INTRODUCTION TO FUNDAMENTALS OF PD DIAGNOSTICS

E. Lemke P. Schmiegel

LEMKE DIAGNOSTICS GmbH, Germany

Preface

The electrical detection of partial discharge(PD) can be regarded as an important tool forboth, quality tests on HV equipment in thelaboratory and diagnosis tests on site.For a better understanding of the existingprocedures for PD diagnosis tests somefundamentals on it shall be presented basedon simple assumptions. The paper isespecially intended as a guidance for thebeginner in this subject in order to introducehim in a simple way to the very complexmatter of PD diagnostics. In the followingimportant questions, which are mostly askedby the testing engineer dealing with PDdetection, shall be answered. In this respectthe reader is recommended, to studycarefully also the relevant IEC-Publication270 “Measurement of partial discharges“ [1].

1. How a PD is defined

Referring to IEC 270 [1] “a partial discharge,within the terms of this standard, is an electricdischarge, that only partially bridges theinsulation between conductors. Suchdischarges may, or may not, occur adjacentto the conductor. Partial discharges occurringin any test object under given conditions maybe characterized by different measurablequantities such as charge, repetition rate, etc.Quantitative results of measurements areexpressed in terms of one or more of thespecified quantities.“

The most important quantities, to be deter-mined within an electrical PD test, are theapparent charge and the inception /extinction voltage.

2. Why PD ignite

Let us consider a simple electrode arrange-ment, which may be presented by a polyethy-lene-insulated power cable (Fig. 1). Ifsubjected to a high voltage (HV), then theinsulation inside the coaxial electrodearrangement is stressed by an electrical field.Because of the non-homogeneous fielddistribution a maximum field strength

Fig.1: Field distribution in the presence of agas-filled void in the solid dielectric ofa cable

Em exist, which appears usually at the innerelectrode. The field strength is reduced moreand more as larger the distance to the innerelectrode is. The qualitative field strengthdistribution is drawn in the left part of Fig. 1.

HV equipment are designed in such a way,that the local stress never may exceed acritical strength, at which a breakdown occurand which is therefore called breakdownstrength Eb. Of course, in this respect all

possible voltage stresses in service have tobe taken into account, including transientover voltages.

Despite careful dimensioning technologicalimperfection may occur in the insulation, likestrange inclusions as cracks and gas-filledvoids. In the following a gas-filled void shallbe assumed in the insulation, which maygiving rise to the ignition of partial dischargesdue to the following reasons:

On one hand, the permittivity of the gaseousinsulation is always lower then those of soliddielectrics, which causes a local fieldenhancement in the air filled void, as drawnin the right part of Fig. 1. On the other hand,the breakdown strength of the gaseousinsulation Ebg is much lower than those of thesolid material Ebs. In Fig. 1 it is assumed, thatthe local maximum strength Emg in the gas-filled void exceeds the breakdown strengthEbg, giving rise to gas discharges. If theboundary to the solid insulation iscontinuously exposed to those discharges, itmay finally be deteriorated and lead topremature breakdown of the totalarrangement.

Conclusions

PD are mainly caused by a local fieldenhancement, due to imperfections in theinsulation, as for instance gas-filled inclusionsas voids and cracks. With continuedexposure to PD the insulation may fail. Anearly warning can be given if dangerous PDevents are detected.

3. Why PD pulses appear

Let us again consider the example accordingto Fig. 1. The critical field region in thesurroundings of the gas-filled void is drawnenlarged in Fig. 2. The assumed local fieldenhancement inside of the void, due to thelower permittivity of the gas compared tothose of the solid dielectric, is shown in Fig.2a. Under the assumed condition, that themaximum field strength in the gas-filled voidis higher than the breakdown strength Ebg, asudden ionization may be ignited. Thenumber of the generated positive charged

carriers is equal to those of the negativeones.

Due to the electrostatic field force thenegative particles move into the direction ofthe inner electrode, if its polarity is positive asassumed in Fig. 2. Finally they areaccumulated on the left boundary of the solidinsulation, as shown schematically in Fig. 2a.Accordingly the positive particles move intothe opposite direction and are accumulatedat the right boundary. The so separatedcharge carriers produces its own partial field.This is superimposed on the static field,determined by the electrode arrangement. Inthis context it is important to note, that thepolarity of the space charge regions in thevoid is opposite to those of the originelectrode arrangement, so that also thepartial field strength is inverted, as evidentfrom Fig. 2a. This results in a local fieldreduction inside of the void.

