You are on page 1of 7
Partial Discharge: Overview and Signal Generation Steven A. Boggs, Underground Systems, Inc. is paper is the first of a series of papers by I various authors on the subject of partial di charge. The initial papers will treat the funda- ‘mental aspects of partial discharges to give the reader some background. Subsequent papers will treat PD detection and measurement for specific types of equip- ‘ment, such as transformers, solid dielectric castings, GIS solid dielectriccables, laminar dielectric cables, ete ‘The present author will write several ofthe papers and will act as editor for the series Why is partial discharge a “hot” topic? Partial dis: charge can occur in solids, liquids, and gases in ways which will not affect the breakdown voltage during a high voltage” proof test. Yet partial dischargewilleven- tually cause failurein most systems, possibly years after the system is placed in service. Partial discharge (PD) ‘can be generated by many phenomena. Froma practical point of viow, partial discharge occurs when some form of electrical activity within the system results in a rapid change of the electric field configuration which causes a current to flow ina conductor connected to the exter- nal world. To be recognized as PD, the current induced in the external conductor must be sufficiently large to be detected and must occur with a sufficient repetition rate tobe recognizableas something other than random Partial discharge tests are applied to an extremely wide range of high voltage apparatus, from miniature power supplies for night vision goggles to multi-ton power transformers. A wide range of detection techniques are employed. For some types of apparatus and under some circumstances, acoustic detection isthe technique fof choice. In other situations, optical detection is most sensitive, whilein most applications electrical detection is employed. Through discussions of partial discharge testing undera wide range of circumstances, the reader can get an overview of the technology and appreciate the wide range of options available, Such “cross fer tilization” in so important and ubiquitous a subject area can only be to everyone's advantage. ‘This first paper will deal with how a partial discharge ‘within a test sample generates a signal which can be measured outside the apparatus under test. As will be demonstrated, the transfer function between the phenomenon within the apparatus under test and the The most common sources of PD are floating components, corona, and voids. The PD-in- duced current in an external circuit is a complex function of the nature of the dis- charge-inducing effect and the geometry of the system. signal which can be measured in the “outside world can vary widely. Thus, measured partial dischargemag ritudes rarely have absolute meaning and are often referred toas “effective partial discharge magnitude" to acknowledge this uncertainty. ‘The most common sources of PD are floating com- ponents, corona, and voids. The following will consider tach of these phenomena in turn and explain the fun- ddamentals of how the phenomenon generates a current in the external circuit. For the time being, the external circuit will be thought of as a voltage source atthe far end ofa connection witha characteristic impedance, Z: Floating Components ‘Thecase ofa floating component within a high voltage system is considered first, not because it is most impor tant, but because it is simplest and can be used to illustrate some of the most important “realities” related to partial discharge analysis. Consider, for example, a spacer in a coaxial bus duct, as employed in gas-insti- lated substations and isolated phase bus. In this ex ample, the system is designed so that the conductor can, slide through the spacer to accommodate differential ‘thermal expansion of the enclosureand conductor. The spacer is cast onto an aluminum insert which supports, theo" rings on which the conductor sides, Fig. 1. Ifno connection is made between the aluminum spacer in- sert and the conductor, the spacer insert will be "float ing," ie, tied to no well-defined electrical potential. If the spacer inset is floating and the potential difference (voltage) between the spacer insert and conductor is, (0883-7554/80/0700-0033$1.00 © 1990 IEEE IEEE Electrical insuaition Magazine July/August 1990-Vol6, No.4 33 sufficiently lange to break down the gas gap between them, then a discharge will occur which (approximate ly) equalizes the voltage between them. ‘One obvious question is "Under what circumstances will a significant voltage develop between the spacer insert and the conductor?” At frst glance, one