cure International Conference on Large High Voltage Electric Systems 112, beuavard Houssmann ~ 75008 Paris 1974 Session — 21-29 August 12-03 ° A CRITICAL ASSESSMENT OF METHODS OF MEASURING PARTIAL DISCHARGES IN EHV TRANSFORMERS J.L,DOUGLAS F.C. PRATT —_—F, RUSHTON, Centra Electricity Contra lectciy Ferrand Generating Bosra Generating Board (Unitas Kingdom) SUMMARY The paper reviews past and present practice in partial discharge calibration and measurement techniques. Comment 1s made on these practices in the light of experimental work which has shown the existence of long and short partial discharge pulses and which has investigated, over a range of measure ment frequencies, the responses of ri. meters and a transformer winding to such pulses In order to minimise calibration and measurement errors which are shown to be very large at IMHz, when measuring such pulses, the use of measurement frequencies in the range 50 to 100 kHz is recom ‘mended in conjunction with narrow-band instruments. The application of partial discharge measurement during reduced level, induced voltage tests on transformers is discussed. Bandwidth, Calibration, Measuring frequency, Partial discharge, Pulse shape, Transformer, Voltage test. REPORT 1. INTRODUCTION ‘The use of partial discharge measurement as one criterion of goodness during voltage testing of transformers has been studied during the past twenty years. The methods of measurement and their interpretation have not been firmly agreed and it is only recently that charge measurements in picocoulombs have become generally accepted as more soundly based than the previous almost total use of microvolts. Recent developments in specifications admit that the very high level induced voltage test may now be superseded by an induced voltage test of much lower level for 2 longer time, However, this lower voltage test removes from the purchaser the assurance, however nebulous, that a transformer which has passed a test at high voltage level is likely to give good service experience. As yet there is no firm evidence that the longer term test is equivalent to the short term, high voltage test. It follows that12-03 -2 some other assurance, such as partial discharge measurement, must be given to the purchaser as to the soundness of the transformer insulation, The extended test time permits this measurement to be more significant than during the one minute test. ‘Thus, there is a need for meaningful partial discharge measurements to be agreed before the introd- uetion of these longer time tests. Recent recommendations by Working Group 01 of CIGRE Study Committee 12 [1] have gone some way towards establishing soundly based methods of measuring partial discharges in transformers, The present paper contains some supplementary data which it is hoped may lead to a standardisation of the basis of measurements so that valid comparisons can be made between transformers of different sizes and constructions. 2. REVIEW OF LITERATURE Although sensitive partial discharge detectors and measuring techniques were developed over twenty years ago [2], it was 1957 before results of corona (partial discharge) measurements in transformers were first published by Adolphson and Vogel (3]. Their method of measurement was based on a modified detector described by Quinn [4]. In the discussion of Adolphson and Vogel's paper, questions were raised by Gehrig, Oliver and Evans et al relating to damaging effects of discharges and comparison of charge and RLV, measurements, which have still not been answered satisfactorily. About that time, radio interference ‘meters were beginning to be used for partial discharge measurements in transformers. A NEMA Specification [5] standardised the measuring procedures for assessing radio noise and, apart from minor modifications, the method is largely unchanged today [6]. Essentially, the circuit employs a coupling capacitor of around 1000 pF to connect a specified resistance between line terminal and earth ; the Ril. voltage is measured actos this resistor. Where it differs from a conventional partial discharge detector is in the method of calibration. The NEMA method of calibration relates the values measured at the bushing tap to thoss which ‘would be obtained at the sample terminals in the absence of a coupling capacitor and with a pure resist- ance of specified value, Reference to the use of the transformer bushing as a coupling capacitor was made in 19$7 [7] but it was not until 1965 that publications by Brown [8] and Narbut [9] established the validity of the bushing tap method for transformer partial discharge or R.ILV. measurements. This method effectively eliminates the need for additional coupling capacitors and it is