High-Voltage Insulation

BY J. L. R. HAYDEN
Member, A. I. E. E.

CHARLES P. STEINMETZ

and

Fellow, A. I. E. E.
Chief Consulting Engineer, General Electric Co., Schenectady, N. Y.

I. General
T HE most important chapters in electrical engineering are those dealing with efficiency, heating,
magnetism and insulation.
In the field of insulation, our knowledge is most
backward.
In regard to efficiency, with efficiencies of electrical
apparatus of 90 up to over 99 per cent, nofurtherradical
progress appears feasible.
In magnetism, the losses in an alternating field have
been reduced so far, that they have ceased to limit, by
their heating effect, the size of apparatus, but are merely
a factor in the efficiency. An increase of the saturation
density would decrease the size of apparatus, but is
excluded by its inherent chemical limitations.
Heating specifications are not made for other classes
of apparatus, such as prime movers, etc., and if an
important subject in electrical engineering, it is almost
entirely because the insulation of the apparatus is
destroyed by the higher temperatures. Thus the problem of the heating of electrical apparatus is essentially
one aspect of the insulation problem.
In our high-voltage apparatus, cables, etc., we operate
the insulation at voltage stresses which rarely exceed
much the disruptive strength of air, though laboratory
tests often show this insulation to have a disruptive
strength of 10 to 20 times that of air.
All phenomena of nature are very complex. Therefore, in calculating a phenomenon or designing an apparatus, we must approximate by neglecting "secondary terms," and take care of these by an allowance,
a margin or a factor of safety. Obviously, the more
completely a phenomenon is known and understood,
the closer it can be calculated, in other words, the less
is the margin or safety factor required in order to allow
for the unknown stresses, etc. The margin or safety
factor, which experience shows as necessary, thus is an
indication of the exactness of our knowledge of a phenomenon. For example when dealing with magnetic
phenomena, with efficiency, with heating, etc., we have
to allow a margin of a few per cent only. In testing
the insulation of apparatus however, the A. I. E. E.
standards specify a test voltage more than twice the
delta voltage, though the normal stress is the Y voltage.
That is, we require a safety factor of over 3.46, a
margin above normal of over 246 per cent.
The insulation problem has become of increasing
importance with the rapid advance of electrical engineering into higher voltages. Not many years ago
44 kv. was the highest transmission voltage for reliability of operation of overhead lines. Now we have
Presented at the Pacific Coast Convention of the A. I. E. E.,
3 .
DeZ Monte, Cal., October 2-5, 192

reached 220 kv. Then 12 kv. was the highest satisfactory cable voltage; now we have reached 22 kv.
and a few cables of 33 kv. and higher, but cables at
33 kv. are still semi-experimental. The comparison
shows that the advance in pushing cable voltages up to
higher values, has been slower than with overhead
lines, and our knowledge of liquid and solid insulation,
such as come into consideration in the cable wall and
the machine insulation, is materially less advanced
than that of air as dielectric. With regard to air, a
good working theory has been established, by considering the dielectric strength of air as analogous to
the mechanical strength of structural materials. The
theory recognizes a definite dielectric strength of air,
or a disruptive breakdown gradient, of 30 kv. per cm.
at normal air density. Puncture occurs when this
dielectric strength is exceeded, just as mechanical
disruption occurs, when anywhere in a mechanical
structure the stresses exceed the elastic limit of the
material. This conception of a definite breakdown
strength then was extended to liquid and solid dielectrics, but with these, it failed to give a satisfactory
explanation of the mechanism of the breakdown, more
particularly of the all-important feature of the time
lag of disruption. And even with air, the theory of a
constant breakdown gradient, as modified by the conception of the energy distance, is satisfactory only
within a certain range.
.
.
II. Air as Dielectric
The present practically universally accepted theory
of a dielectric strength of air at and near atmospheric pressure, as most completely developed by
Mr. F. W. Peek, Jr., is:
Air has a definite and constant "dielectric strength,"
at which it ceases to be an insulator and becomes a
conductor, that is, breaks down electrically.
The dielectric strength of air is proportional to the
air density, and is 30. kv. per cm. at normal air density
of 0 deg. cent. and 76 cm. barometer.
The dielectric breakdown (or puncture) of air does
not occur as soon as the voltage gradient in the dielectric field exceeds the dielectric strength at any
point, but the voltage gradient in the field must exceed
the dielectric strength over a finite distance, the socalled "energy distance."
The energy distance depends on the convergency of
the electric field at the place where the breakdown
occurs, and is the less, the more convergent the field.
The energy distance between parallel cylinders
(wires) of radius R? iS 0.3 ub R; between_spheres of
radius R it is, approximately; 0.54 Vx,6 1R, where
uS= air density (with normal air density as unity).

1029

1030

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

Transactions A. I. E. E.

For gap lengths less than the energy distance, the

Thus the breakdown gradient at the surface of two
stand.ard theory does not apply.
parallel cylinders of radius R is:
For gaps in which corona precedes the spark dis0.3
+
charge, the law does not seem to satisfactorily agree
90 = 303 (1
with experience. Between spheres, corona appears
only at gap lengths materially larger than those in
at the surface of spheres it is
which it should appear according to the theory. Also
0.54
the disruptive voltage at which spark discharge occurs
4
9o = 30 ( 1 +
between needle points, is materially lower than calV3I ib
For gaps of a length less than the energy distance, culated. This may partly be accounted for as follows:
(a) The space filled by corona is not a perfect cona higher breakdown gradient is required, the more so,
ductor, but has a finite resistance, which depends on
the shorter the gap. The law is unknown.
The space in the dielectric field in which the dielectric the corona density, thus on the voltage, frequency, etc.
(b) It is practically an impossibility to calculate
strength is exceeded, becomes conducting, and, if adjacent to the terminals, becomes a part of the terminal. accurately the final dielectric field and its voltage graThis changes the configuration of the dielectric field dients, as resulting from the partial short circuit by the
and if thereby the dielectric strength is exceeded in corona, of regions of the initial dielectric field, even if
other parts of the dielectric field, these also break down the resistivity of the corona regions were neglected.
(c) The nature of the corona-filled space is that of
and become conducting, until either the breakdown
an
unstable third class conductor, which tends to
punctures
the
air
gap
from
terminal
to
terminal,
extends
local oscillations and other transients, and
produce
to
and short-circuits by a spark discharge (which tends
variations of the resultant field and
thereby
rapid
develop into an arc), and the current rises and the
breakdown.
Incidently, when high-frevoltage drops to the short-circuit values of the supply premature
other
and
low-power transients of
source. Or the breakdown in the dielectric field limits quency oscillations
in
itself, producing a new configuration of the dielectric unknown origin appear a system, it is advisable to
field, containing conducting regions, in which the air look for the existence of corona somewhere in the ciris broken down and luminous by "corona," separated cuit, as possible cause.
The theory of the dielectric strength of air at and
by insulating air spaces below the breakdown gradient.
In this case the discharge current is limited to the near atmospheric pressure, outlined above, is based on
small "corona current." A further increase of voltages the distribution of dielectric stresses and therefore
then may extend the breakdown across the terminals depends on the configuration of the dielectric field. As
and change the limited discharge into an unlimited the dielectric field is essentially determined by the
shape of the terminals, the latter thus is of fundamental
spark discharge.
Between needle points, corona always precedes the importance. At the same time, in the "conduction of
spark discharge; between spheres at moderate distance gases in a vacuum" a theory has been developed, based
(less than 3 to 4 times the radius of the spheres), on convection phenomena by carriers, the ions, and in
this theory the configuration of the dielectric field and
spark discharge occurs without corona.
The disruptive discharge of single gaps in air, is the shape of the terminals is practically not considered,
instantaneous, that is, no time lag is produced in the and the two conceptions thus apparently are out of
spark gap, if no corona precedes the spark discharge, line with each other, though the phenomena of the
and the gap length is greater than the energy distance. Geisler tube and those of air at normal pressure should
When with rising voltage corona precedes the spark differ quantitatively only, by the different value of air
discharge, a time lag exists. With gaps of less than the density 3.
Approximately, and within a limited range, the
energy distance, and multigaps, a time lag also seems
to exist. The combination of resistance in series and voltage e required to break down a gap of length 1 at
capacity in shunt to an instantaneous spark gap gives air density 3, is constant as long as the product 13 is
the gap a time lag owing to the time required to charge constant, that is, for constant quantity of air within
the capacity by a current limited in value by the the gap. In other words, for different air densities 3,
resistance. Without a capacity in shunt, an otherwise the gap length 1 at constant voltage e varies inversely
instantaneous sphere gap in series with a resistance proportional with the air density 3. The air densities
lags by the time of charging the capacity of the spheres in vacuum tubes usually vary from one thousandth to
themselves, through the series resistance. For high one millionth of normal air density and less, and even
values of resistance, this may become quite consider- the longest vacuum tubes usually therefore correspond
at atmospheric pressure to gaps of very small length,
able, especially with large spheres.
This standard theory of dielectric strength of air at greatly within the energy distance. In other words,
and near atmospheric pressure is fairly satisfactory for the phenomena of vacuum tube conduction correspond
gaps ranging between the energy distance and a larger to the phenomena occurring in very small gaps at
atmospheric pressure, where the conventional theory
value where corona begins,

