IEEE Transactions on Power Delivery, Vol. 9, No.3.

July 1994 1669

TEE SOIL IONIZATION GRADIENT ASSOCIATED WITH
DISCHARGE OF HIGH CURRENTS INTO CONCENTRATED ELECTRODES
Abdul M. Mousa, Senior Member, IEEE
B r i t i s h Columbia H y d r o
B u r n a b y , B . C . , Canada V 3 N 4x8

-
Abstract It has been known for many yeam that the grounding electrode, the ionized zonewould be a hemisphere as shown in Fig.
resistance of a concentrated electrode d r o p when it is subjected to 1A. In the case of a ground rod, Liew ct al. [18] took the ionized
a high c u m dircharge. This helps reduce the ground potential rise. zone to be of the form shown in Fig. lB, whilc Bellaschi ct al. [3]
The degree of the resistance reduction depends on the magnitude of adopted the alternative form shown in Pi.1C. Liew’s version has
the ionivtion gradient of the soil pb. Based on both a theoretical been adopted in thir paper becauseit identifits the ionized zone by
analysis and a critical review of the large number of available the single parametet a, and because no definite rules exist for
mcaruranaks, thir paperrtcommendrthatE.be- equal to 300 dckminhg the length I+ asrociatcdwith Bellaschi’s method.
kV/m for typical soils. This is signi6cantly )CISthan the loo0 kV/m In practice, the soil is not a homogenous medium mainly because
value used by lome authors. Graph arc also given describing the of the variation of water content and also because of the v a m n in
behaviour of the rod electrodes which arc used in many k l d grain size and the existence of organic and man-made debris. Such
inotallationr. inhomogeneity a h d t s , though to a much smaller degree, under
laboratory conditionswhere sample? are made of sifted material and
-
J(evwords GROUNDING, ELECTRIC DISCHARGES: In Soil, the water in the sample k reasonably well-mixed. As a result, the
LIGHTNING, ELECTRIC LINES: Lightning Performance. current bea”concentrateddong scvcrd discrete channels and the
assumed uniform shape of the ionized zone d o a not materialize.
1. INTRODUCTION This war observed in the laboratory by Petropoulos [27l and he
visualized the shape of the ionized zone to be of the form shown in
whea a current is dwharged into a ground electrode, voltage Fig. 2A. Yet researchers, includmg Petropoulos, accept doing
g d i a l r a will begencraud within the roil in accordancewith the calculations based on the assumption that the ionized zone will take
equation: a &form shape. It h~ not been racognizcd before, however, that
E = p , J. .. (1) the v a l k of Emto be used in the calculations might have to be
where p , = roil resistivity significantly different from the value obtained from puncture tats.
J = the currcnt density at the point under consideration. Erkr et al. [9] recently c o n b e d the effccts of inhomogeneity of thc
Wherever E exceeds the ionization gradient E,, breakdown will roil by tutr on a roil sample in the form of a cylindrical sector with
occur. Thin basically converts the affected poition of the soil from the center electrode sectionalizedas shown in Fig. 2C. They then
an insulator to a conductor. Breakdown would start at the surface of found the currents flowing into individual segments to vary in the
the electrode, where the currcnt density is highat, and it continues range 3: 1. Oettle [U] similarly found wide variations in the cumnts
ouhvuda up to the points at which the currcnt density d r o p below flowing into the different portions of a segmented horizontal ring
the critical value given by: eltctrodc. In nature, the shape of fulgurk which form when
J, = E, / p,... (2) lightning striku rand ia further manifatation of the effects of
The breakdown of the roil described above is basically equivaht to inhomogeneity of the roil, plepre (cc Fig. 2B [33].
an hcmh the dima18ionsof the elecbode. Hencethe grounding Ionization of the roil should be tpkur into consideration when
rerkturoeR en~~untcrcd during the dwhargcwill be less than the dckminhg the resistance of a concentrated ground electrode for the
“meggercd” value q. Thh f.ct has been known for many d d e n following t‘tuons:
and the first major tutr for dckminhg the ratio (R,tRJ were done 1. The portion of the roil in the immediate viciniiy of the
by Towne in 1929 [32]. S ince then several rweprchefl conducted electrodeg- a lnrge p o h n of the total resistance becamthe
similar tab and rome of thean used their results to estimate the crou d o n of flow of cumnt in that zone is small. This is shown
related roil ionization gradient. With E, kpown, it then become? in Tablea 1 and 2 for the canen of a hemispherical elccrrode and a
pouible to ertimpte the resistance& for other CISCS without need for ground rod, terpectively. It follow that shorting out a part of the
fiuth.r tertr. Lim et al [lS] advanced the theory further by roil in that zone (by ionization) would significantly reduce the
accounting for the d u a l nristivity of the ionized zone;unlike the groundingresieturcc.
othen who urumedtheionized zoncto bea perfect conductor. “lis 2. The cumntr nccdcd to initiatt ionization arc quite small.
.Ilowadthantodetarmne ’ the vuiation of resistancewith time for These arc given in Table IU for the following casea:
a gival impulse Waveform. a) The case of a single rod (16 mm diameter, 2.5 m long)
CalculatingR, i s usually bucd on the urumption that the roil is a repreocntingthe grounding of dirtribdon linen at a transformer
homoparour medium urd hcncc the ionized =ne would take simple 1oc;rtion or at the junction between underground cable and M
forma which arc e u y to analyze. In the case of 8 hemispherical overhead feedcr.
b) ThC cplc of two of the above rodn in parallel representing the
94 WM 078-6 PWRD A paper recommended and approved grounding of a typical H-frame wood pole transmission tower,
by the IEEE Transmission ,and Distribution Committee Fig. 3A.
of the IEEE Power Engineering Society for presentat- c) The case of a steel tower in easy terrain using 4 grillage footings;
ion at the IEEE/PES 1994 Winter Meeting, New York, design as shown in Fig. 3B & C.
New York, January 30 - February 3, 1994. Manuscript d) The case of a steel tower in mountain terrain using 4 rock
submitted September 1, 1992; made available for footings; design as shown in Fig. 3D.
printing December 7, 1993.

