oscillate independent- 74 Iy of the test coil. A powerful oscillation is required which w ill produce a strong high frequency magnetic field. A fte r .a number o f experim ents w ith several d if f e r e n t types of oscillator eircuits, it was decided to employ a series- fed Hartley oscillator. This is a simple oscillator to operate due to the strong out of phase coupling between anode and g rid . It. tends to employ v i r t u a ll y a ll- o f th e v o l tage swing a v a ila b le from the supply and is not e a s ily affected by outside influences. The theoretical circuit is shown in figure b.4. A description of the mode of operation may bo found in most electronics textbooks [4l] ■ A practical Hartley oscillator is shown in figure 6,5. The test coil i , .ice d in the proximity of the oscillator c o il and a su i ta b 'e vo lta ge induced in to i t . The o s c illa t io n frequency is controlled by changing the- coil (coarse con trol) or varying the variable capacitor (fine control). The induced voltage in the test coil may be varied by moving the. c o ils or more co n v e n ie n tly by a d ju s tin g the supply v o l tage v ia a V a ria c. The c i r c u i t shown is s u f f i c ie n t to ex- oito the normal type of induction motor col I to 2 KV, which for reasons outlined in Chapter 7 later, is more that sufficient to adequately stress the coil insulation. In all the circuits so far discussed the main difficul ties likely to be encountered are those connected with the measurement of voltage. What is required Is a device which presents little or no load and which is simple to operate, preferably of a go/no go variety. An oscilloscope with a high voltage attenuating probe fu lfils this condition partly and acts as a means o f checking any o th e r de vice. As in the pre vio us ch a p te r, i t was th ou gh t th a t using a spark gap would provide a ready means of achieving this detection. Unfortunate Iy at normal atmospheric pressure the distances involved tended to be rather small. This meant that any change in the e f fe c t iv e gap dimensions had a very marked effect on breakdown voltage. This included temperature effects, dirt on 1 he electrodes or electrode erosion. After several unsuccessful attempts to produce a reliable spark gap the idea was dropped and a d if f e r e n t approach adopted. A glass tube was designed with an electrode at either ! 75 Fig. 0.4 lltii’tley osci I I uhor Fig. 6,5 IKirtlvy osuillutor Pop coil testing 76 end connected to wires which passed through the glass. The tube was evacuated to a low pressure sufficient to ensure that breakdown occurred at the correct voltage. This could be varied by changing the dimensions or internal pressure. These figures could be roughly determined from Paschen's Curve (figure 6.6) taking into account that the curve applies to uniform f ie ld s . Such a i:ube was c o n s tru c te d as shown in fig u r e 6 .7 . The end e le c tro d e s were made o f pure nicke l sheet d e rive d from a vacuum tube anode and the lead out wires were platinum to ensure a reasonable gas seal. The gas- in the tube was air at a pressure of linm of mercury. When the tube was tested it was found to work well, with certain desirable properties. Instead of Breakdown at a set , voltage, there was a far more gradual effect. At a certain lower voltage, 1,6 KV in this case, a pink glow would be observed starting at the electrodes. It was found that thw pink glow spread with increasing voltage until at 2,2 KV it met in the middle whereupon a brighter glow was observed from th e gas as though more c u rre n t was being passed, breakdown being complete. This was a typical corona glow, as is often experienced at Ivi gh frequencies. Corona tends to occur readily at high frequencies due to the effect of ions moving backwards and forwardc under the influence of the alternating electric field. In a nor mal D.C. corona such ions may cause one o r two fu rt h e r io n is a tio n s be fo re being swept away o r c o lle c te d by th e e le c trode. In the case of a high frequency A.C, field the field would reverse before t h is cou ld happen and an ion can read ily move back to its original position. If the energy it acquires is sufficient to cause further ionisation then a belt of ionisation builds up readily around the conductor as the o n ly means o f ion loss is then by d r i f t or recombin a tio n . As the e le c tro n s are more m ob ile and th e re fo re s t i l l readily lost, recombination is not usually significant. Obviously ionisation will only build up where the field is sufficiently intense to cause the production of further i on pairs by collision. This is the case close to the electrodes. In the middle of the tube, although there are ions oscill ating in the field, they do not acquire enough energy to SWtf r-L ^ =•-< Fig Fig 6.6 PfiS> fiiesiOAe X SpAtLiMb «-»1AI Fic, 6 . 6 asz P .H£.n 's fo * /)/ a 78 € 4. Fig. 6 .7 Di echm-ge tube 1500 1000 500 5 10 15 20 Length oF glow cm F ig . 6 . 8 Discharge tube behaviour at 1,0 MHz 79 cause ionisation until the voltage is raised sufficiently. The discharge tube connected across the test coil thus acts as a voltmeter with the length of the discharge giving a measure of the voltiye according to figure 6 . 8 . This a llo w s a range o f measurement on each tube and lessens the number o f tubes re q u ire d fo r a large, range o f te s t vo lta g e s. There is also much to recommend d e lib e ra te ly design ing a tube with a very non- 1i near f i e l d arrangement so as to utilise this effeef and increase the range of measurement. U n fo rtu n a te ly t h is was no t po ssib le due to Iach o f tim e and fa cilitie s for glass blowing, this type of work having to be done by another deportment o f the U n iv e rs ity . The tube may be used to detect breakdown os in the prv.vi ous chapter. If an internal breakdown occurs then i n- suffi ci cut voltage will be available at the coil ends to cause th e tube to glow and Hie c o il is re je c te d . Such a means o f d e te c tio n lends i t s e l f tin ra p id te s tin g o f c o ils by relatively unskilled personnel. Provided that an effort is made to eliminate direct con tact with the oscillator circuit, this method is a safe one to operate. Contact with high voltage radio frequency vol tages leads to a flow of current through the outer layers of the body only, due. to skin effect. Thus the current is confined to the outer layer of the subject's akin, which has a high resistance. It does not pass through vital organs. At the point of contact a rather painful surface burn is experienced. To test this hypothesis tests were, done on a typical coil tost arrangement of the Hartley Oscillator ty p e . It was found that with the radio frequency energies required to test coils, voltages of up to 1 KV peak could be su staine d w ith o u t me h more than a small burn on the skin. Although .small, it was nevertheless painful and would cause any operator to immediately drop a live coil. Despite all the advantages of coif testing using high frequency voI I ages th e re s t i l l remains one m ajor stum bling block and that is that high frequencies interact with the insulation tested and transfer energy to it during the tost. This fransfer of energy may be in the form of electrical hysteresis or partial discharges. Either way sound i nsu- I a t ion may be damaged causing F a ilu re im m ediately or a t a much la te r date. T his means th a t high frequency te s tin g must be done very curcFuI Iy to avoid any s o r t o f damage to the insulation. Contrast this with the performance of solid insulations under impulse conditions where only the i n- trinsic strength of the insulation is of importance. With high frequency voltages, the breakdown is because of thermal o r discharge phenomena. For t h is reason the te s t method becomes c r i t i c a l as i t is n fiv o te d by th e a b i l i t y o f the insulation to eliminate heat. This will be dealt with prac tically in the following chapter. A theoretical prediction of R.F, heating is beyond the scope of this study. It would appear from these reservations that a radio frequency means o f te s tin g is not to be advised where the tests are applied to critical components. It may be con veniently applied to coils in a routine testing rolo where the R.F, voltages used are insufficient to damage coil in sulation. It should never be applied where the effects of such volta ge s on the p a rtic u la r in s u la tio n are unknown. 