1. Introduction
Partial discharges (PDs) being one of the main assessment indicators of high-voltage insulation integrity, still offer a potential for new approaches in detection, mechanism interpretation, and analysis. This paper describes a novel observation related to partial discharges which occur despite the absence of an applied voltage, within a chopped sequence. The discharge pulses, called Partial Discharge Echoes (PDE), appeared in time intervals where no voltage was being applied, immediately after the specimen had been exposed to a base waveform (for example sinusoidal) voltage above the partial discharge inception level.
The chopped PD sequence opens new opportunities for analysis of dynamic phenomena inside dielectric materials, such as charge accumulation, the build-up of an internal field, including the role of the remnant field (due to dielectric polarization) that remains after the external voltage has been removed, space charge, time lag, as well as the processes of charge decay, related to the surface material conductivity [
1,
2]. The chopped timing is composed of a multiple series of packets consisting of base waveforms, with each packet being separated in time by a defined delay period. Whole epochs are repeated periodically throughout the measurement period [
1]. The presented experiments were performed on specimens with an embedded gaseous void in thermosetting insulation, and glass as a non-polar material for reference. In polar dielectrics, the remnant electric field plays an important role within the PD Echo voltage-less period. Additionally, the purpose of this paper is to analyze various scenarios of PDE mechanism, including multiple PD Echo (dependent on the coincidence of remnant polarization field), and the accumulated field from previous discharges, which is completely new with respect to the echo phenomena and have not been published before this.
Comparison of the PD, in polar and non-polar dielectric material, was also performed around the transition point from the high-voltage driven phase to the chopped period. The occurrence of discharges in this phase range requires special attention since they impact the PD Echo mechanism and the timing.
Acquisition of the echo signal was carried-out in a phase-resolved mode, modifying the synchronization path and the settings. The controlled high-voltage source could be implemented with Trek HV amplifier. In contrast to the continuous sinusoidal PD measurements, the chopped approach may provide deeper insight to key PD phenomena, such as inception, propagation, time lag, neutralization, post-discharge time decay, effective surface area, etc. [
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20].
This paper contains examples of the PD echo obtained in specimens composed of a void, in both a thermosetting material and in glass. The various events of echo mechanism, such as late (even tens of milliseconds after transition point) or multiple echoes, resulting from the interplay of remnant polarization field and accumulated field, have been discussed.
2. Partial Discharge Echo in Chopped Sequence
By applying a chopped sequence, aspects of the partial discharge process, not measurable via traditional methods, might be observed. An illustration of a chopped sequence is given in
Figure 1. In contrast to a standard continuous sinusoidal run, a discontinuous approach is characterized by time intervals of duration
td where no voltage is applied [
1]. The base waveform with a period
T (sinusoidal in the presented experiments, however, other shapes of a base waveform may also be applied) and the time delay, together form an epoch lasting for time
te. Each epoch may contain multiple base packets. The proportion of the duration of active waveforms to the duration of the whole epoch is called the fill factor
ff. The transition point from the sinusoidal part to the echo interval is extremely important and denoted as
T0 (red dot), in
Figure 1. For clarity, the time interval, between consecutive sinusoidal periods is denoted in the patterns, in time units, in contrast to the phase units of the sinusoidal part [
21].
Dielectric properties are determined by free and bounded charges. Free charges in the form of ions and carriers can move in the electric field. The mobility, whilst very small, is sufficient to observe the current flow, which is measured by the conductivity. Bounded charges are defined by the polar groups of atoms found in the main and the side chains of the polymer. These groups can rotate in the applied external electric field, which is equivalent to the flow of electric current, called displacement in the dielectric. In the electric field, dipoles subject a partial orientation that is the result of the equilibrium of the orientation, and a focusing on the external field, as well as, the confusing influence of thermasl movements. The relaxation time between different solid dielectric materials differs by a few orders of magnitude [
22]. The effective electric field inside the void
Ev is a superposition of applied external field
E0 (corresponding to the test voltage), which is transposed according to the field distribution and the shape of the void, by the factor
f, and the internal field
Eq, created by the charges accumulated on the void surface, during a previous discharge event:
where
τq represents the cumulative decay time constant of the
Eq field in the void, due to surface recombination, deeper trapping, and bulk conduction, neutralization, and other drifts. The
Eq field can be expressed as a decaying function of time:
where
Eq0 is the initial field value.