October 1923

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

does not apply, and the configuration of the field and

the shape of the terminals become of lesser importance.
Inversely, the voltage phenomena of the Geisler tube,
may also appear in the dielectric strength of large gaps
at atmospheric pressure, as terminal phenomena,
though then secondary in magnitude to the gradient
voltages of the dielectric field.
The conventional theory considers only the voltage
gradient or intensity of the dielectric field as determining the dielectric strength, but not the direction, that
is, assumes the same breakdown strength for either
polarity of the impressed voltage. This resulted from
the use of alternating voltages in determining the laws
of the dielectric strength of air, due to high direct volt-

1031

Investigation shows that it is the form and intensity

of the dielectric field at and near the positive terminal
which has the greatest effect on the disruptive voltage
of the air gap, while the field at and near the negative
terminal is of secondary importance. Thus if at constant gap length the positive terminal is gradually
changed in steps from a sharp point to a flat plate,
while the negative terminal remains unchanged in form,
the disruptive voltage changes correspondingly over a
wide range. A corresponding change in the physical
form of the negative terminal however, with the positive terminal remaining unchanged in form, varies the
disruptive voltage relatively little.
In Fig. 1, curve I approaches the striking distance
curve between needles, as modified by the. different
distribution of voltage gradients due to the large
100
negative terminal, and curve II similarly approaches
AC.l Maximum , |
the striking distance curve between 12.5 cm. spheres.
90
/ ____
This suggests the explanation that the disruptive
III X> H
r Il Z/
I
discharge is due to carriers of current produced by the
/'
/
80
field at the positive terminals, the positive ions possibly.
//
Sphere Positive
Tests made at atmospheric pressure on very small
to fractions of microns-show that with
Grounded & High
|I
Igaps-down
70
decreasing gap length the voltage does not indefinitely
/ Sphere Negative
decrease, but reaches a finite minimum value, of
/ Needle Grounded I/
E
Shere Negative
about 320 volts, and becomes apparently constant at
U60
I /'
this value. This would be the minimum voltage re_
and Grounded
/
quired to disrupt an air gap, no matter how small. It
- _/ll
50
means
that with decreasing gap length, the voltage
zI
3:
gradient of the air film rises to very high values, and
40
|
|tl1such
1t' thin
4 air films have an extremely high dielectric
strength; gradients of 6.2 million volts per cm. have
}
l
Ai | | /
been reached.
Co330
/
Inversely, with increasing gap length, the voltage
with the
increases
thereto,
gapoflength,
;
I
20L /,/lX
320 proportional
volts. The increase
about and
an initial
value
jI < j
from
with increasing gap length, however, is not at the rate
___
of the dielectric strength of air, 30 kv. per cm., as
10
might be expected, but is at approximately twice this
rate; about 60 kv. per cm., up to voltages of about
2
0
4500. Then the increase of voltage with increasing
12
4
6
o
8
2
10
14
GAP IN CENTIMETERS
gap length decreases and approaches the normal value
of the breakdown strength, 30 kv. per cm.
FIG. 1
The voltage e required to break down an air gap of
age of considerable power being unavailable until length 1 and uniform dielectric field intensity thus
recent years. It is very far from correct, however, and would consist of
in an unsymmetrical field, the disruptive strength with
(1) A constant voltage el = 320 volts = 0.32 kv.
the voltage in one direction, may be more than twice
(2) A voltage gradient g2 of about 60 kv. per cm.
that with the voltage in opposite direction. As illus- over a short length lo = 0.07 cm., giving a constant
tration, Fig. 1 shows the striking distances between a voltage e2 = 60 lo kv.
(3) The voltage gradient of dielectric strength,
needle and a 12.5 cm. sphere, for various gap lengths.
Curve I gives the results winth the needle positive, g0 = 30 kv. per cm., for the rest of the gap, I-lo~or
curve II with the needle negative, and curve III with g3 = 30 (1I-lo ); thus giving a total voltage, apalternating voltage. As seen, I and II differ greatly, proximately:
e = e1 + e2 + e3
II being more than twice the voltage of I. The alter= 0.32 + 30 1o + 30 l
nating curve approximates I, as is to be expected. In
or, if 1< 10, simply
the voltage range between I and II, such an unsyme = 0.32 60 1
metrical gap rectifies.

HAYDEN AND STEINMiETZ: HIGH-VOLTAGE INSULATION

1032

In a non-uniform field, the breakdown would occur

from the positive terminal up to the distance, to which
the breakdown gradient g = 30 extends, and this space
is filled with corona.
In Appendix I is given a suggestion of-the mechanism
of the dielectric breakdown of air by carriers issuing
from the positive terminal.
III. Liquid Dielectrics
Oil and similar materials (as petrolatum in the highpotential cable impregnation) are the most important
liquid insulating materials and are depended upon for
the insulation of the highest voltage electrical
apparatus, the transmission power transformer. The
fairly satisfactory agreement of the behavior of air as
insulating material, with the theory of a definite
dielectric strength, modified by the conception of the
"energy distance" led to the application of the same
theory to liquid dielectrics. Here, however, the agreement with experience is not satisfactory; and tests
show that oil does not have a practically available
definite dielectric strength, but successive tests made

30 r

l
l

_
.

z20
I

lo

SamplefOil

v V III

~20

20

'*I_
O

1
o-

_
___K
Ai

Transactions A. I. E. E.

sphere gap in oil, at moderate voltage between the

spheres, and observing the resultant motions and
deformations. It is even somewhat questionable
whether liquid dielectrics have a definite dielectric
strength at all, or whether disruption does not always
occur through gases produced from the dielectric by
the electric stress. A discharge through oil seems
always accompanied by the appearance of gas bubbles,
and while the gas bubbles may not be the cause, but
the result of the discharge, there are some differences
in the shape of the striking distance curve of an oil
gap from that of an air gap, which look as if the disruptive discharge in air is the result of the action of the
dielectric field on the air, while the disruptive discharge
in oil is the result of the action of the dielectric field on
a material produced from the oil by the dielectric
field, that is, the dielectric field produces dissociation
or ionization of the oil, and the disruption occurs
through the gases of ionization

IV.

