621.396.61 : 621.317.374 : 621.3.017.143 Monograph No.

27
MEASUREMENTS SECTION

AN APPROXIMATE METHOD FOR DEDUCING DIELECTRIC LOSS FACTOR

FROM DIRECT-CURRENT MEASUREMENTS
By B. V. HAMON, B.Sc, B.E.
{The paper was first received28th August, and in revised form 9th November, 1951. // was published as an INSTITUTION MONOGRAPH 15/// March, 1952.)
SUMMARY can be obtained direct from the current at a time t equal to
An approximate method is developed for deducing the dielectric 0-1//seconds after applying a direct step voltage. Since the
loss factor of a solid dielectric from the charging current that flows absorption currents are most easily measured if / is large (e.g.
after the sudden application of a direct voltage. It is shown that, greater than about 10 sec), the method is best adapted to the
subject to certain conditions, the loss factor at a frequency / cycles rapid evaluation of loss factors at frequencies below 0-01 c/s,
per second can be deduced from the charging current at a time 0-1// where a.c. measurement would be tedious or impracticable.
seconds after applying the direct voltage. The range of application
of the method is discussed, and its use in correlating d.c. measure- The approximation was developed as an aid in the experimental
ments with a.c. bridge measurements on the same sample is illustrated study of the reversible anomalous charging current and low-
by experimental results. frequency power-loss in solid dielectrics at low field-strengths.
It has proved to be of considerable value in the comparison of the
LIST OF SYMBOLS results obtained by d.c. methods with the more usual a.c. test
/ — Frequency, cycles per second. results, since both sets of data can be plotted to the same co-
/ r z Time, seconds. ordinates on one sheet. The approximation has not been applied
o» = Angular frequency, radians per second, = 2TT/. to liquids, because only rarely do liquids show a reversible
O(0 = Relaxation function. anomalous charging current.10
K' — Relative permittivity.
K" = Dielectric loss factor. (2) THEORETICAL DEVELOPMENT
8 = Loss angle. The theoretical equations for K' and K" in terms of the current
Co = Capacitance of sample at very high frequencies, produced by the application of a step voltage1.2 are:
farads.
Ca — Capacitance of measuring electrodes when sample is
replaced by air, farads. = c \ O(0 cos (jotdt (1)
Go = Steady-state d.c. conductance of sample, mhos.
/?, n,a,A — Constants.
«"(«) = 7r\ — + 0 ( 0 sin uitdt . . . ( 2 )
i — Charging current, amperes.
V = Applied direct voltage, volts. Jo J
T = Time-constant, seconds. where
Co = the capacitance of the sample at very high frequencies.
(1) INTRODUCTION Ca — the capacitance of the electrodes when the sample is
replaced by air.
Two methods are in general use for measuring the electrical Go = the steady-state d.c. conductance.
properties of dielectric materials at low field-strengths. The O(0 = the current flowing in the sample after applying unit
first is the measurement of the relative permittivity K' and dielec- voltage at t — 0.
tric loss factor K" (or their ratio K"/K' = tan 8, where 8 is the co = the angular frequency, 2TT/.
loss angle) over a range of frequencies. The second is the <J)(/) does not include the steady-state conduction current from
measurement of the anomalous charging current or absorption which Go is calculated; thus O(oo) = O.
current as a function of time measured from the instant of O(0 is frequently called the relaxation function or decay
application of a direct potential. It is well known that the function for the material. In deriving the equations, it has been
quantities measured in these two methods are interrelated,1-2 assumed that the Hopkinson principle of superposition holds
and that, in particular, the loss factor at a particular frequency good; its validity for solid dielectrics at low field-strengths has
can be calculated from measurements of the absorption current been checked experimentally on many occasions, e.g. by
within a suitably chosen time-range. The calculation involves Voglis11 and von Schweidler.3
an integral, and has in the past been carried out by expressing
the absorption current as an approximate function of time Many workers have found that O(0 can be approximated over
either by a suitable mathematical formula3.4.5.6 or by the sum a wide range of values by the empirical equation O(0 — fiCQt-",
of a number of exponential terms.7.8.9 Good agreement where /? and n are constants for a particular material at a
between observed and calculated values has been obtained, particular temperature. This empirical equation is assumed
especially by J. B. Whitehead and his collaborators,7- 8>9 but the initially in the following sections. It will then be shown that the
calculation is too tedious to be used in routine work for the approximate result obtained is acceptable even when 0(0 departs
rapid correlation of data from the two types of measurement. considerably from this empirical form.
Putting 0 ( 0 = PCot-n in eqns. (1) and (2) gives
In the present paper, a simple method is developed for con-
verting the absorption current into dielectric-loss-factor values.
It is shown that the loss factor at a frequency/cycles per second - if) cos ^ (3)
Correspondence on Monographs is invited for consideration with a view to
publication.
Mr. Hamon is in the Division of Electrotechnology, Commonwealth Scientific and *"(«) = i - T^° + j8C0co»-ir(l n) cos (4)
Industrial Research Organization, Sydney, Australia.
[151]
152 HAMON: AN APPROXIMATE METHOD FOR DEDUCING DIELECTRIC
Eqn. (3) holds only for 0 < n < 1, but eqn. (4) is true if magnitude of the anomalous charging current FO(0 and the
0 < // < 2. Eqn. (4) often appears in a different form, obtained direct conduction current VG0, so that none of these quantities
by substituting Tr/Y(n) sin mi for r ( l — «) in the form given here. needs to be determined separately.
As derived so far, the result depends on the assumption that
(2.1) Dielectric Loss Factor 0(0 is of the form fiCrf-", when 0- 3 < n < 1 • 2. The effect of
n
For the present consider only the loss factor K". This can be departure from the t~ law will now be considered, first where
expressed in terms of the anomalous charging current O(^) at a / is either small or large relative to 0-63/co, and then where /
particular time tl as approximates 0-63/co.
It does not appear possible to give a general treatment in
« " - [ C 0 + # , ) K . . . . ( 5 ) either of these cases. The treatment is therefore confined to
provided that co and t1 are related thus examining the effect of using certain hypothetical relaxation
functions in which the departure from a t~n law at particular
-\ln values of / is more marked than is usually found in practice.
cot, (6) For values of t small or large relative to 0-63/co, the two
hypothetical relaxation functions chosen are shown in Figs. \{a)
The expression on the right-hand side of eqn. (6) is almost and 1(6). In each case the function consists of two sections each
independent of n in the range 0 - 3 < « < l - 2 , and, to an following a / - " law, but with different values of n. The sections
accuracy within ± 3 %, can be taken as having the mean value intersect at / = afco, where a is a parameter.
0-63. The range 0-3 < n < 1 -2 covers very well the range of The effect of the sudden change of slope can be expressed
n found in practice when a / - " law is fitted to observed values most easily as a percentage change in the value of the integral
of O(/); the values of n are usually less than unity and most in eqn. (2) when O(0 is altered from the form shown in Figs. \{a)