Fig.2: Changing of the field distribution dueto the formed space charge of asingle PD event

If it becomes lower than the critical value Ebg,as assumed in Fig. 2b, then an furtherionization is impossible, i.e. the chargegeneration stops. The time interval betweenignition and extinction of the ionization isextremely short, mostly below 100nanoseconds. Hence each PD event isaccompanied by an electromagnetic transientin the order of 100 ns.

Conclusion

Due to the formation of a space charge fieldwhich is directed opposite to the electrodefield, the ionization processes causes its ownextinguishing. The time from the ignition tothe extinction of charge formation is ingeneral below 100 nanoseconds. Theappearing electromagnetic transients canadvantageously be used for electrical PDdetection.

4. Why sequences of PD events appear

After a space charge is formed and thereforeafter the mean field strength in the void isreduced drastically, a new PD event mayignited again, if the field strength in the voidis increased again and exceeds thebreakdown strength. In principle two reasonsare responsible for subsequent individual PDevents:

Fig. 3: Equivalent circuit for PD ignition

1. If the stressing voltage remains constant,i.e. a DC voltage is applied, then thenecessary critical electric field can only bebuilt up again, if the usually formed spacecharge dissipates. This is mainly causedby the recombination and diffusion ofcharge carriers. In general this requires acomparatively long time, especially forvoids in high polymeric insulation, aspolyethylene (PE). Hence for suchdielectric material the PC repetition rate is

extremely low, which seems to be adisadvantage for PD detection under DCstresses.

2. Subsequent PD events are also ignited,if the to the test object applied voltage isincreased further so that the spacecharge field in the void may be againcompensated. This is the case foralternating voltages and transientvoltages like switching and lightning overvoltages.

Because mostly AC testing is done, theappearance of sequences of PD pulses shallnow be considered more in detail for alter-nating voltages. For this let us consider againa solid insulation having an imperfection likea gas-filled void, shown in Fig. 1. From thisconfiguration an equivalent PD circuit can bederived. The in Fig. 3 presented circuit baseson a proposal of GEMANT and PHILIPPOFF[2], developed already more than 60 yearsago.

Here the gas-filled void is substituted by acapacitor C1 ‚ to which a small air gap G1 isconnected in parallel, characterized by thebreakdown voltage U1 of the origin void. Theresistor R1 represents in connection with C1the transient process of charge formation,which can be expressed by a characteristictime constant:

T1 = R1 * C1

The both dielectric columns, which are placedbetween the void and the both electrodes,are substituted by the capacitor C21 and C22(Fig. 3a). Those can be summarized to thecapacitor C2‚ which is connected in serieswith C1 (Fig. 3b). The total capacity of the testobject is represented by C3. According topractical experience the capacity of all threecapacitors is extremely different, where thefollowing conditions are mostly fulfilled:

C1 » C2

C1 « C3

C2 « C3

If an alternating voltage u3(t) is applied to C3,then u1(t) occurs across C1, which is reducedaccording to the divider ratio C2 / C1 (Fig. 3c).The gap G1 across C1 breaks down, if thecritical voltage U1, is obtained. This is thecase, if the applied voltage u3(t) exceeds thecrest value of:

Uc3 = U1 * C1 / C2

If the gap G1 breaks down, i.e. the partialdischarge is ignited, then the voltage acrossC1 drops suddenly down within less than 100nanoseconds to a very low residual value.Therefore, the voltage jump across C1 ischaracterized approximately by thebreakdown voltage U1. After that the air gapobtains again its insulation performance.Hence, if the applied voltage u3(t) isincreased further, also the partial voltageacross C1 is increased again. The changingof the voltage u1(t) is again determined by thedivider ratio C2 / C1. Finally the critical voltageU1, across C1 is again obtained and a secondbreakdown appears. In this way a sequenceof PD occurs in both half cycles (Fig 3c). Thepulse repetition rate is mainly determined byboth, the frequency and the magnitude of theapplied voltage.

Conclusions

Due to the capacitive coupling of the PDsource to the terminals of the test object apartial voltage of the applied AC test voltageis transmitted to it. If the breakdown voltageis obtained, a suddenly voltage drop takesplace, i.e. a PD occurs. The partial voltage U1can be rebuilt again, if the applied voltage isfurther increased. In this way a sequence ofPD events appears.

From this follows also, that time intervalbetween PD events is increased more andmore as lower the frequency of the appliedtest voltage is. So at VLF or DC voltages therepetition rate of PD pulses becomesex t reme ly l ow , wh ich may bedisadvantageously with respect to PDdetection.

