Distortion oi Travelin g Waves by Corona

By H. H. SKILLING P. de K. DYKES
MEMBER AIEE NONMEMBER AIEE

Synopsis 2. Shortly after its formation space charge becomes relatively

immobile. This is doubtless due to the formation of relatively heavy
In order to make possible the prediction of the attenua­ ions, for which the mobility is almost negligible as compared· to
electrons
tion and distortion of lightning waves on transmission
lines an equation is proposed that relates the change of 3. The supply of space charge to the region about a conductor
commences when the critical gradient is reached, and continues as
shape of a traveling wave to the corona loss on the line at
long thereafter as the applied voltage continues to increase. During
normal operating frequency. Results obtained by com­ this time there is loss of energy from the conductor
putation from this formula are compared with experimen­
4. After t h e crest of a voltage wave is reached, and the voltage
tal results of various investigators. begins to decrease, the space charge remains (approximately) con­
stant in magnitude and fixed in position. During this time there is
little loss of energy from the conductor, what loss there is being due

T
HE ATTENUATION of high-voltage traveling waves to diffusion of the ions in the electric field at a slow rate. This
interval ends when the critical gradient of opposite sign is reached
by corona is of importance because of the protection
and a new interval of active ionization commences
it affords to terminal equipment when lightning
strikes a transmission line at some distance from the near­ Space charge may be considered as forming a protective
est transformer station. The traveling wave that results sheath about a conductor—a sheath that receives and
from the lightning discharge has been found by many ob­ terminates much of the incoming dielectric flux, so that
servers to lose energy quite rapidly as long as its crest volt­ no part of the air about the conductor is subjected to elec­
age is well above the corona-forming voltage of the line on tric stress greater than the critical value. If voltage were
which it travels. A line of several miles length may usu­ raised so that the electric stress at any point exceeded the
ally be depended upon, therefore, to reduce the initial critical value, the gradient at the overstressed point would
voltage of the lightning wave to a safe value. immediately be reduced by further ionization and the
Various equations have been proposed to express the production of additional protective space charge. The
rate of attenuation of the wave. These are reviewed by action is analogous to the relief of mechanical stress in a
Bewley.3 All of them, including an equation proposed by piece of overstressed steel by inelastic deformation of the
one of the present authors, 2 are purely empirical and have material.
one or more adjustable constants which cannot be pre­ When the voltage applied to a conductor is alternating
determined. I t is now proposed to relate the distortion it is necessary to locate a protective sheath of space charge
and attenuation of a traveling wave to the loss of power about the conductor as the voltage rises to a crest, and
in corona at the line's normal operating frequency. then to neutralize that charge and replace it with one of
opposite sign as the voltage changes to a crest of opposite
The Mechanism of Corona Loss polarity. Energy is consumed in the production of the
space charge; the amount of energy consumed per cycle
The energy loss that accompanies corona is due to mo­ depends on the crest voltage of the wave but, as the amount
tion of free electric charge in the space surrounding con­ of space charge required is independent of the duration of
ductors at high potential. This was pointed out by Ryan a cycle, the energy per cycle consumed in its production is
and Henline 4 in 1924, and the study of space charge which independent of the frequency of the applied voltage. The
has since been carried on leads to the following conclu­ power, consequently, or energy per second, is proportional
sions: to frequency. But this is not the total loss.
There is another kind of loss that is due to migration of
1. There is a critical electric gradient for air t h a t cannot be ex­
ceeded. Any a t t e m p t to increase the gradient above the critical the ions of space charge in the electric field. The protec­
value results in profuse ionization of the air, and t h e charges liberated tive sheath of space charge can never be stationary, for it
by ionization take up such positions in space t h a t the gradient is is at all times acted upon by the electric field of the applied
maintained at (or below) the critical value voltage and that of other parts of the space charge. When
A paper recommended for publication by the AIEE committee on power trans­
space charge moves in response to the electric field of the
mission and distribution. Manuscript submitted February 25, 1937; released applied voltage it consumes power. So there is a con­
for publication M a y 25, 1937.
tinuous drain of power due to the motion of the space
H. H. SKILLING is assistant professor" of electrical engineering at Stanford Uni­
versity, Calif., and P. DE K. DYKES is an instructor in electrical engineering a t charge, quite apart from that due to its formation. Both
Cambridge (England) University.
are dependent upon the amount of applied voltage, but
The analytical work reported in this paper is by H. H. Skilling. The experi­
mental work was done by P . de K. Dykes, while. a graduate student at the the loss due to migration differs from that due to forma­
Ryan Laboratory, Stanford University. The aid of Doctor J. S. Carroll, tion by being a nearly constant loss that (at low frequency)
director of the Ryan Laboratory, is gratefully acknowledged.
1. For all numbered references, see list at end of paper. is independent of the frequency. At moderately high fre-

