Eng. V. K. Shcherbakov.

On the Influence of Corona Discharge Present at the Electrodes of Insulators on the Safety Margin with Respect to Atmospheric Overvoltage Waves

I.

As is known, the majority of cases of flashover of insulators on a line are caused by atmospheric overvoltages, which most often have the character of impulses with a duration not exceeding 100 microseconds. The ability of insulators to withstand the action of impulses without discharge is characterized by the value of the minimum discharge voltage. This concept presupposes that we are dealing with a definite form of the voltage curve as a function of time. This dependence may be expressed as follows:

\[ V = V_a f(t) \]

where \(V_a\) is the amplitude of the impulse, and \(f(t)\) is some function of time. Among the whole set of impulses which possess one or another value of \(V_a\), but a definite expression \(f(t)\), we have a certain definite value of \(V_a\) lying at the boundary between values that cause and do not cause discharge. This quantity is called the “minimum impulse discharge voltage.” It is therefore associated with a definite course of the function \(f(t)\). The course of the impulse and, correspondingly, \(f(t)\), usually consists, as a number of cathode oscillograms taken on lines show, of a part rising from zero to \(V_a\), called the “front” of the impulse, and a part descending after \(V_a\), called the “tail” of the impulse, and sometimes passing over a certain length through the zero line. If an impulse acts on an insulator whose amplitude somewhat exceeds the minimum discharge voltage, then at some instant on the tail of the impulse a discharge occurs over the insulator, the actual voltage at the moment immediately preceding the discharge being lower than the amplitude of the impulse. The time elapsed before the actual discharge is called the “time lag,” and it is now customary to measure it from the instant on the impulse curve corresponding to the amplitude of the 50-period discharge voltage, although it would be more correct in principle to measure it from the instant,

corresponding to the discharge voltage under rectified voltage, which is somewhat higher than the preceding one.

In recent years several works have been carried out in America concerning the measurement of impulse discharge voltages and time lags. In these works the question of the influence of the voltage that existed on the object before the action of the impulse on the magnitude of the discharge voltage and on the time lag was not yet considered. In most of these experiments the object was not under voltage at the moment preceding the action of the impulse, while in others it was under a rectified voltage of arbitrary magnitude. In the present work the question of the influence of the initial voltage is set for study.

In the operation of high-voltage apparatus one sometimes has to encounter the phenomenon of corona on certain elements of devices. There are designs—such as the corona rings adopted on AEG bushings—in which this phenomenon is deliberately made use of in order to achieve a definite positive technical result, namely an increase in the discharge voltage observed in the 50-cycle test; or, for example, rings of flat iron used as protective fittings for strings on American 220 kV transmission lines, where it may be expected that in unfavorable weather corona will appear on them at operating voltage. In practice, the discharge voltage with such rings in the 50-cycle test is only slightly lower than for strings protected by rings of German design. Naturally, from this arises the question whether the presence of corona affects the minimum discharge voltage.

The basis for expecting such an influence is as follows: under the action of an impulse, upon the latter reaching the value of the corona voltage, partial impact ionization and space charges arise in a certain zone of the electric field. If, even before the action of the impulse, some ionization and space charges are already present in the field, one might suppose that certain preliminary stages of the discharge have already, as it were, been passed through, and as a result its time lag is shortened; therefore, in the ionized gap the minimum discharge voltage should decrease.

Another circumstance also compels one to think of a significant role of preliminary corona on the electrodes of insulation subjected to impulse overvoltages. This circumstance consists in the fact that the discharge between spheres, when the distances between them are small, proceeds without delay or almost without delay when impulse voltages are applied. The same is observed for discharges under the action of impulses in the case of plate electrodes with rounded edges. That is, where the electrodes form a more or less uniform field, the discharge under impulses proceeds without delay. This phenomenon can be explained simply: under the influence of the electric forces of the field, impact ioni-

ization arises simultaneously in all parts of the field, and in this case no time is required for an ever greater and greater expansion of the zone of impact ionization.

From this point of view it follows that if certain conditions make some zone around electrodes, with a nonuniform field between them, conducting, thereby equalizing the field between the electrodes, then breakdown of the spark gap under these conditions will require considerably less time, or no time at all on the scale of our time measurements.

It may be assumed that the corona is precisely such a condition. The corona, as it were, expands the volume of the electrode, equalizes the field to a certain extent, and brings this field closer to the field between spheres or cylinders of large diameter.