0885-8977/94/$04.00 Q 1994 IEEE

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.
 1670
Table I. Contribution of the Soil in the ImmOdrrte
' Viinity of a H e m i r p h d Electrode to ita Total Rerutancc

Table II. Contributionof the Soil in the Immedrrrte

' Vichitv of a Ground Rod to its Total Rerirt.noe

A ) HEMISPHERICAL ELECTRODE B) GROUND ROD: LIEW'S METHOD C) GROUND ROD: BELLASCHI'S METHOD

Pi. 1. Shape of the ionized zone rssumcd in calculation,

A) H-FRAME WOODPOLE 0 ) ELLVATION O f C R I U G C

PLEXIGLASS SIDES

SEGMENTED
1

Fig. 3. Typicaltower punding rystcrm.

Table Ill indicates tltat the punding rystanr of distributionh e r ,

wood pole t " hea and a t c d transmission towers in rocky
terrain will experience ionization under all practical current
IGH VOLTAGE discbargeoonditions. Considcringthatthemcdiilightningourrent
is about 30 U and the currents expected to oausc backflash have
even larger amplitudes, rteel transmission towers with griUage
c) -8190- footings arc .Is0 cxpectad to cxpcrhce ri,gnificurt mil ionization.
Of course,the results of Table III would be d i e d ifcountapoirC
Fii. 2. Illusmating i n h o m o g h y of the mil. is conncctad to the footings of the tower.

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.
 1671
It is worth noting here that the current IEEE method for calculating dielectric constant of the soil, the average field across the whole soil
the lightning tripout rates of power lines [14] does not take soil gap at the time of the breakdown inside the voids can be much
ionization into consideration. The tripout rate is sensitive to the smaller than the breakdown field for an quivalent air gap. Another
magnitude of the footing resistance and the IEEE method has been proposed explanation is that the initiation mechaniim is primarily
found to overstate the tripout rate for lines with high footing thermal (Snowden et al. [29]) . According to this model, when the
resistances[13]. Taking soil ionization into considerationshould help voltage pulse is first applied, current starts to flow through the soil
resolve thii discrepancy. The recent IEEE Working Group Paper and this current is conducted mainly by the water which coats the soil
[13] does not recommend a value for E. and only quotes the 1000 particles. Aa the temperature of this water starts to increase due to
kV/m value suggested by O d e [2q. The CIGRE Lightning I% heating, the resistivity of the water decrcaaa slightly. Then, due
Performance document [5] uses E, = 400 kV/m in some of its to local non-unifonnitier in the heating rates, a thermal instability
calculations, but no supportingevidence is given regarding selection occurs as current is funnelled toward the higher-temperature(lower-
of the above value. In thispaper, an attempt is made to estimate the resistivity) regions. Eventually all of the current ends up flowing in
soil ionization gradient E, based on the following: a few narrow channels, thus W i n g to vaporization of the water and
1. A critical review of the mechanism of breakdown of the soil ita subsquent breakdown at locations of highest electric field.
and the governing factors. The evidence supporting the theory of breakdown by ionization
2. Examining the results of earlier impulse resistance of the air in the voids of the roil is quite convincing and is given
measurements and systematicallycalculating the assochted values of below. On the other hand, the 1983 work by Snowden et al. in
E, a b r excluding unreliable data. support of the water vaporization theory involves many simplifying
3. A critical review of the possible sources of error in the assumptionswhich make its validiiy rather questionable. Also, their
method of calculating E, from impulse resistance measurements. resulta do not really prove their theory since the calculated values
4. A review of other relevant test data and information. were 1.74.0 times higher than the measured ones.
Graphs are also given describing the behaviour of the rod electrodes
which are used in many field installations. 2.2 EVIDENCE SUPPORTING BREAKDOWN BY AIR
IONIZATION
2. BREAKDOWN MECHANISM OF THE SOIL Flanagan et al. [ll] listed several findings which support the air
ionization version of soil breakdown. The review by the writet
2.1 POSSIBLE MECHANISMS OF SOIL BREAKDOWN fu&er revealed that Flanagan’s data i&so in agreement with the
At the micro-structural level, most soils consist of basically non- findings of Oetrle [24], Petropoulos [27‘Jand Liew et al. [18]. More
conducting particles coated with water in which some salt.& details on this point arc given in [21]. Tho most ir;lportant proof was
dissolved, with air filling the voids between the soil particles. The that Leadon et al. [17] did testa in which the air -in .the voids wps
water coating provides i n t e ~ ~ e ~water
t e d paths which provide the replaced by sp6; a gas which has a higher breakdoh pdient. Thb
low field conductivity of the soil. Such conductivity (and hence also resulted in an increase in the breakdown gradient of d e soil, thus
the resistivity) will depend on both the amount of water and the proving that breakdown is initiated by ionization of the gas in the
amount of salt present in the roil. The size of the particles of a soil voids.
sample usually varies withii a wide range. The average size of the It should be noted that Snowden et al. PO] subsequentlydid other
air voids within the soil will depend on the frequency distribution of testa (using different equipment) which suggest that use of SF6instead
the size of the particles. For example, a soil consistingof very fine of air does not change the breakdown gradiint except in soil samples
dust-& particles will have smaller size voids, while a sandy soil which are completely dry. A critical review by the writer rev&
with coarse particles (or gravel) will have larger size voids. The the existence of several factors which might have led to Snowden’s
shape of the voids is usually highly irregular especially if the failurt to observe a diffemce when air was replaced with SF,.
surrounding pmticles have sharp edges. This makes the maximum Those factors an as follows:
electric field within the voids of a soil gap significantly higher than a) The procodureured by Snowden to replace air by SF6was not
the maximum electric field within an air gap having the same rigorous. Hence the exchange of air by SF6 may not have been
configuration and dimensions. complete. It is well known that contaminating SF6 with air lowers
Two different mechanisms have been suggestad for the its ditlechic strength [ l q .
breakdown of the soil when it is subjected to a high voltage. One b) Unlikc air, the dielectric strength of SF6can drop drastically
proposed explanation is that the initiah? process is primarily when moisture is present [U]. The sF6 gas used by Snowden
electrical and the initiation begins when the electric field in the voids appears to have contained a relatively high percentage of moisture.
between the soil grains becomes large enough to ionize the air in the This is because Snowden introduced the gas into the soil by bubbling
voids (Leadon et al. [lq). Because of field enhancement due to the it through water.
irregular shape of the voids and the effect of the relatively large