81 7 P ra c tic a l breakdown te s ts perform ed on a number of i iisu 1 a t i nci m a te ria ls . 7.1 I ntro d u c t i on A knowledge of the performance of real materials is required to effectively test coils using the five basic methods. Com p a ra tiv e te s ts have been done on a number o f samp I os using the standard impulse (1,2/50), a standard capacitive dis charge wave form and a continuous R.F. sine wave. Samples used included muter i a I in sheet form as we I I as portions of reaI coI I s. 7.2 Tests on real coil samples. Due to their high cost only a limited number of induction motor coils were available for test, these being supplied by a local motor re p a ir fir m . I t was decided to te s t two such c o ils , one in s u la te d between tu rn s w ith epoxy re s in and gla ss f ib r e , th e o th e r w ith a Ikyd cnamoI and gla ss f ib r e . To o b ta in the maximum use ou t o f such a lim ite d number o f coils it was decided to secti onaIi sc the coils, producing short lengths of conductors stiI 1 bound together by the ground in s u la tio n . These were then te s te d by a p p ly in g th e variou s tests between the conductor lengths, care being token to ensure that the inter-turn insulation remained intact and un damaged by t l v sample p re p a ra tio n . Core was taken to separate overhang and s lo t p o rtio n s of the coils as these receive differing amounts of hoot curing in the coil preparation and can be expected to hove different p r o p e rtie s . T h is was found to be so in p ra c tic e . Each sample (as detailed in Appendix E) was approximate Iy 25 cm long. 2,5 cm of ground insulation was trimmed from each end to facilitate the connection of test voltages to the con ductors. To avoid fIaahovcrs at the ends of the sample, the the conductors in this region were gently fanned out, one at a time as tested and insulated with a blob of insulating grease (Penetrol). Such a prepared"samp 1e is shown in fig ure 7.1. Light insulating oils (transformer oil) may not be used for this purpose as penetration of this between the con ductors will influence the results obtained. Although the use 1/200 4x HSK 10000 2 Ha. ^"aac Clreu!u bpcokcr Vorlao H V Tranoformof Ck'Oti'ostatic II V Capooltor 5*0,0125 CKO VOl tfflOtOP FIGURE 7.2 STANDARD IMPULSE CENERATOR 4xllSR10000 ->t-vWWv Clhtiult lirofikoi* EkotPooLotlc >-* vn/tmetop (V / Tp .mw Foriiii. Copnaltnp 83 o f t h i s grease makes the sample p re p a ra tio n messy and te d io u s it is necessary if consistent meaningful results are to be ob ta i ned. The disccted coils provided sufficient samples to allow for a standard 1,2/50 impulse test, an oscillatory impulse test, and a high frequency test at two different frequencies. One piece of overhang was available for each of these tests, along with cither one or two portions of slot section. In this way it was possible to compare these differing regions of the The sta nd ard impulse t e s t was perform ed on th e samples using the impulse generator shown in figure 7.2. As the sample capacity was a variable quantity, it was decided to swamp it out with an external capacitor of a suitably higher value i.e. 0,0025 microfarad. The resistances were constructed out of non-inductive zig-zag type resistance tape to the values shown* The t a i l r e s is t o r Rg was te rm in a te d in a small resistor which allowed the impulse voltage to be measured directly on an oscilloscope as we I I as permitting the wave shape and tim es to be a d juste d to the c o rr e c t valu es. The main c a p a c ito r, C^, was one o f those c o n s tru c te d and used by K r itz in g o r fo r h is work [38] . The sphere gaps, permanen tly attached to the capacitor, came from the same source. The te s ts were done by s e t tin g th e vo lta g e on th e capac itor to the wanted value and then closing the spark gap until it sparked over discharging the generator. This was done with an arrangement of linen tapes which rotated the movable sphere, screwing if in or out as needed. The initial tests were done s t a r t in g a t a low va lu e , 5KV, but were la te r s ta rte d a t 14KV as breakdowns were found to be o c c u rrin g a t above th is va lu e . The procedure used was to g ive two disch a rg e s, a p p ro x i mately ten seconds apart, then raise the voltage by one kilo volt and repeat the two discharges after waiting at least a minute from the previous discharges. This was done until the sample finally failed, care being taken to see that this was not just an external discharge through the insulating grease at th e sample ends. The insulation thickness was obtained by measuring, with a micromotor, the thickness of the copper and insulation com 84 binations as the sample was stripped down at the end of the te s t . T his e n ta ile d many measurements and was te d io u s . A la te r method adopted was to s t r i p the con du cto rs o u t, measure them with and without insulation and average the difference to give an average figure for insulation thickness in that sample. The o s c illa t o r y impulse t o s t was perform ed on th e samples according to the recommendations of Chapter 5. The circuit used is shown in fig u r e 7.3- The value o f th e t a i l r e s is to r shown gave waveforms o f the c o r r c c t decnv tim e . The ge n e ra tin g c o il employed was a spare motor coii with a high strength between tu rn s . I t was found th a t th e c o il and c a p a c ito r used gave a test frequency within the allowable tolerance outlined in Chapter S. The peak p o te n tia l o f th e impulse was v i r t u a ll y that of the stored value on the capacitor. The test procedure was s im ila r to t h a t o f th e p re vio u s te s t. The sample was d is charged twice at each value before waiting a minute and then dis c h a rg in g a t one KV h ig h e r. The in s u la tio n th ic k n e s s was measured as be Pore, The samples were ra d io -fre q u e n c y te s te d w ith th e apparatus shown in figure 7.4. A variable II,V, D,C. supply fed a series Harts Iy oscillator which supplied on R.F. voltage to the sam ple. The entire arrangement was housed in a Faraday cage made of copper sheet to prevent a leakage of radio Interference. Considerable expo ri men to t ion tvaa noccssary to obtain the correct combination of col I, capacitor and coil tapping p'-int for efficient oscillations to occur. Furt'i-sr adjustments were needed to obtain the correct frequencies used namely 1 MHz and 2,5 MHz. Care was also required to avoid 'squeegi ng' (unstable bursts of oscillation) due to too high a grid ca pacitor resistor combination. It was also found that an amount of corona discharging took place from exposed small radii su rfa c e s . These had to bo covered over w ith crumpled aluminium f o i l to ra is e t h e ir o f f e e t Ivc d ia m e te rs. The t r i ode used required a supply of cooling air which was provided by a high du ty fa n. Initial tests indicated that the test circuit provided a t e s t waveform which had a peak valu e o f alm ost th e same leve l as the vo lta g e pro vid ed by th e D.C. sup ply. This