Insulation coordination for the purpose of protection against overvoltages is now beginning to be widely applied. The impulse characteristics of insulation are, in this connection, the basis of coordination. Impulse characteristics represent the relation between the amplitude of the impulse discharge voltage, or the impulse coefficient, and the time lag. The excess of the amplitude of the impulse discharge voltage over the value of the discharge voltage at 50 periods is due to the necessity of expending a certain time on the discharge process. By acting, therefore, on the discharge time lag, one can completely change the impulse characteristics of insulation.

Under this condition, there might be a different approach to the selection and design of the insulation of electrical installations, which in operation are usually subjected to the action of impulse voltages through various kinds of overvoltages.

II.

To investigate the question posed above, the author, using the equipment of the High-Voltage Department of the All-Union Electrotechnical Institute, carried out a considerable number of experiments.

The main content of the experiments amounted to determining the impulse coefficient during discharge on electrodes subjected to preliminary corona.

The influence on the impulse coefficient of direct-current corona and of alternating sinusoidal-current corona (50 periods/sec) was considered. In this, special attention was given to determining the value of the phase of the alternating-current corona at the moment of application of the impulse. The intensity of the corona was also varied.

The experiments were begun with the simplest object—a point. Then AEG bushing rings and protective rings of suspension strings were tested. The preliminary testing of points had the aim of clarifying the most correct direction of experiments with the subsequent objects.

The impulse voltage was obtained from surge circuits connected according to the Marx scheme. The amplitudes of the impulses were measured with spark voltmeters. The desired form of the impulses was achieved by selection of a definite value of leakage resistances.

The basic circuit of the experiments is shown in Fig. 1. At the electrodes of the insulator, by means of a DC or AC source, a potential difference is created that is greater than the so-called initial corona voltage. By changing the potential difference at the electrodes of the insulator, the intensity of the corona at the electrodes is varied.

Fig. 1.

Impulse voltages of the form shown in Fig. 2 are generated by an impulse circuit which, during its charging, is separated from the object by the spark gap \(Z\). Otherwise, the charging voltage of the circuit would exert an additional corona-producing action on the electrodes of the insulator.

The impulse shape was checked by calculation and by oscillograms. The calculations and oscillograms showed that the front of the impulse, during all the experiments, remained sufficiently steep (1–2 \(\mu\)sec), and that, consequently, the process on the front of the impulse could be neglected.

Fig. 2.

The influence of the tail of the impulse (see Fig. 2) may be expected to be more significant, and therefore, throughout all the experiments, control of the shape of the tail was ensured.

The circuit of the impulse generator makes it possible to vary the shape of the tail over wide limits. For this it is necessary to change the magnitude of the leakage resistance. In fact, the law of variation of the voltage on the tail of the impulse is expressed by the formula:

\[ V_x = V_a' e^{-\frac{t}{Rc}} \]

(discharge of the capacitance through ohmic resistances),

where \(V_a\) is the amplitude of the impulse,

\(V_x\) is the instantaneous value of the impulse in the region of the tail at the time \(t\), which is counted from the front of the wave, taken, for simplicity of operation, as vertical.

The time of fall of the impulse voltage on the tail to one-half of the amplitude value (see Fig. 2), which is accepted in the American

\[ t_0=\ln 2\,CR \]

III. CORONA FROM RECTIFIED VOLTAGE

The impulse voltage is obtained from a surge circuit consisting of two elements \(C_1\) and \(C_2\), each with a capacitance of \(0.0154\) microfarads. The circuit is fed from a rectifier installation with a maximum—

Object: points (Fig. 3).

Scheme for obtaining a negative impulse voltage of the corona from a direct voltage; sign of the corona: “—”.

Operating diagram (Fig. 4).

—voltage of 100 kilovolts. The greatest voltage of the circuit (amplitude) is 180 kilovolts.

In this case

\[ t_0=\ln 2\,CR=0.6931\cdot 0.0154\,R\ \text{microsec}, \]

i.e., in order to obtain impulses of duration: 10, 20, 30, etc. microseconds, it is necessary to take leakage resistances approximately: 1000, 2000, 3000, etc. ohms. The voltage drop of the capacitors may depend on the resistance \(R_1\). In the case of an insufficiently large resistance \(R_1\), the transformer may feed the capacitors at the moment of discharge and thereby increase the duration of the impulse. In the present case

the resistance \(R_1\) was chosen to be on the order of several hundred thousand ohms, and its influence was thereby excluded.