Table III. Ionization I n i o n Current I.

~ ~~ ~~~ ~~ ~~

Po JO
’ Io Amperes
Om amp/m*
lRod 2Rods 4 Rock Footings Four Grillages
100 3000 377 754 -* 27,000
500 600 75.4 151 -* 5,400
1000 300 37.4 75.4 755 2,700
5000 60 15.1 15.1 151 540

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.
 1672
c) Tht elactric field in the Snowden apppnrtur waa highly non- 1. The 1929 tats done by H.M.Towne p2].
d o r m s h one of the two elcctrodea waa a needle. It is well 2. The 1941 teats done by Behachi PI.
Lnorm that the percentagedrop in the dkk&k Of SFS is 3. The 1942 tab done by Bellanchi et al. PI.
higher than that of air when the degree of non-unifonnity of the 4. The 1948 tub and analysis done by Petr0poulOucl[271.
dcchic w is incrusadp2]. 5. The 1953 tub done by Annrtfong [I].
6. The 1974 teats done by Lim ct al. [18].
2 3 IMPLICATIONS REGARD1NGCHARAC"ERISTICS OFE, 7. The 1978 teats done by Dick et al. [a.
The soil breakdown mechanism dercribed above i m p k that Other tcdt~~ were done by Kawai [ l q , Fagan [lo], and Rogers PSI.
quantity E.Should d i t the following ch0r;rCtcnr 'ticr: However. tho= tc& wcn done using &her complex elcctrodc
1. E, for any soilwill dwayr be leu than that of the air. confiourationr or rods din concrctc. Hence they can not be
2. E,, should drop with increase in water content because of the uscdto ertimatethegrpdientE,.
resulting increase in the dielectric constant of the soil. In the The t a t s in it an^ 1-7 above have betn carefully analyzed to exclude
application under considerationin this paper. the selected value of E, unrcliablcdata and the nmaining onea wcrc uscd to cstimatCE0. A
should be that corresponding to a reasonable level of moisture summary of the findings is given in Table IV. Note the following:
content rather than for dry soil. Gc4bchnic91 mu~~urun& 1. Tome's measurements wcrc made on a loose gravely soil
available to the writer give water contents in the mnge 3-996 (by partly made of artificinl fill.
weight) for soda having medium to ooaflc grains, and up to a 2. Wota content of the soil waa generally not known and it
whopping 60% for ash (has very h e grains). probably varied with depth along the length of the ground rods.
3. For the same watcr content, s o h having very line (durt-likc) 3. Pctropoulor'r mcsrurunenta were excluded because he was
grains should have somewhat higher valua of 4 comparcd to sods not able to trace the full oscillograms and hence significant errors
having medium or coarse grains. Considering the irrcgularitics and may have occurred. However. we have included his analysis of the
inhomogeneity which might exist under field condiinr, lab tats done by Bcrgcr and by Norinder.
measurementson roil samples consisting of fine grains should not be 4. Berger'r measurements were done on a single sample; the
used when selecting a nprcsentative value for E, water content of which is believed to have been low.
4. Some p p l e hold the notion that a &g correlation exists The procedure used herein to detcrrnine the values of Eo is as
between resistivity of the soil po and its breakdown gradient: if p , is follows:
high then E,, would aLS0 be high! The breakdown mechanism 1. The grounding resistance of the electrode under low current
indicattr that such a notion is not valid. Thin is because Eo is conditions R,, wan available and this waa used to the
governed by ionization of the air in the voids while po is govemed by resistivity of the soil p,. For the CPSCI of rods and hemispheresthe
flow of cumnt in the water which coats the soil partick. It is true cquationr take the following forms, respectively:
that increasing the watcr content would decrease both p , and E,. = (Po/ 2 * L) F (4L /rJ - 1.01 (3.1) ...
However the relation is not a direct one because E, would be R,, = po / 2- ...(3 .2)
constant for a given soil sample if the water content waa k e d , while 2. In the subject tats, their authors obtained oscillogramsof the
po can still be varied by varying the amount of salt dissolved in the cumnt and voltage of the electrode during the impulse, used these
water. S ince the salt contcnt of natural s o h varier, the above to plot the variation of resistance with the, and reported the
conclusion is applicable to p d c a l casu. This is proven by the minimum obsmed $. We used these to determine the radius of the
large scatter in the relation behueen E, and po measured by oectlc ionized zone. Referring to Fig. 1, the equations for the cases of rods
1241. and hemispheres arc, respectively,
$ = (Po/ 22L) In [&+ai) / aJ ... (4.1)
3. REVIEW OF IMPULSE RESISTANCE MEASUREMENTS & = po / 2- ...(4.2)
3. The current density Je at the surface of the ionized zone is
A large number of rtsistance measurements under high cumnt thendetennined:
discharge condition8 have been done in the past, usually using J. = I / (2*a,L + 2 2 azJ ... (5.1)
impulse currents. These include: J, = I /2ra2i ...(5.2)
4. Finally, the ionization gradient is calculated:
E, = po Jo (6) ...
Table IV. Soil Ionization Gradients Obtained h m Im~ulseRcaistanceMeasurements