850 Skilling, Dykes—Traveling Waves ELECTRICAL ENGINEERING

 Analysis of Oscillograms
Oscillograms of waves attenuated by corona are shown
in figure 2. The figure shows at the top, left, the record
of a 320-kv wave impressed upon a line of number 10 cop­
per wire, and at the top, right, the record of the same wave
as it was received after traveling 836 feet along the line.
Below this is a similar record of a wave of negative po­
larity, with an equal initial voltage but showing rather less
attenuation as received. The time scale of the record is
given by the 5-million-cycle wave at the bottom of the
figure; the initial waves rise from zero to maximum volt­
SYNCHRONOUS COMMUTATOR age in 0.3 microseconds, the wave fronts being practically
straight lines. These are typical of a collection1 of several
Figure 1 . Impulse generator and oscillograph
dozen of such oscillograms.
The network shown in figure 1 was used in obtaining
quencies it becomes a negligible part of the total loss, these records. In producing a wave, the sphere gap set to
while at 60 cycles per second it is about one-third of the flash at 285 kv is first to trip, the polarity of discharge
total. 5 being determined by a synchronous commutator. This
The energy required to produce a protective sheath of immediately actuates the trip circuit of the cathode-ray
space charge about a conductor is dependent upon the oscillograph, and then, with a slight time delay, because
height to which the applied voltage rises. Each increase of inductance, the gap set for 200-kv sparks. One set of
of voltage requires an increase of space charge, with a cor­ condensers is charged by the transformers to 285 kv, and
responding consumption of energy. As the applied volt­ the other set to 570 kv; operation of the first 2 sphere gaps
age is raised from e to e + Δβ the energy which has been connects the condensers in series and so trips the main gap
used to produce space charge increases from E to E + AE. which is set at 850 kv. The condensers now commence to
The ratio of the increments is (for very small increments) discharge into the line, producing the test surge, but when
dE/dey which may be found by differentiation. The loss the voltage across the line has risen to 300 kv the surge is
corresponding to a given increment of voltage is of course "chopped" by a sphere gap, and the surge voltage drops
not constant, but is in general greater when the increase almost immediately to zero.
of voltage takes place at a high voltage than when the By adjusting the values of inductance, waves were pro­
same increase takes place at a low voltage. duced having wave fronts of 0.3 to 1.63 microseconds
In developing the above concept of corona loss the duration. All waves were substantially triangular in
mechanism of corona formation has been greatly idealized, shape. The oscillograms
and many small factors have been entirely neglected. One show a superimposed ripple,
of these is a difference in loss on the positive and negative but this originated in the os­
half-cycles of corona. This particular factor requires cillograph circuit ; it couid be
special mention, for it will arise again in the discussion of removed by introduction of
traveling waves. damping resistance in the
oscillograph circuit, but it
Loss From a Traveling Wave was found best not to do so
because of resulting loss of
A wave of electric energy traveling along a transmission accuracy of the record.
line will necessarily produce ionization about the line as it The generated surge was
passes and, since the only source of energy is the wave sent out as a traveling wave
itself, it will lose energy as it travels. There is ample di­ on the test line. The line
rect evidence of corona produced by traveling waves, for was laid out as a large loop,
it may be both seen and heard. returning to the laboratory,
Moreover, the voltage of a wave is related to its energy, and terminated in a resis­
although the relationship is not at all simple in a wave tor having the surge resis­
which is changing in shape as it travels. On a line with­ tance of the line. Both
out loss, or a distortionless line, energy is proportional to outgoing and returning
the square of the voltage for each element of the wave. waves were recorded on a
This simple relationship is an approximation when applied cathode-ray oscillograph.
to waves that are producing corona, but under any cir­ Examination of the oscillo­
cumstances the voltage of a wave becomes less as its energy grams yields the following
is decreased. And oscillographic measurements and sphere- information. The received
gap measurements both show a rapid reduction of the volt­ wave has a wave front that is Figure 2. Oscillograms of
age of traveling waves as long as their voltage exceeds the less steep than was that of traveling waves, showing dis-
corona-forming voltage of the line. theinitialwave. Thetimere- tortion by corona