Control of the circuit consisted of the following. With the switch “\(B\)” open, capacitor \(C_1\) is charged through resistance \(R_1\), and capacitor \(C_2\) through resistances \(R_1\), \(R_2\), and \(R_3\). Then switch “\(B\)” is closed and connects capacitors \(C_1\) and \(C_2\) in series.

The rate of rise of the impulse voltage is determined by the resistance of the arc of switch “\(B\).” Switching was performed by the fall of a massive rod which, at the moment the arc arose, possessed a considerable velocity, ensuring a rapid drop in the arc resistance. As was already indicated above, calculations and oscillograms confirmed the assumption of a high rate of rise of the voltage at the front of the impulse (of the order of \(1/2\)—\(2\) microseconds).

It was, of course, to be expected that the sign of the impulse and the sign of the corona would have a considerable influence on the results of the experiment. Therefore the investigation was carried out with different signs of the impulse and of the corona. The circuit of Fig. 4 gives impulses of sign “\(-\)” (throughout, polarity will be understood relative to ground) and corona also of sign “\(-\).”

Impulses of positive polarity were obtained according to the circuit of Fig. “5.”

With switch “\(B\)” open and with the distance between the balls of the spark gap “\(F\)” somewhat greater than the breakdown distance at the maximum voltage of the rectifier installation, capacitors \(C_1\) and \(C_2\) are charged, the first through resistances \(R_1\) and \(R_4\), the second through \(R_1\), \(R_2\), \(R_3\), and \(R_4\).

Fig. 5.

By closing switch “\(B\),” we double the potential difference on the balls of the spark gap “\(F\),” and thereby cause breakdown of this gap and the connection of the capacitors in series. It is evident from the circuit that in this case a charge with sign “\(+\)” arrives at the object (the polarity of the voltage source is determined by the kenotron). The corona in the circuit of Fig. 5 has positive polarity. In order, in both circuits, to obtain a corona of the opposite sign, it is necessary to separate the object from the impulse generator by a spark gap and to feed the corona at the electrodes of the object with voltage of the opposite polarity.

The results of the experiments are presented in the form of curves in Fig. 6. These results amount essentially to the following:

a) The presence of corona on the electrodes, in the form of points, at the moment when an impulse voltage is applied to the electrodes, has only a slight effect on the magnitude of the impulse discharge voltage.

b) Coincidence of the polarity of the impulse and of the corona lowers the discharge voltage, as compared with that without corona.

The results of the experiments would seem to have given a quite definite answer to the question of how significantly preliminary corona affects the strength of a spark air gap; moreover, the answer is negative.

However, the conditions of these preliminary experiments had considerable limitations. Thus, 1) during the experiments the corona voltage was always no more than one-half of the impulse amplitude; 2) the points (needles), while producing an intense corona directly near the electrodes, nevertheless have a small range of propagation of this corona, and consequently only a small zone of impact ionization; 3) the influence of corona from an alternating potential difference—which in practice is the case mainly in the experiments under discussion—is not taken into account at all.

Fig. 6.

IV.

CORONA FROM A FIFTY-PERIOD SINUSOIDAL VOLTAGE.

Corona from an alternating potential difference presupposes a change with time both in the degree and in the sign of coronation. Therefore, for a detailed study of the influence of individual moments of alternating corona, it is necessary to be able repeatedly to observe the influence of one and the same phase of the corona on the magnitude of the impulse discharge voltage.

The latter is achieved in the present work by means of a special device for firing (bringing into action) the surge circuit, in the form of a synchronous switch, which makes it possible to apply an impulse to the object at any desired instant (phase) of the alternating corona. The synchronous switch, Fig. 7, is set in rotation by a synchronous motor. The motor is connected to the alternating voltage in parallel with the transformer feeding the corona on the object.

Fig. 7.

Each time the rotating copper spoke of the switch closes the circuit for firing the circuit, the impulse voltage, with a certain very small delay, is applied to the object. By setting the position of the switch spoke differently, but fixed at a given instant, relative to the motor rotor, one can obtain closure of the circuit-firing circuit at any instant on the curve of the sinusoidal alternating current feeding both the corona and the switch motor.

Fig. 8.

The operating circuits with alternating corona and impulses with a maximum amplitude of up to 180 kv are shown in Figs. 8 and 9.