Solid Dielectrics

The dielectric strength of solid dielectrics is often

expressed by the disruptive voltage gradient in a
-r
uniform field, in kilovolts per cm., or per mil. Such
_- - an expression however, is meaningless, unless accompanied by a statement of the thickness of the
dielectric, to which it applies, as the dielectric strength
-LH 4Sof solids varies with the thickness, sometimes by many

- -.

hundred

44444
60 70
30 Wo
40 50 Wlv+498
90Ilo10 20ffio
NUMBEROFBREAKDOWN

cent.

Besides

this,

the

disruptive

strength

previous history
under
and even with
~~of
the ally
sample,
etc., conditions,
differences
widemade
just astests
the same
identic
the voltage application, the humidity,

occur between successive tests of solid dielectrics, as

shown by Fig. 2 to occur in liquids. The "disruptive
strength of a solid dielectric" thus has only a limited
with all precautions with the same oil in the same gap meaning, and that, when the conditions are fully given.
Disruptive strength tests of solid dielectrics are
(using such a large quantity of oil as to exclude the
effect of deterioration) differ from each other by values usually made on sheets between metal plates as termany times greater than the possible errors of observa- minals, -to produce a uniform dielectric field. This is
tion. The upper curve on Fig. 2 shows the deviation open to the objection that at the edges of the terminals
from average of 100 successive tests of the disruptive the field is not uniform. However, experience shows
voltages of an oil gap, and the lower curve shows the that puncture does not always occur at the edges of
corresponding results with an air gap. Plotting the the terminal plates, but often inside, and by throwing
number of times each deviation from average occurs, out the tests in which the disruption takes place at the
in A, as function of the value of deviation, gives a edge of the terminals, and averaging those in which the
,disruption occurs well inside of the terminals, fairly
probability curve.
The most satisfactory explanation of the mechan- representative results are secured.
Two phenomena come into consideration in solid
ism of the breakdown of oil and other liquids seems
to be that based on the assumption of lack of homo- dielectrics: The electric conductivity and the dielectric
geneity, namely that the breakdown of an oil gap losses in an alternating field.
The dielectric has an appreciable though usually
is the result of materials of different conductivity or
low conductivity. That is, under an impressed
the
by
gap
into
sucked
being
capacity
very
the
specific
dielectric field, or being produced in the gap by it, and voltage, a slight current is conducted through it,
so causing a distortion of the field, with local high causing, alo of wer and thereby a heating of the
densities,- at which disruption begins. This can be dielectric. The power consumed by ohmic conduction
visually shown by dropping a minute quantityoflinstheudielectriousualy is extremely small, at least at
ahorizontal atmospheric temperature. The conductivity usually
water or other foreig nm ial i
FIG. 2

October 1923

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

1033

increases very greatly with the temperature, often

MECHANISM OF THERMAL BREAKDOWN OF THE
approximately exponentially, (for instance, increases
DIELECTRIC AS THIRD CLASS CONDUCTOR
ten-fold for every 25 deg. temperature rise, so that at
high temperature the dielectric may become a good
in some range of
conductor). Often, the conductivity also increases third class conductors, that is, have
such
a high negative
with increasing voltage. In an alterating field, the their voltampere characteristic
resistance loss usually is very small compared with temperature coefficient of conductivity, that in temperature equilibrium the voltage decreases with inthe other dielectric losses.
Under an alternating voltage, losses occur in the
g curent
Suppose a uniform sheet of solid dielectric is exposed
nature of a dielectric
dielectric, more or less of the
. proportional. to ~~~~to
t aa constant
c
ddirect vvoltage between two conducting
hysteresis. While probably.
mainly
the frequency, some >of these losses may increase
terminals. Due to the slight conductivity of the
. . in
.faster
than the frequency, some at a lesser rate, giving
the e~~~~~~~dielectric, a current then flows through it at a uniform)
though very low current density. This consumes
latter case a dielectric loss at zero frequency.
energy, converting
heat, and o,n
gives a
lTtercse
slgh teprtr
ieoof itthe
h intodielectric.
ilcrc
These dielectric
dielectric losses
losses may be
be representedbyanelectric
represented
by an. slight
rise
..
no
temperature
Now,
energy current and an effective conductivity of di- material can be absolutely uniform and homogeneous,
electric hysteresis, and lead to a power factor of the and thus in this sheet of dielectric there will be some
dielectric. If the energy current of dielectric losses is spots of a slightly higher conductivity. However
proportional to the frequency and to the voltage, and slight the difference, at such a spot the current density
the dielectric power loss therefore proportional to the must be slightly greater, thus the energy consumption
frequency and the square of the voltage, then the and the heat production slightly greater, giving locally
power factor of the dielectric circuit is constant and a slightly higher temperature rise. However little
independent of the frequency and the voltage. An this may be, if the conductivity of the dielectric inincrease of power factor with increasing voltage shows creases very rapidly with increasing temperature, it
losses increasing faster than the square of the voltage, will lead to a slightly higher current density, a corressuch as losses due to ionization of gas spaces inclosed pondingly higher energy consumption and heat proin the dielectric; a decrease of power factor with duction and thus temperature rise, and so on. Two
increasing frequency shows losses increasing less than p bilities the
or constant lossesisuchhn
eithera the
which
be
proportional to the frequency, orconstantlossess
exist, such
conducted away from
due can
to its
"hotheat
spot"
as ohmic conduction losses.
temperature rise, is more than the heat produced by
The observation of the power factor thus is one means the increased conductivity due to the temperature rise.
of judging the nature of the losses in the dielectric. Then the temperature of the hot spot finally limits
A dielectric of high disruptive strength in general itself, and stable thermal conditions pertain. Or the
has low conductivity and low dielectric losses, thus a low heat which can be conducted away from the hot spot
power factor in an alternating field; but the reverse into the dielectric, by the temperature rise, is less than
is not necessarily the case, that is, a dielectric may the heat produced in the dielectric by the increased
have low conductivity or low power factor, and still conductivity due to the temperature rise. Then the
be of poor disruptive strength.
conditions are unstable, that is, the temperature conIt is doubtful whether a true dielectric strength of tinuous to rise and the conductivity to increase insolid insulation exists, that is, a definite voltage definitely, until the energy concentration at the hot
gradient at which the dielectric is disrupted directly spot becomes destructive, and the dielectric is destroyed
by the intensity of the dielectric field; or if it exists, by "puncture."
it is so far beyond the disruptive strength observable,
The successive steps of this phenomenon have been
as to be of no practical consideration. That is, observed by limiting the current density by suitable
puncture of a solid dielectric probably in practise always terminals and can be calculated from tests made on the
occurs as the result of a more or less rapid progressive conductivity of the dielectric at various temperatures
deterioration, far below the true dielectric strength, and voltages.
and the latter may be of moment only under direct
As the energy consumption by the conduction current
lightning condition, for instance, in the puncture of a varies with the square of the voltage, there is thus a
sheet of insulation by a powerful Leyden Jar discharge definite voltage-for a given set of conditions-at
(giving the characteristic appearance of an internal which instability is reached and puncture results, and
explosion),
this voltage is the "breakdown voltage," its voltThe mechanism of the breakdown of a dielectric age gradient the "dielectric strength"' of the
under electric stress, may probably be any of a number material under the conditions of test. This voltage
of possibilities, thermal, mechanical, chemical, physical, depends upon the initial temperature, and considerably
etc.
decreases with increasing temperature; at higher
may