Rekrence No. of ElectrodeQp soil Qp Gradient ;Vim

casu
Range Average

5 Rod8 Gravel -
130 686 29 - 104 75
28 RodI shale, clay 93-100 160 519 - 333
27 Rod8 Gravel, Sand 77-90 50 - 350 160
4 Rods and hemiaphcrw Clay, Sand, Gravel 50 - 310 110 - 300 230
2 Concentric cylinders - - 330 360 - 343
6 Hemispherm - lo00 490 540 - 513
6 Rod8 Sand, Clay 52 - 495 330 - 480 432
19 Rods - 12 - 25 13 - 221 71
I I I

*Berger's and Norinder'a data were taken from the analysis in Pctropoulor'r paper [27J

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.
 1673

4. ERRORS IN ESTIMATING THE BREAKDOWN

GRADIENT FROM IMPULSE RESISTANCE I ~ M o t a l l i cRadius a.
MEASUREMENTS
4.1 ERROR DUE To INHOMOGENEITY OF THE SOIL
The ckdrodc behnviour shown in Pi.2A occurs because of the
possible v r r i . t i 0 ~in both Eo and po at thet’fdi points of the soil
in the Vicinity of the electrode. To analyze the impact of such
variations on the perceived value of the ionization g r a d i i Q, we
considtrcdthe simplified case shown in Pi.4 in which only two Ktr Fig. 4. Configuration used in analyzing *

of conditions exist cpch prevailing uniformly in the vicinity of one effect of inhomogeneity of the soil.
half of a hunisphcrical electrodt. E, and E,and p 1 and p z denote the
ionization gradients and the resistivities for the two segments. The As a practical case,le€ us assume the variations in po and E. to be
equivalent resistivity p., U determinbd from low current due to variation in water content. Data on such variations are
mcasuremmts, i.e. in the absenee of ionization. By expressing the available from the paper by mPnagan et al. [12]for two soils: a soil
total resirtpnce in terms of those of the two segments we get, designated SVBO consistingof very fine grains, and a medium grain
P., = P A 1w.5 ( P I + p-31 *.* (7) soil d e s i DWP25. p , and E, were taken an the values for the
Let V denote the voltage of the electrode under high current soil when wptcr content wan low, i.e. p , and El are the higher
discharge conditions, and J denote thc cumnt density at the edge of values. Several rets of pz and were then taken, each
the ionized zone. Then, corresponding to a specific water content. The valuer of the
El / p1 = JI = 1, / nzI (8) ... ionization gradients were not explicitly available but quantities
ByexpressingI,intennsofVandR, ,andRl intermsofp,anda, proportional to them were available from measurements of the
we e+% minimum breakdown voltage of a fixed .mil gap. If the
a, = VIE, ... (9) proportionalityfactor applied to E,and E,is K,then (16)would give
RI = plEJ*V ...
(10) KE,. Results of the calculationsam given in Table V. The amazing
I, = r v / p , E , ...
(11) finding is that the equivalent or perceived g r a d i t closely tracks the
Similar equations apply to the quantitiuirq, R, and &. It follows that lower of the two ionization gradients existing on the two sides of the
the total resistance i given by: hemisphere; the error in doing so being le-ss than 2% for both types
s = PIE1 W[* V(P1EI + p113111 *..(12) of soil. In the general case of inhomogeneity depicted in Fig. 2A,
But: it would hence be expected that impulse resistance measurements
R, = p&r c ...(13) would yield an ionization gradii which is close to the lowert value
It follows that: occurring anywhere along the electrode and in the vicinity of it.
%q = V(PlE, + P”IE, (PI + d l (14) Sice the degree of inhomogeneity in the field far exceeds that in the
From (ll), the total current i given by: prepared samples used in laboratorytests,the value of Eo to be used
*
r, = v2(p,E, + Pze3/(P,El AE3 (1s) e.. in calculating the impulse resistances of electrodes need to be lower
From I, and c, J., can be deterdned. The product Jq pq then gives than the values obtained from the puncture tests done on short
E4 uniform gap, e.g. Oettle’s test results PI. The above plso explains
Q = EiE2 (PI + pz) 1(PIEI + (16) e-. why Tome’s data in Table N are exceptionally low: his soil was
If = pz ,(16)reduces to: loose, gravely piutly made of paificial fill. Hence air pock& and
+
E,, = El&/ [0.5 (E, E-31 ... (17) organic and man-made debris appear to have been present thus
This i similar in form to (7). In prooticot cu#I, howeve€, soil creating spots of low strength.
inhomogcnCity involves variation in both pmand Eo and hence (16)
should be used.

Table V. E f f d of Soid Inhomogeneity on the Perceived Value of the Ionization Gradient

soil Type Wotcr P KE, KE, &/PI &/El Q/&

wt% om kV kV
SVBO 3.8 1250* 188* - - - -
5.0 357 170 173.7 0.286 0.905 1.021
6.0 118 146.5 149.3 0.094 0.782 1.019
7.2 33 114 115.2 0.026 0.607 1.010
0.9 4ooo* 192* - - - -
1.5 370 152 154.7 0.092 0.79 1.018
2.6 100 136 137 0.025 0.71 1.007
4.5 33 123 123.4 0.008 0.64 1.003

* These arc tplrtn PI the basevaluea (PI and E,) which arc maintained on one side of the hemisphere, while the other side
is assigned one of the Kt of values in subsequenth e r .