now generally accepted as a reliable and convenient method, whatever type of measuring instrument is used. In Europe, ri, meters have been used extensively, but investigations with partial discharge detectors indicated that the latter, calibrated in terms of charge in picocoulombs, have some advantages [10]. A few years earlier, the quadratic mean method of charge measurement was developed by Renaudin [11] jin France where much experience has been gained on transformers and other equipment. The main advantage claimed for the quadratic mean method is that it responds mainly to the large discharges but that numerous small discharges may contribute to the reading. At present there are insufficient data to make a strict comparison between the quadratic mean, peak and quasi-peak methods of measurement in transformers, However, it is now generally accepted that the loss of charge at the transformer terminals is the most informative partial discharge parameter to be measured, even though there is no simple relationship between the apparent charge at the terminals and the damage to the dielectric involved at the discharge site, is not It is felt that some of the difficulties in interpretation of results have arisen because it clearly established which discharge quantities are being considered. There are four different discharge ‘quantities which are of interest in a transformer12-03 1) the charge dissipated at the source 2) the charge at the winding adjacent to the discharge site 3) the charge at the appropriate transformer terminals 4) the charge measured at the bushing tap. ‘The relationship between (3) and (4) can be established by the calibration procedure specified by the LE.C, [12]. Attenuation along windings accounts for the difference between (2) and (3). It has been claimed that the magnitude of (2) can be estimated from results of measurements at different terminals [13], [26]. In the past, specifications have generally referred to partial discharge measurements at transf- ormer line terminals, but WGO! of CIGRE Study Committee No. 12 have proposed that measurements should be made at more than one terminal [1]. The actual charge quantity (1) cannot be measured, although energy dissipated in a cavity can be estimated in a simple case [14] but is never used in the complicated case of a transformer. This Working Group also recommend that the measuring impedance should be essentially resistive and within the range of 50 to 200 ohms. Previous work has shown that high sensitivities can be achieved by the use of high input impedances [15] but care must be taken that long time constants of the input circuit do not influence the results, NEMA Publication 107A, 1964, specifies that the measur- ing impedance should be 150 ohms ; if the ri, meter has a lower input impedance then some form of high impedance coupler is required to meet the specification. The advantages of this additional complication are obscure, since neither sensitivity nor signal/noise ratio are increased by the higher input impedance. ‘The question of bandwidth has led to some confusion and the impression is gained from the literature that true charge measurements can be made only with a wide band detector. In fact, charge in individual pulses can be measured over a wide range of bandwidths [16] provided that the measuring circuit is calibrated correctly [17]. Mole [16] has shown that the percentage error difference between narrow and wide band detectors is small over a wide range of frequencies. One of the advantages of wide band measurements on transformer windings is improved resolution of pulses but this effect is relatively small ; another advantage is said to be that variations of meter response with frequency are averaged out, but results on transformers indicate that an extremely wide band would be necessary to achieve this and extrancous noise could then be a problem. ‘The measurement frequency is another factor which is controversial. IMHlz is specified by NEMA whereas a much lower frequency has been advocated [18]. Both theoretical and practical investigations indicate that partial discharge measurements on transformers should be made at less than 200 kHz [19] although the precise frequency has never been specified unambiguously. Other authors [1] have susgested that any measurement frequency is satisfactory, provided that high background disturbances and high at- tenuation points are avoided. There is little doubt that the measurement frequency has a greater effect on the measured values of partial discharges arising from sources within a transformer winding than for sources near the measuring terminal ‘The types of instrument used to measure discharge magnitudes can vary in their