asopohmiconductionth losesuenc,

1034

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

temperature, a lesser temperature rise is necessary to

reach unstable conditions, and at the higher initial
temperature the phenomenon starts with a higher
conductivity, that is, greater power consumption and
heat production. The heat conductivity of the dielectric and of the terminals and other surrounding
objects, and their ability to quickly absorbing heat,
that is, their specific heat, are essential factors in
determining the value of the puncture voltage. An
essential factor also is the relation between temperature
and conductivity of the dielectric, etc., so that the
"dielectric strength" thereby determined, is very far
from a constant for the material.
Suppose the "hot spot" is filamentary in shape.
That is, the diameter of the spot of slightly higher
conductivity (or temperature) at which the breakdown
starts, is small compared with its length in the direction
of the lines of force, so that heat conduction from it is
essentially into the dielectric. Then a change of thickness of the dielectric will not affect the heat conduction, and in this case it follows that the disruptive
voltage is proportional to the thickness of the dielectric,
its dielectric strength independent of the thickness.
Suppose however the "hot spot" is plate-shaped,
that is, its dimension parallel to the lines of force
(from terminal to terminal) is small compared with
its diameter parallel to the terminals, so that the heat
conducted from it flows into the terminals. Then an
increase of thickness of the dielectric gives an increase
of heat to be eonducted through a longer distance
through the same area, thus instability is reached at
a lower voltage gradient. In this case it is found that
the puncture voltage is proportional to the square root
of the thickness of the dielectric, and the dielectric
strength inversely proportional to the square root of
the thickness.
In general then, in dielectric breakdown by temperature instability, depending on the path of the heat
conduction, the puncture voltage varies between the
first and the 0.5th power of the thickness of the dielectric, and the dielectric strength or breakdown
gradient varies between independence of the thickness,
and the-0.5th power of the thickness of the dielectric.
This phenomenon of dielectric breakdown by thermal
instability is the result of the increase of energy loss
in the dielectric, with increasing temperature, and
occurs wherever the losses greatly increase with the
temperature. While its mechanism has been illustrated above on ohmic conduction in a direct-voltage
field, the same phenomenon occurs in the same manner
in an alternating field, and as the result of the specific
losses of the nature of a dielectric bysteresis, as far as
these losses increase with the temrperature. The increase-and usually very rapid increase-of the power
factor with the increase of temperature of most dielectrics shows that the losses in the dielectric in an
alternating field, increase with the temperature, and

Transactions A. I. E. E.

thermal instability leading to puncture at "hot spots"

thus results.
MECHANISM OF DISRUPTION DUE TO MECHANICAL

INSTABILITY

Suppose a dielectric encloses a particle of a higher

conductivity or higher specific capacity. An electric

field then exerts a mechanical force on this particle,

which tends to elongate it in the direction of the

electric field, and to compress it at right angles thereto,
that is, tend to form it into a filament short-circuiting
the dielectric field. With liquid dielectrics, this can
be observed visually by dropping a little water into a
horizontal sphere gap in a transparent oil. Each
droplet, as it is sucked into the field between the spheres,
lengthens into a filamentary conductor which bridges
between the terminals, and then is destroyed with a
flash by the heat of the current conducted by it. A
similar phenomenon occurs with moisture in a solid
dielectric, except that the puncture usually is permanent; occluded moisture moving under electrostatic forces through the pores of the solid dielectric,
may form a conducting bridge between the terminals
and by the heat of the current in the moisture filament
start destruction of the dielectric. Or the moisturethread may partially bridge between terminals and
locally short-circuit the field, which will cause excessive
dielectric stresses in the part of the gap in series with
the moisture filament. In these over strained portions
of the field, destruction starts by thermal or chemical
instability, etc.
In impregnated insulation, 'where the impregnating
material is liquid or viscous, motions of the impregnating material through the impregnated material may
result from the mechanical forces caused by differences
in specific capacity or conductivity of the materials,
leading to a redistribution of the impregnating material.
Conductivity and specific capacity here act in the same
direction, in general. The differences in specifi ccapacity
of differentt dielectrics are however relatively small;
rarely more than 1 in 6 in materials which are suitable
for apparatus insulation; while the same constituent
materials may differ in conductivity over an enormous
range. The final effect of conductivity as compared
to specific capacity is likely therefore, to be much
greater. On the one hand, the mechanical forces due to
differences in specific capacity appear instantaneously,
and thus are present also in alternating fields. On the
other hand the differences due to difference in conductivity appear gradual, being accompanied by the
formation of local internal electrostatic charges, and
may require seconds or even minutes for their completion. These local changes would not be present or
only partly present in alternating fields, and the
mechanical actions of alternating fields thus differ from
those of continuous fields. This can conveniently be
observed visually on sphere gaps in oil.

October 1923

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

1035

action, and such ionization due to enclosed air probably

is a frequent cause of disruption. It is guarded against
by excluding the air by impregnation with a material
of higher dielectric strength and higher specific capacity,
such as oil. It means however, that the process of
to be effective must be perfect.
impregnation
to note that some of the mechanisms
It is interesting
of breakdown, as thermal instability and mechanical
h c
forces, are reversible, that is, at the withdrawal of the
oi of 1 ielectricof0
megohms
.
1000
voltage, the original condition may gradually return,
megohms
of
diTion
the dielectric undamaged, while the chemical
fre.ths
side leaving
tric.atin
es
theelinodl
of the sphericaltycof
deterioration
by the electrostatic cutting edges or by
th
sheicapricl
rounding dielectric, and-at least temporarily, until ionization is irreversible, that is, what damage has been
a redistribution of internal charges has occurred-the done is permanent and remains at the withdrawal of
density of the lines of dielectric force and thus the the voltage, and at the next applica-tion of voltage,
voltage gradient in the dielectric outside the poles of the deterioration progresses further.
the spherical particle would be three times normal.
In non-homogeneous dielectrics, such as laminated
if
the
times
specific
three
than
normal,
It would be less
capacity orconductivity of the sphesinsulation, due to the differences in the ratios of the
capacities and the specific conducspecific
respectiveof the
less than given above from the surrounding delectrc.tivities
component dielectrics, gradual changes
But it would be greater if the particle is not a sphere, may occur in the distribution of the voltage gradients
but more irregular in shape, and would assume much through the dielectric, with the formation of interal
higher values at the edges and at points of the charges, which continue for some time, thus resulting
particle. Thus in the dielectric adjacent to edges or
points of an enclosed particle of higher specific capacity
V. Time Lag of Insulation
or conductivity, very high-voltage gradients occur, and
imporan optie onsulation
T t
may be far beyond the dielectric strength of the malg Thelforemas
terial, and lead to local breakdown of the material by dieltricpsrtant
h
'which
that trenislii in the lage gae uprto
what may be called "electrostatic cutting edges.
u
the
a
voltage
,and
bc
hold
insulaio can hold back the voltage, and when
With organic insulation, the effect usually would be anan insulation
when
by hihlcltmeaue,aceia
co by hthis voltage gradient, the "dielectric strength" or
s
change which in general increases the conductivity. "
of the insulation, is reached,
The mechanical electrostatic forces brought about "'disruptive strength"
hereby are in the direction of the field, so that the shape becomesia cuctor.
of the product of chemical deterioration tends towardbecomes
Experience shows the following condition: At
the form of a conducting needle, with excessive voltage
many insulations, the voltage at which
gradients in front of it, gradually piercing the dielectric least with very
breakdown
or
occurs, depends on the time
disruption
ei
until..
fia.untrocrbtenh
.. of voltage application.
terminals. ~~~orduration
between
until final puncture occurs
The lower the
fl
.. the
Thus in a laminated insulation consisting of very many
There
voltage, the longer the time it has to be applied. applied,
layers, a foreign particle in one of the layers though is
a minimum voltage, which continuously
originally forming only an insignificant part of the still just disrupts the insulation, and inversely, the
total thickness of the dielectric, may gradually but h
cumulatively, in the course of time, pierce and destroy hithas toe vaped.
the insulation by its electrostatic cutting edges, the ithastobeapplied
This time, during which a given voltage has to be
.
voltage gradients within the dielectric,
being
to cause disruption, is called the "time lag" of
average
..applied
still very low compared with the tested "dielectric
strength" of the material. The destruction will be his vlt
atio o iparticular
voltage of
the faster, the higher the local voltage gradient. This brief apiation the voltag n,wic anently
seems to be the main reason why such excessive margins
impulse bratio" dof the time of voltage application.
arins api
Time lag bears to dielectric strength, in electrical
i
alr Irequird
itfnh
engineering, a relation similar to the one between
elasticity and mechanical strength in mechanical enIONIZATION
The specific capacity of the common solid dielectrics gineering; if it were not for elasticity, there would
is from 2 to 8 times that of air. Thus if air pockets be no mechanical engineering. A pebble dropped on
are contained in the dielectric, the electric stress in the an armor plate would shatter it,-since theoretically,
air is much higher than in the solid, and as the break at the point of impact, without elasticity infinite medown gradient of air is low, it breaks down with the chanical forces would be produced.
So without the phenomenon of time lag, there would
formation of "corona," giving heat and chemical