JULY 1937 Skilling, Dykes—Traveling Waves 851

quired for the received voltage to rise from zero to maxi­ The rate of energy loss as voltage increases is then found
mum is equal to the time required for the initial voltage to by differentiating equation 2:
rise from zero to maximum and then to descend to a value
equal to the maximum received voltage. In other words, — = 2 — (e - eo)
de n
(3)
if the received wave were superimposed on the initial wave
its crest would fall upon the "tail" of the initial wave, as so that for a differential increase of voltage the loss is
shown in figure 3. At the same time the received wave dE = 2(k/n)(e - e0)de) (4)
has acquired a longer tail than was possessed by the initial
wave. The appearance of the whole is as if the wave When an electric wave is traveling on a line the voltage
front had been cut down to a gentler slope as a result of at a point of the line rises as the front of the wave passes.
having traveled along the line, so that the front and crest During the passage of the wave front, space charge is
of the wave have been reduced, while the tail has some­ placed about the conductor, and energy is lost from the
what gained. wave. Consider a specific point of the line P at a distance
Various attempts to deduce a law of action by a pos­ which may be called 5 from some stationary reference axis.
teriori reasoning from the osciUograms failed, but they (See figure 5.) Consider a short section of the traveling
served at least to indicate the nature of the differential wave at the point P, which section of the wave may be
equation from which the law would have to be derived. specified as being at a distance x from a reference axis
It was seen to be related to the "quadratic law" of corona which is moving along the line at the speed of the traveling
loss as promulgated by Peek, 5 and so the quadratic law was wave. It follows that for a given point on the line s is
adopted as a starting point for the a priori deduction of a constant, while for a given section of the wave x is con­
law of distortion by corona. stant. The 2 systems of co-ordinates are related by s =
x + y. By definition, then, y is the distance between the
reference axes, so y — vt wherein v is the velocity of propa-
Deduction of a Law of Action

Π
Power loss from the experimental transmission line was Figure 4. Corona loss from
determined by measurement, and is shown in figure 4. a number 10 American wire °0
By plotting on a logarithmic scale it was found that the gauge copper conductor, 836 /
curve may well be expressed analytically as feet long, 12.5 feet above I5

Loss = k'(e — eo)2 ) (1) ground. Voltage: 60-cycle /

I0
sine-wave, wire to ground
in which e is voltage, and e0 and k' are constants. The «,
constants as so determined are in good agreement with
0
those computed from the coefficients of Peek's law. 50 I00 I50 200 250
Corona loss is measured as power; that is, as energy per KILOVOLTS (RMS)
second. After that part of the power loss which is not

gation of the wave (practically that of light) and t is the

MICROSECONDS time that has elapsed since y was zero.
0 0.5 I.O
0 The energy loss at the point P from a section of wave of .
length dx while the wave is traveling a short distance dy is
V-
(/) 2θομ I00
b k
o Loss = 2 — dx (e — e0) de (5)
g ΙΟΟΙ—
( n
200

in which de, the change of voltage at the point P, is

0 0.5 I.O
MICROSECONDS
dß — — dx -f- — dy
Figure 3. Initial and received waves of osciUograms, super­ ox oy
imposed. Computed curves are also shown But since at a fixed point s is constant, dx = —dy and

Loss = 2 — dx (e — e0) (— — dy -f- — dy) (6)

proportional to frequency has been subtracted, the re­ n ox ày
mainder may be divided by the number of voltage crests
It is now necessary to express energy of a section of
per second, and by the length of the line. The result, loss
traveling wave in terms of voltage. In the absence of dis­
per unit length per half cycle, is expressed by an equation
tortion the energy of a section of wave of length dx would be
similar to equation 1 but with a different value for the
\
coefficient which (for reason^ that will appear later) may E = cdxe21 (7)
be written as k/n:
c being the capacitance per unit length of line. Half of the
energy is in the electric field, and half in the magnetic. 13
(Loss as voltage rises to e) = — (e eo)2 (2)
n When distortion of the wave is taking place this equation

852 Skillingy Dykes—Traveling Waves ELECTRICAL ENGINEERING

becomes an approximation only. Although of uncertain Use of Derived Equation
accuracy in the presence of corona, it will nevertheless be
used in the following discussion, for it must be of the right This form of the equation is easily applied to any travel­
order of magnitude. ing wave. The only information regarding the wave that
Differentiation of equation 7 gives the rate of change is needed is a curve of voltage and time; an oscillogram,
of energy of a section of wave as its voltage changes, so that is, of the initial voltage (or if this is not available, of
the change of energy corresponding to a differential change the voltage at any point on the line). There is no neces­
of voltage is sity to express f0(t) or F0(e) analytically, for the equation
has been arranged to allow a graphical solution. The
dE = 2 c dx e de ) (8) equation gives the information that at a point at distance
But loss due to corona, as expressed in equation 6, must
be supplied from the energy of the wave, so
k I àe be\
2 c dx e de 2 — dx (e — e0) dy (9) 1.53 MICROSECONDS
n \ bx by)
Since de of equations 8 and 9 is the change of voltage of a
section of the wave it is the partial differential with con-