response to repeated pulses. During the past decade, much has been written about the advantages and disadvantages of the various types of r.i, meters or partial discharge detectors in common use but no clear decisions can be ‘made from the published data, Apart from different calibration methods, data are often incomplete and there have been no straight comparisons made between the various instruments. The conflicting statements made, for example, on the relationship between picacoulombs and microvolts [20], [21] are due, at least in part, to different measurement frequencies and bandwidths, different calibration methods and even to different points of measurement on the transformer rather than the measuring instruments themselves. The effect of the type of winding on the propagation of partial discharge pulses has been studied by Kawaguchi and Yanabu [22] and by Bertula et al [23]12-03 4 3. EXPERIMENTAL WORK ON PULSE SHAPES OF PARTIAL DISCHARGES 3.1 Transformer Insulation Samples - As part of 2 programme of experimental work on discharge measurement, the waveshapes of current pulses emanating from transformer insulation samples were investig- ated. The pulses produced across a 1,000 ohm non-inductive resistor in series with the test sample were recorded oscillographically. Of the many types of sample tested, three forms are shown in Figure 1, which also includes some typical results of pulse shapes obtained. The SO Hz test voltage was of a level just above inception of partial discharges. Long pulses of up to 16 usec duration were observed but many pulses were of less than 1 usec overall time. In most cases, both short and long duration pulses were observed on the same samples. This confirms observations by Bertula et al (23] and Izeki et al (24). ‘Typet asoanee uecrione ammcenewr PULSE mavens ; 5.2000 KX. covouern se Perea Eh anmsee Te, raneT PAPER CoveRNG : tee 1 ae a qi me, 2Son120 x20 Ca) Te, es Figure 1 - Typical electrode arrangements and waveshapes ‘There was a tendency for long duration pulses to be associated with observed surface discharges and it is noteworthy that during tests on paper covered conductors without oil duets, any pulses observed were of short duration, In general, the short pulses were an order of magnitude lower in charge quantity than the longer pulses and this may account for the first and second corona pulses which have been described previously [25]. ‘The partial discharge pulses recorded during the above tests were too random and variable to directly measure the frequency spectra accurately. For subsequent measurements, simulated pulses were used to produce the steady repetitive signals required. Pulses of 0.1/0.5 usec and 4/12 psec, representing pulses towards the extremes of the measured values, were employed for the tests reported in the following sections. These pulses were obtained from a high-impedance source and injected either across the meter or at appropriate points in a transformer winding. This procedure is the same as used in conventional charge calibrations. 3.2 Effect of Pulse Shape on Detector Response - The response of ri. meters to the two pulse shapes was ‘measured over a frequency range from 10 kHz to 1.5 MHz. The results shown in Figure 2 were obtained at a pulse repetition rate of 100 pulses /s ; measurements at other rates between 50 and 10,000 pulses/s confirmed earlier reports [20], [21], that the maximum error due to pulse repetition frequency would not exceed 6 GB for an r.i, meter complying with NEMA Specification 107 A. 1964. Tests were repeated with the ri. meter connected to a simulated bushing tap (a 780/220 pF capacitance divider). Results of pulses of both shapes fed to the divider were identical to those shown12-03 » uns os Puss n 12s PULSES ‘a L.METER READING, 2/1 .¥ w ie i MEASURING FREQUENCY, Figure 2 - Etfect of pulse shape on rl, meter reading Figure 2 up to a measurement frequency of $00 kHz. Above this frequency there was a slight reduction in meter reading due to the shunting effect of the 2200 pF capacitor. When an additional capacitance of 5000 pF (representing transformer terminal capacitance) was connected in parallel with the simulated bushing, the shapes of the curves were similar to those obtained previously but the level was about 16 dB down throughout the measurement frequency range, which is in accordance with theory. All results quoted in this section refer to a bandwidth of 4.5 kHz and a measuring impedance of 150 ohms. It should be noted that the results shown in Figure 2 are frequency dependent which would not be the case if calibration was made using the conventional pulse calibrator of short front and long tail (16) Measurements with two partial discharge