was 85 9A Vnrlae Trdnaformer FIBURE 7.5 RADIO FREOUENCY TEST SET (LOW VOLTAGES) 86 determ ined by exam ining th e te s t waveform usin g an o s c i l l o scope with a high voltage probe, used up to 10 0 0 volts, the lim it of safe operation for the probe. A consistent difference o f II0 v o lt s between peak value and supply v o lts was in d ic a te d . For values o f vo lta g e ove r 1KV t h is was assumed s t i l l to ho ld and the peak o f th e te s t waveform was assumed equal to the supply voltage less this value of 50 volts. Core had to be taken to avo id o ve rh e a tin g th e sample be fore breakdown occurred. The method adopted under those con ditions was to raise the test voltage at a specified constant rate until breakdown took place. The rote of increase is approximately that to give the average breakdown value in te n seconds. Breakdown is in d ic a te d by a sudden drop in supply vo lta g e as well as noise from th e sample. I t was found th a t little separation of the ends of the conductors was required as the breakdown of the solid insulation matched that of air. The th ickn e ss o f th e in s u la tio n was measured as b e fo re . The condensed results of the tests on the coil samples are shown in ta b le 7.1, These are discussed la te r . The o r ig inal readings arc presented in Appendix E. 7.3 Tests using shoot samples of insulating materials It was decided to tost two modern materials, often used for in t e r - t u r n in s u lo 'J o n in sp e cial a p p lic a tio n s such as on hand toped overhangs. These wore Kapton, a polyimidc film and Nomex M, a crushed mica reinforced felted poI am I do fibre paper. Both materials ore available in the foi m of thin sheets and are typical of high performance modern insulating materials. The Impulse te s ts were a p p lie d as b e fo re , usin g th e same equipment. The sample, in sheet form, was placed between two brass electrodes. The bottom electrode was a fla t plate. The top electrode was a cylinder of 25 mm diameter and 25 mm long with its edges rounded off to a radius of approximately 2 mm. This rested under its own weight on the sample. The rad io freq ue ncy te s t was perform ed usin g a lower voltage version of the surios-fed Hnrtcly oscillator used before. As the currents and voltages involved were lower, t h is cou ld use a sm a lle r c o il a n d power supply as we I I as a low power n a tu ra lly coo led t r io d c . The c i r c u i t is shown in fig u r e 7.5. In t h is case a reasonable number o f samples 87 were avai la b le and a much la rg e r number o f fre q u e n c ie s were used. This gave the results in the form of a curve (figure 7.6). The rest of the results are presented in Table 7.2. The original readings are available in Appendix D. TABLE 7 .1 Breakdown results from tests on dissected motor coils. COIL 1 - 6 , 6 KV EPOXY DOUBLE GLASS 7 tu rn s , co i I uut to 12 p ir c c s . C oiI breakdown = 35 KV REGION IMPULSE OSC. IMPULSE R.F. TEST R.F. TEST TEST TEST 1MHZ 2 ,5 MHX CORE 378 340 49,0 44,7 KV/cm KV/cm KV (pk)/cm KV (pk)/cm OVER 239 256 28,7 31,3 HANG KV/cm KV/cm KV(pk)/cm KV(pk)/cM COIL 2 - 6,6 KV ALKYD DOUBLE GLASS 9 turns, coi I cut to 10 pieces. CoiI breakdown = 31 KV REGION IMPULSE OSC. IMPULSE R.F. TEST R.F. TEST TEST TEST l-MH., _ 2,5 MHk CORE 643 58 8 76,9 75,1 KV /cm KV/cm K V ( p k ) / c KV (p!<)/cm OVER 505 413 58,6 75,0 HANG KV/cm KV/cm KV(pk)/cm K V ( p k ) / c TEMPERATURE - 23°C - 27"C I TABLE 7 - 2 Breakdown o f d is c re te in s u la tio n samples. 'KAPTON' p o lim id e f i hi. Average thickness = 0 ,03mm. IMPULSE TEST OSC. IMPULSE R.F. TEST R.F. TEST TEST 1MH_ 2,5 MU, MM 2300 180 165 KV/cm KV/cm KV(p!<)/cni KV (pk)/cm 'NOMEX M' mic i leaded felted polamide fibre Average thick less = 0 , 14 mm IMPULSE TEST OSC. IMPULSE R .F . TEST R.F. TEST TEST IM I^ 2,5 MHz 650 596 122 108 KV/cm KV/cm KV(pk)/c.n K V (p k)/cm TEMPERATURE = 20°C - 25°C. (sjeod |e!)Uoq.o,j u^op^eajg 90 7.4 Analysis of the- tost results A number of trends are obvious from the results of the break down te s ts , 7,4.1 Impulse tests it can be seen that the core sections of the coil have better breakdown values than the overhong sections. This is due to th e b e tte r c u rin g and com paction t h a t fc/i/s region receives when the c o il is m anufactured. There is also more chonce o f damage to the'overhang region during coil forming. It can be seen that little difference in effect exists between the two impulse tests though it would appear that the oscillatory impulse test is harsher. Breakdown values with it are about 10 per cent lower than with a standard impulse test. It is of intei-est to notice that both coils would only sta nd about 30 to 35 KV when su b je cte d as com plete c o ils to the oscillatory impulse test. When individual sections of coil are subjected to the impulse tests they appear capable of withstanding about 20 KV between individual adjacent con ductors. If it is presumed that the test voltage per turn is almost the same irrespective of turn position then it would appear to be an anomaly that the entire coil is weaker than any of its components. There are perhaps two explanations o f t h is phenomena. Firstly account must be token of the statistical proba b i l i t y th a t a weak p o in t e x is ts somewhere in th e c o il which w ill cause a breakdown to ta ke pface a t a much Zoiver vo lta g e . Obviously just one such weak point is sufficient to fail the entire coil. The probability of this weak point having a c e r ta in breakdown value depends on two fa c to r s , normal s ta tistical variation of breakdown values and the chance of an unusual weak point in the insulation. The latter woufd.be the case with on occluded metal particle or void or some such similar fa u i t , and is d ifficu lt to predict re)i ably. To predict the normal statistical variations of coil sec tion breakdowns the results of all the impulse tests on coil segments wore lumped to o b ta in a s in g le histogram . T h is could be done as th e samples wore a ll a p pro xim a tely th e same length and i t was f e l t th a t as many readings as p o s s ib le were needed to obtain a reliable curve to predict the statistical proper 91 ties of coil segment breakdowns. The individual results of the breakdown tests were normalised by dividing them by the average for that test, taking into account that the overhang and core sections were different, For the purposes of drawing a histogram an interval of 0,05 of the normalised value was chosen. The histogram is shown in fig u r e 7>7* To p r e d ic t th e behaviour of coil segments it is necessary to know mathemati cally how the histogram varies about the average value. For this purpose the root mean square of this variation about the mean was computed as in the Appendix E. This allowed a nor malised Gaussian probability curve, with the same interval (0 ,0 5 ) to be p lo tte d over th e histo gra m . From t h is can be seen that there is a fa ir amount of correspondence between the two and i t is assumed th a t th e breakdown o f c o il segments fo llo w s th i s di s t r i b u ti on. Assuming that the samples obey this distribution, a sta tistical prediction can be made for certain failure rates (see Appendix E). For coil 2 the one per cent failure voltage is 54/4 KV with a 50/% failure at 8 6 , S KV. As these are much higher than those in practice, a di fFerent explanation of break downs is fo rth co m in g . A far more successful approach is that based on the field situation which exists within the coil and on the voltage distribution between coil conductors where flaws exist in the insulation. Th i s has already boon, partially discussed in Chapter 5, where it was observed that breakdowns tended to occur between conductors which were not adjacent. The vo lta g e d i s t r ib u t io n between con du cto rs and i t s in fluence on breakdowns in the region down the sides of the conductors is complex depending on conductor shapes and in s u la tin g m a te ria l d is t r ib u t io n s . This becomes even more com plex when voids in the insulation are considered especially as such voids are unpredictable in extent. The situation down th e side oF the conductors may be reco gn ise d as a hazardous one, it being a region oF mechanical discontinuity which is at right angles to a voltage gradient. Such situations are avoided in high v o lta g e p r a c tic e as exp erien ce has shown them to be dangerous. This dangerous situation may be analysed by making some 93 assumptions concerning the nature of the coil edge region. 