MOVING Figure 5. Co­

AXIS (C)
ordinates of travel­
,0.80 MICROSECOND
ing wave
/
- - ^ 0 . 3 0 MICROSECOND

PLOT OF
k e-ep,
^ 30.30 MICROSECOND

(9)\
TIME
INITIAL WAVE FRONT
stant x, and is equal to (àe/ày) dy. Introducing this into COMPUTED FINAL WAVE FRONT
equation 9, arid performing obvious simplifications, OBSERVED FINAL WAVE FRONT

be k /be_ _ be\ Figure 6. Comparison of computed and recorded waves

c e = _ (,( (10) after attenuation and distortion by corona. A l l initial waves
by n \bx by)
have crest voltage approximately 310 kv
Rearrangement gives the partial differential equation
k(e — eo) be
(11) 5 there will correspond to each value of voltage e a value
k(e — e0) + ^ce bx by
of time /. This will differ from the value of time that cor­
for which the solution is an arbitrary function responded to the same voltage at the initial point on the
/ _ Ke - eo) + nce \ o line by the last term of equation 14, which is merely the
(12)
time required for the wave to travel the distance s, and
by the next-to-last term which alone describes the dis­
(Elimination of x and y between equations 11 and 12 has tortion of the wave due to corona.
been accomplished by use of the expressions defining 5 and
To determine the shape of the wave at distance s, plot
y.) the initial voltage-time curve. Then, at each value of
The form of function to be used in equation 12 is de­ voltage increase the value of time by an amount Ks(e —
termined by some known relationship between e and / at e0)/e. The result is the wave as distorted by corona. (It
some known value of s; for example, at the point where must be noted, however, that this procedure applies only
the voltage wave is applied to the line. If the initial wave on the front of the wave, while the voltage is rising to a
is the function of time e = f0(t), it follows from equation higher value than it has before attained. In the absence
12 that at a distance 5 the wave will have become of further information it must be assumed that the tail of
k(e — eo) -j- nce a wave is not affected by corona. This will be discussed
/o(* - (13)
in a later paragraph.) Reference to equation 14 gives a
value of k/ncv for K,
The initial voltage as a function of time may also be ex­
pressed t = F0(e), and corresponding to this is the equa­ Symbols have the following meanings:
tion (derived from equation 13) which applies at distance t — time in seconds
e = voltage in volts
eo — corona-starting voltage in volts (crest)
k (e — eo) s c = capacitance of the line in farads per foot
(14)
t = FM + — J + - v — velocity of the wave in feet per second (practically 0.985 X 109)
ncv e v