/A.t" 0 '

t
m CO
3 !\
DO
o
it ~? \
,_ Case II
^XCl^O-63/u)

loff t logt
(b)
Fig. 1.-—Hypothetical relaxation functions for estimating the effect of departure from a /—« law where / is either much greater or much
less than 0 • 63/™.

frequently lie between 0-5 and 0-9. S. Whitehead12 has given or \(b) to the form At~n for 0 < / < co, A and n being given the
an expression equivalent to eqn. (6), but does not appear to have values they had for the section of the curve containing / = 0 • 63/co.
investigated its relative constancy as n is varied.
To an approximation accurate within ± 3 %, eqn. (5) can now
be written Thus, for a < 0- 63, the values of O(0 sin u>tdt for the functions
/,^<VhO(0-63/co)
shown in Figs. l(a) and 1(6) are compared with A2t~"2 sin cotdt
Jo
or, if /(/) be defined as the total charging current as a function (cases I and II), and for a > 0-63, the values are compared with
of time after the application of a step voltage V, i.e. the sum of
the anomalous charging current and the direct conduction sin cotdt (cases III and IV). Table 1 shows the results
current, so that i{t)IV ••= Go + O(0» this equation takes the
simple form of comparisons of this kind for different values of a in each of
the four cases shown.
To evaluate the integrals for the hypothetical functions, it
or alternatively
/(01//) was necessary to have values of \ u~n sin udu for 0 < x < 8
(8) Jo
2rrfCaV
and for n = 0-3 and n = 1-2; these were computed numerically.
It will be seen that eqn. (8) gives the value of K" at any Modification of an assumed t-» function by altering part of
frequency / in terms of the total charging current at a time it to a negative exponential was also tried as an alternative to
/= 01//sec after applying a sudden direct voltage V. The the method just described. The negative exponential and the
expression is independent of Co, j8 and n, and of the relative t~" function had the same value and the same slope when
 LOSS FACTOR FROM DIRECT-CURRENT MEASUREMENTS 153
Table 1

f oo
Percentage change in <t>(/) sin <s>tdt
a J 0
(co/)
in = 0-3, m = 1-2 n\ - 1-2, n2 = 0-3