As illustrated in
Figure 1, at the transition moment
T0, at the beginning of the echo part, there would be a “quasi-DC” remnant field
Er, present in the void, due to the capacitive polarization of the dielectric
Ep, superimposed with the charge accumulated
Eq field (
Figure 2a). Hence, in the echo interval,
Er would coexist with the decaying
Eq field, as shown in
Figure 2b, assuming there is a much slower decay of the polarization field
Ep. As a result, both
Er and
Eq fields would be present in the voltage-less time interval, due to the fact that the driving field
fE0 will stop abruptly at the transition point
T0. In addition, the
Er field is related to polar dielectrics. After transition moment
T0, the void field would be a superposition of the remnant field and the accumulated field, taking into account their polarity:
where function
polEq controls the polarity of
Eq field at PD events.
A demonstrative illustration of this situation is presented in
Figure 2a. Just before the transition
T0−, the polarization field
Ep, based on the
fE0 but phase-shifted due to the discharges, would polarize and positively orient the dipoles in the dielectric. This is indicated by the yellow waveform, in
Figure 1 (the void field “frozen” as a remnant, quasi-DC field, is also marked in the chopped part, where the PDE signal appears).
Hence at
T0+, the void sees the remaining field that is memorized in dipoles and the field from the accumulated past events, which have compatible directions only once, after the polarity reversal (marked by a red arrow in
Figure 2b). The remnant field, except for the depolarization, is quasi-stable. After the first discharge, the directions of the
Er and
Eq field are opposite to each other, thus,
Eq is a weakening remnant field. However, a post-discharge decay of the accumulated field
Eq leads to the restoration, and an increase, of the resultant field, until a subsequent discharge event occurs, as is illustrated in
Figure 2b. The various scenarios and polarizations of this field are analyzed in
Section 5 of this paper. The presented experiments were carried out on a void, in a thermosetting material and, for comparison of non-polar behavior, on a void encapsulated in glass.
3. Measurement Setup and Dielectric Specimens
The experimental setup for a PD Echo measurement in a phase-resolved mode is presented in
Figure 3a. The partial discharge signal was detected using a measuring impedance
Zm, connected in series with a coupling capacitor
Ck = 1100 pF, then filtered and preamplified in the signal-conditioning unit SCU (signal conditioning unit and preamplifier). The acquisition was synchronized with the zero phase angle of the testing voltage and discharge pulses were accumulated, in a 256 × 256 matrix, in an ICM system (Power Diagnostix, Germany) connected to a host computer via GPIB (General Purpose Interface Bus) bus, during a
tm = 60 s measurement period. Thus, for the base period equal to
T = 20 ms (50 Hz), the total number of periods, within the measuring time
tm, was 3000. Depending on the fill factor in the chopped sequence, the number of active periods would be reduced accordingly.
The main difference in instrumentation and acquisition of the chopped patterns was related to the necessity of a high-voltage source and synchronization modes. The measurements were performed in a phase-resolved, wideband PD acquisition system. The chopped pattern was defined in a waveform generator driving the Trek Model 20/20B (Trek Inc., New York, NY, USA) high-voltage amplifier. The period of the sinusoidal base waveform was 20 ms. The synchronization was obtained from a compensated voltage divider connected to a high-voltage terminal. The measurement time for all experiments presented in this paper was 60 seconds. Partial discharge detection was provided from coupling impedance, followed by a filter and a preamplifier. The PD pattern including the echo was stored in a 256 × 256 matrix.
Following the frequency of the base waveform, at 50 Hz, the synchronization was adjusted to 25 Hz to visualize both the PD sinusoidal part and the echo interval. The position-ing of an acquisition window is illustrated in
Figure 3b. The two synchronization frequencies
fV and
fS are also marked. The first 128 channels in the acquisition matrix were devoted to recording the sinusoidal excitation and the remaining 128 channels were dedicated to recording the PD echoes [
1,
2].