The corona is supplied from a 110/65000-volt transformer connected to the object through the protective resistance \(R_5\).

With the transformer 110/65000 V disconnected and with the addition, indicated by the dotted lines, the circuit is valid for corona from direct voltage.

Fig. 9.

1. Object—points.

Preliminary experiments with alternating corona and points do not make it possible to draw any conclusion, since the character of the curves (Fig. 11) compels one to assume the intervention in the experimental process of some factor not previously taken into account. An analysis of the causes of the distortion of the experimental results is given below.

2. Object—pin-type insulator (6 kV operating voltage).

In this case (see Fig. 12) the corona acts in the same way as in the experiments with corona from rectified voltage. Coincidence of the polarities of the impulses and of the corona leads to a reduction in the strength margin of the pin-type insulator with respect to impulse voltages. However, this reduction is relatively small.

Before proceeding to experiments with higher voltages and other objects, the causes of the distortion of the experimental results in the case of points and alternating corona are clarified. It was necessary to carry out this investigation, since otherwise a similar distortion of the experimental results might also occur subsequently.

In switching to experiments with alternating corona, an additional spark gap \(Z\) had to be introduced into the impulse-generator circuit (see Fig. 8). The purpose of the spark gap \(Z\) is to separate the object from the action of the source charging the surge circuit, since the corona is now supplied from a special transformer.

With this circuit the object comes under impulse voltage only when the isolating gap \(Z\) has broken down.

At the same time, it turns out that breakdown of the gap “\(Z\)” occurs under the action not only of the impulse voltage, but also of the voltage of the transformer feeding the corona.

Fig. 10.

The ball of the isolating gap, connected to the transformer, at the moment when the impulse voltage arrives has a certain potential \(V_{\text{cor.}}\), depending on the phase at which the couplings of the synchronous switch are set, and on the magnitude

Fig. 11.

which definite potential \(V_{\text{cor.}}\), depending on the phase at which the couplings of the synchronous switch are set, and on the magnitude

of the established effective alternating voltage of the transformer feeding the corona.

The other sphere of the same separating gap, connected to the circuit, receives at the moment of application of the impulse the potential of the latter, \(-V_{\mathrm{imp.}}\). If \(V_{\mathrm{cor.}}=0\), then breakdown will occur at the moment when the potential \(-V_{\mathrm{imp.}}\), increasing on the front of the impulse, becomes equal to \(V_{\mathrm{br.}}\), the breakdown voltage of the separating gap. If \(V_{\mathrm{cor.}}\ne 0\), then breakdown will occur at the moment when the potential \(V_{\mathrm{imp.}}\), increasing on the front of the impulse, reaches such a magnitude that the potential difference on the two sides of the gap becomes equal to its breakdown voltage, \(V_{\mathrm{br.}}\) (Fig. 13), i.e.

\[ V_{\mathrm{imp.}} - V_{\mathrm{cor.}} = V_{\mathrm{br.}}, \]

whence:

\[ V_{\mathrm{imp.}} = V_{\mathrm{cor.}} + V_{\mathrm{br.}}. \]

Fig. 12.

Therefore the discharge will occur at a smaller \(V_{\mathrm{imp.}}\), in other words, at an earlier instant, the smaller the algebraic value of \(V_{\mathrm{cor.}}\) is in comparison with \(V_{\mathrm{imp.}}\). For \(V_{\mathrm{cor.}}\) having the sign opposite to \(V_{\mathrm{imp.}}\), breakdown will occur at an even earlier instant than in the absence of voltage on the second sphere (i.e., in the case \(V_{\mathrm{cor.}}=0\)).

Fig. 13.

All that has been said is illustrated by the experimental data of Fig. 14. The above reasoning and the experimental data show that, in order to avoid non-simultaneity and uncertainty of breakdown of the separating gap, it is necessary to take larger ratios of the impulse amplitude to the corona voltage than was done in the first series of experiments.

But it is not only the presence of a 50-cycle voltage on one of the spheres of the isolating gap that leads to uncertainty in the amplitude—

Fig. 14.

(In the figure: “Dependence of the impulse voltage beyond the isolating gap on the magnitude of the corona voltage”; axis label: \(V_{\text{corona}}/kV\); some curve annotations are partly illegible.)

—of the impulse beyond the isolating gap; as experiment has shown (Fig. 15), it is also the circumstance that two spark gaps operating in series give certainty of the ra—

Fig. 15.