01 + JjCaseII
0-2 = io} C a s e I
1-57 + 171
3-14 + 7
4-72 - 8 ^Case III + 1 yCase IV
6-28
7-86 + SJ -?J
log io u)
/ = a/co. The values of percentage change due to a modification Fig. 3.—Dielectric loss factor K" as a function of angular frequency
of this type were less than the values shown in Table 1, for for the relaxation functions shown in Fig. 2.
corresponding values of n and a.
By numerical integration.
It will be seen from Table 1 that the part of <P(0 that needs By using eqn. (8).
to be known with reasonable accuracy if K" is to be calculated
at an angular frequency co is the part corresponding to values of
/ in the range from 0-2/w to 3/cu. The limits of this range are Fig. 2 shows the two hypothetical relaxation functions chosen
not rigid but both are appreciably lower than the limits (0-1 and in this case, and Fig. 3 shows the corresponding true values and
5 periods, corresponding to the range 0-63/to < / < 31/a>) given approximate values of log K" for different values of log co.
by Benedict.4 Eqn. (8) can therefore be expected to apply with Both the curves in Fig. 2 are of the t~n form for / < 1 and
reasonable accuracy if the t~n law is a good approximation to for / > 10. For values of / between 1 and 10, the curves have
the observed values of <D(0 in the range 0-2/w < t < 3/to, and been made circular arcs, to the ends of which the t~n sections
if the value of n for the approximation in this range is between are tangential. In curve (a), the value of n is 0- 3 for low values
0-3 and 1-2. It may be noted that <D(0 must still satisfy some of / and 1 -2 for values greater than 10. This type of relaxation
restrictive conditions for t < 0-2/o> and t >3/w, but it is to be function leads to a maximum in the curve of log/c" against
expected from a knowledge of the general behaviour of dielectric log co [see curve (a) of Fig. 3]. In curve (b) of Fig. 2, the value
materials that these conditions will always be satisfied in practice. of n is 1 -2 for / < 1 and 0-3 for / > 10; this leads to a minimum
The possibility of using the approximation expressed in in the graph of log K" against log co [see curve (6), Fig. 3].
eqn. (8) when the / - " law is not a good approximation to the Curves (a) and (b) of Fig. 3 give the values of log/c"
observed values of <P(0 in the range 0-2/o> < / < 3/w will now corresponding to the functions plotted in Fig. 2 when the
be considered. For this purpose, hypothetical functions have approximation expressed in eqn. (8) is used. The true values
been assumed in which the departure from a / - " law over one of log/c" for several values of co have also been obtained by
decade is more marked than is usually found in practice, par- numerical integration and are shown by circles and squares in
ticularly for values of t greater than about one second. Fig. 3. It will be seen that the approximate formula for K"
gives fairly satisfactory results, in spite of the unusually curved
form of the log/log plots of the relaxation functions. The
greatest error is in the region of the peak in curve (a) of Fig. 3,
where the approximate expression gives a value too high by about
15%. Even in the two extreme cases just considered, it is
doubtful whether the errors introduced by the approximate
\ formula are large enough to affect conclusions that might be
based on the general shapes of the curves, on the position of the
maximum or that of the minimum.
It is of some interest to compare curve (a) of Fig. 3 with
Fig. 4, which shows the relative values of the true and approxi-
mate loss factors as functions of frequency when the behaviour
of the material is characterized by a single time-constant process.
In this case, if the time-constant is denoted by r, and if A is a
\ constant, we have
(bit) = Ae-' (9)
\
s so that the approximate expression in eqn. (7), neglecting Go,
\\ gives
\ y^T C OJ/tUt
V
\ Ca LOT
\
\
1 0 1 The true expression for K", however, is
log l0 t
A T COT
Fig. 2.—Hypothetical relaxation functions for estimating the effect of K" -
departure from a t~n law where / approximates 0-63/co.
VOL. 99, PART IV. 11
154 HAMON: AN APPROXIMATE METHOD FOR DEDUCING DIELECTRIC
(3) EXAMPLES OF THE USE OF THE APPROXIMATION
In experimental work on the problem of dielectric after-effect
in the Division of Electrotechnology, C.S.I.R.O., measurements

/
/
/

\ have been made in the frequency range 50-105 c/s, using a

bridge circuit with thermionic oscillator and cathode-ray-
oscillograph detector. An improved bridge circuit permitting