One specimen contained an artificial void filled with air at atmospheric pressure. The thermosetting embedded void had a radius of 15 mm and a thickness of 0.1 mm. The thermosetting material was a composition of mica, estrofol foil, and epoxy resin. For comparison, a void (radius 15 mm, thickness 0.1 mm) placed between the two glass plates, each 2 mm thick, was used. On both sides, plain stainless steel electrodes were used. The electrodes were 40 mm in diameter, with an end curvature of 8 mm radius. During the experiments, the specimen was placed in an oil tank. Furthermore, the measurements were performed at room temperature.
5. Selected Scenarios of PD Echo Mechanism
Depending on the dielectric material (polar/non-polar) and the interplay of the remnant
Er field with the field from accumulated charges
Eq, inside the void, various PD Echo scenarios could be considered. Additionally, the time lag was also taken into account. This could last from hundreds of milliseconds for the low frequency (1 Hz), to individual milliseconds at a power frequency of 50/60 Hz [
2,
24,
25].
The internal remnant field
Er was obtained due to the polarization effect of the external field, in a polar dielectric. The dipoles were oriented partly along the field as an effect of the balance of orienting the action of the external electric field and disorienting the effect of thermal movements. Due to the voltage-less part of the sequence, the dipoles returned to a statistical average of disordered placement (the equilibrium state), which is usually described as an exponential relaxation process (in solid dielectrics, in the range 10
−4–10
4 second) [
18,
23]. In the presented experiments, the
Er relaxation time in a solid dielectric was assumed to be long, relative to the high-voltage period. The PD pulses, prior to the transition point
T0 and the PD Echo pulses in the voltage-less part, were observed in both cases of the void being embedded in glass and that of it being embedded in the thermosetting insulation. Following is an explanation of the potential mechanisms and scenarios for this behavior, depending on the presence and interactions of the remnant and the accumulated internal field, in the void.
First, consider the case, visualized in
Figure 8a, where an individual PDE pulse was recorded. This PD echo impulse was triggered due to the coinciding of both
Er and
Eq fields, after passing the inception level. The time delay was due to the time lag (
Figure 8a). The
Eq accumulated field is represented by a green curve in
Figure 8b, and the field inside the void is given in blue.
In the sinusoidal phase, the PDE mechanism was related to the difference of the external and the accumulated field, whereas, in the echo interval it was primarily related to the remnant field. The waveforms of the corresponding electric field components are shown in
Figure 8b. Visualization in the plot has not been done to scale, to pinpoint the discharge mechanism.
Cases where multiple echo pulses followed each other in a sequence, were also observed, as shown in
Figure 9a. In this case, after the first PDE pulse, as described above, the second one would be triggered, when the following condition was full filed:
Er − Eq > EPD_INC. After the first PDE pulse, the
Eq field had an opposite direction to the
Er field counteraction.
Eq field was found to decay with the time constant
τq, as indicated in
Figure 9b, and the moment when superposition of those fields exceeded the inception level (marked by a red dot in
Figure 9b), the subsequent echo pulse appeared. This was called a “late” echo, because it was observed often after tens of milliseconds, from the transition point
T0, depending on the decay time constant of the
Eq field. The visualization of this process is shown in
Figure 2b. In a positive polarity sequence, the PD Echo pulses also had a positive polarity. Thus, the second PDE pulse must have been caused by the
Er field as the
Eq field had flipped, and hence would result in a negative PDE, which was not observed.
Another scenario refers to the case when prior to the transition point T0, PD pulses occur at the end of the sinusoidal part. Here we can distinguish between the two cases observed in our experiments for polar dielectric (thermosetting specimen) and non-polar material (glass specimen).
