Setup for Pulsed Breakdown Testing of Dielectrics

G. A. VOROB'EV

(Presented by Professor Doctor A. A. Vorob'ev)

At the present time, pulsed tests of solid dielectrics for breakdown are being widely carried out. In pulsed tests it is possible to determine the true electric strength of solid dielectrics, when such factors as thermal effects and the influence of space charge have been reduced to a minimum, and to obtain the volt-second characteristic, which makes it possible to determine the time lag of the discharge in the dielectric.

The performance of pulsed breakdown tests of dielectrics has a number of special features. For a long time, in pulsed tests, precise determination of the breakdown voltage and of the duration of voltage application was difficult. In some works the amplitude of the pulse, measured by a sphere gap, was increased in steps (by 10%) until breakdown occurred. Consequently, before breakdown the dielectric was subjected to the action of several pulses, and its breakdown voltage, owing to the cumulative effect, could have decreased, as was shown in the works of B. M. Vul [1], A. A. Vorob'ev [2], and others.

It is known that, when short pulses are measured by a sphere gap, considerable errors may occur. Experiments by A. F. Walter and L. D. Inge [3] using an electronic oscillograph showed that the breakdown voltage of a sphere gap irradiated with ultraviolet light increases, for a pulse duration of \(10^{-6}\) sec, by 2%, and for a duration of \(10^{-7}\) sec—by 10–12% in comparison with the static breakdown voltage.

Often the waveform of the pulse voltage was calculated from the parameters of the generating circuit. The calculated pulse shape may differ greatly from the actual one, owing to failure to take into account the parasitic inductance and parasitic capacitance of the generating circuit and the capacitance of the specimen.

In the works of Jost [4], Strigel [5], and others, the pulse was formed in a long line. The voltage was measured and the specimen was connected. The oscillograms presented in the work of Dittert [4] show that connecting a dielectric specimen may significantly distort the shape and reduce the amplitude of the pulse.

The time of voltage application up to breakdown of the specimen was determined with still greater errors.

A. F. Walter and L. D. Inge [6] point out that in the works of Handreck, Jost, and others it was not possible to determine precisely the time of voltage application ...

Since no electronic oscilloscope was used, during the breakdown of the specimen it was possible to say only that the time to breakdown was no greater than the wave duration.

The problem of accurately measuring the pulsed breakdown voltage and the time of action of the voltage was solved with the improvement of high-voltage electronic oscilloscopes.

There are individual cathode-ray tubes with a voltage on the deflection plates of 100 kV [7]. However, the majority of cathode-ray high-voltage tubes have a maximum voltage on the deflection plates of 1–2 kV. For breakdown of dielectrics, as a rule, higher voltages are required. In order to apply a reduced pulse voltage to the deflection plates, voltage dividers are used.

For calibration of the oscilloscope along the horizontal axis (the time axis), a high-frequency generator (H.F.G.) is needed.

Thus, an installation for pulsed breakdown testing of dielectrics must contain the following elements: a pulse generator, an electronic oscilloscope, a voltage divider, and an H.F.G.

In the laboratory of the High-Voltage Engineering Department of the Tomsk Polytechnic Institute, in order to obtain volt-second characteristics of solid dielectrics of small thickness (tenths of a mm), an installation was constructed and several circuits were investigated. Separate elements of this installation are described in this article.

Pulse Generator

There are various circuits for forming a pulse: pulse-forming lines, circuits with capacitances and resistances, pulse circuits in combination with a pulse transformer, and some others.

Long lines make it possible to form rectangular pulses with a short front duration. However, the maximum pulse duration obtainable by means of long lines under laboratory conditions is limited to a value on the order of \(10^{-6}\) sec, since, to form pulses of greater duration, lines several thousand meters long and more would be required.

A pulse transformer does not make it possible to obtain at the output pulses with a front duration of less than tenths of a microsecond.

Circuits with capacitances and resistances make it possible to obtain pulses of various durations.