5. How PD faults can be classified

With respect to a simple consideration an

inclusion in the solid dielectric, consisting ofa gas-filled void, has been assumed aboveas an example. In practice, however, a lot ofdifferent PD faults may happen. The mostimportant kind of PD sources can berepresented to a limited number, representedby typical fundamental electrodearrangements, as shown in the following.

As previous discussed, PD are caused bylocal field enhancements and appear in weakregions, as in gas-filled voids. Of course, PDmay appear also in pure gas insulation aseither in ambient air or in SF6 insulation.Hence, the local field enhancement can belimited by a point electrode in air.Furthermore, dielectric interfaces can besimulated insulating barriers, if placedbetween the electrodes in a point-to-planegap, and will be shown below.

Let us first consider a PD source in theambient air. lt may be caused, for instance,by a small protrusion at the surface of an HVshielding electrode (Fig. 4.) This PD sourcecan be represented by a simple point-to-plane gap.

If such arrangement is subjected to an ACtest voltage, which exceeds the PD inceptionvoltage, then the first PD pulses are ignited inthe negative half cycle, which appear in thesurroundings of the crest of the appliedvoltage. Those PD events are known asTRICHEL pulses. If the test voltage isincreased more and more, then finally PDmay happen in the positive half cycle. Themagnitude of positive pulses is much highercompared to those in the negative half cycle,but the pulse repetition is much lower. In thiscontext it should be mentioned, that the pulserepetition rate of the TRICHEL pulses atincreasing test voltage may increase up tothe MHz range. Furthermore, they may alsodisappear, because of transition into a DClike current, so called pulse-less discharges.

Another typical PD fault may appear, if ametallic of floating potential approaches theHV electrode. Such constellation may ariseby corrision of connecting structures, asassumed in Fig. 5.

Fig.4: Protrusion at a HV electrodesimulated by a simple point-to-planegap

Under such condition the capacitive potentialdistribution may cause bursts of sparks. Theyappear at every half cycle in a region, atwhich the changing of the applied AC voltagebecomes a maximum, i.e. in the surroundingsof the zero-crossing between the half cycles.This arrangement can be simulated by apoint-to-plane gap, having an additional thirdelectrode between the intermediate space, asshown in Fig. 5.

Fig.5: Potential-free metallic part simulatedby a point-to-plane gap and a metallicdisc inside

A further kind of PD source is shown in Fig. 6. lt refers to surface discharges, which mayappear in the end region of coaxial electrode

arrangements, as PE cables. Thecharacteristic PD behavior can be simulatedby a point-to-plane gap, where an insulatingbarrier is placed between. This configurationis well known as TOEPLER arrangement.The appearing sliding discharges arecharacterized by extremely high PDmagnitudes, ranging between some 1000 pCand more than 100000 pC. They appear ineach half cycle before the crest of the appliedAC voltage is obtained and disappearimmediately after that.

Fig 6: Simulation of surface discharges by aTOEPLER arrangement

Finally also inclusions in solid insulation, asgas-filled voids and cracks, can be simulatedby a point-to-plane arrangement if aninsulating barrier is put between theelectrodes. Different from the previouspresented classical TOEPLER arrangementit has to be looked for, that between the pointelectrode and the insulating barrier a small airgap in left, as evident from Fig. 7.

Furthermore, the applied test voltage has tobe limited to such a magnitude that no sliding discharges may occur. Under thiscondition the PD pulses appear in the regionof the zero crossing at each half cycle.

Conclusion

The in practice known characteristic PDfailures can be simulated by simple point-to-plane arrangement. In this context it has tobe noticed that the presented examples areintended only, to give the beginner in PDdiagnostics a preliminary intention oncharacteristic PD events. In practice,

however, extremely complicate cases mayarise, which makes the recognition of criticalPD sources more difficult. lt should benoticed, that recent computer-aided PDmeasuring systems are available, which maybe helpful for further identification of PDevents. Two powerful PD diagnosis systems,called “LDS-4“ and “TEXpert“, are alsooffered by LEMKE DIAGNOSTICS.

Fig.7: Simulation of PD in gaseousinclusions

6. How the apparent charge ischaracterized

In order to detect PD events according to therelevant standard a coupling unit has to beconnected to the terminals of the test object,as specified in IEC 270 [1]. In this respect itseems to be of interest, which part of the“inner“ space charge, generated by a PDignition inside the origin PD source, can bedetected “outside“ at the terminals of the testobject, where the latter is quantified asapparent charge.