might think that the voltage of the spacer insert remains fixed while the conductor voltage waries, and thus a large potential difference between them is inevitable. How- fever, this isnot correct. Let us assume that the system starts at zero voltage with no charge on any component. Now we raise the ac voltage slowly to operating volt- ‘Toddemonstrate concrete example of the relationship between the (normally || unmeasurable) charge transfer within the PD test object and the detectable charge transfer Capparent PD mag nitude’ betwoen the high voltage test Source and PD test object, we will con- sider the following numerical example | For the floating component shown in Fig. 1, we assume an average radial slstance between the spacer insert ard ‘condictor of | mim (40 mils) in SFe ata pressure of about 400 kPa (bars ab- Solute or 44 psig). Further, we will as- sume 350 V (1550 KV BIL) system ‘geometry witha conductor diameter of 180 mm (7 in) an enclosure ID of $40 ‘mm (2 in), and a spacer insert length ‘of 10 cn (in). Forte range of spacer {ert radial thickness from Oto 20 mat ‘and spacer deletric constant from 1 to 40, we can calculate a peak potential difference between the spacer insert and conductor for an applied voltage of 425 kVpeak as shown in Fig. 2. For the above geometry the breakdown volt lage between the spacer insert an con- “actor might bein the range of 8 KV, taking into account probable stress en- hhancerents in the system, Thus even fora 2cm thick spacer insert, at about SkV thevoltagedeveloped between the {inset and conductor would be insu dent to cause breakdown and PD at a spacer dielectric constant of 1. "The charge transferred between the spacer insert and conductor at each PD {just he potential difference between the conductor and spacer insert at the time of the PD multiplied by the ‘capacitance between them, which is bout 500 pF for the present case. For 34 IEEE Electrical Insulation Magazine kV potential difference at breakdown, the. charge transferred is about 4000.00 pC oF 4 HC. However, this dlischarge transfers buried within the System. All that can be detected from the outside is the current which flows om the high voltage test ead energizing thetestpeceorthe curent which lows between the testpiece and ground. Just ‘because an electron leaves the conduc tor does not mean thatan lectron flows past every point inthe circu. Indeed, Anclectron movinga distance AR across the conductor surface does not cause any current to flow in the conductor ‘The current in the conductor is caused by the motion of the electron from the conductor to ground, not by the electron leaving the conductor [1] ‘The charge transfer between the con- ductor and the spacer insert (which {generally takes place in about Ito 2ns) uses the voltage to drop on the co- Guctor, This creates a potential dif {erence between the conductor and the voltage source. A current then flows to climinate ths voltage diference. This ‘currents capactively coupled through the test object to ground. Only thiscur- rent, wich floes from the external G- cuit, the “effective PD magnitude” can bbemensured from theoutside world. To 2 first approximation, the equivalent ‘Srcuit ofthe system i as shown in Fi, 1. At each discharge, a. momentary sort occurs across C1. C2 has a value fof about 2.8 pF. Thus with 425 kVpesk Applied tothe conductor, the voltage on ‘the spacer insert would have a peak value of I8KV ifn discharges were to ‘occur. For breakdown of C1 at 8 RV, C2 ‘would support about I85 KY. ARer the age. Imagine, for the moment, that the spacer has a dielectric constant very near -I(thesameas the gas) and that the spacer insert hasa small radial thickness. In this, case, the potential of the spacer insert is determined by capacitive voltage division between the conductor and the spacer insert, and the spacer insert and the enclosure. Asthe thickness ofthe spacer insert becomes small, the distortion of the electric field becomes small. Thus the presence ofa floating thin metal spacer insert does not, yer se, imply afield distortion and PD. As the Spacer insert becomes thicker, the capacitive division becomes distorted and the electric field in the gas gap ea PD Calculations - A Numerical Example breakdown of 1,C1 willbe supporting ‘no voltage and all the voltage (093 KV) trill occur across C2. After the break: down, weassumeC! supports no vol ‘age, so the circuit can be analyzed just {as if CL were sl shorted, eventhough itis not ‘The capacitance of the system to ground before shorting of Cl is C (er! + Cx'V" = 20908 pF The charge fon this capacitance at 193 KV is $410253966 1C- After the shorting of Cl, C2 supports