detectors of the ERA type [2], one of 5 kHz bandwidth at 500 KHz centre frequency (Model 2) and one with 10-60 kHz bandwidth (Model 3) confirmed that short pulses were more accurately measured at higher frequencies and long pulses required a lower measurement frequency 3.3 Effect of Transformer Winding - Pulses of the same waveshapes as used for the determination of meter response were injected at various points along the winding of an e.h.v. single phase transformer (Pig. 3). An ri. meter was connected across the lower end of a capacitance divider representing the bushing tap 4as shown. Pulse shapes of the injected pulses and those at the simulated bushing tap were recorded for injection points at 100 %, 75 %, SO % and 25 % of the winding length ; ri, meter readings were noted over the frequency range of 10 kHz to 1.$ MHz for each pulse shape and injection point. With pulses injected across the complete winding, the results of bushing tap measurements shown in Figure 4 were obtained. The results of measurements at 100 %, 75 % and SO % injection points are illustrated in Figure 5 for long pulses and Figure 6 for short pulses.+ L toupee Et Faoopr [soe 3 soooaRr Figure 3 - Transformer winding measurements FL METER READING, «6 sees pases ie 18 100 requeucr, mo Figure 4 - Effect on r4, meter reading of pulse shape injected at 100 %e position LLMETER READING, FREQUENCY, Figure 5 - Effect of pulse injection point - 4/12 psec pulsesa 12-03 L,WETER READING, 2 FREQUENCY, Me Figure 6 + Effect of pulse injection point - 0.1/0.5 pisee pulses 4. DISCUSSION 4.1 Experimental Results - The results reported in Section 3 were obtained when injecting long and short pulses such as observed on samples, into a full size transformer winding. No work has yet been done in the programme to examine the actual pulse shapes emanating from a transformer undergoing voltage test but it was. observed that considerable distortion of the injected pulses occurred between the injection and ‘measuring points. It is reasonable to suppose that similar distortion would occur for any partial discharge pulses initiated during a voltage test and that minimum distortion would occur in the case of pulses emanating from @ source at or near the line end. Pulses emanating from a source further down the winding would be expected to have reduced rates of rise when seen at the bushing tap but the extent to which this would occur would be dependent on the a value of the winding and on its propagation characteristics. It is possible, therefore, that long and short pulses of the type found on samples would only be detectable during a voltage test if the source of discharge was very close to the line end. The lack of evidence concerning pulse shapes occurring during a transformer test does not invalidate the conclusions to be drawn from the Section 3 results since any slowing of the rate of rise would mean a smaller high frequency component, Examination of Figure 2 shows that if a meter is calibrated using a given pulse shape and the pulses then measured are of a different shape, then the indicated values of pulse charge may be consider ably in error. Assuming that the action of a winding is to distort pulse shapes so as to lessen the h.f. content, it follows that the practical case of partial discharge measurement requires good Lf. sensitivity and therefore a low measurement frequency. Furthermore, due to the long pulses being an order of ‘magnitude larger than the short pulses, it is believed that the former may be more indicative of insulation damage and therefore should be the more accurately measured, 4.2 Methods of Measurement - If the acceptance criterion is no discharge at working voltage, then the choice of measuring instrument is unimportant, provided that the sensitivity can be shown to be adequate. If specifications are worded such that a certain discharge quantity is tolerable at a specified test voltage, then calibration of the circuit is essential in terms of discharge quantity. It is assumed that by the time the long term lower voltage test is introduced it will be generally accepted that calibration in terms of picocowlombs is more logical than microvolts, irrespective of the type of measuring instrument employed. The difficulties in changing from microvolts to picocoulombs are well appreciated but it is hoped that the transition from one to the other will not extend over too long a period of time. At the present time, there are no published data to indicate any one instrument has particular advantages over others. Until comparative tests have been made on transformers with the various12-03 ~8— detectors of interest, it must be assumed that laboratory experience is typical and that, provided