Figure 1 shows the circuit of a pulse generator by means of which pulse voltages with durations from \(10^{-7}\) to \(10^{-4}\) sec were obtained. As the switching and synchronizing element, a trigatron \((P)\), proposed in the 1930s by I. S. Stekolnikov, was used.

For synchronizing the operation of the pulse generator and the electronic oscilloscope, a delay circuit \(R_1 C_1\) was used. Changing these parameters makes it possible, over wide limits, to vary the relative switching time of the electronic oscilloscope and the pulse generator.

When the trigatron operates, the generating capacitance \(C_1\) discharges through the resistance \(R_1\) into the resistance \(R_2\) and capacitance \(C_2\). A double-exponential voltage wave arises on the capacitance \(C_2\).

In our experiments, volt-second characteristics were investigated. In this case, the change in voltage exposure was achieved by changing the length of the pulse front. Since in our experiments the wave duration was always much greater than the front duration, the latter was determined approximately by the value \(3 \div 4 R_1 \cdot C_2\).

To obtain the shortest front length, it is necessary to reduce the values of \(R_1\) and \(C_2\) to a minimum. In this case the front length may be different.

1. In the case of large values of inductance and capacitance with respect to ground in the generating circuit, the front length is determined by the half-period length of the oscillations arising in the generating circuit.

Fig. 1. Diagram of the setup for breakdown testing of dielectrics at voltage exposure times from \(10^{-7}\) to \(10^{-4}\) sec

1. In the case of small values of the parameters indicated in item 1, the length of the impulse front is determined by the voltage fall time across the discharger \(P\) during its breakdown.

Figure 2 shows the electrical circuit of the setup by means of which impulses with the shortest front length were obtained and recorded \((3 \div 4 \cdot 10^{-8}\ \text{sec})\). On the flat part of the impulse there were weak oscillations caused by the parasitic parameters of the generating circuit.

Fig. 2. Diagram of the setup for breakdown testing of dielectrics at a voltage exposure time of the order of \(10^{-8}\) sec

In view of the difficulty of synchronizing the impulse generator and the time sweep of the oscilloscope by means of a trigatron when recording phenomena of duration on the order of \(10^{-8}\) sec, the generating circuit and the time-sweep circuit were rigidly connected. The resistances \(R_1\) and \(R_2\) (Fig. 2) were calculated approximately from the condition of obtaining the necessary recording speed. As a first approximation it may be assumed that a wave with a sloping front acts on the chain \(R_1\), \(R_2\), and the sweep circuit. Then

one obtains the calculation circuit shown in Fig. 3, in which the transient process can be easily calculated.

Fig. 3. Circuit for calculating the discharge resistance of the pulse generator

Knowing the voltage \(U_0\) required for breakdown, the voltage on the time-sweep plates, their sensitivity with respect to deflection of the electron beam, the diameter of the screen of the cathode-ray tube, and the required speed of the time sweep, it is possible to determine the values \(R_1\), \(R_2\), and \(R\), if the value \(C\) and the ratio \(R/R_2\) are specified. In this case, somewhat overestimated values of \(R_1\), \(R_2\), and \(R\) are obtained, since the length of the pulse front is comparable with the sweep time. More accurate values of \(R_1\), \(R_2\), and \(R\) can easily be established experimentally.

The circuit of Fig. 2 has the significant drawback that, after breakdown of the dielectric specimen, the sweep speed may change abruptly, and the phenomenon (for example, the voltage drop across the specimen at its breakdown) will be distorted. Therefore we proposed another circuit, experimentally investigated by K. K. Sonchik [9] and shown in Fig. 4.

Fig. 4. Circuit of an installation with two circuits for testing dielectric breakdown with a voltage rise time of the order of \(10^{-8}\) sec

For producing and recording pulses with a front length of about \(0.1\) sec, the circuit shown in Fig. 5 was adopted. Large values

the resistances \(R_{\mathrm{я}}\) and \(R_{\mathrm{в}}\) caused a weak spark in the switching element. Under these conditions it was impossible to count on its stability. Therefore, in this circuit a mechanical switch is used instead of a thyratron. The elements of the circuit \(R_{\mathrm{я}}, C_{\mathrm{я}1}, C_{\mathrm{я}2}, R_{\mathrm{в}}, C_{\mathrm{в}1}\), and \(C_{\mathrm{в}2}\) are readily calculated if the following are known: the charging voltage of the capacitance \(C\), the required voltages on the plates for the time sweep and for the phenomenon, their sensitivity to the deflection of the electron beam, and the diameter of the cathode-ray tube.