For this let us consider again the equivalentPD circuit according to Fig. 3. With respect tosimple calculations the following quantitativevalues shall be introduced:

- Breakdown voltage of the gap C1:U1 = 1 kV

- Capacity of the PC fault:C1 = 1 pF

- Capacity of the solid column:C2 = 0.01 pF = 0.01 * C1

- Capacity of the test objectC3 = 100 pF = 10000 * C2

lf an alternating test voltage u3(t) is applied tothe terminals of the test object, having thecapacity C3, then only about 1 % of the testvoltage occurs across C1 because it isdivided according to the capacitive ratio C2 /C1.

lf the crest voltage of u3 (t) exceeds a criticalvalue of 100 kV, then the breakdown voltageU1 = 1 kV across C1 is obtained. Thebreakdown event causes a voltage jump ofabout U1 = 1 kV. This is transmitted via C2 tothe capacity C3 and therefore reduced by theratio C2 / C3, which is previous assumed as 0.01 / 100 = 10,000. Therefore, the voltagejump U1 = 1 kV causes only a step of 0.1 Vacross the test object capacity C3.

Based on this example the “inner“ pulsecharge Q1 and the “outer“ pulse charge Q3can be determined in a simple way:

Q1 = U1 * C1 = 1 kV * 1pF:Q1 = 1000 pC

Q3 = U3 * C3 = U1 * C2 = 1kV * 0.01 pF:Q3 = 10 pC

Q3 / Q1 = C2 / C1

Q3 / Q1 = 0.01

That means, that the at the terminalsappearing pulse charge Q3 in essentiallyreduced, if compared to the origin pulsecharge Q1. The attenuation ratio is deter-mined by C2 / C1, which is in general notknown. Hence, the at the terminals of the testobject detectable charge is called “apparentcharge“ [1].

Conclusions

Compared to the in the PD source generatedcharge, the at the terminals of the test objectdetectable “apparent charge“ appearsextremely reduced. The attenuation is mainlydetermined by the ratio between the capacityof the solid column C2 and the equivalentcapacity C1 of the origin PD source. Bothvalues are not known generally. This fact canbe considered at the most weakest point with

respect to PD diagnostics.

7. How to calibrate the PD circuits

If we consider Fig. 3a and 3b, then it seemsevident, that the at C3 appearing pulse chargeresults from the voltage jump across the PDfault which is transmitted via C2, representingthe capacity of the solid column between thePD fault and the test object terminals.

This internal charge injection can also besimulated by an external charge injection intothe terminals of the test object, as shown inFig. 8. Here the gap C1 is substituted by afast voltage step generator Gc, which has tobe driven potential-free, i.e. it must bebattery-driven. In accordance to U1 a voltagestep of Uc is generated. The capacitor C2 issubstituted by a calibrating capacitor Cc.Under the condition

Cc « C3

the to the terminals of the test objecttransmitted calibrating charge can beevaluated in a simple way as:

Qc = Uc * Cc

Taking into account the charge formation inreal PD sources the duration of the injectionof calibrating pulses must be limited below100 ns.

Fig. 8: Calibration procedure for the apparentcharge

Based on the above presented procedure thePD quantity “apparent charge“ is defined inaccordance to IEC 270 [1] as follows:

“The apparent charge q of a partial dischargeis that charge which, if injected

instantaneously between the terminals of thetest object, would momentarily change thevoltage between its terminals by the sameamount as the partial charge itself. Theapparent charge is expressed in pico-coulombs.“

In addition to this the following notes aregiven:

1. The apparent charge is not equal to theamount of charge locally involved at theside of the discharge and which cannotbe directly measured.

2. In practice, the waveform of the voltageappearing across the terminals of the testobject due to the partial discharge itselfmay be different from that produced bythe calibrating pulse. The apparentcharge is considered to be such chargewhich, if injected between the terminals ofthe test object, will give the same readingon the measuring instrument as thepartial discharge itself. Special cases arethose in which the test object includetravel ing wave or attenuationphenomena.

Conclusions

The at the terminals of the test objectdetectable PD quantity “apparent charge“ isa measure of the PD event, but it is notequal to the amount of charge locallyinvolved in the origin PD source. In order toquantity the apparent charge, calibratingpulses have to be injected into the terminalsof the test object. The calibrator consists inprinciple of a battery powered step pulsegenerator, which is connected to the testobject via a small calibrating capacitor.