MECHANISM OF DISRUPTION BY CHEMICAL

DETERIORATION
Suppose the dielectric contains a particle of higher
caact or hihrcnutvt. Le
assume, at first, this particle to be spherical in shape,
and of relatively infinite specific capacity or conductivity, that
-thefield(forinstan
ta is,
feld for istanc
isshort-circuiting
rt-irtingth
ductivity,

specific.

us1

resistivity)woul
aition inhad.itr
fresinstiv
isathr ieast
rofndthe

dapieleceri

carbonization~ ~~ ~

mrrv

1036

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

be no electrical engineering, as there would be no

possibility of insulation, since every insulation, even a
low-voltage lighting circuit, is theoretically constantly
exposed to transients of infinite voltage (though
negligible energy), that is, far beyond its possible
dielectric strength, by the inductive and capacity
effects of any change of circuit conditions, and thus
saved only by the time lag of its insulation,
Just as the relations between mechanical strength
and elasticity give the wide variety of structural
materials, on which the mechanical engineer depends,
and which we denote by brittle, tough, ductile, elastic,
rigid, flexible, yielding, etc., etc., so the relation of
time lag to dielectric strength gives us insulating
materials of widely different properties and correspondingly widely different uses-but our knowledge in this
field is unfortunately still very limited.
To illustrate the importance of the time lag: Experience as well as calculation shows that in 2300-volt
primary distribution circuits during thunder storms,
potentials of short duration of the magnitude of
hundred thousand volts are not infrequently produced
by the setting free, by the lightning flash, of the
bound charge of the atmospheric electrostatic field.
The lighting transformers distributed over these circuits are not, and economically cannot be insulated to
stand this voltage continuously. Hence, they must
depend on the time lag of their oil insulation to stand
the lightning voltage until it is dissipated or discharged
by the lightning arrester. Inversely, the time lag of
the lightning arrester must be so short, and its discharge rate so high, as to discharge the lightning voltage
in a time less than the time lag of the transformer,
bringing the voltage down to values safe for the
transformer.
From the pnenomenon of time lag it results that the
rate of voltage application has a discriminating effect.
Suppose two insulations are used in parallel, the one of
lower dielectric strength but higher time lag than the
other. (As for instance, the oil insulation of a transformer
and the surface air insulation of its entrance bushings).
At very rapid voltage application, the voltage may
rise beyond the dielectric strength of the stronger
insulation of shorter tirre lag, in less than the time lag
of the weaker insulation, and the former thus punctures,
while with a slower voltage application the weaker
insulation of greater time lag would puncture. Thus
under lightning conditions, the transformer bushings
may flash over, short-circuit and blow the fuses without
any damage to the transformer, while under high
potential test the oil insulation may puncture far
below the voltage at which the bushings flash over,
To illustrate the importance of this discriminating
effect of rate of voltage rise: The insulation of highvoltage transmission lines depends on sectional or
string insulators. With these, it is very desirable that
in case of a failure the insulator disks should flash over
rather than puncture. The transmission insulators

Transactions A. I. E. E.

thus are designed for a puncture voltage much higher

than the flashover voltage, and 60 cycles tests as well
as high-frequency tests show such insulator strings to
flash over and not to puncture. Nevertheless lightning
punctures them not infrequently. The time lag of
flashover is greater than that of puncture, due to the
relatively high capacity of the insulator string. Therefore, under lightning- conditions, that is, very rapid
application of voltage of considerable energy, the
voltage reaches puncture values in less than the time
lag of flash over, while in low-frequency tests flashover
limits the voltage, and in high-frequency tests the rate
of voltage application is reduced, that is, the wave
front flattened by the capacity of the insulator string,
unless there is very great power back of the voltage,
so as to maintain it against the short-circuiting effect
of the capacity of the testing appliance (so-called
"lightning generator").

TIME LAG OF SOLID AND LIQUID INSULATION

When the mechanism of the dielectric breakdown of
the insulation consists of a cumulative thermal, mechanical or chemical effect, as discussed above, it
inevitably involves a time lag, and usually a considerable one.
This time lag may be as short as a fraction of a
second, such as occurs in the electrostatic field of an oil
gap sucking in a moisture droplet, stretching it into a
filamentary conductor, bridging between the electrodes
and flashing over. In the formation of hot spots in
solid insulation, it may take minutes and hours, until
the process leads to the final disruption, and may extend
to days and years in the chemical action of ionization,
etc., so that here the time lag of the insulation breakdown gradually merges into the aging or deterioration
of the insulation.
For instance, under overload a cable may get overheated and some hot spots form in the insulation and
finally, after some hours, lead to a puncture. If
however the load is taken off before the final breakdown,
the cable cools and the hot spots disappear and leave
the insulation undamaged, and we say that the cable
has been saved by the time lag of insulation and no
harm done by the temporary overload, because in this
case of approaching thermal breakdown the process is
reversible.
If however, under high-voltage test, ionization occurs
in the cable and by chemical action begins to destroy
the insulation, we also may say that the cable is saved
by the time lag of chemical breakdown, if the overvoltage is taken off before failure has occurred. However, as this process is not reversible, some damage has
been done, and at the next overvoltage further damage
is done and adds itself, until final disruption occurs.
In some respect, we may thus consider the gradual
and slow aging and deterioration of the insulation
during the years of use, which limits its final life, as a
progressive breakdown, and the entire life of the

1037

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

October 1923

insulation as the time lag of breakdown; however, the final condition at disruption. Then the phenomenon would be exponential, that is, start at maximum
this rather extends the meaning of the term.
rate and gradually decrease, thus theoretically take
infinite time. It is customary to consider as the duraELECTRICAL TIME LAG
Air apparently has no time lag, at least, no appreci- tion of such an exponential phenomenon (for instance
able time lag, and the dielectric breakdown of the air the duration of an exponential transient) the time
gap between spheres at a distance greater than the which would be required if the phenomenon conenergy distance and less than the corona distance, tinued to its end at constant initial intensity. Aswith a negligible impedance between spheres and the suming this for the time lag, then the duration T would
source of voltage supply, is as nearly instantaneous as be the time lag at impulse ratio U = e.
can be measured. Time lags (or impulse ratios) with, In the adjustable time gap consisting of an inair gaps therefore are due either to the configuration of stantaneous sphere gap in air, shunted by a capacity C
the gap, or due to the conditions of the supply circuit. and in series with a resistance r, the duration T, as
A sphere gap with a considerable resistance in series defined above, then is:
T = r C,
has an appreciable time lag, the greater, the higher the
resistance, due to the time required to charge the thus has a physical meaning as the rate of condenser
charge.
capacity of the spheres over the resistance.
The capacity of two spheres of ratio R, at distance 1
By shunting an instantaneous sphere gap by a small
from each other, is, if:
160