1. An air discontinuity exists between the turns in s u la tio n and th e ground in s u la tio n and extends along this entire region. 2. The s o lid in s u la tio n has the same p e r m i t t i v i t y and dielectric strength over the whole region. 3- The tu r n in s u la tio n has the same s tre n g th as th a t ascertained in the tests, i.e. two separate layers of conductor covering have a combined strength equal to that of an adjacent double layer. 4. The field distribution is that which would occur in s o lid in s u la tio n . The d is c o n tin u itie s are assumed to be insignificant in size and unlikely to dis- .t o r t th e f i e l d a p p re c ia b ly . Likew ise i t is assumed that stress concentration does not occur due to the differences of perm ittivity that occur between in s u la tio n and voids. A coif of cross-section si m i far to that of coiI 2 was considered as an examp Io. Only four turns i.e. five conduc tors were considered. A potential plot was obtained in two dimensions using a con du ctive paper model, T h is is shown in fig u r e 7*8. Any d is c o n tin u itie s between tu rn in s u la tio n and ground insulation would occur along the dashed line which indicates the join between those insulations. The conductors were raised to progress!voly higher voltages as shown. From t h is analogue model may be deduced a number o f fe a tu re s : - 1. High stresses occur in the regions marked A. Break downs are l i k e l y to bo in it ia t e d in void s in t h is re g io n . 2. Low stresses are present at poi nts marked B. Breakdowns are unlikely to start at these points. It may precede through these regions if the stress pattern is distorted by brcokdowns in region A. As the conductors arc subjected to pressure when curing occurs, it may be surmised that a certain amount of material fills the region C where the conductor coverings join. The actual join between turn insulation and ground insulation is thus likely to follow the dotted lines in this region and 4xV 3xV 2xV ScXlE 1:20 Figure 7.8 Field plot 95 avoids the cleft at C, which could give trouble. Measuring along these new joins gives a stress of three potential lines in 0,75 mm for the highest stressed regions. A, and about one in 1,0 min for the low stressed region. This indicates a maximum stress, o f appro xim a tely ^ V over 0,75 "iM and a minimum s tre s s o f Coil 2 broke down fit 31 KV for 9 turns, giving an inter- turn voltage of 3,44 KV = V Thus the maximum stress is 2,1 KV over 0,75 mm Examining Paschen's curve for air (Figure 6 . 6 ) yields a breakdown vo lta ge o f 2,2 KV when pressure = 625 mmHg and d = 0,75 mm It would thus appear that local breakdown in the regions of high stress would lead to a flashover between conductors due to f a ilu r e down th e in s u la tio n j o i n t . E v e n tu a lly a high enough voltage would bo imposed across two of the conductor insulation layers to couse them also to fa il. As this is like ly to occur at a value close to that of the impulse tests, i t may be surm ised th a t a number o f c o il tu rn s must be spanned to a IIow th i s. In the case o F co i I 2 thIs wouId be at 16,1 KV corresp on din g to th e v o lta g e across a minimum o f 5 tu rn s . The f i e l d p lo t done did not ta ke in to account th e lim ite d thickness of the ground insulation around the coil. This is, however, at least ten times the turn insulation thickness and o n ly has e f f e c t a t the edge o f th e f i e l d p lo t where the stress is very low. The field at the stress points is virtually unaffected, Similarly with the coil placed in position in the motor core the field plot would appear different. It would not, how ever, greatly affect the separation of the potential linos. Even though they would appear in different positions they would s t i l l tend to have the some s e p a ra tio n and hence the same stress relationship. From the f i e l d p lo t a s o lu tio n to th e in t e r - t u r n weakness problem is obvious, if the i ntcrfoce between the two i nsu I a>- 96 tioiis could lie mode, to occur furthor envoy from the conductors, say 1 mm away, then it would occur in a region of virtually uniform stress. As the stress in this region is half of that at A, the inter-turn breakdown could be doubled and it may even be found to be limited by the statistical probability of failure directly between turns. A practical method of achiev ing this is not obvious. The answer may bo found in a technique of impregnation of a relatively loose fibre covering around the conductors. Such 'a covering (glass fibre), being loose, could easily be compressed to form the direct insulation be tween turns of the correct thickness, but would retain its looseness at the edges. Those edges could then be made elec t r i c a l l y stron g by s u ita b le i mprcgnati on and c u rin g . 7.4.2 Radio Frequency tes.s. Although a limited amount of information is available on R,F. breakdown in gaseous dielectrics [ 4 4 ,4 5 ,4G3 , an exten sive search for information on R,F, breakdown in solids revealed o n ly one ra th e r dubious re fe re n ce by Peek [ 4 7 ] , I t would appear that the two breakdown processes are completely d iffe r ent. In the gaseous dielectric case the breakdown is charactor- i sod by the behaviour of mobile ions. When o certain transition frequency is reached, correspond!ng to the transit time of the ions, then ions are allowed to build up in the i ntcsr-o I ccerode space and breakdown occurs more r e a d ily . As p o s itiv e and nega tive ions have different mobilities there are two such tran sitional frequencies, a f a i r l y low frequency one due to p o s i t iv e ions and n hig h Frequency one due to ne ga tive ions, Tho ease with which R.F. voltages cause corona discharges is a further man I fei,tat i on of this effect. In solid insulation the ions are not mobile and do not observe these, transitional effects. As can be seen from the results for Kapton and Nomex M, the R.F. breakdown voltage is very much lower than the impulse strength. This can be . best explained by the ability of the. R.F, voltage to release ions from tho normally stable lattice of the dielectric ma t e r i a l . T h is cun be done in two ways, d ir e c t e x c ita tio n and temperature effects. Nothing is known of tho first effect. The second e f f e c t is from losses in th o d ie le c t r ic due to hysteresis. Raising the temperature of a dielectric causes 97 its breakdown value to drop appreciably. It can thus be seen that the R.F. breakd '> . of a sample is great Iy affected by its ability to gonorote and dissipate heat. This depends great Iy on the physical arrangement of the specimen as well as th e presence oF p o la r im p u r itie s , such as m o is tu re ,.' The exa ct p r e d ic tio n o f R.F, breakdown i s d i f f i c u l t and is best done empirically. This is a further disadvantage of th e R.F. method o f i nt o r - t u r n t e s tin g . Also th e mechanism of breakdown is not a natural one and leads to a situation where, breakdown is dependent on fa c to r s such as heat d is s i pation. More important is the fact that as the relative R.F, breakdown strengths of air and solid are sirnilor the effect of voids is not critical in producing breakdown. For this ' reason a coil passing an R.F. tost can never bo relied on under surge conditions if it is likely to fail in the manner described previously, namely via voids in the insulation interface region. 