Fig. 5. Diagram of the setup for dielectric breakdown testing with a voltage action time of the order of \(10^{-1}\) sec

It is important to note that one should not take excessively large values of \(R_{\mathrm{я}}\), since the resistance of the specimen and of the surrounding medium (for example, transformer oil) may be comparable with the value of \(R_{\mathrm{я}}\), and the capacitances \(C_{\mathrm{я}1}\) and \(C_{\mathrm{я}2}\) will not charge to the full voltage.

Cathode-ray oscilloscope

The setup uses an electronic oscilloscope with a schematic diagram kindly provided to us by the Laboratory of High-Voltage Gas Discharge of the Power Engineering Institute of the Academy of Sciences of the USSR. At the suggestion of Prof. Dr. A. A. Vorob’ev, work on mastering the operation of the oscilloscope and on its design was carried out in the laboratory of the Department of High-Voltage Engineering of TPI by I. I. Kalyatskii, A. M. Trubitsyn, and the author.

The oscilloscope circuit is shown in Fig. 6. The 13LO2S cathode-ray tube has constant power supply (20–25 kV). By means of the spark-gap relay \(S\), synchronization is accomplished of the opening of the beam (upon breakdown in the middle gap of relay \(S\)), of the time sweep (upon breakdown of the left-hand gap), and of the operation of the pulse generator (upon breakdown of the right-hand gap).

For the time sweep, the voltage obtained across the resistances \(R_9\) during discharge of the capacitors \(C_3\) through them is used. This produces a nonlinear time sweep.

When recording short phenomena according to the circuit of Fig. 2 and pulses with a front duration of about 1 sec, the time circuit of the oscilloscope was disconnected.

from the plates in time. In the latter case (pulses with a duration of about 1 sec) the electron beam was constantly open and had sufficient brightness for recording the phenomenon, which, on the other hand, was insufficient to cause damage to the screen before the event under investigation was applied. The parameters of the oscilloscope circuit can be calculated comparatively easily [8].

Fig. 6. Circuit of a high-voltage electron oscilloscope and pulse generator

A photograph of the oscilloscope is shown in Fig. 7. The height of the oscilloscope is 1.55 m, its length 0.9 m, and its width 0.55 m. These dimensions of the oscilloscope were chosen for reasons of insulation, as well as for the convenience of making individual switchings in the oscilloscope.

If necessary, the dimensions of the oscilloscope can be reduced considerably.

Typical oscillograms are presented in Figs. 8, 9, and 10.

Divider of Pulse Voltages

In our experiments, the breakdown voltages reached values of 60 kV. Voltages of no more than 1.5 kV could be applied to the deflection plates of the oscilloscope. Therefore a divider of pulse voltages was necessary.

Since it was intended to record very short phenomena, a capacitive type of divider was chosen, the circuit of which is shown in Fig. 11.

The first design of the capacitive divider is shown in Fig. 12. The capacitance between the spherical electrodes corresponds to the capacitance \(C_1\) in the cir-

Fig. 7. Photograph of the electronic oscilloscope (side view).
1 — D.N. 1; 2 — spark-gap relay S; 3 — tr. 2; 4 — C1;
5 — filament transformer of rectifier B1; 6 — tr. 1; 7 — H.V.G.,
attachment for photographing from the screen

Fig. 8. Oscillogram of the breakdown of rock salt

Fig. 9. Oscillogram of breakdown of rock salt. Frequency of the calibration oscillations: \(1\ \text{MHz}\)

Fig. 10. Oscillogram of breakdown of rock salt. Frequency of the calibration oscillations: \(50\ \text{Hz}\)

Fig. 11. Circuit of a capacitive voltage divider

[[unclear: beginning of sentence]] of Fig. 11. The distance between the spheres was calculated for the full operating voltage (up to 60 kV). The capacitance between the disk electrodes and the capacitance of the phenomenon plates correspond to capacitance \(C_2\) in the circuit of Fig. 11. By changing the distance between the disks, it was possible to adjust the division ratio of the divider.