JULY 1937 Shilling, Dykes—Traveling Waves 853

s = distance of travel in feet to n for positive and negative waves; n+ (for positive
k — the constant of equation 2 which relates crest voltage in volts waves) would on this basis be about 2*/2, while n" (for
to energy loss in joules per foot per half cycle, and which m a y
be found from Peek's quadratic law or otherwise
negative waves) would be about 4.
n = a factor which differs for positive and negative waves, and as
discussed below Comparison With Oscillograms of Waves
In figures 3, 6, 8, and 9 a comparison is shown of dis­
The Factor n tortion as predicted by equation 14 and actual oscillo-
graphic voltage records. The oscillographic curves of
The factor n is a more or less constant factor which is figure 3 have been described above ; superimposed on them
needed to account for 3 differences between the corona of are computed curves based on equation 14. The values
traveling waves and the corona at power frequency. These of n used in making the computations were 21/2 and 4 for
are (1) the effect of mobility of charge, (2) the fact that positive and negative waves, respectively. These are the
when voltage is alternating there is space charge left over predicted values of n, and they give reasonably good
from one half cycle to the next, and (3) the difference be­ agreement.
tween positive and negative corona. Figure 6 shows a comparison with the fronts of the same
The effect of mobility of charge has already been men­ waves and with longer ones also. Each curve shows the
tioned; at power frequency there is considerable loss due average of data from several oscillograms; in most cases
to mobility that is essentially constant regardless of fre­ 5 or 6 photographic records were averaged. The term
quency. Peek 5 shows that this is 25/85 of the 60-cycle from equation 14 that expresses wave distortion due to
loss. Other investigators give a slightly higher value, and corona is plotted in the lower left corner of the figure : in
it will be approximately correct to consider that the loss computing the shape of the distorted wave the abscissas
a t 60 cycles due to cyclic reversal of space charge is 2/3 of of the appropriate curve are added to the values of time of
the total loss. It was partly to take care of the division of the initial wave front, at the corresponding value of
power loss into cyclic and constant parts that the factor voltage.
n was introduced, and this consideration alone gives n a It will be seen that agreement with measured waves is
value of 3/2. (See equation 2.) But this value of n may within the probable experimental error. However, com­
be altered to take care of other considerations also. putation for this figure was done with n+ = 4 and n~ =£#,
Each crest of alternating voltage leaves the conductor which values are about 50 per cent higher that the pre­
surrounded by a sheath of space charge. Some of this re­ dicted ones.
turns to the conductor from which it came, some reaches It was also possible to compare the results of computa­
the other conductor or is lost in space, but most remains tion with equation 14 to data obtained some years ago. 2
until it is cancelled by a discharge of opposite sign as the A series of measurements with sphere gaps gave crest
applied voltage is reversed. During each half cycle of values of very short traveling waves as they were attenu­
voltage, therefore, it is necessary not only to supply space ated by corona. Only maximum voltages were measured,
charge of the same polarity as the voltage, but also to re­ so the comparison must necessarily be with computed
move (or cancel) space charge of the opposite polarity. maximum values. It is known that the waves employed
Consequently the total loss per half cycle is greater than in the experimental work had practically uniform rise of
would be the loss if there had been no previous opposite voltage from zero to maximum, at which point they were
charge by a factor of uncertain magnitude that is some­ "chopped/' so that they were essentially triangular waves
thing between 1 and 2. The use of n may be extended to without tails. The slope of the impressed wave front was
care for this factor, and the value of n therefore becomes of the order of 2,000 kv per microsecond, and the crest
approximately 3. voltage varied from 100 to 220 kv, so the duration of the
Moreover, it is found that the loss of energy from a posi­ wave is computed to be from ΧΙ^ microsecond for the
tive conductor is greater than that from a negative con­ short waves to Vio microsecond for the longer ones. It will
ductor. This is observed even with alternating voltage, be seen that these are of a shorter order of magnitude than
although in the steady state the loss is largely equalized by the waves of figure 6. Comparison of experimental and
the production of a rectified space charge. It is impossible computed results is shown in figure 7; the agreement is
for the charge that escapes from a conductor in corona to surprisingly good. The value of n used for the computed
be continuously unbalanced, for the resulting excess of curves was 5, but the polarity of the experimental waves
charge of one sign will accumulate until the voltage gradi­ is not known.
ent about the conductor is so adjusted that the production The line used for the data of figure 7 was similar in
of charge on alternate half cycles is equalized. But many size and height to that used later, so its loss at 60 cycles
observers of impulse voltages and traveling waves have was taken from figure 4. Its total length was 1,000 feet
found that the loss is greater from a positive conductor. but the first 500 feet were most significant and data for
It is, of course, natural that there should be a difference that part of the line only are shown in figure 7. The
due to polarity because of different details of action of complete experimental curves appear in figure 1 of a
positive and negative ions. The ratio of loss from a posi­ previous paper, 2 and agreement with computed values
tive wave to loss from a negative wave appears to be about is good throughout.
3 to 2. It is helpful, therefore, to assign different values For comparison with completely independent experi-

854 Skilling, Dykes—Traveling Waves ELECTRICAL ENGINEERING

 250i Figure 7. Com­ Behavior of waves on the Bushkill-Roseland line is
parison of com­ of particular interest because it is typical of 220-kv lines
> puted and meas­ in general dimensions, and therefore in corona loss.
,200
c DMPUTE D ured wave crests The conductors are 795,000-circular-mil-steel-reinforced
M EASURE as traveling waves aluminum cables, some 60 feet above ground. Con­
150 are attenuated by ductors are spaced 28 feet, 6 inches, in a horizontal plane.