(In the figure: “Dependence of the impulse voltage at the object on the magnitude of the isolating gap”; additional circuit and curve annotations are partly illegible.)

are suitable only for definite ratios of the magnitudes of these gaps and of the voltage applied to them.

In work with subsequent objects (a suspension string and a bushing), which require higher impulse amplitudes, the shortcomings indicated above are almost entirely eliminated.

B. Object—a suspension string, 8 elements PP 4, 5.

The operating circuit is shown in Fig. 16. Here the generator circuit has a maximum impulse amplitude of 1500 kilovolts and is fed from a rectifier installation with a maximum voltage of 250 kilovolts. The circuit is fired through the synchronous switch “S.B.” The impulse amplitude is measured by the sphere gap “\(F_{100}\).” The suspension-insulator string is separated from the circuit by the gap “\(Z\),” in order to avoid the influence of the charging voltage on the corona. The corona is supplied from a transformer group of 1000 kilovolts maximum effective voltage.

Fig. 16.

The corona voltage is measured by the sphere voltmeter “\(F'_{100}\).” The 1000-kilovolt transformer is protected from the action of impulses by a special high-ohmic resistance.

Before the tests were carried out, the circuit was checked for the stability of the impulse amplitude beyond the isolating gap. The results are summarized in the table below.

Checking the stability of the pulse amplitude beyond the isolating gap.

Pulses “−”
Corona voltage 200 kV/max.

The pulse shape was photographed with a cathode oscilloscope. One of the oscillograms is shown in Fig. 17.

Fig. 17.

For characterizing the string of suspension insulators as an object of investigation, the impulse characteristics are given (Fig. 18). The curves show that the impulse discharge voltage of the gar-

Fig. 18.

...strings with flat shielding rings is higher than that for a string with round rings, which, of course, is already known from Torok’s experiments.

As a result of the experiment, curves were obtained both for the case of a string without rings and for the case of a string protected by flat rings (Fig. 19). These results (Figs. 20 and 21) confirm the preceding experiments, i.e., for all combinations of the signs of the corona and of the impulse, with and without shielding rings, the influence of the corona manifests itself in the direction of lowering the discharge voltage. In the case of like signs of the corona and of the impulse, and conversely—an increase in the discharge voltage in the case when the signs do not coincide. In percentage terms, this decrease and increase of the impulse discharge voltage, even for the case with impulses of 10 microseconds’ duration, does not exceed 5–6%.

Fig. 19.

Fig. 20.

Visible inscriptions in Fig. 20:

• \(U_{\text{disch.}}\), kV
• \(U_{\text{cor.}}\), kV
• \(U_{\text{cor.}}\), kV
• \(U_{\text{cor.}}\)
• \(U_{\text{imp.}}\)
• \(1\) — object: a string of five insulators [[unclear: type/designation]], length 550 mm; corona, negative; impulse, negative, 300 [[unclear: units]].
• \(2\) — object: a string of five insulators, with flat rings; corona, negative; impulse, negative, 300 [[unclear: units]].

Fig. 21.

The string with round rings (Fig. 22), evidently because of the weak corona, showed an almost complete independence of the impulse discharge voltage from the sign and magnitude of the corona voltage.

Fig. 22.

A string without rings has almost the same behavior as a string with rings.

In continuation of all the experiments with the string, the corona voltage was maintained at

\[ \frac{220}{\sqrt{3}} = 127 \text{ kilovolts.} \]

4. Object—a bushing. (Operating voltage 110 kV, Fig. 23).

It is known that the AEG-design bushing is provided with corona rings in order to raise the fifty-period discharge voltage. At present it is sufficiently well known that so-called corona electrodes raise the discharge voltage very substantially and, on this basis, are a desirable element of high-voltage insulation.

However, the electrical strength of insulation in the operation of electrical installations is determined not by the fifty-period discharge voltage, but by the impulse voltage. Therefore, in applying them to the AEG bushing, two points must be clarified:

Fig. 23.

a) how the corona rings affect the impulse discharge voltage, and

b) how the preliminary corona on the bushing rings affects the impulse discharge voltage.

In Fig. 24 the results of experiments are presented under the following conditions: impulse sign “+,” bushing with protective rings, bushing without protective rings. From this it is seen that the protective rings also raise the discharge voltage in the case of impulse voltages; however, this increase is small and, it seems to us, has no practical significance.