the full 193 kV, s0 the | charge onthe system is 41998846 jC. ‘Thus the charge which flows through the external citeuit when a partial dis charge equalizes the voltage across C1 is 41998846 - 40253965 UC or 174,488 PC. Although 4,000,000 pC flows ‘within thesystem this causes.an effec tive" (ie, externally detoctabe) PD of only 174488 pC, a reduction of 9%. Using other geometric factors (ex, a {gap of 0.1 mm and breakdown voltage FT KV), this reduction could haveboen made more than 99.9%, "This simple calculation clearly demonstrates the highly variable relationship between the charge tans- fer within the testpece, which i some how related to thedamagetherein,and | theexterally detectable signal,"theap- parent PD magnitude,” on Which any fest judgements must be based. The situation s very simula for PD in other situations, such asthe more important ‘ase of Void (gas-filled eavity) within ‘atest objet. However for that case, the ‘aleulation of charge transfers is far ‘more comple and willbe dealt within 4 paper devoted solely tothe subject of void discharge July/August 1990-VoL6, No.4 |- Fig. 1 Typical geometry ofa discshaped epoxy spacer sup- porting the conductor in SF,-ineulated bus duct If the n- tended connection between the spacer insert and the conductor is ineffective, then partial discharge between the spacer insert and the conductor is probable. The PD is nor- mally sufficiently well shielded that it will not affet the breakdown voltage of the system, which makes this defect particularly insidiousas itcan be missed during high ve and toxio byproducts which, over a period of months, will ead to system failure. Asa result of the difficulty of making simple, inexpensive low current sliding contacts in SF, this was a common problem in a few carly designs. The obvious model for the system is two capacitors in series. between the spacer insert and the conductor becomes larger than the electric field elsewhere in the gas. However, when we increase the dielectric constant of the spacer from 1 to about 4, which is a more typical effective dielectric constant fora filled epoxy spacer, this has a dramatic effect, as the capacitance between the spacer insert and the enclosure is increased by a factor of 4-which thereby increases (by 4) the potential dif- ference (voltage) between the spacer insert and conduc- tor,as demonstrated in Fig. 2. Even at zero spacer insert radial thickness, a spacer dielectric constant of 4 is sufficient to cause partial discharge under the condi- tions outlined inthe sidebar. Thus, the cause of PD from this floating component isnot simply "the fact" ofa gap, butrather the gap combined with the dielectric constant ‘of the spacer, When the voltage between the spacer insert and con: ductorbecomes sufficiently large tocausea breakdown, sufficient electric charge is transferred to bring, the ‘spacer insert to approximately the same potential as the conductor. If we assume a conductor spacer insert breakdown voltage of 8 kV for the above system, the IEEE Electrical nsualtion Magazine é o Fig. 2 This three-dimensional plot shows the peak vollage between the conductor and spacer insert (without PD) as 2 function ofthe inert radial thickness and spacer dielectric ‘constant. Note thatthe spacer insert radial thickness has relatively litleeffect, while the dielectrieconstanthasalarge ‘effec. The third set of lines in addition tothe Xand Y grids) are lines of constant potential difference at an interval of 4 KY, with the first ine at 6 kV. The parameters on which the ‘calculation i based are given in the sidebar potential difference (voltage) between the conductor and spacer inserts shown in Fig.3 which also includes, for reference, the potential difference which would occur without PD. Note the pattern of breakdowns (partial discharges), with three or four on each slope of the sine wave. This patter is characteristic of floating ‘components, The pattern tends to be fairly steady on a PD detector, and the number of discharges perhalf cycle increases in steps as the voltage is raised. Ifthe PD is audible, one can often hear the change in pitch as the number of discharges per half cycle increases with in- creasing voltage. ‘As demonstrated in the sidebar, the relationship be- tween the measured effective" PD magnitudeand what actually goes on in the system depends on geometric factors. I the location within the system is known, then ‘one can often compensate for such factors. However, if these geometric factors are not known, whichis usually the case, then one must proceed blindly on the basis of the externally measured "effective" PD magnitude. In ‘many systems, the externally