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.
 1674
4.2 ERROR DUE TO THE NON-ZERO RESISTIVITY OF instant no (n +
l), i.e. just after the instant of maximum current.
THE IONIZED ZONE Also, the resistance versus time curve waa flat in that zone. "his is
in agreunent with the oscillogramsproduced earlier by Lim ct al.
The impulse resistance calculation method presumes the ionized [18] and by Bellaachi d al. PI.
zone to be a perfect conductor (a simple increase in the metallic 2. Referring to Pi.lA, let R,be the mistance (at low current)
dimensions of the electrode). The actual residual resistivity of the of the portion of the roil within the ionization radius a,, and let & be
ionized zone causes the calculated radius of the ionized zone to be the reoistance of the portion of the roil ojtsidc 4. The true value of
less than the actual value. Hence the calculated current density at the E, correspondsto & whilc the perceived value componda to & +
edge of the ionized zone and also E.would be somewhat larger than BR,) where ia a fraction reprcsmting the incompleteness of the
the actual values. The resulting error was calculated by the writer transformation into a conductor. When 4 is increased, the ratio &
for both the cases of rods and hemispherical electrodes. Regarding +BR,)/R, would increase, thus giving 1 larger error. Radius
the waveform of the input current, both the double exponential and increases when the resistivity is higher or the c u m t is higher and
linear rise and decay cases were considered. The difference between hmcc the error ia larger under these conditions. Thir expected
the two cases was found to be negligible. Hencc the w e of a 2 x theoretical aspect has been confinned by the calculationa.
50 ps impulse having linear rise and decay as shown in Fig. 5 will 3. For a single ground rod (16 m m dip., 2.5 m long) subjected
be used in the illustration. Waves having larger durations (6x50ps to a 5 kA impulrc, the crmr waa only about 1% for reristiw up to
and 6x100~s)were also tried and were found to give smaller erron. 5000 Qm. Thir shows that thc subject crmr is insignificant even for
Hence the analysis was based on a 2x5044s wave. The calculation the ampllert practkd electrode. This single rod is equivalent to a
method used is as follows: hemisphere having a diameter equal to 0.815 m. For such
1. For the given po and based on the surface area of the given hemisphere, the maximum error was also found to be only about 196.
electrode, the current i.needed to initiate ionization waa calculated. Both the above values were obtained using a number of soil segments
From the wave paramttera, the ionization initiation time b (scc Fig. equal to 70.
5 ) was then determined. 4.Referring to Fig. lA, if the c u m is constantthen 4 and Rz
2.The radius of the ionized zone r, correspondingto the peak of would be constant. Decreasing the radius of the hemispherewould
the impulse I- was then calculated. increase RI. From the discussion in item 2 above, it follows that
3. The resistance (at low current) of the portion of the soil located decreasing the electrode size would increase the error. If the
between ro (radius of conductor) and r, was then calculated. This diametet of the hemisphere was gradually decreased from the above
was then divided into n equal segments and the corresponding radii 0.815 m, the error would initially increase at a small rate but it
were calculated. becomes subatantialif the diameter reaches amall values of 0.1 m or
4. The currents needed for ionization to reach the outer d i u # of leas. Such error, however, im not of concern becam it docr not
each of the aegments were then calculated. Also the corresponding apply to practical electrode sizcs; a single rod being thc minimum.
time instants on the front of the wave were calculated.
5 . Referring to Fig. 5 , ionization covefp aement no. K at instant 4.3 EFFECT OF ERROR IN MEASURING V, I AND p.
and the segment remains in a case of increasing ionization as the
current rises to I- then gradually drops until it reach= ikagain at In terms of the relative error, (6)gives:
instant h.During that period, resistivity of the segment decays; the AEi, = Ap, + U , (20) ...
value at instant t, being: For the case of a hemisphere, (5.2)and (4.2) give AJ, in terms of
~j = P. EXP 1 - Ctj 4&11 (18)
e.. AI , Ap, and AR, . R, is obtained by dividing the voltage and
where y1 is the ionization time constant and ia equal to 2ps 1183. current mcaaurements. It follows that:
6. At instant &, resistivity of the mil within segment no. K will AE, = 2Av AI - 4, ... (21) -
have dropped to a Certain minimum value &-. From that instant If the errors in measuring V & I arc f 596, and the m r in
onwards, resistivity of segment K recovers according to the equation: measuring po (or %, is f lo%,then:
4.t = &,mu (Po - [l-(Jj.k 1J3l29 (19)
0 . . Maximum AEo = 25% ... (22)
9 = 1.0 - EXP [-(t, - +.,tJ / 7J ... (19a) Applying a similar procedure to the case of ground rods gives:
where y2 = the deionization time constant and is qual to 4.5 ps
1181.
J, = the value given by (2)
J,,k = the current density in segment no. K at instant no. j.
For each of the n segments, a logical variable was assigned for each
of the 2n instants showingwhether the segment was in a state of d e
ionization or increasing ionization. This gives a matrix of 2nxn
elements. Similarmatricea showing the ionization duration tima and
the deionization duration times, whichever ia applicable at the
subject instant for the subject regmeat, were plso built. Using
equations (18) and (19), a 2nxn matrix giving the resistance of each
element at tach instant was thar calculated. The total r e s h c c of
the electrode at each of the 2n inatants waa then calculottd, the
results tabulated, and the minimum value dettrmined. Thu latter
value was then used to estimate both the perceived value of the
ionization gradient and the error between the perceived value and the
true value. The findings of the study arc as follows:
Fig. 5. Analysis of effect of non-zero resistivity of
1.With the number of segments of the ion* zone taken up to 70
(the maximum tried), the minimum resistance always occurred at the ionized zone.