\
measurements down to 5 c/s has also been used. Measurements
/ 1

/
/
•
•

1 of the charging current following the sudden application of a

direct voltage have been made, using an electrometer in a
circuit similar to that described by Doust and Sulston.14 These
measurements cover the range from / = 15 sec to t = 200 sec
and in some cases higher values. Using the approximation
given in eqn. (8), these d.c. measurements have been converted
Fig. 4.—Dielectric loss factor K" as a function of angular frequency co to values of loss factor at frequencies from about 0 007 to
for a single time-constant process. 0 • 0005 c/s or lower. To reduce the gap between 5 and 0 • 007 c/s,
Exact expression. a bridge circuit supplied with a sinusoidal potential from a
• • Approximate expression. motor-driven potentiometer has been used. This bridge gives
loss factor (and relative permittivity) at frequencies between
The approximation in this case breaks down completely for
0-16 and 4 c/s, and can be used on some samples at frequencies
COT less than unity, but it is interesting to note that it shows a
below 0-16 c/s, although the process of attaining a balance in
peak of the right order of magnitude close to the correct fre-
that region is tedious. The gap between 0 • 007 and 0 • 16 c/s can
quency, and that the shapes of the curves near the peaks are
be filled in cases where K" is fairly large by adapting the electro-
similar in Fig, 4 and curve (a) of Fig. 3.
meter circuit to a point-by-point method of measurement,5-15«16
but in most cases the variation of loss factor with frequency is
(2.2) Relative Permittivity so uniform in this region that graphical interpolation appears
There does not appear to be a corresponding simple procedure justified.
for deriving the relative permittivity K as a function of frequency Fig. 5 shows some examples of measurements made by the
from measurements with direct current. Clearly, the steady- three methods described, and illustrates the ease with which the
state conduction current must first be separated from the results of charging-current measurements can be assessed in
reversible anomalous current, since it cannot contribute to K. relation to measurements in the audio-frequency range.
S. Whitehead12 has given for the capacitance of a sample an In two of the examples given, a pen-recorder with a flat
expression similar to eqn. (6), but it is not so nearly independent frequency-response up to 80 c/s has been used to obtain charging-
of n, and is in any case valid only for 0 < n < 1. The greater current values for t as low as 0 08 sec. The values of loss
difficulty in deducing a.c. capacitance values from d.c. tests has factor calculated from these records fell within the frequency
been pointed out by Banos.'3 range of one of the bridge circuits, so that a direct experimental
Since the capacitance and that part of the loss factor not due check on the accuracy of eqn. (8) was possible. The shape of
to d.c. conduction are related to one another,1-2 the single curve curve (a) of Fig. 5 in the region of 0- 5 c/s is very similar to that
of K" as a function of frequency over a wide frequency range will of one of the hypothetical curves dealt with earlier [Fig. 3,
reveal the important properties of the material in this range, curve (b)]. It will be noted that in both cases the approximate
and little more would be gained by plotting K as well as K". expression gives values of loss factor that are too low.
1000
\
\
\
\
100 \