it mects the CIGRE recommendations, the instrument may record peak, quasi-peak, r.m.s. or mean values. ‘The differences are small compared with other possible large sources of error. It is generally agreed that an oscillographic display is very helpful to identify spurious signals and in interpretation of results. Experience indicates that the choice of measuring impedance is relatively unimportant, although it has been recommended [1 ] that values should lie between 50 and 200 ohms and be essentially resistive in character. Measurements with r.i. meters having input impedances between 50 and 150 ohms gave similar results, apart from the question of sensitivity. The need for a purely resistive impedance is questioned, since the authors’ experience with narrow band detectors does not support this. The requirement for 2 purely resistive impedance implies that with a high frequency measurement at a bushing tap, added inductance must be used to compensate for the capacitance in parallel with the instrument input imped- ance. This appears to be an unnecessary complication unless measuring sensitivity is inadequate and is unacceptable if measurements are to be made at more than one frequency. Wide band measurements are often recommended for transformer partial discharge tests. The main advantage of a wide band detector is in the resolution of pulses and this is considered to be of limited value in a transformer partial discharge test. Also, the practicability of making wide band measurements in factories where the background noise level is high is doubted. On the other hand, although doubts are often expressed about the validity of narrow band measurements, instruments of the ri, type have some ‘net advantages including the ability to select low background noise level, calibration over a range of frequencies to select maximum sensitivity, and the capability of measurements over a wide frequency range which can be informative about pulse attenuation. Many of the criticisms which have been levelled at the use of ri, meters are considered to have been associated with the method of calibration and the insistence on employing a measuring frequency of 1 MHz, ‘The use of such a high measurement frequency has been criticised by many authors but the recom- mendations currently proposed adhere to a frequency range between 0.1 and 1.2 MHz. The main reason advanced for a lower frequency is based on the response of transformer windings which attenuate most frequencies above about 200 kHz. This present paper goes further and shows that the larger (and possibly most damaging) pulses are of slow rise time even when the source of discharge is at the measurement terminals. Therefore the high frequency components of such pulses are small and large errors exceeding, 20 dB may be introduced when measurements are made at high frequencies (Fig. 2). These graphs indicate that there is no ideal frequency at which both short and long pulses can be accurately measured but at frequencies in the range 100 to 200 kHz, the errors would be relatively small ; less than 6 dB. However, the existence of these short pulses at transformer terminals is dubious, because the 0.5 sce tails measured in section 3.1 were very dependent on the measuring impedance (Fig. 7). Therefore, it is considered that the decrease in meter reading at frequencies below 300 kHz, because of the short tail, would not be observed in practice. In any case, results show that the 0.1/0.5 psec pulses are extreme values and the sm TAR DECAY TE, 6 Row se TE 2 cy ry iw 1 ‘mm DETECTOR RESISTANCE, cbs Figure 7 - Bffect of different values of detector resistor on a fast pulse.9 12-03 actual charge in these pulses is small. For these reasons, a value of between 50 and 100 kHz appears to be the most suitable frequency for measuring the majority of pulses at transformer terminals. Results illustrated in Figure 4 confirm that this frequency range is suitable for measuring partial discharges occuring at transformer terminals. If the discharge source is within the winding, the magnitude cannot be accurately measured, no matter at how many terminals the measurements are made. Again it is seen from Figure 4 that the frequency range of 50 to 100 kHz appears to give the minimum error for this particular transformer if the pulses are of the long type. But, because of the resonant peaks and troughs, it is impossible to specify one particular frequency for a narrow band instrument. A wide band instrument covering this range would be suitable but for the problem of extraneous disturbances. Therefore, it is suggested that measurements should ideally be made at a few selected frequencies and the mean value quoted as the estimated discharge magnitude. However it is recognised that there may be practical difficulties in such a procedure. 