sts of the ‘early streamer emission’ principle for

rotection against lightning

[Link]
K.J. Corn ick
D.C. Faircloth
[Link]

Indexing terms: Early streamer emission principle, Lightning protection, Simulation testing

to the fabric, since upward corona may be initiated at

Abstract: Experiments are described which are parts of the structure more favourably placed in rela-
designed to test two devices based on the ‘early tion to the downward leader. However, if the corona
streamer emission’ (ESE) principle, for lightning can be activated at an earlier time in the downward
protection, against the traditional Franklin rod. progress, development of the upward leader may be
In all three cases, the device was subjected to a advanced sufficiently, by such an activated rod, to
steady negative electric field from a sphere, overcome the distance disadvantage and effect an
simulating the field beneath a thundercloud, prior attachment before corona at other sites can develop
to application of a superimposed negative impulse sufficiently to compete with it. Extension of this argu-
field, simulating the field due to the downward ment suggests that the corona set up at the active rod,
leader. The first device consisted of a vertical rod placed centrally, for example, could replace the coronas
to which a subsidiary 1150ps positive impulse from the separate passive rods placed in a conventional
voltage, variable up to 40kV peak, could be system around the roof of the building.
applied with varying delays from the start of the This principle forms the basis of the so-called ‘early
negative impulse field. Energising of the rod was streamer emission’ devices which have been developed
thus independent of the applied, negative, field. in recent years. The success of such a device depends
The second device was a commercial product, on the timing of the corona initiation in relation to the
energising of which was controlled by its own downward leader approach and the rapidity with which
power supply. Sparkover voltages in the sphere/ the leaders can attach compared with the time that
device gaps and times to breakdown were would have been taken with passive rods.
measured. It is shown that the ESE devices In this paper, tests are described in which the charac-
showed a small advantage, in time to breakdown, teristics of the Franklin rod have been examined, under
over the Franklin rod. simulated lightning storm conditions, and compared
with those of two ESE devices, one in which a rod has
been energised by an independent pulsed voltage, the
1 Introduction other a commercial device with its own power supply.
The traditional ‘Franklin rod’, used in lightning protec- 2 Test techniques
tion, depends for its effectiveness upon development of
a corona discharge at its tip as the result of high elec- Prior to initiation of a lightning discharge, the average
tric fields developed in a lightning storm. With the ‘steady’ electric field at the ground increases to several
approach of a downward leader, the resulting rapid tens of kilovolts per metre, due to the charge on the
increase in field augments corona activity. Under cloud above [I]. The field at the tip of an exposed light-
favourable conditions, one of the streamer filaments ning conductor terminal can thus be expected to be
which constitute the corona may undergo sufficient much higher. These conditions can be simulated in the
heating to develop into a highly conducting, arc-like laboratory by application of a DC voltage to a large
‘upward leader’ which can then propagate for consider- object suspended above a conductor. The subsequent
able distance in a comparatively low electric field. It descent of the leader is simulated by super-imposition
may thus progress towards the downward leader, effect of an impulse voltage to the gap, of rise time approxi-
an attachment and allow the subsequent high-current mating to that of the field produced by an approaching
discharge to pass down the conducting path so formed. leader. A critical test requires not only measurement of
A simple passive Franklin rod, on the roof of a large the respective probabilities of striking to active and
building, may not give full protection against a strike passive rods, but also information on the time during
0IEE. 1998 the impulse at which the strike occurs.
ZEE Proceedings online no. 19982209 These conditions were partly satisfied by Bouqueg-
Paper first received 9th February and 111 revised form 30th April 1998 neau [2] in which the numbers of strikes were meas-
N.L. Allen, D.C. Faircloth and C.M. Kouzis are with the High Voltage ured, out of two groups of 100 trials, to an active and a
Laboratory, UMIST, Manchester M60 lQD, UK passive rod mounted l m and 2m apart and symmetri-
K.J. Cornick was with UMIST and is now with Power Grid Ltd, cally placed relative to an upper rod electrode sus-
Singapore 118485 pended vertically above the mid-point of the line
200 IEE Proc -Sei Meas Technol, Val 145, No 5, September 1998
joining the bases of the two rods. An impulse voltage depends upon the velocities of the upward and down-
( 1 . 2 / 5 0 ~ was
) applied to the upper rod. The active rod ward leaders up to the time of their attachment, it
was excited by a steady 25kV voltage applied from a might be expected that a measurement of the time to
separate supply; active corona is assumed to have been sparkover, in identical plane-rod gaps, would be critical
set up. The results showed no significant difference in making comparisons between them.
between the rates of striking to the active and passive This question has been examined in a critical review