The same is obtained also for impulses of negative polarity, Fig. 25.

It is necessary to note an interesting phenomenon: an increase, and a very considerable one, of the discharge voltage for impulses with the minus sign. In Fig. 26 this increase reaches 90%.

Fig. 25.

Visible labels in the graph:

• Object: AEG bushing; impulse sign “−” ([[unclear: parenthetical abbreviation]]); tail ~ 10 ms.
• 1. Bushing with protective corona.
• 1. Bushing without protective corona.
• Axis label: \(V\)-[[unclear: cm?]], \(KV\)-([[unclear]]).
• Scale markings: 950, 900, 850, 800.
• \(90^\circ\), \(\varphi\), \(210\).
• \(\dfrac{V10^{-\lambda}}{d-d\lambda}\), kV.

Fig. 24.

Visible labels in the graph:

• Object: AEG bushing; impulse sign “+”; tail — 32 ms.
• 1. Bushing with protective corona.
• 1. Bushing without protective rings.
• Axis label: \(V\)-[[unclear: cm?]], \(KV\)-([[unclear]]).
• Scale markings: 550, 500, 450, 400, 350.
• \(V_{pr}=0\).
• \(V_{pr}=2.9\) [[unclear]].
• \(90^\circ\), \(\varphi\), \(\alpha=50^\circ\).
• \(210\).
• \(\dfrac{V10^{-\lambda}}{d-d\lambda}\), kV.
• 150.

Fig. 27.

Object: ABC through bushing.
Sign of the impulse: \(+\) (110 kV — operating voltage).
Bushing without protective rings.

1 — impulse tail \(32\,\mathrm{ms}\)
2 — impulse tail \(\sim 10\,\mathrm{ms}\)

Axis label: \(V_{\text{stat}}\), kV (cm)

Visible markings: 400, 450, 475, 500, 525, 650; 0, 90, 270; 150, 150; \(d=\) [[unclear]].

Fig. 26.

Object: ABC through bushing.
Impulse tail \(\sim 10\,\mathrm{ms}\) [[unclear: parenthetical note]],
bushing without sharp angles.

Axis label: \(V_{\text{stat}}\), kV (cm)

Visible markings: 400, 500, 600, 700, 800; 0, 90, 180, 270; 150, 150; \(d=\) [[unclear]].

Sign of the impulse: “−”
Sign of the impulse: “+”

The influence of the preliminary corona, in the case of impulses of negative sign, is quite insignificant. In the case of positive impulses this influence is more significant, but since it is directed toward an increase of the voltage, and not toward a decrease, this phenomenon can have no practical importance. The decrease of the discharge voltage in this case proves to be very slight.

An experiment was also set up to determine the influence of the length of the tail in conjunction with the influence of the corona. The results are presented in Fig. 27. The experiments were carried out with impulses having lengths of 32 microseconds and 10 microseconds. As is seen from the curves in Fig. 27, the influence of the impulse tail proved to be quite insignificant.

Note: In all cases of operation with the bushing, the corona voltage was 110 kilovolts eff.

GENERAL CONCLUSIONS

1. The results of the experiments basically confirmed the assumption stated earlier, that preliminary impact ionization should facilitate discharge under impulses.

In all cases there was a decrease in the impulse discharge voltage if the application of the impulse was preceded by a corona of the same sign.

However, this decrease, as a rule, is of such magnitude as cannot appreciably alter the characteristics of the insulation.

1. With positive impulses, the discharge voltage in all cases is lower than that with negative impulses. A particularly significant decrease in the positive impulse discharge voltage was found for the bushing.

2. The tail of the impulse, within the range from 10 to 50 microseconds, did not exert any appreciable influence on the results of the experiment.

3. It may be assumed that, with an increase in the intensity of the corona, the noted influence of the corona should increase and will evidently lead to a very considerable decrease in the impulse discharge voltage. Experiments with corona voltages higher than the operating voltage of the insulation are difficult, owing to the instability of the impulse across the isolating gap.

However, investigation of the influence of corona voltages considerably exceeding the operating voltage of the insulator has no practical meaning, since in service insulators may be subjected to prolonged loading by a fifty-cycle voltage of \(\sqrt{3} V_{\phi}\), and then only in the case of a system with an ungrounded neutral.

1. The experimental results obtained may be of great importance in the further study of the physical processes in discharges in air, and especially under impulse voltages.