IN
corona It will be noted that in all cases of comparison the
100 agreement of recorded and computed distortion is good,
Sv but that different values of the factor n have been em­
ployed. The values used for n are shown in table I.
50
** It is not improbable that the apparent variation
of n with duration of wave front is entirely fictitious,
being due to experimental difficulty. On the other hand,
100 200 300 400 500
DISTANCE ALONG LINE IN FEET it may be due to an actual decrease of loss when the dura­
tion of the wave front under consideration is very short.
mental work, computations were made of the change of If the decrease of loss is real it may be due to inability
shape to be expected in waves on one wire of the 5-19 line of ions to move with great enough speed, although this
of the Consumers Power Company (Michigan). Dimen­ is improbable because the motion required for the forma­
sions of the line are given in a paper 8 by McEachron, tion of space charge is largely electronic and electrons
Hemstreet, and Rudge; corona loss on the line was com­ in a critical field travel several meters per microsecond;
puted according to the quadratic law, and the change of it may more probably be attributed to ground resistance
shape as predicted for traveling waves by equation 14
was determined on that basis. The experimentally
lOOOi lOOOi
recorded distortion of waves on the 5-19 line is shown in
a paper7 by Brune and Eaton, and one set of their oscillo- ,- NEGATIVE POSITIVE
I \INITIALWAVE
grams is particularly fitted for comparison with computed
distortion. Experimental and computed waves are
shown in figure 8. The initial wave is shown, and also
the received waves at 3 different distances along the line.
In computing distortion of the negative wave of figure
Sa, a value of 4 was used for n~~, and for the positive wave
of figure Sb a value of 2XA was taken for n+. See table I.
The general form of the computed and observed wave
fronts of figure 8 is quite comparable. It is not at all 5 10 15 5 10 15
surprising that there is difference of detail, for the com­ MICROSECONDS MICROSECONDS
puted curves consider only the effect of an ideal distri­
Figures 8a (left) and 8b (right). Comparison of computed
bution of corona and completely neglect resistance of
waves (solid lines) with oscillograms by Brune and Eaton
line and ground, the presence of other conductors, insula­
(dash lines) as a traveling wave is distorted and attenuated
tors and towers, variation of ground level, and other
by corona
irregularities of the line. It is natural that the shorter
the distance of travel, and the more rapid the attenuation
by corona, the better is the agreement.
Several of the other curves shown by Brune and Eaton and a consequent decrease of gradient about the con­
offer further interesting comparisons, but will not be ductor due to what may be loosely spoken of as a lowered
reproduced here. equivalent ground plane.
A study of the change of shape of traveling waveskpn Imperfect agreement of the values determined for n+,
the Bushkill-Roseland line of the Public Service Electric and for n~, is not at all surprising when it is remembered
and Gas Company (New Jersey) is reported by Cox and that the duration of the waves varies from about 10
Beck9 and by Conwell and Fortescue.10 The latter microseconds for Brune and Eaton's waves to less than
paper gives data that make possible the computation of V20 microsecond for Skilling's, or a factor of over 100 to 1,
corona loss at power frequency (using the quadratic law) and that a half cycle at 60-cycle frequency is nearly 1,000
and both papers show a number of oscillograms of waves times the duration of the waves of Brune and Eaton.
affected by corona. Comparison of a wave from the It appears, on the contrary, that there must be some ele­
latter paper with a curve of wave form computed from ment of truth in the theory proposed since the discrepancy
line dimensions appears in figure 9. The agreement is in n is less than a factor of 2 ov^f a time range of 1 to
gfcen to be satisfactory. The value of n+ used in com­ 100,000, particularly considering that the crest voltage
putation was 21/2. There are several other sets of oscillo­ varied from 100 kv (Skilling) to 900 (Conwell and Fortes-
grams of waves in these papers for comparison with com­ cue), the line from a number 10 wire (Dykes and Skilling)
puted distortion, but unfortunately there are no records to a one-inch aluminum cable (Conwell and Fortescue),
of waves of negative polarity. and the distance of travel from 50 feet (Skilling) to 2.8

JULY 1937 Skilling, Dykes—Traveling Waves 855

miles (the greatest distance for really significant com­ TabU
parison with Brune and Eaton).
Data From