98 8 . CONCLUSION It is recognised that a need For surge resisting motors exists. It is possible to protect ordinary motors using capacitors and surge arresters as outlined in the references (3,4 and 12). This is not p ra c tic e d as much as i t should be, find would be even more effective if used with motors which are designed and tested to resist surges. Such motors are indicated for applications where surges are l i k e ly to be troublesom e as is th e case when th e motor is in an exposed p o s itio n , or when vacuum c i r c u i t - breakers are employed. When th is is combined with the need to maintain continuous operation the need is most critica l. This includes such operations as pumping, winding, ventilating and material handling at remote stations. Even within a safe system such m otors should p ro v id e more p r o te c tio n where used fo r p ro cess, refrigeration or production applications and for essential s e rv ic e s . Where a • ;.i =re m otor is on hand i t u s u a lly ta kes a prohibitively long time to change over if one should fail elec t r i c a l I y. The decision to apply i ntcr-turn tests to induction motors during manufacture is-, always, a financial one. The tests cannot be applied to each and every motor produced without raising the general price of the motors. This w ill place one at a disadvantage with regard to one's competitors in appli c a tio n s where surge re s is ta n c e is not demanded. Likew ise the limited production oF surge tested and resistant motors must be accompanied by a campaign to inform customers of the advan tages of using such motors. This supposes the possession of statistical information on motor failures, something which the manuFecturers are better able to gnther than anyone else. It may also be of advantage to produce a number of different classes o.! surge resistant motors, th- aim being to minimise costs in relation to the application intended. The most ex pensive and r e lia b le motor would be one using a ll te s te d c o ils , with the fu ll 20 KV per microsecond per KV of motor supply as detailed earlier. This would give a certain test voltage depen dent on surge p ro pa ga tion v e lo c it y , size and number o f tu rn s of the particular motor coil. An intermediate class of motor may be arrived at by specifying tested coils only at the c riti- 99 cai Ii ne end of the motor. Alternatively full rated coils may be used in t h is p o s itio n w ith h a lf ra te d (10 KV) c o ils used elsewhere, or half rated coils throughout. The full rating referred to is based on a study of currently 'iv i i e information on lightning surges. Most of bhis i nfor- f.kV •o n is American in origin, it is fe lt that a thorough study of fast lightning impulses on South African power systems would be useful in clarifying the position. Work is also indi c ated on vacuum c i r c u i t breaker, induced surges and t h e ir e f fe c t s on m otor wi ndi ngs. I t is f e l t th a t m anufa ctu re rs have l i t t l e to ga in by an exh a u stive stu dy o f surge p ro pa ga tion phenomena in motor c o il assemblies. As is shown, such a study is complicated and can be liable to errors out of proportion to the effort involved in com puting them. A s im p lif ie d model, as proposed, when com bined with a practical study of the actual coil yields useful results which are able to be utilised in design. A sound basic principle is to take a course of action which limits the trans mission line behaviour of the coil and which encourages the energy of the surge front to be dissipated uniformly and sw ift ly over the entire coil. On this basis the coil requires low capacity to ground, high inter-turn capacity, low self induc tance and high mutual inductance between turns. These properties are in conflict with the use of thick insulation between turns itind, as in any in s u la tio n design, th e re is an optimum th ic k n e s s of Insulation which gives the maximum potential withstand with in certain physical consbraints. The radio frequency method of coil testing is subject to the limitations outlined In Chapter 6 and should on ly be used where p e rd itio n s demand a safe, s w if t and non-exact method o f inter-turn testing. It is fe lt that by applying a little thought the: impulse method may be s im p lif ie d and speeded up t o make i t ;• tit as easy to ap ply as th e ra d io frequency method. Where a number o f c o ils are t o be te s te d then t h is becomes j u s t i f i e u, To aid effective design of coils and to prevent failure due t o surge Induced breakdowns between tu rn s , more is re q u ire d t o be known about th e actua l breakdown process. As has been demonstrated, knowledge of the breakdown strength between turns does not imply a knowledge of the surge strength of the coil. This can be much lower than the inter-turn strength would lead one to believe, due to the stress concentration in the critical interface between turn insulation. The effects of differing insulating materials may be particularly significant. The only effective method of analysing this is by field plotting fol lowed by a practical breakdown test to confirm one's findings. One o f the aims o f any p r o je c t such as t h is one is to provide design materiat and improved methods of construction and testing of motors. Although hampered by a lack of material i t is hoped th a t th e methods employed here are an a id t o th e production of improved induction motors for South African con d itio n s . 101 APPENDIX A Characteristics of Lightning Arresters 1, American P ra ctice The fo llo w in g ta b le was taken fro m :- AI EE Lightning Arrester Subcommittee "Station-type Lightning Arrester Performance Characteristics" AI EE Trans, Volume 59, June 1940. pp 347-348 AI EE wave is a wave r is in g a t the r a te o f 100 KV/m i crcrecynd for every 12 KV o 4 rrester rating until sparkover occurs. A rre s te r Sparkove r Voltage 1 ,R Di scharge Voltage Rating 1,5x40 wave A1 EE wave KV KV rms KV KV 1500A 3000A SKA 10KA 20KA 3 10 13 9 10 10 11 12 6 19 23 18 19 20 22 24 9 27 35 27 29 30 33 35 12 39 43 36 38 40 44 47 15 49 53 45 47 50 54 59 20 66 72 60 64 67 72 78 25 78 89 74 79 83 90 10 0 2. European P ra ctice I EC Publication 99/1 2nd Edition A rre s te r Sparkover R ating KV KV 10 KA rate d a rre s te rs 3 13 6 22 ,6 10.5 38 12 43 Note T his is fo r wave s im ila r to AI EE wave. 102 3. A u s tra li an P ra ctica From A u s tra lia n Standard 0338 - 1965 Surge D iv e rte rs A rre ste r 1.2x50 wave Wavefro nt test Rated KV KV Sparkovei' Front steepness Sparkover KV/msec KV 3 13 25 1.5 22.6 50 26 10.5 38 87 44 12 43 100 50 This is for 10KA rated arresters. 4. South African practice There is no S.A.B.S. standard as yet although one is i n pre paration. The ESCOM specification EED 8/3 is given, which w ill be the same basis of an eventual S.A.B.S. specification. Arrester 1.2x50 wave Maximurn Rated sparkover permissabie KV per unit of lead drop rated volts KV 25 3.3 15 25 to 150 2 . 