__

__

14012

g8r

120

;0 l

100

1
9 t 2 R1 +R

1
{~~~~~~~.13

10

0b

20
40

la_f_.0_m eoecns

I_

60

________

20

o_

-4

104

-3

-2

-1

LOG OF TIME
i03 0.01 0.1 1
TIME IN SECONDS

10

100 1000 10,000

FIG. 3

capacity C, with a non-inductive resistance r in series,

and a sphere gap of 12.5 cm. spheres, at 5 cm. distance,

in series with 1000 ohms resistance, would have a time
lag of 0.05 microseconds.
Two 25-cm. spheres in air, at 0.7 cm. distance, have
a time lag of about 0.5 microsecond. (Within the
energy distance).
Two 1-cm. spheres in air at 3 cm. distance, have a
time lag of about 1 microsecond. (Corona)
A needle gap of 5 cm. in air has a time lag of the
magnitude of 1 to 2 microseconds.
A sphere gap of 2 mm. in oil, between 2.5 cm. spheres,
has a time lag of about 20 microseconds.
Whether, and how far and in what manner the time
lags depend on configuration, size, voltage, etc. is
still largely unknown, and the entire field, though of
high importance, is almost entirely unexplored.

any desired time lag can be produced by the proper

choice of C and r, and such a combination thus forms
an adjustable time gap, convenient for the testing of
time lags.
This raises the question of measuring the time lag.
The usual way of expressing a time lag is by the impulse
Appendix
ratio. This however is not a constant, but a function
THE DIELECTRIC BREAKDOWN OF AIR AS A
of the duration of the voltage application, as shown by
the curve Fig. 3, which gives the impulse ratio of a
CONDUCTIVITY PHENOMENON
cable, for duration of the voltage application from continuous down to a microsecond.
-1. Assume that the dielectric breakdown of air, and
An approximate expression of the electrical time lag the conductivity produced by it, are due to conducting
by a constant may sometimes be given as follows: particles or carriers, that is, particles which carry
Assuming that proportionality exists,-a usual as- electric charges and thereby conduct the current.
We may assume that in free air, there are always some
sumption with such phenomena-that is, that the
change of the condition in the dielectric is proportional such carriers present. In an electrostatic field, they
to the difference between the existing condition and are set in motion with an acceleration a proportional

Transactions A. I. E. E.

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

1038

to the dielectric field intensity g, and therefore in the or in general, in a non-uniform field:
time t acquire the velocity v, in a uniform field:
c
(g-go) dl
(1)
v = gt

(10)

In a non-uniform field, the velocity would be given is the total production of conducting particles in the
by:
distance 1, or the ionic density or conductivity produced
the
g in the distance 1.
(2) byIf wegradient
=
Jw
v
gdt
assume that the conducting particles or car. riers are produced from the gas molecules by collision,
us asm ththro
it v the ionic density or conductivity c must reach a finite
Let us assume that there isLeta critical
vvelocity
at which these carriers produce additional carriers by maximum value, at which "complete ionization'' of
collision with the atoms or molecules of the air or gas. the gas has occurred, and the conductivity thus reached
The time t1 required to reach collision velocity vo its maximum or "short circuit" value co.
Dielectric breakdown thus would be characterized
then is,

=i

vO

g
and the distance, within which this is reached:
t1 i

21vVoO

V02

o2g

b
V(3) by

cO = (g - go) lo = constant = e1
or, in a non-uniform field

(11)

= (g -go) d = constant = el
cO1(2

(4)

(12)

thus:
vO2
(5)
11 g = 2 - const.
or, in a non-uniform field:
v2
const.
(6)
g d l = 22
This is of the dimension voltage. That is, a definite
voltage,
vo2
(7)
eo = 2

is required to raise the free conducting carriers of the

air to the collision velocity vo, at which they produce
additional carriers from the atoms or molecules.
The voltage eo may be considered as of the nature of
a terminal drop, which must be present before the formation of conducting carriers can take place.
It then would be the minimum voltage, approached
for minute air gaps:
eo = 320 volts.
2. As the conductivity of the air rapidly disappears
with the withdrawal of the dielectric field, these conducting carriers must be very short lived.
Let then go be the voltage gradient, at which the rate
of production of conducting particles equals their rate
of decay, and the number of conducting particles, that
is, the conductivity of the air, remains constant (whatever may be its value, whether hightor low). /8=

This quantity el is of the dimension of an e. m. f.

and may be called the "corduction voltage."
To cause dielectric disruption of an air gap, there
must thus be available a dielectric field giving
(a) A terminal drop eo = 320 volts.
(b) An excess gradient g, above the critical gradient
go (the so-called "dielectric strength of air"), extending
over such a distance lo as to integrate to a definite
voltage el, required to produce complete ionization or
maximum conductivity of the air.
(c) The normal or critical gradient go, required to
maintain the conductivity, and extending as far as the
breakdown extends.
3. Applying this to the air gap between spherical

electrodes:

With spherical electrodes of radius r, assuming first,

that the other electrode is of such shape or so far away
as not to distort the field, if we denote
gr the gradient at the surface of the sphere of radius r.
ro the radius, at which the gradient has decreased
to go
we have, at radius x,
r 2
0 ) go
(13)
gx =
x

and

ro

ro

co = ) (g - go) dxz = go )qi(

r

__

1+,

\2

- 1L d x

____+r2o)=g(r-)

then is the rate of increase of conducting particles, or it follows, solving for r0

the "ionizing force,'' and
c
=r
c =
go r
(9)
=(g-go)lI

GlIr

rO

)(14)

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

October 1923

m(

1039

is, the gap length 1 is less than the length lo, within which

)
k

the density of conducting particles increases from the

low density existing in the free air, to the maximum
and the voltage gradient at the surface of the electrode density, where maximum conduction is produced, and
the dielectric gradient thereby has dropped to go, so
sphere is
2
that no further increase of ionization occurs.
The conductivity throughout the gap then varies
Ur = go ( )
from
the low initial value to the maximum value of
\rI
)2
gradient go, and its average is half the maximum value,
f
=
1+
(15) and the average gradient throughout the gap thus is
twice the minimum value go.
/r
I
That is, in a short gap, the average gradient must be
or approximately
2 go, and the voltage consumed by the gradient (which
might be called "stream voltage," in distinction from
2m
+
the
(16)
voltage eo of (7) or the "terminal drop") thus is:
gr= g
e, = 2 go I
(22)
where
*this gives as the total voltage consumed by the dielecCO
(17) tric breakdown of a very small gap of length 1,
m =
go
e = e0j+-e1
go= goo, where 3 = air density and goo equals
(23)
= eo + 2 go l
critical gradient at normal air density 3 = 1, at 0
b. Assume the gap length 1 to be greater than the
deg. cent. and 76 cm. barometer.
ionizing distance lo, that is, the distance in which equaIt thus is
tion (1) produces maximum conductivity. Dielectric
{ 1 + MO
breakdown of the gap then requires the voltage
=r
ro= r I+ r0r}(18) to 1 rdcsmxmmcnutvt.Deeti
a/r6
eo = (g - go) lo
conducting stream, or the
(1MO as the voltage producing the
+
not
limited to the terminal
(though
(19)
drop"
9 = goo d
"terminal
P
+
3
x
r6
/)
tr~~~~~~alone).
r t 1