The "Tau" of the Wave Predicted from 60-cycle corona 2.4 . . . . 3 * . .3.6
Brune and Eaton (figure 8) 2 1 /* .4
While the front of a wave is passing a short section of Conwell and Fortescue (figure 9) 21/z —
Dykes (figure 3) 2Vs 4
line it is surrounding that section with a sheath of space Dykes (figure 6) 4 6
Skilling (figure 7) 5|
charge. This reduces the potential gradient in the im­
* Average of polarities.
mediate vicinity of the conductor. But after the crest of f Polarity unknown.
the traveling wave has passed, and the line voltage has
dropped to well below its instantaneous maximum, there
will be some return of space charge to the conductor. Other Theories
Because of the presence of space charge, the gradient about
the conductor will be reversed in polarity even before the So far as is known to the author, there has been no
wave has completely passed, and it is this that will previous attempt to account in detail for the change in
return some of the inner part of the charge to the con­ shape of traveling waves due to corona loss. Boehne11
ductor. As the charge returns, the tail of the passing in 1931 suggested that the change in shape is due to a
wave is built up and lengthened. This effect is visible change of the capacitance of a line during corona (in this
in all oscillograms. (See figures 3, 8, and 9.) Whether connection, see also a paper by Gardner 12 ). Bewley6 in
an important part of the energy of the wave is thus re­ 1933 discussed the change of shape of waves due to
gained depends on the shape of the wave and the relation ground resistance, and offered a theory of traveling waves
of its maximum voltage to corona-forming voltage of the which involved 2 or more velocities of wave components.
line. In general the tail of the wave is not greatly altered He proposed the extension of this discussion, which ap­
from point to point of the line. It will be seen from any pears to give beautiful results at voltages below the corona-
of the oscillograms that voltage at a given point on the forming voltage, to explain distortion due to corona on
tail is continuously but slowly increasing, but that the the basis of a change of capacitance of that part of the
total increase is small compared to the crest voltage of line on which there is corona. There would be no sharp
the initial wave. It seems entirely practicable, therefore, distinction between such a theory and the one that is
to neglect any change of the tail in computing the change used in this paper if it were possible to express the postu­
of shape of a traveling wave. lated change in capacitance as a function of the voltage
and to include the loss of energy involved. But when loss
This is done in figures 3, 6, 8, and 9. It will be seen that
of energy is omitted, and only distortionless wave com­
the wave form at any point of the line is determined by
ponents of different velocities are considered, the extremely
large attenuation which is actually produced by corona
Figure 9. Comparison of does not appear.
computed wave (solid line) The following comment should be made in this connec­
POSITIVE
\INITIAL WAVE with oscillogram by Conwell tion. Distortion due to corona becomes less as the voltage
and Fortescue (dash line) as a of a surge is diminished, and the wave tends to flatten out,
traveling wave is distorted with the voltage of its flattish top equal, generally speaking
and attenuated by corona to the corona-forming voltage. This is illustrated by
RECEIVED WAVE
s AT 2.65
both waves of figure 8, at the 14-mile position. But
MILES many factors which are of negligible importance when
the surge voltage is high make themselves felt when the
surge voltage is low. Energy lost due to line resistance
2 4 6 8 10 and ground resistance, and energy transferred inductively
MICROSECONDS
to nearby conductors, are overshadowed at high voltage by
the great loss of energy due to corona. At lower voltage
extending the computed curve (based on equation 14) they have much to do with determining the shape of
until it intersects the initial wave* and thereafter follow­ the wave. It can be seen, for instance, by reference to
ing the tail of the initial wave. This does not give an the original oscillograms of the paper by Brune and
accurate representation of the tail of the distorted wave, Eaton,? that the waves of figure 8 at the 14-mile position
but it does give a satisfactory curve for the front of the do not have a simple flat top, but appear, in detail, to
wave and a good approximation of the wave's maximum have 2 rather distinct humps. It is interesting to consider
voltage. It is this last value that is usually of greatest this shape in the light of the theory of multivelocity
practical importance. components. But it will also be seen from Brune and
The shape of the computed wave will usually more Eaton's curves that this is a minor effect which does not
nearly approach the shape of the actual wave if its sharp appear while rapid attenuation by corona is taking place.
corners are slightly rounded. Physically this is due to In other words, the reduction of lightning voltage from
the resistance of the line and ground, particularly as in­ a dangerous value to a safe value is the result of energy
fluenced by skin effect,7 loss in corona.