8. How PD events can be detected

As presented previous, the by the breakdownof the gap G1 caused voltage jump U1 istransmitted via C2 to the terminals of the testobject, having the capacity C3 .Hence it isreduced according to C2 / C3. This statement,however, is only valid, if the test object is notsimultaneously discharged. In practice thiscondition can not be fulfilled, as discussedbelow.

With respect to practical PD detection let usassume, that the test object represented bythe capacity C3, is grounded via measuringimpedance, which is simulated by a resistorRm in Fig.3. Then the equivalent timeconstant Tm of the PD measuring circuit canbe expressed as:

Tm = Rm * C3

According to physical studies of PD pheno-mena the charge formation in the PD sourcecan be approximated by an exponentialfunction, characterized by a pulse decay-timeconstant Tp. Then the voltage drop across C1is given by:

u1(t) = U1 * exp (-t / Tp)

In most cases the charge formation due to aPD event is finished after about 100 ns, asmentioned previous. This time corresponds toapproximately 3 * Tp .

Hence, the PD pulse decay-time constant of

Tp = 30 ns

shall be introduced for further quantitativeconsiderations.

In order to avoid a remarkable discharge ofthe test object capacity C3 via the measuringimpedance Rm (Fig. 3) during the spacecharge formation, the measuring circuit timeconstant should be chosen much higher thanthe pulse decay-time constant Tm . Thisrequirement is well fulfilled for

Tm > 10 Tp ; Tm > 300 ns

lf a measuring impedance Rm = 50 Ohm isapplied, which corresponds to thecharacteristic impedance of usual measuringcables, then this condition is only fulfilled fortest object capacities C3 larger than 6000 pF.

In practice the test object capacity is oftenmuch lower than 6000 pC. Under thiscondition the integration capability of C3 isreduced more and more as lower the testobject capacity is. To overcome this disad-vantage, commercially available PD detectors

for measuring the apparent charge aredesigned like quasi-integrators, as reportedbelow.

Conclusion

Due to the integration performance of the testobject capacity the apparent charge can bederived directly from the voltage jump, whichis detectable at the terminals of the testobject. For this, however, the time constantTm of the measuring circuit must be muchlarger than the pulse decay-time constant Tpwhich characterized the duration of thecharge formation inside the PD source.

For test objects of low capacity, say below5000 pC, this condition is not longer fulfilled.Hence, the amplifier of PD detectors isusually designed as a quasi-integrator,having a band-pass filter characteristic.

Fig.9: Characteristic step responses of thewide-band PD detector LDS-5obtained at the output CHARGE (a)and those of a narrow-band amplifier(b)

9. How PD detectors can becharacterized

As mentioned above, PD detectors formeasuring the apparent charge must have asuitable integration capability. That means,the characteristic integration time constant Timust be much larger than the pulse decay-time constant Tp . This condition is in generalwell obtained, if the condition

Ti > 10 * Tp

is fulfilled. With the previous introduced valueTp = 30 ns follows the requirement

Ti > 300 ns

Because the usual amplifiers can becharacterized by the invariant condition

Ti * f1 = 0.15

The upper cut-off frequency f1 can beexpressed as

f1 < 0.015 / Tp

For the assumed value of Tp = 30 ns follows,that the upper cut-off frequency for amplifiersmeasuring the apparent charge should belimited below f1 = 500 kHz.

The choice of the lower cut-off frequency f2seems not so critical. lt is mostlydetermined by the capability for elimination ofthe influence of harmonics due to the AC testvoltage. Practical PD test circuits arecharacterized by f2 ranging between 50 and200 kHz.

Depending on the used bandwidth

Bf = f2 - f1

lt can be distinguished between wide-bandand narrow-band PD measuring systems.The wide-band PD processing ischaracterized by a bandwidth close to f2 ‚which obtains in practice several 100 kHz. Asan example Fig. 9a shows the step responseof the PD detector LDS-5.

The narrow-band processing is characterizedby a band-width which is much lower than f1and causes an aperiodic oscillating responseas evident from Fig. 9b. The damping of theoscillations becomes lower as lower thebandwidth is. Hence, the advantage ofselective amplifiers with respect to theelimination of radio interference (RIV), maycause fatal superposition errors if subsequentPD events occur. Hence, attention has to bepaid, if narrow-band amplifiers are used as

also remarked in IEC 270 [1]. Furthermore,the pulse polarity can not be recognized,disadvantageously for PD fault location.