=oo

t1

2m2m

ei = 2 go

where

/O
m

lo

(20) as the voltage consumed by the average, gradient 2 gO

in the distance lo, in which the conductivity is produced by (1)

co
g oo

21

90

(1- lo)

(21) as the voltage consumed in the rest of the gap (l - lo)

the gradient go required to maintain conductivity.
(olouibyThe
and
it
voltage,
and co iS the (total) conduction or ionizing
total voltage thus is,
e = eo + e1 + e2
go0 the critical gradient at unit air density, and 3 the
= eo + 2 go l'o + go (1 - lo)
air density. Equations (14), (16), (18) and (20) are
(24)
the equations of energy distance and of surface gradient,
= eo + 9O to + g0 l
given by Peek, and are here derived from the concep= Eo + go I
tion of the dielectric breakdown as a conductivity
where
phenomenon.
(25)
4. With small gaps between plates or spheres of a
Eo = eo + go lo
length, the. s
gap legth,the
compared with
largeargecompred
diameter
he gp
diamter
ith the
t total excess voltage over that consumed by the
dielectric field and the gradient are uniform throughout s the
the gap, and the maximum gradient, or gradient at the critical gradient go in the gap l.
c. Assume now that the gap is larger and the field
electrode surface, equal to the mean gradient through- iS not
uniform.
As Uo
is the minimum gradient of ionic conductivity,
OUt the gap.
co

g =e/l
and the conduction voltage required to produce maximum conductivity is by (11) given as
(11)
el= (g- go) I

that is, the gradient required to maintain constant

conductivity, the conductivity of the gap space can
extend only up to the distance from the electrodes,
within which the gradient g is greater than Uo.
The dielectric breakdown thus extends up to the
(a) Assume the gap length is so small, that the distance of the gradient Uo, and within this distance 1',
ionic density increases throughout the entire gap, that the voltage consumed must exceed the value of break-

Transactions A. I. E. E.

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

1040

6. The polarity of the charge of the conducting

down gradient go times distance l', by the total ionizing
particles or carriers, which give the dielectric breakvoltage Eo of (25), that is, must be
down of air gaps at atmospheric pressure, can be detere' + Eo + go 1'
mined
by the action of unsymmetrical gaps at direct
or, since
voltage. Consider an unsymmetrical air gap such as
that between a needle point and a plate, or between a
e' = g d l
small and a large sphere. In such a gap, with increasing
J
applied
it is
of go is first reached
thethe critical gradient
and voltage,

production

thereby

Eo =f(g-90go) dl

eo + go

conducting particles

326)

10

2e--oo(6
that is:
The voltage c in paragraph 3, which gives the
constant m in Peek's equation (14) etc., is not merely
the excess voltage eo required to produce the conducting
stream, but is the sum of this voltage plus the excess
voltage Yo lo required by the excess gradient in that
part of the field, in which the conductivity has not yet
(
reached full value.
F0 =

~~~~~

5. Consider now numerical values.

From test values of small gaps between large spheres
and between a large sphere and a plate, we find,
e = 320 + 59 1
= eo + 2 go l
This gives
eo = 320 volts
=
2 go 59 is in good agreement with go = 30, as the
temperature correction has not been applied.
For medium gaps, (case 4b,) the tests give,
e = 2370 + 30 i
Thus
Eo = 2370= co
This gives by (17)

30j

50

~200
>150

100

10-'305

I - 20__CENTIMETERS

.
~~~~~(27)
0-3xlO

I o

D.C
A. C. Maximum

,.Peek A. C.

20

20005

1.0

-1

30

NIG 4

40

started at that terminal, which has the greater curvature and thus the higher gradients, the point or small
sphere. If then the polarity of this terminal is that
of the conducting carriers, these move outwards away
from the terminal, towards the opposite terminal, and
thereby short circuit the gap and cause disruption as
soon as the average gradient is sufficiently high to
maintain the conducting stream across the gap. If
15,000-

inMgood4m 2)7 .VO079

igodagreement with Peek's value:
g

= g

= go{i+

0.

0. 27 or

10 10,000
>

CENU0.56

5000
~~~~~~~~~~5

while Peek gives-

$7

28

mn

90{l+

0.54
-

J
The limit lo between equation (23) and (24) is
e = eO + 2 g0 l = 0. 320 6010o
e =m F0 go I = 2.370 3010o
X 10-3
a68X1 68cm.
which is in close agreement with the experimental data
given in the tables and figures, of lo =m 69.
In Figs. 4, 5 and 6 are given numerous test values
taken under various conditions, as indicated, with the
theoretical curves shown by the drawn lines,
Vr

00

0
0.2
0.1
0.3
(CENTIMETERS

0.4

0.5

FIG. 5

however, the larger terminal-plate or large sphereis of the polarity of the carriers, a higher voltage across
iS exbefore
the critical carriers
gradientstarted
gapand
IS required
~~~~~~the
the swarm
of conducting
ceeded
at this larger terminal and disruptive discharge occurs.
At the smaller terminal, in this case, the formation of
conducting carriers has started already at lower voltage, but these carriers do not move away from the
terminal and across the gap, and thus do not cause

October 1923

1041

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

unlimited current, and when it begins, thus short-circuits the gap and drops the voltage. Electronic conduction however is limited in current, and when it
occurs, the voltages can be maintained and still further
raised. At such high vacuua, where the voltage of
ionic conduction has risen beyond that of electronic
conduction, and the conduction become electronic, the
voltage can be raised beyond the value of ionic conduction through the traces of residual gases, and then
ionic conduction again begins and "short-circuits" the
electron tube, thus limiting electronic conduction.
Therefore a practically perfect vacuum is needed for
pure electronic conduction in the modern high-power
electron tube.
the other side, with pressures higher than atmosloo
pheric, the disruptive voltage increases with increasing
pressure, and approximately proportional thereto, as
shown by experiment. It may be expected then, that
at some high pressures the voltage of ionic conduction
increases beyond that of electronic conduction, and the
latter limits the increase of disruptive strength of gases
C0
Whileindicate
little
this field
10
at high pressures.
2.5
dielectric
thathasthebeen
experiments
investigated,
CENTIMETERS 7.55
FIG. 6
gradients of air and gases with increasing pressures
reach a limiting value somewhere at 1000 kv. per cm.
gradient of air, the disruptive voltage of such an
Discussion
unsymmetrical gap should be about twice as high with
the voltage of the carriers on the large electrode, than
R. W. Sorensen: My first point is in connection with7the
with the voltage of the carriers in the opposite direction. statement that we use our insulations under stresses which rarely
Fig. 1 shows this, and also shows that the disruptive exceed the breakdown voltage of air, though tests show a strength
voltage is lower with the small terminal (the needle of 10 to 20 times that of air for many of the insulations used.
This plea for a more strenuous use of insulating materials is
point in this case) positive.
It follows from this, that the conducting particles or
carriers, which carry the disruptive discharge in air at
1i
- -I
0 I
atmospheric pressure, are positively charged, that is,
A
1
they may be the positive ions, but cannot be the nega0 80
tive electrons.
Y
<
60
- 7. Dielectric conduction thus may be either ionic,
/|

break down, but move towards the terminal near which

they started, and form a local corona in the space at
and near this terminal.
Thus with unsymmetrical gaps, the direction of the
polarity of the applied voltage should make a difference
in the disruptive voltage, and this voltage be higher
with the voltage in one direction, than with the voltage
in the opposite direction. As the average gradient
in the space in which the conducting particles are
formed, was shown to be about twice the breakdown

ttSOn

or electronic. The relation possibly is the following:

At atmospheric pressure the voltage required for
ionic conduction, that is, disruptive discharge, is much
lower than the voltage required for electronic conduction, and the conduction, that is, the disruptive discharge through air, is ionic, that is, by positive carriers.

a
z

40
20 -

- 1 1 14
8
10
12
2
4
6
c0
GAP LENGTH IN CENTIMETERS
FIG. 1-ARCING POTENTIAL VS. GAP LENGTH
Between Needle and Sphere, 12.5 cm. dia.
Curve I-Direct Current, Sphere Negative
Positive
II- ts

With decreasing air pressure, the voltage required for

ionic co-nduction decreases, approximately proportional
to the air pressure, and the conduction thus remains
III-Alternating Current, Maximum Values.
ionic. At some low pressures a minimum value of
voltage (or rather voltage gradient) is reached. Then,
with further decrease of air pressure, the voltage of interesting in the light of a request in another paper given this
in the
thetransformers
author recommends
a decrease
morning,
which to
th aecease
,eces test
agi lnceases
inrass eueto
du to tn
ioi
codcto
onlc onclctlonagal
grounded.
with one terminal
voltagein a.pplied
II

Of carriers, and finally passes beyond the voltage of

It

In Fig. 1 of the paper the curves show the interesting fact

electronic conduction. The phenomenon then changes that a given potential will cause spark over between a point and
its character, ionic conduction ceases and electronic a sphere for much greater spacing when the sphere is negaUtive,
conduction begins. At high vacuua, the conduction than will be the case writh the needle points negative and the
thu

is

elcroi.
tnus ISelectronlc.

There iS a difference between ionic and electronic

conduction: Ionic conduction permits practically

sphere positive. The results shown in Fig. 3 have been dupliby Messrs. Otis and Mendenhall, two students abt Calia~~~~~ated
fornia, Institute of Technology a.s shown by the curves in Fig. 1

accompanying this discussion. In making these tests the high-

1042

HAYDEN AND STEINMETZ: HIGH-VOLTAGE INSULATION

voltage alternating current was rectified by means of a twosegment commutator driven by a two-pole synchronous motor.
The a-c. curve is very much like that obtained by Hayden
and Steinmetz, but the d-c. curves differ in shape because of the
pulsating current obtained with the commutator, whereas
Hayden and Steinmetz had very steady current delivered by a
four kenetron rectifier. An explanation of the dip in these
curves would be very interesting.
In John S. Townsend's "Ionization of Gases by Collision" is
found this statement "When the point is negative, the strong
field is near the negative electrode, so that the potential required to produce a discharge is less than when the point is
positive." Also in "Conduction of Electricity through Gases,"
1906 edition, by J. J. Thompson on page 498 we find this statement ".this minimum potential depends upon the
sharpness of the point, the pressure and nature of the gas, and
the sign of the electrification of the point, being less if the point
is negatively than if it is positively electrified." Hence we have
from these authorities statements which, at first reading, appear
quite contrary to those given in this paper.
There is, however, some confusion as to the definition of the
term "spark discharge," in the texts referred to the term does
not seem to apply to an arc current and the voltage required
to cause it, but means the point at which a leak discharge only
and not a complete. are over occurs between -the electrodes.
In fact in one treatise on the subject of ionization this definition
appears: "Sparking potential may be defined as the potential
which is required to maintain a very small current in the gas."
Dr. Millikan has explained the Hayden-Steinmetz results on
the basis of the increased difficulty experienced in extending the
ionization envelope with negative points, as compared to that
phenomena for positive points, hence the required higher potentials for a breakdown over a given distance.
In discussing the mechanism of "thermal breakdown of
dielectrics" the paper follows the work of Mr. Wagner as published in the JOURNAIT for December 1922, but Wagner does not
deduce the same law. Also some tests we have made in our
laboratory do not conform the deduction that puncture voltage
is proportional to the square root of the thickness of dielectric.
To emphasize "time lag" is indeed worth while as our experience shows it an important factor in making an analysis of voltage stresses on insulations, and indicates that an intensive study
should be made of the laws which govern it. It may be of
interest to note here that in testing thousands of porcelain
insulators I have found that very few which stand a potential
test at minimum arc over voltage for fifteen seconds, fail when
the potential is applied for a longer time. In testing apparatus
insulated with organic insulating materials very often potentials
apparently harmless when applied for short periods will cause
breakdown when applied for longer periods of time.
At the bottom of the last page: "It follows from this, that
the conducting particles or carriers, which carry the disruptive
discharge in air at atmospheric pressure, are positively charged,
that is, they may be the positive ions, but cannot be the negative

electrons."

If an atom is ionized and you have your positive ions turned

loose what becomes of the negative ions, where do they go?
Then follows: "At atmospheric pressure the voltage required for ionic conduction, that is, disruptive discharge, is

Transactions A. I. E. E.

much lower than the voltage required for electronic conduction."

This is explained further on in the paper. If you have a perfect
vacuum, as stated, how would you have the pure electronic
conduction? I presume tubes in which there are heated filaments have been used.
C. P. Steinmetz: The paper deals with a subject which has
assumed, in the last year, a still greater importance than it had
before, that is, the problem of high-voltage insulation and
mechanical breakdowns.
This is such a vast field and so much work has and is being
done, and can be done, that the paper must necessarily be only
a general part of the preliminary announcement of the results

bandnoediatiso futureocaon.
There is, however, one feature which begins to get clear,
namely, that our conception of insulation and of breakdown of
insulation again begins to change and to be subjected to multiplication. To members in the early days, insulation was merely
a boundairy bar. We knew, by experience, that a tenth of an
inch of insulation of a conductor would protect it against 2300
volts. Then, when it came to higher voltages we realized that
there is something occurring within the boundary bar, of importance, and it is not merely the material, but there is a dielectric field with potential radius and other things within the
boundary bars which require consideration and study, and which
we are studying.
Now, it seems that our views are just beginning again to get
a multiplication with respect to at least the failure of insulation
and the mechanism of breakdowns. It seems to be clear that
the mechanism of breakdowns, under the failure of insulation
on high voltage is a phenomena of instability. In other words,
it is not that insulation fails, that dielectric breaks down, when
electric stresses are beyond limits and value, but it is said that
under conditions very much lower than those gradients in
lightning conditions of instability occur which gradually brings
about the multiplication and changes leading ultimately to a
breakdown of insulation. It is, therefore, a condition of instability of constants of material which instability brings about,
largely upon lesser changes, which leads either to destruction
or breakdown. Therefore, the mere reduction or stress on the
insulation is not a factor which saves breakdowns, but that the
new problem of insulation seems to assume the shape of arranging
of designs in the dielectric field of insulation, so as to get the
condition of stability and not instability. That is the future
which seems to impress itself upon us. The more we study the
problems of insulation we find that it is not existing stresses that
cause this, but largely it is the result of stability, or instability.
Now, that is not only true in solids, but probably in the air.
In this respect we could point out the same idea that the discharge makes its own gradient. Now, you have the same conception there, that it is the discharge which is taking place in the
dielectric field about conditions which are unstable. We all
know that if we had 2,000,000 volts spread over, I don't know
how many thousand inches, there could be a gradient that
would be so low there would be no puncture and for that discharge under such conditions. By the discharge making its own
gradient there would be produced a dielectric condition of
instability which would finally lead to self-destruction.