8 25 150 35 Also given is the Impulse Coordination Specification. It is generally fe lt that those are slightly high. 103 IMPULSE COORDINATION- ESCOM SPECIFICATION Nornina1 A rre s te r A rre s te r Protect!ve Required Past Proposed system sparkover externa 1 Ei.COM ESCOM ai r o s t t r BIL, KV e x te r BIL ^ K V 696 V°KV J impu1se na 1 KV BIL, swi tc h in g KV s urge, KV 3.3 3.6 27 41 54 45 6.6 7.2 24 39 59 80 75 11 40 55 84 102 95 22 24 80 95 144 160 150 29 82 107 163 215 200 33 30 101 120 191 110 205 44 270 250 i : 135 160 243 58 188 285 163 375 350 75 204 229 347 78 408 380 88 219 (97) ( 2 7 2 ) (297) (450) (485) (450) 132 110 325 592 550 (145) (406) (431) (653) (700) (650) 154 136 382 407 617 700 650 220 196 510 545 826 900 825 275 240 624 659 1000 1130 1050 330 290 754 789 1200 1300 1300 400 336 873 908 1380 1550 1425 500 420 1093 11 28 1710 - 1800 725 612 1590 1025 2470 - 2550 NOTES: 1. Diita Tor* p o s s ib le T'uturo v o lta g e s oF 500 and ■ 725 KV is pre I I mi nary and is given fo r guidance in considering possible applications of these voltages. i 2. Bracketed values are fo r f u ll in s u la tio n and i 100,’'. a rre s te rs and wi I I not be used on now t equipment unless the system is not effectively ea rthed. - j J 1 0 4 APPENDIX B ;. 3. Theoretical capacitor divider chain -ihPfirWx T T T T K = capacity between turns per turn C = capacity to ground per turn & - Sim i Ia r I y d l = JVwCdx dx ™ JVWV Therefore ' i f - 10 5 This is a differential equation with general solutio of the form V = Ae9x + Be"9X ..... 7 where A and 8 are constants Using the fact that V = 0 at x = 0 and V = U at x = n, where n is th e t o t a l number o f c o ils , and s o lv in g fo r A and B y ie ld s a s o lu tio n U 2 5 inh ng „ Si nh Si nh ng If the coils are numbered from the high potenti a I end then the potential on coil m is V = I] SI nh (n-iii)g Q Si nh ng •«■•• B.2 Theoretical turn potentials due to electrostatic field Coi I 1 has t L . • ng param eters k .94 pF C » 219 pF g ^ 0 , 8 6 allowing tho approximation U1 “ U — ~ U e " 0 , 8 6 - C.43U S im !la r iy This does not agree w ith the measured potential. . 10 6 APPENDIX C Cl Model c o il dimensio:-- and param eters Coil consists of ten turns of 21 SWG copper wire sus pended in a ir above an aluminium f o i l e o rth p la in as shown. Mode I coi I No 1 ’*f~ Figure C,1 Cross section of corf Figure C,2 Plan of coil Coil parameters as measured on General Radio bridge 219 . 5 5/«H - Pf L( o - l) 218 pF . 5 5yUH Cl- 2 /E ' 1 ( 1 - 2 ) 270 PF 172/uH C O -I/E " "-(0-2) 50 Z H °0 -1 0 /E " 395 PF L(0-l)+(2-ir 294 Pf 8 2 5 ^ H C0 - l / l - 2 " "-(o-s) 2 , 9 mH C0 - l / l - 10* 342 Pf L(0 - 1 0 ) Surge impedanc< i o f s i n g le tu rn = 500 ohm s ( c a l c u l = 500 o h m s Mode I c o i l No 2 Figure C.3 Cross section • Mode 1 c o i I No 2 an as in figure C,2 Coil parameters ( GR b r id g e ) = 221 55 C0 - 1 / E PF L0 --1 = « 225 pF . 1 7 5 /-H ° l - 2 / E Lo-•2 “ 4 0 5 PF 83 5 C0 .-1 0 /E Lo-•5 = = 3 4 2 C0 - l / l - 1 0 pF l 0 .■10 = 2 , 8 4 »H = 29 4 PF C0 - l / l - 2 Surge impedance of s i n g 1e t u r n = 500 ohms ( Iculated ) 360 ohms ( measured ) Model coi 1 No 3 •in n irrrrm m w - Figure C,4 Cross section - Mode I coiI No 3 Pi an as i .i f i gure C. 2 Coil parameters ( GR bridge ) 0-1/E "• pt' 1 - 55/uH = 3 , 2 5 i»H °0-10/E “ 349 PF 1 Surge impedance of single turn = 500 ohms { calculated ) M od oI c o l I No 4 Figure C.4 Cross section - Model PI an as in figure C .2 Coil parameters ( GR bridge ) C0-1/E “ 932 pF LL0-1 - - 53, C0-10/E * 8360 pF 1-0-10 " ?" W "H Surge impedance of turn «= 127 ohms (calc) = 130 ohms (meas) C . 2 Rea I m o to r c o l I Manufacturer - GEC Motor - 1272 KW ( 1705 HP ), 6600V, 134A, 990 RPM, 50 Hz Slots - 72, 13,5 mm x 77 mm Conductor - copper, 9,0 mm x 4,5 mm ( bare ) 9,5 mm x 5,0 mm ( covered ) Insulation - Class B - Polyamide enamel + double glass and epoxy resin Coi I - turns - 6 per coi I length - slot = 930 mm overhang = 420 mm turn = 2,70 m coi I - 16,20 m Inductance = 260 /jH ( in motor ) = 67 / j H ( i n air ) Capacity ( cut I to stator ) = 2400 pF Four coils used wore adjacent and overlapping each other 10 9 APPENDIX D Analysis of sim plified impulse generator v T h e r e f o r e Oi fPereni-iai-i ng th is express! o + d t of the form D* + CF + TC ' a n d h a s a so I x it i on 0 - » - / h L ^2 - is In tho underdamped case IF L e t t : n g a *l,d b - |< h ) 2 + t o 110 gives a solution of the form V = ea*(, Kj. Cos bt + Kg Sin bt ) If the voltage on the capacitor i s Vc at t == 0 then * 1 - ? . and i f V = 0 at bt = then -Kg = 0 T h u s V = Vc eat Cos bt In terms o f the frequency of osci I I at I on b = 27ff Therefore if V 4 ^ LC APPENDIX E 1, Reading of Impulse Teats C o N 1 Di amond coi I - 6, (> KV 7 turns of 3r5xb,2 mm copper Insulated w ith cpoxy-s •' ass between turns and epoxy-mi ca- glass to ground. Inductance — 72,5 j j H O scillatory impulse test Failure = 35 KV Coil cut into 12 pieces as shown 1 1 2 1(a) Standard impulse Test I EC Standard 1,2/50 impulse Temperature = 2‘)<>C B re a kd o w n A v e ra g e Norma 1 i sed Di F f 2 T h i c k - (KV) B re a kd o w n B re a kd o w n 2 2 ,0 0 , 9 5 0 ,0 0 2 5 22 , 5 0,99 0 ,0 0 0 1 o v e r 2 4 ,5 2 3 ,2 KV 1 ,0 3 0 , 0 0 0 4 A v . ha ng 2 3 ,0 = 2 3 9 K V 4, 0,99 0 ,0 0 0 1 «a 0 , 97mm 2 1 ,0 0 ,9 1 0,0081 2 0 ,0 1 ,1 2 0 , 0 1 4 4 2.1,0 0,97 0 ,0 0 0 9 2 3 ,0 0 , 9 7 0 ,0 0 0 9 2 1 ,0 0,89 0 ,0 1 2 1 2 1 ,5 0 ,9 1 0,0081 2 o , 0 1 , 1 0 0 ,0 1 0 0 e o iv 2 7 ,0 2 3 , 0 5 KV 1 , 1 4 0 , 0 1 9 6 A v. 14,0 *378 KV/cm 0,62 0 , 1 4 4 4 =» 0,61mm 2 0 ,0 0,89 0 ,0 1 2 1 1 0 ,0 0,84 0 ,0 2 5 6 32,0 1,42 0 , 1 7 6 4 3 1 ,0 1 ,3 * v , i4 -r4 1 0 ,0 0,84 0,0 2 .5 6 1(b) O scillatory impulse test f = 180 KH= T, = 50 sec Temperutti re = 26°C Reg i on B re a kd o w n Bt'i'dkvlown Normali sod Di Ff2 Tlxi ck- KV C ivcfocjo b re a k d o w n n e s s (mm) 2 ^ 5 1,11 0,0121 1,06 27,5 1 , 0 4 0 ,0 0 1 6 1 ,0 0 o v e r - 32,5 2 0 .4 KV J .2 3 0,0529 1 , 0 4 a v = h „ . 1,03mm 2S,0 1,05 0 , 0 0 2 5 0,98 0 , S5 0,0225 1,00 1 ^ 5 0,70 0 ,0 9 0 0 1 ,1 0 25,0 1 ,0 8 0 ,0 0 0 4 0 , 7 4 39,0 1 ,2 5 0 ,0 0 2 5 0,73 1 0 ,0 0,k3 0,0324 0 , 6 5 2 4 ,0 1 ,0 3 0 , 0 0 0 9 0 , 6 6 ?J , o 0 , 0 0 0 , 0 1 0 0 0 , 5 9 1 8 ,0 o o r,. 0 ,7 7 0 , 0 5 2 9 0,68 3J,1KV 0,68mm 2 1 ,0 0,90 0 ,0 1 0 0 0 , 6 7 W40KV^, 2 1 ,0 0,1)0 0,0100 0,68 2b,0 1 ,1 2 0 ,0 1 4 4 0 , 7 4 20,0 0 ,S o 0,017l) 0,70 32,0 I.3X 0 ,1 4 4 0,63 2 1 ,5 0,1)0 0 , 0 0 04 0,68 11 4 t(c) R,F, test Frequency — 1,0 MH_ Temperature = 22°C R e g io n B re a kd o w n A v e ra g e Tb i c k n e s s (p e a k KV) b re a k d o w n .1,0 1 ,0 7 1 ,9 O .k i) 2 , 2 2,7 b K V 0 ,*9 L ir 3 . 2 3«,7KV^ 0 , 8 9 0 , 97mm 3 .3 0,77 3 , 0 1 ,2 b 2 ,3 3 , 5 2 , t l 3,0 2 . 4 2 ,8 !)k V 0 , 59mm 2,9 4> ),0 K V ^m 3 . 4 2 ,N 3 , 0 3 , 6 ------11 5 1(d) R.F. test fr e q u e n c y =2,5 MHz T e m p e rd tu re « 23°C B re d kd o w n A v e ra g e Tlti ckness (p o a k KV) b re a k d o w n 2 , 0 2 , 0 2 ,3 o v e r - 2,b 5 K V o v = 3 ,1 ( p e a k ) 0 , 91mm 2 , 9 3 , 0 3 , 0 2 , 5 3 , 4 2 , 9 o o rv 5? kv a v « 3 . 0 0,574m m = 4 4 ,7 K V 4 i 2 . 7 3 . 0 2 .8 2 , 9 2 , 8 116 Diamond coil for 6,0 KV motor Insulation: Alkonex enamel plus double glass and a I k y d enam el Inductance “ 119 mi croHcnry Copper cross section 2,0mm x 7,3mm O scillatory impulse tost failure at 31 KV Coil cut in 10 pieces - 4 overhang and 6 core 117 2(a) Stundcird Impulse Test. I EC Stdndord Impulse 1,2/SO microseconds TempePdturu = 24°C + 2°C ovepKang potion B r e d k - A v e ru g u N orm a 1 - ( v u r ^ Thick- AvePdge Di e 1 oc t r i c i s e d t(u " cfc.- s t r e n g t h KV 25 1 ,2 0 0 , 2 0 0 ,0 o 7 b 25 l , 2 o 0,2n 0 ,0(> 7o 21 1 ,0 0 0,0 (1 0,003b 23 1 y ,7 k V 1 ,1 0 0 , 1 b 0 ,0 2 5 b 0,37mm 14 0 ,7 1 0 , 2 ') 0,0N4I 12 0 , o0 0 ,j0 0 ,1 0 0 0 IK 0 ,D 1 0 , 0 ') 0 ,0 0 S I uoPti region | 2W,0 1 ,2 5 0 , 2 5 0,0b35 18,9 0,S4 0 , 1 h 0,025b 1 7 ,1 0,7u 0 , 2 4 0 ,0 5 7 0 1 , 0 2 0 , 0 2 0 ,0 0 0 4 3 0 ,0 1,J4 0 , 3 4 0 ,1 1 5 4 2 0 ,3 0 ,9 1 0 , 0 9 0,00X1 . 2 5 ,5 M 4 0 , 1 4 0,0196 1 5 ,7 0 , 7 0 0 , 3 0 2 3 ,5 k V 0 ,0 * 4 ...... 