First consider the case with a polar dielectric, where the remnant field
Er was maintained. A time snapshot of such an instance is shown in
Figure 10a, where three partial discharge echo pulses with positive polarity could be recognized. Prior to the transition moment
T0, the internal conditions in the void acted to support the discharge event (sufficient
Er + Eq field above inception threshold, initial electron available). After the first discharge
Eq, the field would flip, being in opposition to
Er for subsequent events. Hence at
T0, the remnant field was frozen and
Eq had weakened the internal field in the void. The decaying accumulated field would restore the void field, as previously described, and the trigger the event in the case of the fulfilling discharge criteria. Depending on the
Eq field decay time
τq, multiple PDE pulses might have occurred. This time constant could have different origins such as ionic drift, neutralization, conduction along the void surface and recombination, and trapping into the dielectric or bulk conduction. Depending on the trapping level, some electrons might have been de-trapped, in this process, for a subsequent discharge. It should be noticed that in the presented scenarios it was assumed that the
Eq field had always changed its polarity after the first discharge, when both
Er and
Eq were coherent. One could imagine a situation, where the accumulated field was substantial and one discharge event was not able to change the polarity. In such a situation the process would be more gradual.
Now consider the PD echo observed in the glass specimen (
Figure 6b) where discharges, both prior to
T0 and in the echo time interval, are recorded. In the case of non-polar dielectric, the PDE pulses were obtained at relatively high multiples of the PD inception voltage, for sinusoidal voltage (2.4
U0), as the remnant field
Er, supporting the
Eq field in the case of the first discharge after polarity reverse, was not present. This case could be split into two parts. The group before
T0 could be formed in a regular sinusoidal sequence, especially when the voltage was high, with respect to the inception level.
This did not address the question of how the echo pulses were generated. Since glass is a non-polar insulating material, no remnant electric field remained after T0.
The observed echo pulses had a positive polarity, thus, they could not have been caused by a flipping of the
Eq field. The intensity of the PDE versus prior
T0 discharges, whilst relatively low, was still significant and observable. Thus, a possible explanation (illustrated in
Figure 11) might relate to the scenario when no prior
T0 discharges had happened, in some sequences, and there was an accumulated
Eq field, including in the time lag condition, as this would have resulted in an intermittent echo discharges, spread along the entire delay time interval. However, in this case, less probable was the occurrence of multiple echoes in a single sequence. The measurement observation recorded in
Figure 6b supported this hypothesis, as the partial discharge echo pulses covered the delay time period
td. The mechanism visualized in
Figure 11, also shows the interplay of the two effects, i.e., first related to the time lag and the electron availability, and the second one associated with the decay of the
Eq field.
6. Conclusions
This paper reported a novel observation on partial discharges, which occurred despite the absence of an applied voltage, within a chopped sequence. The chopped sequence allowed to separate the consecutive periods of high-voltage and provide insight into the physical PD mechanism. The Partial Discharge Echoes pulses were recorded in the chopped sequence in time intervals where no voltage was applied. Experiments were performed on specimens with an embedded gaseous void in the thermosetting insulation and glass. They demonstrated a comparison of both the chopped effect and the PD Echo, in polar versus non-polar dielectric material, revealing the impact of a remnant field.
Various scenarios of PDE mechanisms, including multiple PD Echo, were analyzed, depending on the coinciding of the remnant polarization field (present only in polar dielectrics) and the field accumulated from previous discharges, which is a novel observation in the echo phenomena. It was shown that in the sinusoidal phase, the PD mechanism was related to the difference of external and accumulated field, whereas, in the PDE interval it was primarily related to the remnant field and accumulated field for polar dielectrics, and only to Eq field, in the case of non-polar insulating materials. In the case of polar dielectrics, multiple echo pulses were observed. A “late” echo, occurring after tens of milliseconds from the transition point was also observed. For the above-mentioned effects, the corresponding PD mechanisms and the interplay of electric fields, in the void, was explained and visualized.
It was detected that the PD pulses occurring just prior to the transition point T0 influenced the PD Echo occurrence and timing. Acquisition of PDE was demonstrated and a mechanism based on the superposition of a remnant polarization field, and an accumulated internal field, was explained.
In contrast to the continuous sinusoidal PD measurements, the chopped approach might provide deeper insight into key PD phenomena, such as inception, propagation, time lag, post-discharge time decay, and effective surface area. Special focus was paid to the transition point between the sinusoidal phase and the echo interval. The various scenarios of echo mechanisms, depending on the coinciding of the remnant polarization field and the accumulated field, on void walls, were analyzed and explained. The preliminary PD echo observations confirmed a high potential for research into its interpretation, further analysis, and application.