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.
 1675

AEo = &, + AI - # (bo + AI - AV) ... (23) the inner electfodc was a rod touching the surface of the soil.
where: Surface flashoven were not reported in any of the testa in which part
(I = [(L + 2aJ / L] In [(L + ai)/aJ ... (24) of the rod was inscrtod in the soil.
Our cplcutations indicpte that pi< < L for proctirpl CPPCI. For A solid Ccrpmic insulator with lots of voids (due to mnnuf.cturing
example, Bellaschi's 1941 data gave a, in the range 1.42 - 12.17 cm defects) is known to be more likely to foil intenrnlly than through a
while L waa 228.6 an. Hence the fonowing approximations may be surfooe flashover. A gran& soil sample b baskdy an cx&ane
adopted: case of the defedivc insulator deecribed above and hence it should
+
(L 24) I L I 1.0 b d down intenrnlly rather than undergo surfacc ionization. The
(L+aJ/pi ULIpi existence of water in the soil further encourages the breakdown to
Thus (23) becomes: occur intcmdy.
+ + -
aE, dp, AI - (boAI AV) In ( Wai ) ... (25) Flanagan et al. [12] did a remakable test which provea that the
If dpo= *lo%, AV = AI = 3396, L = 2.5 m, and 4 = 0.05m. d i s c h r g c p r e h to take place below the surfacc of the mil rather
then In U 4.0, and (25) becomes: than along its surface. kkrring to Fig. 6 [19], a h h l y - c u t tree
AEo = 4AV - 3AI - 3dp0 (26)... branch (about 5 mm in diameter) waa placed at the surface of the soil
M a x i " aE, = 65%... (27) starting at the hanippbcricplebctrodcandexrardingtowithin 28 cm
Thin M significantly higher than the case of 8 hunisphcxe. In most fromthe planar electrode. In this test, the soil level waa only up to
practical cpscs, the trror will probably be within half the values the midpoint of the bemisphericolelectrodcandthe planar electrode
given by (22)and (27). exhdcd above the surface of the soil. The discharge path waa
beneath the surfocc of the roil over most of the distance U shown in
4.4 ERROR DUE TO VARIATION OF SOIL DENSITY Fig.6. In rtnothatest (*ut thc e), the sod level Wps lowered
CAWED BY INSEBTION OF A GROUND ROD 10 that a gap existed between the hemispherical electrodeand the
surfaceof the soil. In thia case, the arc jumped into the soil and
In the testa done by Petropoulos p7J using collcentfic remained beneath the surface until it reached the planar electrode.
hemispherical electrodes the soil waa put in the vessel in successive
layers and pressed by hand to obtain a homogeneous filling.
Pe&opouh noted that the resistivity depended appreciably on the
SOIL LEVEL7 I
compression of the soil. When a rod or a pipe is driven in the - I-

ground, the electrode maka room for itself by compressing the soil
initsilnmdmk ' v i c i i . This may affect the resistivity U well U
the breakdowngrodiart at the surfaceof the ehtrodc. This change
M expected to be limited to a small soil volume around the electrode,
say a layer having a thickness equal to about twice the radius of the
electrodt. In the t8- pmcticd casea when po. I, or both are Fig. 6. Test showing that surface ionization doer not govern
high, the edge of the ionized zone will be far beyond the soil breakdown phenomena 1191.
"comprcased" soil wne. Hencc Eo would not be affected.
Considering the effcct explained in Tabla I and II, however, the
effective value of pmcalculated from R, (the resistance under low 5. DISCUSSION
currtnt conditions) will be somewhat less than the "natural" value
appearingat the edge of the ionized zone. This factor may betreated In the impulse tests done by Octtle on short uniform soil gaps,
U a contributor to the crror in measuring po and h effcct cm be the puncturestrength wu found to be in the range600 - 1850 kV/m.
allowed for when applying (21) and (25). Those testa covered several typea of soils. Based on our discussion
of the breakdown mschanism of the soil, the lower part of the above
4.5 ERROR DUE TO SURFACE IONIZATION? range appliu for the highet watez contart. Under thuderatom
conditions, rain would takep h and the moisture content of the roil
Bued on the data available at that time. Sun& pl] suggested that will garerolly be on the high side. Hence it M the lower range of
the result8 of impulse reristMce measurtmenll might be mostly Oettle's puncture &rea@ which is of interest in lightnhg tripout
controlled by surhceionizotiOnrather than by ionilntinnthrough the CP1CU1PtiOns,MY600- 800 kV/m. Unda field c~nditionr,the degrce
soil. SundebasedhiscommenConthetestadonebyEaton~. A of inhomogeneity will be much higher than under laboratory
careful CXBminatiOn of Eaton's data by the writer rev& that condition#. Heace it M reoronabkto expectthat mmc roil spot8 will
Sunde's conclusion waa not juati6ed PI OM be lcfll from the have ionization gradiits equal to about halfthe 600 - 800kV range
following: mentioned above. Hence minimum E, would be in the range 300 -
1. Eaton mted gravel soils, sand soils, and clay soils. Surface 400 kV/m. As shown in Saction 4.1, it is the lowest value of Eo
dir0bUg.r did not occur in the sand and gravel roil tcrtr. Also, they o c c u r r i n g i n t h e v i c i i o f the tlccrrodcwhich #houldbeused. In
did not occur in clay soils having moisture up to 10.3%. Surface the tab done by Liew et al. [lS], they succdedbd in shuloting the
d d g e s o c c d only in the clay sample when moisture waa o b m e d dynamicbehaviour of several electroduby using E . = 300
inc&to 11.1%andthesurfacewaawet. ThatMmptewaacollcd kVIm. The above analysis suppoltr thia value and hence it h M y
"plastic clay" by Eaton. -
rccOmmcndOd for deter"g the impuke resi&mce of p u n d
2. The plastic clay runplc hod an unusually high puncture g r a d i i eleotroder. That value waa used earlier by Erilrsron [SI and also by
becauseit was hrmcd in a mold underpresrureand the median grain t h e e in UI d y S h Of the M f e Of firancn @lUt firainduced
size was very a d . nprhovcrm W].
3. The test @-up whge surfocc flashover o c c d waa conducive Now consider the valua of E. calcubd from impulse
to the initiation of suchphenomenon: the soil was in a mcellic rcrieturCt mcaauremartr given in Table IV. The reoron for Towne'r
bucket, the surfaceof which exteadad above the level of the roil,and valuea being cxccosively low waa explained in Section 4.1 (bone