\
10

•
1
N "+«

01
i
\

001
\
i
00001 0001 001 01 1 10 100 1000 10 000 100 000
Frequency, c/s
Fig. 5.—Experimental results, showing the application of the approximate formula [eqn. (8)].
—o—o— K" as a function of frequency, obtained directly from bridge readings.
1 1 K" obtained from total charging current using eqn. (8).
Curve (a): docosyl alcohol (m.p. 71° C).
Curve (by. paraffin wax, pyrolysed 3 hours at 330° C with restricted access to air.
Curve (c): commercial mica capacitor, 0-001 piF («' assumed to be 7-0 at all frequencies).
 LOSS FACTOR FROM DIRECT-CURRENT MEASUREMENTS 155
Since the values of K" may become very large at the lower (6) ACKNOWLEDGMENT
frequencies, it is necessary to use a logarithmic scale for K" as The author wishes to express his thanks to Mr. T. Pearcey for
well as for frequency. This has the advantage that if the direct the numerical computation of the integrals used in Section 2.1.
charging current becomes constant within a reasonable time the and to Mr. H. W. Stokes for assistance in the measurement of
contribution made by the constant conduction current at any currents.
frequency can be indicated graphically by a straight line with a
slope of — 1, as shown in the dashed line against curve (b) in (7) REFERENCES
Fig. 5. Further, eqn. (4) shows that when the charging current (1) GROSS, B.: "On the Theory of Dielectric Loss," Physical
has the /-« form over an appreciable range of values of t the Review, 1941, 59, p. 748.
curve of K" against w or/in double logarithmic co-ordinates will (2) FROHLICH, H.: "Theory of Dielectrics" (Oxford University
be a straight line with a slope of {n — 1) in the corresponding Press, 1949).
frequency range. Parts of curves (b) and (c) in Fig. 5 are of this (3) VON SCHWEIDLER, E. R.: "Studien iiber die Anomalien im
type, the values of n being 0-4 for curve (b) [0-2-5 c/s] and Verhalten der Dielektrika," Annalen der Physik, 1907,
0-75 for curve (c) [0- 02-200 c/s]. 24, p. 711.
(4) BENEDICT, R. R.: "Behaviour of Dielectrics. A Study of
the Anomalous Charging Current and the Variation of
(4) LIMITS OF THE APPROXIMATION Dielectric Energy Loss and Capacitance with Frequency
The conditions that should be satisfied if the approximation in Solid Dielectrics," Transactions of the American I.E.E.,
is to hold with reasonable accuracy are most easily specified in 1930, 49, p. 739.
terms of the shape of the curve of loss factor against frequency (5) TANK, F.: "Ober den Zusammenhang der dielektrischen
in double logarithmic co-ordinates. Relatively straight sections Effektverluste von Kondensatoren mit den Anomalien
of the curve should have slopes to the frequency axis of between der Ladung und der Leitung," Annalen der Physik, 1915,
0-7 and + 0-2, corresponding respectively to n = 0-3 and 48, p. 307.
n ••=- 1-2 in the t~n expression for the charging current, except (6) GRANIER, J.: "Les pertes d'energie dans les dielectriques,"
where the contribution due to the direct conduction current Bulletin de la Societe Francaise des lilectriciens, 4 me Serie,
predominates. In curved sections of the curve, it is to be 1923, 3, p. 333.
expected that the error due to the approximation will be greater (7) WHITEHEAD, J. B.: "Dielectric Absorption and Dielectric
the greater is the curvature. As a guide, the curvature in Fig. 3, Loss," Journal of the Franklin Institute, 1929, 208, p. 453.
i.e. a change in slope from — 0-7 to + 0-3 or vice versa in one "Dielectric Loss and Relaxation Time in Resin," Physics,
decade of frequency, can be taken as an upper limit. These 1932, 2, p. 82.
restrictions on slope and curvature apply only to the section of (8) WHITEHEAD, J. B., and BANOS, A.: "Predetermination of
the curve obtained by using the approximation. the A-C. Characteristics of Dielectrics," Transactions of
In addition to the above restrictions, the superposition prin- the American J.E.E., 1932, 51, p. 392.
(9) WHITEHEAD, J. B., and KOUVENHOVEN, W. B.: "Funda-
ciple must hold for the material. It is to be expected that this
will be so if, as is usual, the anomalous charge and discharge mental Properties of Impregnated Paper," ibid. 1931,
currents are reversible and also proportional to the applied 50, p. 699.
voltage. It is often found, however, that the steady-state direct (10) JACKSON, W.: "A Resume of Recent Work on the Ano-
conduction current is not proportional to the applied voltage. malous Behaviour of Dielectric Liquids," Journal I.E.E.,
When this is so, the contribution due to the d.c. conduction . 1934, 75, p. 93.
cannot be given an exact meaning, but the use of the approxima- (11) VOGLIS, G. M.: "Uber die dielektrischen Nachwirkungser-
tion is not affected in regions where the anomalous current is scheinungen in festen Nichtleitern," Zeitschrift fur
large compared with the conduction current. Physik, 1938, 109, p. 52.
(12) WHITEHEAD, S.: "A note on the Analysis of Dielectric
With corresponding restrictions, it is possible to use the
approximation to calculate charging-current values for low values Measurements," Transactions of the Faraday Society,
of / from observed values of loss factor. 1946, 42A, p. 66.
(13) BANOS, A.: "A-C. Characteristics of Dielectrics—II,"
Electrical Engineering, 1936, 55, p. 1329.
(14) DOUST, J. F., and SULSTON, W. J.: "Properties and Testing
(5) CONCLUSION of Dielectric Materials," Institution of Post Office
It has been shown that the dielectric loss factor at a fre- Electrical Engineers, Paper No. 165, 1938.
quency / cycles per second can be calculated from the total (15) JORDAN, H.: "Zur Kenntnis des Verhaltens anomaler
charging current 0-1 If sec after applying a constant direct voltage Dielektrika," Berichte der Deutschen Physikalischen
to a sample of solid dielectric material. The calculation involves Gesellschaft, 1912, 14, p. 451.
an approximation in which the error should not exceed ± 1 5 % (16) GROSS, B.: "Zum Verlauf des Einsatzstromes im anomalen
and is generally about ± 5 % of the true value. ;'Zeitschrift fur Physik, 1938, 108, p. 598.