2 8 ,0 2,1,1 0,2.1 3 0 ,0 1 ,2 1 0,21 0 , 0 4 4 i 1 4 ,0 0,5<> 0,44 0 ,1 9 3 b 2 7 ,5 1 ,1 1 0 ,1 1 0 ,0 1 2 1 2 7 ,0 1,09 0 , 0 9 0 ,0 0 X 1 2 0 ,0 1 ,0 5 0 , 0 5 0 ,0 0 2 5 2 7 ,0 i , c y 0,09 0,00X1 1X,0 0 , 7 0 0 , 3 0 0 ,0 9 0 0 1 1 8 2(b) Odin pud Osci I I utory Test f = 180 KH„ Ti = 40 microseconds Temperature = 24°C o v e rh a n g r c g i orJ Average Norma 1 - (var)^ Thick- A v e ra g e D ieleetric d r n v l . . d a t i i n t h ic k ~ s t r e n g t h 1 7 ,5 1 ,0 8 0 , 0 8 0 , 0 0 0 4 0 , 4 2 n , o 0 , 0 8 0,3.: 0,1024 0,3b 1 9 ,0 1 ,1 7 0 ,1 7 0 ,0 2 X 9 0,39 1 6 ,0 0 , 9 0 0 ,0 1 0 ,0 0 0 1 0 , 4 2 1 0 ,1 K V 0,340)11111 413KVL 1 7 ,5 1 ,0 8 0 , 0 8 0 ,0 0 0 4 0,39 17,5 1,08 0,08 0 ,0 0 l> 4 0,39 1 5 ,0 0,93 0,07 0,0049 0,39 1 5 ,5 0 , 9 o 0,04 0,001b 0 ,3 b cor-' region 1 9 ,0 0 7 8 s '" ' 0 , 1 2 " 0 , 0 1 4 4 0 ,3 * 2.1,0 1 ,0 0 0,0b 0,0 0 .1 b 0,34 2 5 ,5 ! , l t 0 , 1 8 0 ,0 3 2 4 0,39 1 9 ,5 0,90 0,10 0 , 0 1 0 0 0 ,3 * 19,0 0 , 8 8 0,12 0 , 0 1 4 4 0,35 2 1 ,0 0,9? 0,03 0,0009 0,37 20,5 0,95 0,05 0 ,0 0 2 5 0,39 23.0 1 ,0 2 0 , 0 2 21 .5 K V 0 , 0 0 0 4 0 ,3 0 0 , 3b0miii SWKV6. 2 5 .0 1,1b 0 , 1 b 0,025b 0 , 3 4 1 0 ,0 0 , 7 4 0,2b 0,01 )7 0 0 , 3 7 2 0 ,5 0,05 0 , 0 5 0,0035 0 , 3 b 23,0 1 ,0 b 0,0b 0,003b 0,3b 1 7 ,5 0 ,8 1 0 , 1 9 0 ,0 3 0 1 0,35 2 4 ,0 1 ,1 1 0,1 1 0,0121 0 , 3 5 23,3 1 ,0 9 0,09 0 ,0 0 8 1 0,39 2 5 ,5 1 ,1 8 0 , 18 0 ,0 3 2 4 0 ,3 7 1 1 9 2(c) Radio Frequency Test F = 1 , 0 MHz T e m p e ra tu re = 25°C overhang region) B re a k d o w n A v e ra g e Th io k n e s s A v e ra g e Di e 1e c t r i c KV (p e a k ) b re a k d o w n t h ic k n e s s s t r e n g t h 3 , b 0 , 3 4 3 , 4 0 , 4 0 2 , 0 0 , 3 7 2 , 0 3,29 KV 0,37 0 , 3 90mm 1 2 ( p e e k ) 0 , 4 2 2 , 0 0 , 4 0 2 ,3 0 , 4 2 0 , 3 7 Lore region .3 ,4 2 , 0 3 , 5 2 , 5 3 , 0 0 KV 3 . 0 0,390m m 7o,9KV(i.k)4 2 . 0 .3 ,4 3 , 0 j 12 0 2(d) Rudio Fruqutiiicy Tt*s<- F = 2,5 MM, 'Vomperiiturti = 24°C o v e rh u n g r e g i on .iroukdown A v e ra g e T li i c k n e s s A v e ra g e D i d e c t r i c KV ( p o u k ) O re u kdo w n t h i c k n o s s s t r e n g t h 2 , 2 2 , 0 2 , 9 2 , 7 2,85 KV 0,3 8mm 75,OKV(pk)4i 4 , 0 3 , 2 2 , 7 2 , 5 c o r e r t- g i on 2 , 0 1 , 9 2 ,5 0 KV 2 , 9 (P "d k ) 0 , 341iiim 7 5 U K V ( p k ) ^ 2 , 9 3 , 0 3 , 2 121 3. T e s t s o n Noiih-x M - mi cu-! osdod polyomide fib re paper 3(a) Radio Fropi'vi.. > Tc/f Tcmpvroturv “ Frctiiioticy Rrt-ukdoivn Th i ckiivss Die! pc kr*i o s fc ru n c itli 1250 V 0 , l vU)inm 92,4kV(pk)4 13 00 V 0 , 14 0mm l. O M I t 1 7 0 0 V 0,139m m m ,3KV(pi04, 1 7 0 0 V 0 , 14 0 mm 0 , 5MHS 1 » 0 0 V 0 , I2 9 ihim 139,8KV(pk)^it 2 0 0 0 V 0 , 1 40mm 12 2 4. Tests on Kopton Fi I in - Po I yimi dc (Du Pont:) 4(=i) St, nd.inl ImpuiMv T«st Tom puriiiui'i' -■ 25°C A w r -u jv N oriiu i 1 - V o r i - ...... ) j 1 li i u k - A v u rtin o D i d c . c t r i c Jow n u t io n t h i csk- s t r e n g t li 7/1 0 ,0 S 0 , 0 2 0 ,0 0 0 4 0,0.10 7,5 7 .4 K V i . o i 0 ,0 1 0 ,0 0 0 1 0 ,0 .1 0 0 , 030mm ;>,47M V^R| 7 ,S j, 0 l 0 ,0 1 3 ,0 0 0 1 0,0.10 A (b) 0«vi 11 ittovy Impulsv Tew, TvmnviMtui'v - 25°C B i'v u k - A v v i'iit)v Nonihil - (v-.r)- Thivk- A w i’ ,u)v lu x itik - T h ic k - \ v H, 0 i ,0 4 0 , 0 .i 0 ,0 0 1 0 0 , 0 2 * 7 ,1 h,DK V JAM 0,0,i 0,000') o,u.u 0,030mm 7,0 1,01 0,01 0,0001 0 , 0 ,u 123 4(c) Sddio Frequency Tests Temperature =■ 25°C F re q u e n c y Bri’dkdown A v e ra g e T h ic k n e s s D i d e c t r i c volts (pcuk b re a k d o w n s t r e n g t h MHs volts (podk) KV(pk%_ ' 475 45 0 2 , 0 4 5 4 0,030 168 450 44 0 41 0 4 , 0 400 410 0,028 164 4 2 0 4 l>0 490 1 , 2 510 4 l>7 0 , 0 3 0 177 500 550 550 0 ,7 0 555 0 , 0 3 0 570 185 550 5t>0 1 ,0 1 500 503 0 , 0 3 2 17b 570 .. .. 5H0 0 ,9 1 580 5«0 0 , 0 3 2 181 580 5% 0 ,7 1 580 503 0 ,0 jl 1.92 010 o()0 0 , 5 2 040 043 0 ,0 3 0 214 (130 (*.10 MO <>2S 0,3.1 h30 0 , 0 2 8 224 020 12 4 F re q u e n c y B re a kd o w n A v e ra g e T h ic k n e s s Di e 1e c t r i c MHz volts (peak) b re a k d o w n s t r e n g t h volts (peak) K V (Pk ) 4 , 540 520 520 5 , 0 0 522 0 ,0 3 0 174 53 0 510 510 4 6 0 400 4 2 0 1 0 ,0 4 4 0 43 4 0 , 0 2 9 14 9 4 2 0 45 0 45 0 480 4 7 0 7 , 5 4 7 0 0 ,0 3 0 157 4 7 0 4 6 0 4 8 0 1 , 8 50 0 48 7 0 , 0 2 9 168 68 00 p o w e r 64 00 F re q u e n c y 6800 6 7 0 0 0 ,0 3 1 21 60 50 Hz 5 9 0 0 7 6 0 0 ...... 125 5. S tatistical analysis of coil sample breakdowns, 5.1 Deterrr: - i.ion of standard deviation, histogram and p r o b o b i s t r i b u t i o n o f s a m p le s . 83 samples v a.lable from impulse tests (of both types) on the coiI .moles. Using the formula for standard deviation a = j S (V-VF yi eIded s = 0,1831 Using the pre-programmed sta tistica l function of a Hewlett Packard HP25C calculator yielded s = 0,1833 It was thus assumed that s = 0 , 1 8 3 2 A histogram w ith an interval of 0,05 was drawn using values of breakdown normalised according to sample, te st and region (core or overhang). B re a kd o w n N um ber o f i n t e r v a 1s r e a d !n g s i n probabi1ity (n o rm a 1is e d ) 1n t e r v a 1 i n interva1 0 , 5 0 - 0 , 5 4 0 0 ,0 0 0 0 , 5 5 - 0 , 5 9 1 0 , 0 1 2 0 , 6 0 - 0 , 6 4 2 0 ,0 2 4 0 , 6 5 - 0 , 6 9 1 0 ,0 1 2 0 , 7 0 - 0 , 7 4 5 0 ,0 6 0 0 , 7 5 - 0 , 7 9 2 0 ,0 2 4 0 , 8 0 - 0 , 8 4 5 0 ,0 6 0 0 , 8 5 - 0 , 8 9 6 0 ,0 7 2 0 , 9 0 - 0 , 9 4 10 0 ,1 2 0 0 , 9 5 - 0 , 9 9 10 0 ,1 2 0 1 , 0 0 - 1 , 0 4 5 0 ,0 6 0 1 , 0 5 - 1 , 0 9 12 0 ,1 4 5 ' 1 , 1 0 - 1 , 1 4 9 0 ,1 0 8 126 B re a kd o w n Num ber o f Samp 1e i n t e r v a 1s readi ngs in probabi1ity (norm alised) i n t e r v a l in interval 1 , 1 5 - 1 , 1 9 5 0 , 0 6 0 1 , 2 0 - 1 , 2 4 2 0 , 0 2 4 1 , 2 5 - 1 , 2 9 4 0 , 0 4 8 1 , 3 0 - 1 , 3 4 1 0 , 0 1 2 1 , 3 5 - 1 , 3 9 2 0 , 0 2 4 1 , 4 0 - 1 , 4 4 1 0 , 0 1 2 1 , 4 5 - 1 , 4 9 0 0 , 0 0 0 A normal probability distribu tion curve can be drawn using the values obtained from tables of erf y, the error function ( V i - y 2 . where y = V-V s J T with V the normalised breakdown voltage and s the standard deviation i.e . 0,1832. As the above expression applies to both.sides of the pro b a b ility curve, ■§ e rf y must be used. V i e r f y P r o b C V ^ V ^ ) 0 ,9 7 5 0 ,0 9 6 0 , 0 5 4 0 , 1 0 8 1 ,0 2 5 0 , 0 9 6 0 , 0 5 4 1 ,0 7 5 0 , 2 8 9 0 , 1 5 9 0 ,0 9 3 1 ,1 2 5 0 , 4 8 2 0 , 2 5 2 0 , 0 7 8 1 ,1 7 5 0 ,6 7 5 0 , 3 3 0 0 ,0 6 0 1 ,2 2 5 0 , 8 6 8 0 , 3 9 0 0 ,0 4 3 1 ,2 7 5 1 ,0 6 1 0 ,4 3 3 0 , 0 2 8 9 1 ,3 2 5 1,254 0,4619 0 , 0 1 7 8 1 ,3 7 5 1 ,4 4 7 0 ,4 7 9 7 0 ,0 1 0 1 1 ,4 2 5 1,640 0,4898 0 , 0 0 5 4 1,475 1 ,8 3 3 0 , 4 9 5 2 0 ,0 0 2 7 1,525 2,026 0 , 4 9 7 9 0 , 0 0 1 2 1 ,5 7 5 2 ,2 1 9 0 ,4 9 9 1 127 5.2 S tatistical estimation of coil breakdown IF a normal distribution is assumed for the breakdown of the samples then the breakdown of a coil made up of a number of such samples can be estim ated. i.e . Prob(coil) = Prob(scmipIe) x no. of samples The number of samples must be estimated to be those that can be fitte d into the weakest portion of the coil (usuo11y the overhang region). Since about 20% of each sample is not act i v being used for end insulation and connection, the actual num ber of samples must be corrected by a factor of 1,25x. Thus for coil no. 2 the actual number of overhong samples is 3 2. This gives a corrected number of 40. 1% probability of coil failure therefore requires a s a m p le probability of 40x less than this. i.e . Prob(saiti|i I e) = 0,00025 This probability in relation to the error function is erf y = 1 - 2Prob(sai»pIe) - 1 - 0,0005 = 0 ,9 9 9 5 since only the lower section of the probability distribution etween y and -in fin ity is of interest. Thi s yieIds y = 2 ,4 6 1 The normalised voltage correspond! ng to th is is Vn = 1 - s / iy if s = 0,1832 Vn = 1 - 0,1832 /T x 2 ,4 6 1 - 0 , 3 6 2 i.e . if average breakdown is 16,1 KV V - 5,83 KV per turn. Thus the total voltage across the coil is 9x this 12 8 i.e . 