4.3 Calibration Procedure - Much hes been written on the topic of calibration of partial discharge measur- ing circuits which, in principle, should be simple but in practice leads to some difficulties and misunders- tandings. Unfortunately, the facts presented in this paper further complicate the subject. This is becauseit has been assumed in the past that partial discharge pulses to be measured have fast rise times of around 100 ns. Although partial discharges in a gas filled cavity behave in this way, breakdowns in oil or at an oil/solid interface have much slower rise times of up to several microseconds. It can be seen in sections 3.2 and 3.3 that if measurements are made at high frequencies after calibrating with fast rising pulses, then large errors can occur if the pulse fronts are much slower. ‘There are two approaches to the solution of this problem. The first is to calibrate the circuit using simulated pulses of the same shapes as the partial discharge pulses, but this is impracticable when there are several different pulse shapes. The alternative is to measure the pulses at a suitable frequency where the response is similar for all pulse shapes ; this method has been adopted here. If it is accepted that this procedure is logical, it should then be possible to calibrate simply by injecting pulses of known charge at a defined repetition rate across the sample terminals and noting the reading on the detecting instrument (meter or oscilloscope). The conversion factor is then given by the ratio Input (pC)/Output (pC or #Y). If, as has been recommended, measurements are made at more than one terminal of the transformer, calibration will be required at the appropriate terminals. Other correction factors are not considered to be essential unless the level of partial discharge is high and some attempt is to be made to locate the discharge sites. Problems concerned with location are outside the scope of this paper. 44 Significance of Results - It is generally accepted that there should be no measurable discharge within a transformer at working voltage and it is a reasonable premise that this criterion should also extend to a limited overvoltage, say 120 %. Between this level and the long duration test level, which will be ap- proximately 150 % of working voltage, some degree of discharge is believed to be permissible with two provisos : 1) that the discharge level does not increase throughout the one hour test 2) that the discharge must extinguish before the reducing test voltage reaches 120 %. Given that these two provisos are met and in the lack of knowledge concerning acceptable discharge levels at the test voltage, it appears that an “act of faith” is required in accepting firstly the long duration test and secondly reasonably high acceptance limits of partial discharge so that the test is not made unduly onerous. Alternatively, if the partial discharges were measured at 150 % down to 120 % voltage levels and could be extrapolated back to 100 % to show that there is negligible discharge, the difficulties associated with measurements would disappear because absolute values would not be necessary. All test specifications must include some proviso as to the action to be taken if the acceptance criteria are not met. Bearing in mind that a transformer designed for long duration voltage testing may not be able to withstand the hitherto normal high voltage test, it is clearly unrealistic to insist that reversion to the old test voltage shall follow a failure to meet the criteria of the reduced level voltage test. Assuming that the source of the discharge cannot be located and corrective measures taken, it is suggested that a further possible test is one involving several hours, probably at a voltage around 120 %, coupled with12-03 —10— continuous measurement of discharge, followed by a repeat of the reduced level high voltage test. Reprocessing may of course render the 120 % voltage soak run unnecessary. Additional evidence may be obtained from the results of gassin-oil measurements made at various times in the test sequence. 5, CONCLUSIONS 1, Partial discharge pulses emanating from transformer insulation samples have been shown to have many different waveshapes and pulse lengths. Pulse lengths varied from less than 1 psec up to 16 sec. In general, there was an order of magnitude difference between the energies in the short and long pulses. 2. The responses of partial discharge measurement instruments have been determined when supplied with similar pulses to those found on samples, It has been found that large errors will arise if long pulses are measured at a frequency of 1 MHz and that the errors arising from different pulse shapes are minimised if the measurement frequency is much tower. 