Conclusion

In principle it can be distinguished betweentwo procedures of measuring the apparentcharge, the wide-band and the narrow-bandPD detection.

With respect to RIV noise suppression theapplication of the narrow-band procedureseems advantageously. In this respect,however, it has to be taken into account, thatfatal superposition errors may occur, if thePD pulse repetition rate exceeds some kHz,because of the comparative long setting timeof usual narrow-band amplifier, especiallydeveloped for RIV measurements. Narrow-band amplifier may not identify the pulsepolarity and can in general not be applied forPD fault location.

Hence, the wide-band-PD detection shouldbe preferred, if possible. The advantage ofthis is, that superposition errors may occuronly, if the repetition rate exceeds several 10kHz. Furthermore, it offers advantageouslypossibilities for PD fault location. In thisrespect also the advantages for computer-aided PD processing have to be mentioned,which is extremely restricted in case ofnarrow-band PD detection. More advantagesof wide-band PD detection are presented inthe following.

10. What are the advantages of the LDS-5

In commercially available PD detectors therequired PD pulse integration is usuallycarried out by amplifiers of a band-filtercharacteristics, as described above.

Fig.10: Bloc diagram and frequencycharacteristics of the LDS-5,compared to those of conventionalPD detectors

In the LDS-5, however, the necessaryintegration is carried out separately from theamplification of the PD pulses. That means,the PD pulses are first amplified, for which awide-band amplifier is used, which upper cut-off frequency in more than 50 times higher,than those required for the PD pulseintegration in order to obtain the apparentcharge. The integration is not made by aband-pass amplifier but by a specialdeveloped electronic integrator, which can becontrolled and which separates the PDpulses according to their polarity. Withrespect to the detected apparent charge,however, there is no fundamental differencebetween the LDS-5 and other commercialavailable PD detectors, if only the PD inputand the output apparent charge like a blackbox (Fig. 10) is considered.

The in the LDS-5 used integrator is charac-terized by an integration time constant Ti ofabout 0.5 µs, which corresponds to anequivalent upper cut-off frequency of about300 kHz.

The main difference between the LDS-5 andother available PD detectors is characterizedby the following:

a) The output “AMPLIFIER“, characterizedby an upper cut-off frequency above 30MHz is advantageously applicable for PDfault location on long lengths of PE-cables, based on the well known PDpulse-reflectrometry. Furthermore, theidentification of PD events according totheir origin occurrence seems useful forstudies in research and education,dealing with the nature of PD. To thisoutput also a spectrum analyzer or apowerful computer can be connected.

b) The output „POLARITY“ offers additionalpossibilities for PD fault location on shortlengths of cables, using the differentialPD bridge procedure.

c) The output “SCOPE X“ and “SCOPE Y“allow additionally to the usual display ofPD pulses on a linear time-base also aconvenient display on an elliptical time-base which rotates synchronously withthe test voltage frequency.

d) The modes for noise reduction increasethe field of application not only in wellshielded test laboratories but even undernoisy condition. The influence of radiointerference voltages (RIV) can bereduced if „RIV RED“ is switched on.Furthermore, stochastically appearingpulse shaped noises, as for instancecaused by thyristors or switching overvoltages, can effectively be suppressed.For this they are picked up by anelectromagnetic sensor, which isconnected to the input „SENSOR“.According to the noise magnitude thedesired trigger level can be adjusted.Under this condition measuring windowsare generated automatically, free ofexternal disturbances.

e) The pulse resolution performance of theLDS-5 can be evaluated as excellentcompared to those of other PDdetectors. No significant superpositionerrors occur down to a pulse distance of

a few of µs, which corresponds to apulse repetition rate of more than 100kHz. Of course, in practicalmeasurements superposition errors arenot recognized, because no comparativemeasurement is done, using anequipment free of superpositionphenomena.

Conclusion

In the LDS-5 the PD pulse processing is donein the following two steps, first the extra wide-band amplification and after that an electronicintegration. This procedure offers additionalpossibilities for evaluation of PD events. Sobesides the detection of the apparent chargea pulse reflectometry can be done withrespect to PD fault location. Furthermore,possibilities for reduction of the influence ofexternal noises are given.

References

[1] Partial discharge measurements. IEC Publication 270 (1981)

[2] GEMANT, A.; v. PHILIPPOFF: DieFunkenstrecke mit Vorkondensator.Ztschr. f. techn. Physik 13 (1932),S. 425-430