measured “effective” PD. July/August 190-Vol6, No.4 38 ie | - Fig. 3. This figure shows the potential difference between the inductor and spacer insert which would occur without par- I discharges, and with partial discharges which occur at 8 KV potential difference between the conductor and spacer insert. Each breakdown (equalization of the potential be- tnvoen the spacer insert and conductor) causes aPD pulse in the external circuit. The patter of PD from a floating com pponent is characterized by a fairly constant aumber of dis: ‘charges per half eyele of excitation. magnitude scales inversely with system size. Thus for many systems the effective PD magnitude is roughly inversely proportional to voltage lass. a specification, states an acceplable effective PD magnitude of 5 pC, ‘exactly the same internal phenomenon that causes an effective PD magnitude of 10 pC in a 138 KV class component would cause only 4 pC in a 550 kV com- ponent, assuming that size scales with the BIL. ‘The purpose ofthe above detail and that in the sidebar is to illustrate some of the fundamental “truths” about PD measurement on a system (the floating componend) which is sufficiently simple that one can conduct the analysis without becoming "swamped" with the details. Corona Corona is a “partial discharge” in the sense that gas breakdown begins at a position of high electric field ‘within a system, but the breakdown “dies out" because the electric field decreases very rapidly as a function of distance from the highly stressed position within the system, The breakdown can die out for two reasons, viz., (1) that the region of high field is too small to generate a fully formed breakdown channel or 2) that the field falls to such a low a value that even a fully formed breakdown channel cannot propagate. 36 IEEE Electrical insulation Magazine Corona-induced partial discharge can be the ‘phenomenon of interest or it can be an irrelevant and interfering signal which tends tomask the phenomenon ‘of interest, A full paper in this series will be devoted to discussing corona in air in these two contexts, The present purpose is simply to review how a corona gen= ‘rates a measurable signal ‘A corona is a (partially) ionized region adjacent to a conductor. It causes a change in the electric field be tiween the conductor and ground. In very"rough" terms, fone can view the corona as an extension of the conduc tor. Assuich, the corona can be thought of as increasing, the capacitance of the conductor to ground. This causes drop in the voltage on the conductor, a potential difference between the conductor and the voltage source, and, therefore, acurrent to flow from the voltage Source to the conductor. This later current is detectable from the “outside.” The above view of corona simply as, {an extension of the conductor leads to some erroneous conclusions, as when such a corona dies, the ‘capacitance would return to the previous value, and no net charge would flow in the circuit. A corona is a partially ionized region of gas created because the tlectric field és suficiently high that whena free electron, ‘occurs, that electron will, on average, generate more than one additional electron (and positive ion). A corona is Full of positive and negative ions (or elec- trons). Thuseven though current ceases to flow into the region (4.2, the corona ceases to gro, the electric field does not immediately return to its previously high value, Before thatcan happen, thepositiveand negative jons must “flow” in the field toward the negative and positive electrodes, respectively, This on flow isa “real (Ge,,"physical”) current and isan additional component of the partial discharge signal. If the negative changes are in the form of electrons, they can propagate sulli- ciently fast to contribute to the measured partial dis- charge signal; however, massive positive (and, for an tlectronegative gas such as SFe, negative) ions flow so slowly that they typically generate a very small current ‘over a long period of time, and this is not detected by most PD measurement systems. ‘Thus a corona can be thought of as generating a PD signal inthe external world through three mechanisms. First the ionization ofa channel which tendsto ook ike an extension of the conductor and, therefore, increases the capacitance of the conductor to ground; second, rapid migration of electrons toward the positive clectrode in a system where negative charge flows as electrons; and third, flow of positive ions which tends to be too slow to be detected by most PD measuring systems, The time scale forthe first two phenomena is nanoseconds to microseconds, while that for the third phenomenon