52,5 KV Actual fa ilu re was at 31 KV F o r 50% f a i l u r e Prob(sample) = 0,0125 e r f y = 0,975 y - 1,585 vn = 0,589 Vt = 9,49 KV T o t a l V = 85,4 KV For col I no. 1 No. of samples = 24 Corrected number o f samples == 30 F o r 1% p r o b a b i I i t y Prob(somple) = 0,000333 e r f y = 0,999333 y = 2 ,4 7 Vfi = 0,360 Vt - 8 , 3 2 KV Total V = 58,2 KV Actual failure was at 35 KV For 50% fa i Iure Prob(ttample) « 0,0167 e rf y = 0,9667 y m 1 , 5 5 ' Vn « 0 , 5 9 8 Vt = 1 3 ,8 2 KV Total V = 96,8 KV REFERENCES 1 O liver, J.A., Woodson, H.H., and Johnson, J.S., #A Turn Insulation Test for Stator C oils' AIEE Trans, on Power Apparatus and Systems, Vol PAS-87, No. 3* March 1968. pp. 669-678. 2 Private letter, from Reid and M itchell (Pty) Ltd. 3 Riidenberg, R. ' E lectrical Shock Waves in Power Systems', 1st Engl edition = 4th German edition. Harvard U niversity Press, 1968. 4 L e w is , W.W. 'The Protection of Transmission Systems against Li flhtning'. 1st ed. Dover Publications, 1965* 5 B e c k , E. ' Liohtninc Protect!on.for Electric Systems? 1st ed. McGraw-HiI I Book Company, 1954. 6 Eriksson, A.J. 'Overvoltage Surges in Industrial Complexes'. Symposium on High Voltage Engineering in South A fri Johannesburg, 18-19 November 1974. 7 Kano, I., Hakamada, T ., Kurasawa, Y., and Sugawara', H., 'Switching Surge Phenomena in Induction Motor Windings and Their Endurance', Hitachi Review, Vol. 24, No. 5, 1975, pp. 225-233. B ritish Standards Specification B.S. 923. 130 9 American Standards for Transformers. Regulators and R e a c to r s , American Standards Association C57.22, 1948. 10 I.E.G. Specification 60. 11 Foust, C.M. and McAuley, P.M. 'Recommendations 'fo r H.V, Testing by Subcommittee on C orrelation of Laboratory Data of EE I and NEMA'. AIEE Trans, Vol. 59, 1940, pp. 5 9 8 -6 0 2 . 12 Al ston,- L,L. et al . 'High Voltaoe Technology? Chapter 6, Section 8. Oxford U niversity Press, 1968. 13 KuffeI, E., and Abdullah, M., 'High Voltage Enqineerino' , 1st edition, Pergamon Press, 1970, 14 Levi is , W.W., Wei seman, R.W. and Rudge, W.J. 'Protect!on of Roteti ng'A.C,"Mach i nes against"' 1 Travelling Wave Voltages due to Lightning' AIEE Trans, Vol. 52, 1933, pp'. 434-440. 15 Griscorn, S.B. et al. 'Five Year Field Investigation of Lightning Effects on Transmission Lines' AIEE Trans, Power Apparatus and Systems, Vol. PAS 84 No. 4, Apri I 1965. pp. 2 5 7 -2 8 0 . 16 MartzI o ff, and Hahn, 'Surge Voltages in Residential and Industrial Power C i r c u i t s ' , IEEE Trans, Power Systems and Apparatus, Vol. PAS 89 No.6, Jul/Aug 1970, pp. 1049-1056. 131 17 Harder, E.L. and Clayton, J.M. 'Lightning Phenomena', Westi nghouse Engineer, Vol. 2, July 1951- PP. 106-111. 18 Mai an, D.J. 'Physics of Liohtnimn' 1st ed., Chapter 11, Engl Ish U niversities Press, 1963. 19 McEachron, K.B., 'L ig h t:.;.to Empire State Building', AI EE Tra.- Vol. 60, 1941, pp. 885-889. 20 McCann, G.D., 'The Measurements of Lightning Currents in D iroct S t r o k e s ', AIEE Trans, Vol. 63, 1944# pp. 1157-1164. 21 Bellaschi, P.L., 'Lightning Surges Transferred from One C ircuit to Another through Transform ers', AI EE Trans, Vol. 62, Dec. 1943, pp. 731-738. 22 Heller, B. and ''everka, A., 'Surge Phenomena in E lectrical Machines', 1st ed. 11i ff" Books, 1968. 23 We 11a u e r , M ., 'The Voltage Stresses in the Entrance C oils of Windings on the Occurrence oP Impulse Voltages of D ifferent Steepness', B ulletin Oerlikon No. 270, pp. 1823-1829 (continued') No. 271, pp. 1843-1848 24 Friodlando". E., 'TraveI Ii ng Waves in H.V. Alternator W indings' 1 EE Journal, Vol. 89 port 2, 1942, pp. 492-508 and Vol. 90 1943, pp. 83-130. 13 2 25 Meyer, H,, 'Behaviour of Rotating Machinery Windings on the Occurrence of Travelling Waves', Brown-Boveri Review, Vol. 30, 1943■ pp. 279-286. 26 Poritsky, Ahctti, P.A. and Jerrard, 'F ield Theory of Wave Propagation along C oils', AI EE Trans, PAS 72, Oct. 1953. pp. 030-938. 27 Blume; and Boyajian. A., 'Abnormal Voltages W ithin Transformer W indings' AIEE Trans, Vol. 38, 1919. P 577. 28 Palueff, K.K., 'The Influence of Transient Voltages on Transformer D e s ig n ', AIEE Trans, Vol. 448, July 1929, P 681. 29 Hodnettc, J.K., 'E ffe ct of Surges on Transformer W indings', A!EE Journal, Nov. 1929, p 829. 30 WaIdvogeI, P. and RouxcI, R., 'A New Method of Calculating the E lectric Stresses in a Winding Subiected to a Surge Voltage', Brown Bovcri Review, Vol. 43, 1956. pp. 206-213■ 31 Ryfander, J.L., 'High Frequency Voltage Test for Insulation of Rotating E lectrical Apparatus', AIEE Trans, Vol. 45, 1926. pp. 459-465. 32 We!Iauer, M., 'A New Arrangement for Testing the Insulation of the Turns of Machine Coils', » Bull : •'■rl ikon, No. 251, Vol. XXIV, pp. T624-1626. 133 33 • Kranke, 0. and Schuler-, R., AA Method fo r Checking the Turn Insulation of Form- wound Coil Windings fo r High Voltage Machines', Rrown Rover i Review, Vo I 57/ April 1970, pp. ‘191-196. 34 Sexton, R.M. and Alke, R.J., 'D etection of Turn to Turn Faults in Large U.V'. "Turbine Generators'’, AIEE- Trans, Vol. 70, 1951. pp. 270-274. 35 C etlin, E.H." and Rohats, N., •'IVindtng 'i lation Tester for D.C. Armatures', AIEE Trans, Vol. 70, 1951, pp. 465-468. 36 Moses, 6,1. and Marten, E.F., 'W inding Fault Detection and Location by Surge Com parison Test!ng', AIEE Trans, Vol. 64, July 1945, pp. 499-503. 37 Reynolds, D .J., Alke, R.J. and Buchanan, L.W., 'Testing Insulation with the Surge Comparator', We$,ti nghouse Engineer, July 1951. 38 Weed, H.R., 'E lectronic Surge Testing of Universal Armatures With Mild Detection', A ICE Trans, Vol. 78 part I I IB, Doc. 1959, pp. 1219- 1 2 2 6 . 39 Strain, R.A., 'Some Aspects o f Surge Comparison Testing of Fractional Horsepower M otor', AIEE Trans, Vol. 75 oart III, Oct. 1956, pp. 917-921. 40 Kritzingcr, J.J., 'The Breakdown Mechanism of Long Sparks in A ir', , , Ph.D. Thesis U niversity of the W itwatersrsnd, Johannes- ' 134 41 -lones, R.F.B, and Ahmed, M,,, 'ImpuNe Testing of Motor C oils'-, D issertation prepared in partial fulfilm ent of BSc(Eng degree, 1970, Dept. Flee. Eng., U niversity of the W it- watersrand, Johannesburg. 42 Thomas, H.A., /Theory and Design of Valve 0sol I Ia to rt for Radio and other Frequencios/. Chapman and H all, 1st edition, 1944. p 32, 37-38, 40. 43 Terman, F.E., ' Electro nic and Radio Enoi neari ng*-, McGraw-Hill, International Student Edition, 4th Editio pp. 489-496. 44 Pimm, J.A., 'The E lectrical Breakdown Strength of Air at V.H.F.', Proc I EE Vol. 96 (III), 1949. pp. 117-129. 45 Fatohchand, R.R.T., 'The E lectrical Breakdown of Gaseous D ielectrics at High Frequencies', P r o c I EE V o l. 1 0 4 ( c ) , 1 9 5 7 . p p . 4 8 9 -4 9 5 . 46 Bright, A.W, and Huang, H.C., 'Form ative Time-lag Studies with High Frequency Dis c h a r g e s ', Proc I EE, Vol. 102 (o), 1955. pp. 42-45. 47 Peek, P.M., 'D ie le c tric Phenomena in Hi oh Voltage F.nni ncer i no' , McGraw-Hill, 2nd edition, 1920. p 177. Author Hopkins MJ Name of thesisTesting of the Inter-Turn Insulation of High Voltage Induction Motor Coils PUBLISHER: University of the Witwatersrand, Johannesburg ©2013 LEGAL NOTICES: Copyright Notice: All materials on the University of the Witwatersrand, Johannesburg Library website are protected by South African copyright law and may not be distributed, transmitted, displayed, or otherwise published in any format, without the prior written permission of the copyright owner. Disclaimer and Terms of Use:Provided that you maintain all copyright and other notices contained therein, you may download material (one machine readable copy and one print copy per page) for your personal and/or educational non-commercial use only. The University of the Witwatersrand, Johannesburg, is not responsible for any errors or omissions and excludes any and all liability for any errors in or omissions from the information on the Library website.