For recording short phenomena lasting several millimicroseconds (such was the duration of the voltage drop across the specimen during breakdown), a capacitive voltage divider was designed, the circuit of which is shown in Fig. 13. In designing

Fig. 12. Capacitive voltage divider.
V. H. — high voltage. 1 — spherical electrodes, 2 — flat electrodes, 3 — phenomenon-plate switch, 4 — phenomenon plates, 5 — oscilloscope cover, 6 — electron-beam tube, 7 — shielding casing of the electron-beam tube.

Fig. 13. Breakdown-oscillographic setup.
V. H. — high voltage. 1 — high-voltage electrode, 2 — lead from the phenomenon plate, 3 — phenomenon plates, 4 — switch for connecting the specimen to high voltage, 5 — specimen holder with specimen, 6 — electron-beam tube.

the divider, we tried to bring the test specimen as close as possible to the voltage divider in order to increase the natural frequency of the circuit—divider–specimen. This was necessary so that the rapid voltage drop across the dielectric specimen during its breakdown would not excite strong oscillations in the indicated circuit. During breakdown of the dielectric, very weak oscillations were observed on the oscillogram. The capacitances

\(C_1\) (Fig. 11) here corresponds to the capacitance between the high-voltage electrode (up to 40 kV) and the lead of the upper plate of the Iavlenie tube; \(C_2\) (Fig. 11) is the capacitance of the Iavlenie plates.

When working with capacitive dividers it is necessary to avoid corona, which can distort the pulse shape and lead to large errors in measurement. The presence of corona was determined from two phenomena:

1. With strong corona in the divider, a sharp vertical deflection of the electron beam is observed.

2. With weak corona, the zero line on the oscilloscope screen shifts upward or downward (depending on the polarity of the pulse and on which electrode is corona-forming) under the action of each pulse.

Fig. 14. Circuit of the experiment for determining the frequency dependence of the divider ratio.

Using the circuit of Fig. 14, the frequency characteristic of the divider was recorded. The beam was opened only enough not to damage the screen. The divider ratio was determined from a quantity proportional to it, equal to the ratio of the high-frequency voltage to the height of the trace on the oscilloscope screen. In the range \(10^5 \div 5.7 \cdot 10^7\) Hz, the divider ratio remained constant within the accuracy of the measurements. This circumstance made it possible, when determining the breakdown voltage of a specimen in the case of a short exposure, to use a calibration graph (the dependence between the voltage on the divider and the deflection of the beam on the oscilloscope screen) obtained under the action of pulses of duration \(10^{-5}\) sec on a sphere spark gap.

This work may be useful for persons beginning work in the field of pulsed breakdown of dielectrics.

REFERENCES

1. Bull B. M. ZhTF, 2, 3—4, 371, 1931.

2. Vorob’ev A. A. and Val’ter A. F. Breakdown of porcelain under surge voltages. Proceedings of the 3rd All-Union Conference on Electrical Insulating Materials, 3, 173, 1934.

3. Val’ter A. F. and Inge L. D. ZhTF, 4, 813, 1934.

4. Binder L. Traveling waves in electrical networks. Translated from German, p. 160, 1935.

5. Strigel R. Electrical impulse strength. Berlin, 1939.

6. Val’ter A. F. and Inge L. D. ZhTF, 3, 6, 840, 1933.

7. Hergenrother R. C., Rudenberg H. I. Proc. IRE, 41, 7, 896, 1953.

8. Vorob’ev G. A. Investigation of the electrical breakdown of solid dielectrics at different durations of voltage application. Dissertation, Tomsk, 1956.

9. Sonchik K. K. On the delay time of discharge in ionic crystals. Izvestiya AN SSSR, Physical Series, vol. 22, No. 4, 1958.