can be milliseconds or more. Corona tends to be repetitive, as once the region is cleared of July/August 1990-Vol.6, No.4 charge, the region returns to the conditions which generated the previous corona pulse. Corona in air is Sensitive toair velocity and other environmental factors, ‘which affect space charge near the conductor. In many gases, including. air, corona generated by positive and negative voltages differs substantially. ‘This is basically because of a fundamental asymmetry in nature: negative charge carriers (electrons) can be very light and mobile, and can, therefore, gain kinetic energy very rapidly froman electric field, while positive charge cartiers (positive ions) are always heavy and much less mobile. Further, the outside surface of molecules is made up of electrons, thus violent phenomena which "shake loose” charge from a ‘molecule free an electron and simultaneously create a heavy positive ion. In corona from a negative conduc tor, the electrons propagate away from the conductor in the direction of corona growth. Thus they can create further electrons through molecular collisions. In corona from a positive conductor, the electrons propagate toward the conductor and away from the direction of corona growth. In this case, electrons are generally detached ahead of the corona tip by photons generated within the corona. This is a very different _mechanism from negative corona and can result in very different properties forthe two polarities. For example, ifthe ac voltage i raised on a wire in ai, negative corona will normally appear frst in the form of very numerous small pulses. The space charge jenerated by this negative corona is very useful in grading the wire and increasing the voltage onset of positive corona, which takes the form of relatively in- frequent very large pulses. Such pulses can have suffi- cient charge and energy that individual pulses can adversely affect the operation of an ac resonant test st. Void Discharge ‘The most technologically important source of partial discharges is probably voids within a more dense dielectric system. Comona can often be detected acous- tically or optically, which has the benefit of both detect- ing and locating the source, but a void buried within a dielectric system generally cannot be seen, Detection of such defects is one of the most frequent uses of partial discharge analysis, Because void discharge isso impor- tant and because of recent progress in the under- standing ofthis subject, a full paper in this series will be devoted to the topic. The purpose here, therefore, will, betoreview only the most basic concepts related to void discharge. Describing why discharge withina void causesa signal in an external circuit is nontrivial. A rigorous explana- tion must be couched in electric fields and Maxwell's, \EEE Electrical insuoition Magazine ‘equations [2]. While such an exposition is essential for analytical work, few can boast so thorough an under- standing of electromagnetic theory that such an expos tion is easily conceptualized and integrated into intuitive reasoning. The following explanation is less precise but more easily understood at a conceptual level Consider a void in a solid between two electrodes. IF ac voltage is applied to the electrodes, an electric field ‘occurs within the void. If the voltage is raised to generate a sufficiently high field, then a free electron ‘within the void can result in a breakdown of the gas ‘within the void, Such breakdown involves substantial ionization within the void, and as with any conductor, this charge will tend to distributesoas tocancel the field within the void. Thus the field within the void will drop substantially (Wo roughly the value necessary to sustain the breakdown). Let us make the very rough assump. tion that the field within the void goes to zero. This is equivalent to substituting a conductor for the void. Placing a piece of metal within a solid dielectric will increase the capacitance between the electrodes, the voltage on the electrodes will drop, and charge must flow from the voltage source to the electrodes o main- tain the potential thereon. Now when the discharge ‘within the void extinguishes, as it must (since once the field drops too low, the mechanism by which energy is. fed from the electric field to the gas disappears), one is left with large numbers of positiveand negative charge carriers (ions). These ions Will end to drift to the void ‘walls so as to cancel the electric field within the void. Thus, even after the discharge-induced "plasma" disap- pears, the electric field within the void remains near zero. Thus the charge which flowed from the