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.
 1676
gravely roil partly made of artificial fill). Thia or a similar value of E, to be used needs to be significantly less than the value
undocumented reason probably caused the similarly low valucr obtained h m puncture testa of sampler having appropriate water
obtained by Dick ct al. Berger's value8 wcrc relatively high because content.
of the low moisture content of the single spmple he tested. The 5. The residual resistivity of the ionized zone is not a significant
@g 5 PUPS Of testa generally supp~rtthe u l l ~of E, = 300 - factor in R, calculations
400kV/m when the error bounds given by eqn. (27) are taken into 6. Typical error levels in measuring the voltage of the electrode,
consideration. the impulse current discharged through it, and resistivity of the soil
For the case of a hemisphere, it can be shown that: can lead to errors in tetimating E, up to about 65% for the case of
ground d s .
(R, / RJ = a , d 2 * E, / p,I ... (28) 7. A large number (97) of reliable impulse resistance
HenCC measurements h m several sources together with the above analysis
/ %) a 6 ... (29) indicate that the value of E, should be taken qual to 300 kV/m. The
Eqn. (29) indicates that R, is not very sensitive to E,; a fact which 400 kV/m value suggcrted by CIGRE is also usable considering that
has bcea earlier noted by Chisholm ct al. [4]. In view of this,the R, is not very rensitive to (R,a -for case of hemispheres).
400kV/m quoted in the CIGRE Working Group document [SI is a h On the other hand, the loo0 kV/m value earlier suggestedby O&
usable. is definitely too high and should not be used.
To facilitateapplication, the reduction in the impulsc resistance for
the cpst of a singlerod is shown in Fig. 7. For more complex CPIUI. REFERENCES
the gawralmcthod developed by Chisholm ct al. [4] should be used.
Note that the shape of the c u m a in Fig. 7 results from the spatial [l] Armstrong, H.R. (1953). "Grounding Electrode
dietxibution of the resistance shown in Tables I and XI. Characteristics from Model Tests", AIEE Trans., Vol. 72, pp. 1301-
1306.
1.1 I
p]Bellaschi, P.L.(1941). "Impulseand 6ocycle Characteristics
1.0 - -
Noto:
R . Is thomlnlmum abaorvod value and It of Driven Grounds", AIEE Trans., Vol. 60,pp. 123-128.
aAcurs shortly a f t o r tno erait o f tho impulse [3] Bellaschi, P.L., Armington, R.E., and Snowden, A.E.
(1942). "Impulse and 6O-cycle Chara&tics of Driven Grounds,
Part XI", D E Trans., Vol. 61, pp. 349-363.
[4] Chisholm, W.A., and Janischewskyj, W. (1989). "Lightning
Surge Responseof GroundElectrodes", IEEETrans., Vol. PWRD-4,
NO. 2, pp. 1329-1337.
[SI CIGRE Working Group on La-tning. (October 1991).
"Guide to Prootdurts for Estimating the Lightning Performance of
0
Transmission Lincr", CIGRE, Paris, France.
10 20 30 40 50 [q Dick, W.K., and Holliday, H.R. (1978). "Impulse and
CURRENT (kA) Ahernating Current Tests on Grounding Electrodes in Soil
Fig. 7. Reduction of grounding rtsistance for the case of a Environment", P E E Trans., Vol. PAS-97, No. 1, pp. 102-108.
single rod (2.5 m long, 16 mm diameter). Eaton, J.R. (Ocbber 1944). "Impulse Characterisitice of
Electrical C O M ~ C ~ ~ OtoMthe Earth", General Electric Review, Vol.
47, NO. 10, pp. 41-50.
The breakdown mechanism of the soil presented herein similarly [SI Erikason, A.J. (May 1981). "Transient Impedance of
implies that the flashover voltage of a long roil gap is significantly Earthing Systems", put 11of "Pammcters Influencing the Lightning
lower e that of a comparable air gap. Intertstcd readers are Performance of Elcctricol Systans", CIGRE Study Committee 33,
r e f t r r c d to [21] where this lpttet topic is discussed. Paper 33.81 (SC) 03.2, Rio De Janeiro, Brazil.
Our eXaminatiOn of available data indicatu that ionization of the [9] Brler. J.W. and Snowden, D.P.(1983). "High Resolution
air trappal in the voids of the roil is the mechanism by which Stud* of the ElcctricPl Breakdown of Soils", IEEE Trans. on
breakdown of the mil occu~s. Howccr, there may be other Nuclear Science, Vol. NS-30, No. 6, pp. 45644567.
unidcntificd mechanism(s) which can better explain the observed [lo] Fagan, E.J. (1970). "The Use of Concrete Enclosed
phenomena. Basic research on this topic wouldbe most welcome. .
Reinforcing Rods PI GroundingElcctrodco" E E E Trans .on Industry
and General Amlications, Vol. IGAd, No. 4, pp. 337-348.
6. SUMMARY AND CONCLUSIONS Ell] k g a n , T.M., Mallon, C.E., Denson, R., and Leadon,
R.E. (1981). "Electrical Breakdown Propertien of Soil",
1.Breakdown of the soil is initiated by ionization of the air trapped Trans. on Nuclear Scien%. Vol. NS-28. No. 6 pp. 4432439.
in the voids of the soil. The ionization gradient of the soil is [12] k g a n , T.M., Mallon, C.E.,Denson, R., and Smith, I.
always less than that of air and it is drastically reduced when the (1982). "Elcchical Breakdown Characteristicsof Soil",E E E T m s .
water content is high. on Nuclear Science, Vol. NS-29, No. 6 pp. 1887-1890.
2. Increasing the water content decreasea both E. and the soil [13] IEEE Working Group Report. (1992). "Estimating
resistivity p,. However, no direct correlation exists betweem E, and Lightning Performance of Transmission Lines U - Updates ta
po becpuscp, is also dependent on the amount of salt in the roil. Analytical Models", Paper No. 92 SM 453-1 PWRD, 14 pp.
3. Surface ionization doer not govem the value of the impulse [14] IEEE Working Group (April 1992). "IEEE Design Guide
resistance R, and it does not occur to a significant degree exccpt in For Improving the Lightning Performance of Transmission Lines",
rare circumstancco. m, PisCOtaway, New Jersey, U.S.A.
4. The value of R, is mainly deteimined by the lowest value of E, [la Kawai, M. (1965). "Studies of Tower Footing Resistance
occurring anywhere along the electrodeand in its vicinity. Hcncethe on TransmissionLinm", B E E ConfemcePaDer no. 31 Cp. 65-704.