3. Injection of pulses of different shapes at places on a full-size transformer winding and measurement of the response at a simulated bushing tap confirmed the findings in (2) above. Such errors will be of great importance in transformer partial discharge measurement if the: discharge pulses arriving at the terminals are of the long duration type as seems likely from consideration of winding propagation characteristics. 4, Partial discharge measurements on transformers should be made with a narrow band instrument in the frequency range of $0 to 100 kHz. More than one measurement frequency is preferred, 5. The choice of measuring instrument, measuring impedance and bandwidth is relatively unimportant compared with the measurement frequency, 6. Because of uncertainties in measurement, no fixed rejection or acceptance levels can be agreed at the present time. 6. ACKNOWLEDGEMENTS ‘This paper is published by kind permission of the Central Electricity Generating Board and the Directors of Ferranti Ltd, 7. REFERENCES [1] CIGRE SC12 WGOI - Measurement of partial discharges in transformers. (Electra, November, 1971, No. 19 p. 13) [2] MOLE G. - Improved methods of test for the insulation of electrical equipment. (Proc. LE.E., 1953, Vol. 100, Part IA, p. 276) [3] ADOLPHSON A.J. and VOGEL FJ. - Corona measurements in oil insulated transformers. (A.LE.E, ‘Trans, 1957, Vol. 76, Part III p. 798) [4] QUINN G, E, - A method for detecting the ionization point on electrical apparatus. (A.LE,E. Trans 1940, Vol. 59, p. 680) [5] NEMA - Methods of measurement of radio influence voltage (RIV) of high voltage apparatus. (1940, Publication 107)ne 12-03 [6] NEMA - Methods of measurement of radio influence voltage (RIV) of high voltage apparatus. (1964, Publication 107A) (7] GEHRIG E.H. - Discussion on paper by Adolphson and Vogel. (A.LE.E, Trans., 1957, Vol. 76, Part Ill, p. 801) [8] BROWN R.D. - Corona measurement on high voltage apparatus using the bushing capacitance tap. (LEE, Trans., 1965, Vol. PAS 84,p. 667) [9] NARBUT P. - Transformer corona measurement using condenser bushing tap and resonant measuring citeuits. (LE.E.E. Trans, 1965, Vol. PAS 84, p, 652) [10] HARTHILL S.R., SMITH L., JAMES R.E., TAYLOR F.W, and RYDER Dull. - Some aspects of internal corona discharges in transformers. (CIGRE, 1962, Paper No. 102) [11] RENAUDIN D. - Instrument for measuring ionization in transformers. (CIGRE, 1954, Paper No. 105, Appendix 2) [12] LE.C, - Partial discharge measurements. (1968, LE.C. Publication No, 270) [13] RAJU B.P., HICKLING G.H. and MORRIS L. - Experience with partial discharge measurements at more than one terminal of a transformer. (I.E.B. Conference, 1973, Publication No. 94, p. 48) [14] MASON J.H. - Discharges detection and measurements, (Proc. LE.E. ,1965, Vol. 112, No. 7, p. 1407) [15] DAKIN T.W., WORKS C.L. and MILLER R.L. - Utilisation of peak reading voltmeters and recorders for corona measurement. (LE.E.E, 1966, Paper 31 p. 66 - 341) [16] MOLE G. - Basic characteristics of corona detector calibrators. (LE,E.E, Trans., 1970, Vol. PAS 89, No. 2, p. 198) [17] DOUGLAS J.L. - Calibration of circuits for measuring partial discharges in EHV transformers. (LE.E. Conference, 1973, Publication No, 94 p. 40) [18] HYLTEN - CAVALLIUS N. - CIGRE Discussion, (CIGRE, 1968, Vol. 1, Group 12, p. 2 [19] KREUGER F.H. - Discharge detection in high voltage equipment, (Heywood, London, 1964, p. 145) [20] HARROLD R.T. and DAKIN T.W. - The relationship between the picocoulomb and microvolt for corona measurements on h.v. transformers and other apparatus. (I-E.E.E. Trans., 1973, Vol. PAS 92 p. 187) [21] DEMBINSKI E.M. and DOUGLAS J.L. - Calibration and comparison of partial discharge and ra interference measuring circuits. (Proc. LE.E., 1968, Vol. 115, No. 9, p. 1332) [22] KAWAGUCHI Y. and YANABU S. - Partial discharge measurement on high voltage power transformers. (LE.E.E, Trans,, 1969, Vol. PAS 88 p. 1187) [23] BERTULA T., PALVA V, and TALVIO E. - Partial discharge measurements on oik-paper insulated transformers. (CIGRE, 1968, Paper No. 12-07) [24] IZEKI N., KURAHOSHI U. and MATSUURA K. - Behaviour of oil corona and damage of transformer insulation, (ILE.E.E. Trans., 1971, Vol. PAS 90, p, 2330) [25] YAKOV S., MADIN A.B., HONEY C.C. and KEIL C, - Corona in power transformers. (CIG! Paper No. 12-06) [26] WHERRY F.E. , TOOTHMAN L.R., YAKOV 8. and PRESTON L.L. - The significance of corona ‘measurements on transformers. (ILE.E.E., 1968, Paper No. 68 TP 59 - PWR) 1968, Extmit de le Conférence Inremationale des Grands Réseaux Electriques Session de 1974