.856 Skilling, Dykes—Traveling Waves ELECTRICAL ENGINEERING

 Any of the various empirical formulas which have been
Relaxation of Copper
proposed can only hope to be true for an imposed wave
of some standard form. This is clearly true when one With the present tendency to employ higher tempera­
considers that the corona loss at any particular voltage tures in various industries, more extensive investigation
depends not only on the value of that voltage but also of the behavior of metals at elevated temperatures has
on the rate of change of voltage at that point; and con­ become important. Under a grant received from the
sequently the rate of decrease of the crest voltage of a Engineering Foundation during the years 1935-36, work
traveling wave depends on the shape of the entire wave. on this subject was done by John Boyd with the facilities
It is perfectly certain that any successful formula for the offered by the research laboratories of the Westinghouse
distortion and attenuation of waves that are affected by Electric & Manufacturing Company in co-operation
corona must take into account the whole wave, and not with the University of Pittsburgh, and reported in the
just the maximum voltage a t the wave's crest. paper "Relaxation of Copper at Normal and at Elevated
Temperatures/' which was presented at the annual meet­
Conclusions ing of the American Society for Testing Materials at New
York, N. Y., June 28-July 2, 1937. The problem was
This paper is a derivation and discussion of a formula suggested by Doctor A. Nadai of the Westinghouse labora­
for the change of shape of a traveling wave that is dis­ tories.
torted and attenuated by corona. The formula, equa­ The phenomenon of the decrease of stress at constant
tion 14 of the paper, is length, such as occurs in bolts, has been called relaxation.
TW N , k (e — e0) s Some investigation of this phenomenon has been made,
i = FM + — ^+ -
ncv e v but few of the methods used have attempted to produce
The first term of the right-hand member is the initial pure relaxation conditions. The tests described in this
shape of the wave, and the second term is the change of paper were made with an apparatus designed to eliminate
shape due to corona. (The third term merely accounts the principal difficulties of former methods. Special at­
for the lapse of time while the wave travels.) tention was directed toward fulfilling the conditions of
All quantities in this formula, except nf can easily be pure relaxation, that is, measuring the decrease in stress
predetermined. The factor k is found from the measured while the length of the test piece is carefully kept constant.
or computed corona loss of the line at normal frequency. Copper was chosen as a test material because of its impor­
Approximate values of n are 2.5 for positive waves and 4 tance in electrical machinery as well as its relative sim­
for negative waves, but more data are needed to sub­ plicity, and was used in the form of wire 0.204 inch in
stantiate these values and to determine their range of diameter.
application. A complete correlation has not as yet been made be­
The derivation of the proposed law involves several tween data for relaxation and creep, which is the increase
approximations, but its agreement with experimental in length with time for material at constant stress. All
results over a wide variety of conditions is reassuring. experimental evidence seems to show that the creep rates
found in relaxation tests are greater than those obtained
References from creep tests under what were assumed to be similar
conditions. As a result, the magnitude of the decrease
1. CORONA ATTENUATION W I T H IMPULSES OF SHORT DURATION, P. De K. in stress in relaxation is larger than at first would be ex­
Dykes (a dissertation), Stanford University, 1934.
2. CORONA AND L I N E SURGES, H. H . Skilling. ELECTRICAL ENGINEERING,
pected from considerations based upon creep rates ob­
volume 50, October 1931, pages 798-801. served in tests made with constant stress. A brief sum­
3. TRAVELING W A V E S ON TRANSMISSION SYSTEMS, L. V. Bewley. John Wiley mary of the results is as follows :
& Sons, Inc., New York, N. Y., 1933.
4. T H E HYSTERESIS CHARACTER OF CORONA FORMATION, H. J. Ryan and
H. H. Henline. A I E E TRANSACTIONS, volume 43, 1924, pages 1118-24. 1. For practical purposes, t h e relaxation of copper a t room tem­
perature and probably u p t o 200 degrees centigrade may be repre­
5. DIELECTRIC PHENOMENA IN H I G H - V O L T AGE ENGINEERING, F . W. Peek, Jr.,
McGraw-Hill Book Company, New York, N. Y., 1929 (third edition). sented b y an expression of the form :
6. ATTENUATION AND DISTORTION OF WAVES, L. V. Bewley. ELECTRICAL σ = σ, [1 - A log (1 + Bt)]
ENGINEERING, volume 52, 1933, pages 876-84. (See also discussion by
J. Slepian, volume 53, pages 472-73; by C. L. Fortescue, pages 596-97; and where σ is the stress a t any time t, <r* is the initial stress, and A and B
closure by author, pages 597-98.)
are constants.
7. EXPERIMENTAL STUDIES I N THE PROPAGATION OF LIGHTNING SURGES ON
TRANSMISSION LINES, O. Brune and J. R. Eaton. AIEE TRANSACTIONS, 2. A marked increase in relaxation takes place when the tempera­
volume 50, September 1931, pages 1132-8.
ture is raised above about 80 degrees centigrade. A t 200 degrees
8. TRAVELING W A V E S ON TRANSMISSION L I N E S W I T H ARTIFICIAL LIGHTNING centigrade the stress is estimated to fall to about 20-25 per cent of its
SURGES, McEachron, Hemstreet, and Rudge. AIEE TRANSACTIONS, volume 49,
July 1930, pages 885-94. initial value in a service time of 20 years.
9. LIGHTNING ON TRANSMISSION L I N E S , J. H. Cox and E. Beck. AIEE 3. Recovery effects tend to decrease t h e creep rate found in t h e
TRANSACTIONS, volume 49, July 1930, pages 857-65.
early part of a relaxation test.
10. LIGHTNING LABORATORY AT STILLWATER, N E W J E R S E Y , R. N . Conwell
and C. L. Fortescue. AIEE TRANSACTIONS, volume 49, July 1930, pages 872-76. 4. The dissimilarity between the speed laws for decreasing and for
11. Discussion, E. W. Boehne. A I E E TRANSACTIONS, volume 50, June 1931, increasing stress gives rise t o rather large amounts of relaxation.
pages 558-59.
12. CORONA INVESTIGATION ON AN ARTIFICIAL L I N E , M. F . Gardner. AIEE Further creep and relaxation tests are a t present under
TRANSACTIONS, volume 44, 1925, pages 897-904.
way and may be expected to add to knowledge of this
13. TRANSIENT ELECTRIC CURRENTS, H. H. Skilling. McGraw-Hill Book
Company, New York, N.Y., 1937. subject.

JULY 1937 Skilling, Dykes—Traveling Waves 857