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.
 1677
[la Laghari, J.R., and Qururhi, A.H. (1980). "Flashover
Voltages of Cylindrical Insulators in Gas Mixtures",Proceedings of
the Sixth International Conference on Gas Discharges and their
Application, pp. -242.
[17] Leadon, R.E., h g a n , T.M., Mallon, C.E., and Denson,
R. (1983). "Effect of Ambient Gas on Arc Initiation Characteristics
in Soil",IEEE Trans. on Nuclear Science, Vol. NS-30,No. 6, pp.
45724576.
[18] Liew, A.C., and Darveniza, M. (1974). "Dynamic Model of
ImpulseCharacteristicsof ConcentratedEarth", prOc. IEE, Vol. 121,
NO. 2, pp. 123-135.
[19] Mallon, C., Denson, R., FWagan, T.M., and Leadon, R.E.
(April 1982). "Electrical Breakdown Characteristics in 0.8 to 1.0
Meter Soil Sampler", Theoretical Note 318, Air Force Weamnp
Laboratory, Kirtland AFB, New Mexico.
[20] M o w , A.M., (1990). " P r o d g Firemen A g W F b
Induced Flashovers",, - Vol. 5., No.
1, pp. 297-302.
[21] Mousa, A.M., (1992). "Breakdown Gradient of the Sod
Under Lightning Discharge Conditions", gs of the
International Aerospace and Ground Conference on Lightning and
Static Electricity, Atlantic City, New Jersey, U.S. A., Paper No. 67,
12 pp.
[22] Nitta, T., and Shibuya, Y. (1971). "Electrical Breakdown of
Long Gaps in Sulfur Hexafluoride", IEEE Trans. Vol. PAS-90, pp.
1065-1071.
[U]Nitta, T. et al. (1978). "Factors Controlling Surfice
Flashover in SF, Gas Insulated Systems", IEEE Trans., Vol. P A S
97, NO. 3, pp. 959-968.
[24] Oettlc, E.E. (December 1988a). "The Characteristics of
Electricpl Breakdown and Ionization Procwsu in Soil",Transactions
of the South African IEE, pp. 63-70.
[2s] O d e , E.E., and Geldenhuys, H.J. (December 1988b).
"Results of Impulse Tests on practical Electrodes at the High-Voltage
Laboratory of the National ElectricalEngineeringResearchInstitute",
Transactions of the South African IEZ, pp. 71-78.
[26] Oettte, E.E. (1988~). "A New General Estimation Curve for
Predicting the Impulse Impedanceof Concentrated Earth Electrodes",
IEEE Trana., Vol. PWRD-3, No. 4, pp. 2020-2029.
[27l Petropoulos, G.M. (1948). "TheHighVoltage Characteristics
of Earth Resistance+",Journal IEE, Vol. 95, Part U, pp. 59-70.
m] Rogers, E.J. (1981). "Impedance Characteristics of Large
Tower Footings to a loop Wade Square Wave of Current,
m., Vol. PAS-100, No. I, pp. 66-71.
[29] Snowden, D.P., and Erlcr, J.W. (1983). "Initiation of
Electrical Breakdown of Soil by Water Vaporization", IEEE Trans.
on Nuclear Science, Vol. NS-30,No. 6, pp. 45684571.
[30] Snowden. D.P., Beak, E.S., and VM Lint, V.A.J. (1986).
"The Effect of Gaseour Ambient on the InitiPtion of Breakdown in
Soil",JEEE Trans. on Nuclear Science, Vol. NS-33, No. 6.. pp.
1669-1674.
[31] Sunde, E.D. (1949). Rrth Conduction Effects in
Transmission Systems, D. Van Nostrand Company, Inc., pp. 267-
269.
[32] Towne, H.M. (November 1929). "ImpulscCharoctcns * ticsof
Driven Grounds", General Elwtric Review, pp. 605-609. Also
published in the AIEE Lightning Reference Book 1918-1935, pp.
259-263.
p3] Vianeister, P.E. (1961). The Liehtnine Book, Doubleday &
Company, Inc., Garden City, N.Y., pp. 137-141.

B w a p h y of Author: Please nee IEEE TrPnsPctions on Power

Delivery, Vol. 6, No. 1, p. 149, January 1991.

Authorized licensed use limited to: Volodymyr Shostak. Downloaded on November 23, 2009 at 17:02 from IEEE Xplore. Restrictions apply.