power supply to the electrodes as a result of the drop in the field within the void caused by the discharge (conver- sion of the void to a “conductor”) remains on the electrodes after the discharge, This charge flow, which can becalculated analytically for many important cases and numerically for any case, produces the externally ‘measurable PD signal. The time scale for void-induced PD pulses tends to be Ito 3 nanoseconds. ‘The mechanism described above for void discharge has much in common with a floating component. In both cases, the electric field within the discharge region builds asa function ofthe applied (ae) voltage through capacitive coupling and goes (more or less) tozero upon, discharge. Not surprisingly, therefore, the discharge pattern for void discharges very much like that shown above for a floating component. However, void dis: charge is normally of much smaller magnitude and is, much less predictable, as the walls of a void may be- ‘come semiconducting as a result of discharge-induced. chemical reactions, and the field in the void after the discharge depends on the nature of the gas, gas pres- July/August 1990-Vol6, No.4 37 sure, et, Thus while “ideal” void discharge is about the same as (although usually of much lower magnitude than) discharge froma floating component, in the “real world,” void discharge can be highly variable. “The onset and extinction voltages of void discharge often differ greatly: This is usually a purely statistical ceffect caused by the absence of an electron to initiate discharge atthe lowest possible vollage. Electrons are generated in atmospheric air by maturally occurring tadiationata rateof about 3 persecond-cm?. Thusifone ‘hasa void of only 1 mm’ (10cm), one would wait,on average, about 5 minutes forthe first naturally occur- ring electron, and this might nat be ereated ata position, within the void which could initiate discharge. As the voltage is raised, the probability of discharge initiating, increases for several reasons. For example, the increased tlectric field within the void lowers the energy required fof any naturally occurring disturbance to generate an electron; and the region within the void from which an electron can initiate discharge increases. Although under laboratory conditions, one may be able to iden tify a single pulse as being PD-induced, in the "real world” of PD testing, PD initiation is normally taken as, clearly identifiable signal on an oscillographic PD display, a least 10 pulses per second, ‘One very important characteristic of void discharge is that the “effective” PD magnitude scales inversely with, system size. If we take exactly the same void with the same breakdown characteristics and we place itat the same relative position (eg., halfway between the electrodes) in twosystems, one of which istwice thesize fof the other, the externally measured PD signal in the larger system will be half that measured for the smaller system, Thus PD testing for voids inthe highest voltage systems least sensitive although “quality” tends tobe ‘most critical for such systems as a result of a cecreased, ratio of BIL to operating voltage, relative to lower volt- age systems. Electrical Treeing in a Solid Dielectric Electrical treeing ina solid dielectric isa phenomenon ‘which results in the growth of a dendritic pattern of tubules which eventually transverse the electrodes and ‘cause failure. Electrical trees can initiate from metallic stress enhancements (as from an asperity on an ‘electrode, a “point” from a semiconducting layer protruding into the dielectric ete.) orfroma void within the dielectric as a result of void discharge. The mechanism by which an electrical tre "grows" is only partially understood and will not be reviewed here. Tn general, electrical tre initiation in most solids takes hours to years at realistic stresses. AS a result, a tree is not likely to initiate and be detected during PD mea 38 IEEE Electrical insulation Magazine ‘urements performed for the purposes of quality as surance, As well, electrical treeing,induced PD can be ‘of very small magnitude until minutes before failure ‘Thus PD-based measurements of electrical tring are fof much greater interest to researchers than to prac titioners, Treeing-induced PD occurs on a nanoseconds scale, often with several peaks ina waveform separated. by less than I ns. These multiple peaks are thought to be the result of the discharge cascading down multiple tubules within the structure ‘The mechanism by which a discharge within the tubules of an electrical tree cause a PD pulse in the ‘external system is essentially the same as for void dis- charge. Corona in Liquids So far as is known, partial discharge isa phenomenon which takes place in gases. For the cases of corona, voids, and floating components, the gas isa natural part of the system. In the case of electrical trees, the treeing ‘phenomenon creates tubules within the solid in which partial discharge takes place as part of the growth mechanism of the tree. Mechanistically, “corona” in highly purified liquids has most in common with clectrical teeing in solids; however, the time scale for growth of the dendritic pattem is much shorter as a result of the much lower viscosity of the material [3 The time scale drops from hours to years for solids, to microseconds to milliseconds for liquids, and to nanoseconds for gases, In another sense, PD in liquids hhas more in common with corona than treeing, because the liquid is self-healing in the sense of a gas, Le, itis fluid so that damage to a specific point within the system isnot necessarily permanent and stationary. The damaged material tends to mix into the overall fluid. ‘Thus PD ina liquid can be repeating as fora corona in a gas, but with a much longer time scale as a result of, the much lower charge mobility inthe liquid, and hence a longer time necessary to clear the PD-induced space charge from the system. In realistic (i.e, "somewhat dirty”) engineering liquids, partial discharge (and breakdown) can beinitiated by a wide range of phenomena, including density changes resulting from electro-hydrodynamic motion and cavitation, alignment of microscopic conducting par- tices, etc Summary In general, the PD-induced current in an external circuit is a complex function of the nature of the dis- charge-inducing defect and the geometry of thesystem. July/August 1990-Vol6, No.4 {As PD testing is often carried out precisely because one cannot see into the system, one often has little JImowledge of the defect-elated geometry within the system. Thus an externally measured PD magnitude has no absolute meaning, and external calibration is, relative rather than absolute. At best, it assures that the same test piece will give the same PD magnitude in, different laboratories. It does litte to relate the external, measurement to the nature of the defect within the system, ‘The nature of the pulse magnitude distribution on the power frequency cycle, the time evolution thereof, and the variation with applied voltage can provide some insight into the nature of the internal defect, if not its position or absolute “size” within the system, These ‘aspects of PD measurement tend to be somewhat sys tem specific and willbe covered in paperson individual ‘components, ‘The next paper in ths series will concentrate on detec tion of PD signals, given that a PD has taken place within the test object and the charge on the electrodes. of the test abject has changed. Issues which will be discussed include the fundamental limits to detection sensitivity the tradcofls between sensitivity and broad applicability normally made in commercial PD detec- tion systems, and novel PD detection methods which allow both PD location and detection within a solid object. Steve Boggs was graduated with a B.A, from Reed College in 1968 and received his Ph.D. in physics from the University of Toronto in 1972. After bref stints in the fields of remote sensing and non-contacting in dustrial sensors, he settled down for 12 years at the Research Division of Ontario Hydro. During this ‘period, his research covered many aspects of gas-insu lated substations, solid dielectrics, and advanced ‘methods for partial discharge testing. In 1987, he joined ‘Underground Systems of Armonk, New York, where he continues consulting and R&D in the areas of substa- tions and underground transmission. References [11 Fora deta scusion ofthispoint ce forexample,M. Wetacr tnd PCT. van dr Lan, “rebrenkdown Caren: Base Interpret rand Tie resolved Messuement IEEE Pascoe Insuton, OL 24, Ne 2, Api 1988, 297308. [2] For an exelent ae dtalled expostn of this sub, se GC. Crichton, PW. Karson, and A Pedersen, "Partal Discharges in tipeotdal and Spor Vide” IEEE Tansatons on Fla Salton, VOL 24 No.2, Ap 96, p 305-342 [BY fothe authors knowl, the concept oF inking growth me to ‘cout as st suggested by Dr EO. Forster, ro- IEEE Electrical Insuattion Magazine FOR A FREE IEEE MEMBERSHIP INFORMATION KIT USE Working with a competitive edge makes the difference. IEEE—The Institute of Electrical and Electronics Engineers, Inc. your personal edge on technology. Join us! MEMBERSINP DEVELOPMENT The institute of Electrica and Electronics Engineers Ine 1 Hoes Lane, PO, Box 18, acta Ona, USA, (01) 802-524 July/August 1990-Vo.6, Nod 39

You might also like