rods. The tests, however, could not be regarded as con- by Mackerras et al. [6],who considered the paths taken
clusive, since the impulse voltage to the upper rod was by leaders in the neighbourhood of simple structures
too fast to simulate the effects of the leader descent and protected by ESE or Franklin rods. 'These authors dis-
no preceding steady electric field was provided. Gener- puted claims that an appropriate leader velocity was of
ally similar tests have been carried out more recently by the order 106ms-' and that a lower value of the order 2
Grzybowski et al. [3] on commercial devices with simi- x 104mss1 is more appropriate whlan estimating the
lar results. attachment position of upward and downward leaders
In large-scale tests at the Les Renardieres High Volt- and hence, the time of attachment, which is simulated
age Laboratory, Berger and others [&6] placed a com- in laboratory experiments by the time to breakdown.
mercial ESE device, based on a vertical rod, beneath a Uman [l], in a review, points out that measured down-
large metal plane 15 x 20m, suspended at a height of ward leader velocities vary between 6 x lo4 to 2.6 x lo6
13m above the ground plane, and to which high volt- ms-l and that upward leader velocities vary from a
age impulses were applied. Comparisons were made minimum of 2 x io4 mss'.
with a passive Franklin rod placed in separate, identi-
cal tests in the same position. The gap between the 3 Test arrangements
plane and the tips of the respective rods was always
9.5m but two experimental arrangements used rods of Test conditions require that lightn mg conditions be
lengths 3.5m and l m standing vertically on the ground simulated by first, a steady electric field between
plane. Steady negative voltages were applied to the 'cloud' and ground, followed by a superimposed
plane to simulate the field beneath a thundercloud and impulse field to provide an apprloximation to the
negative impulse voltages, rising to peak in 500psec increasing field set up by an approaching downward
were superimposed, so simulating the increasing Geld leader. In the present experiments, the cloud was simu-
due to the approach of a leader. lated by a 0.75m diameter sphere, 'to which the high
The mode of operation of the ESE device was not negative voltages were applied, suspended above the air
disclosed. However, image convertor photography termination to be tested, which was mounted on the
showed that whereas the Franklin rod required several laboratory floor. Its role was to generate in the test
ionisation steps before continuous leader propagation space the same electric fields as encountered in nature
was achieved, the ESE device initiated a continuous and that it should be corona free.
leader about 100ps earlier. Thus, the energisation of The conditions were produced by the circuit of
the rod of the ESE device supplied sufficient energy to Fig. 1. A steady voltage was applied to the sphere
convert the early streamers into a leader. The velocity from a DC set producing voltages variable up to
of advance of the leader was shown, however, to be 750kV. A negative impulse voltage, of rise time 5 0 0 ~
exactly the same in the two cases, namely 2 x lo4 m d . to peak, was selected for this work since Berger [4] has
Results were quoted in terms of the mean field shown that this provides a fair approximation to the
between plane and ground, disregarding the perturba- rising field of an approaching leader. The impulse was
tion due to the rod. For the l m and 3.5m rod, the supplied by a conventional Marx generator, of maxi-
mean fields at which continuous leader development mum peak voltage 2MV, coupled to the sphere
was initiated were as in Table 1. through a capacitor of 550pF, which served to block
direct current from the generator circuit. Impulse volt-
Table 1: Mean fields for initiation of continuous leader ages applied to the gap were measured by means of a
development
potential divider created by the series connection of the
Rod length Franklin rod ESE rod capacitance of the large sphere gap, and a low voltage
capacitance C2. The DC set was protected from volt-
Im 285kVm-1 190 kVm-' age impulses by a series resistance and parallel capaci-
(100%) (66.7%) tance, as shown.
3.5m 213kVm-I 127kVm-'
impulse
(100%) (59.6%)
generator DC supply
n n
Lower fields, recorded for the ESE rod, correspond
to earlier initiations during the negative impulse
applied to the plane above. Examination of Fig. 3 of
[5] shows, however, that the leader from the Franklin
rod, despite initial intermittent development, has
reached the same distance from the tip, at the same
time, as the continuously developing leader from the
ESE device. The physical meaning of leader initiation
in relation to the attachment process thus requires care-
ful consideration.
No comparison is given between the respective spark- Fig. 1 Combined impulse and direct voltage circuits for producing a
over voltages and the times to sparkover, in these composite high voltage
experiments. Since the efficacy of both types of rod Rf= 115kP, C,, = 550pF, C, = 200pF, C, = 9lpF

IEE Proc -Sei Meus Technol, Vol 145, No 5, September 1998 201
 For tests with the independently energised Franklin decided to apply a steady voltage to the sphere suffi-
rod air termination (the 'active Franklin rod') a small cient to cause a small visible positive corona at the ter-
impulse generator, providing voltage pulses up to mination. This voltage was -150kV.
+40kV was used (omitted for clarity in Fig. 1). This The first test was carried out without the steady field.
generator could be triggered at any desired time within The 50% sparkover voltage at this condition, was
the impulse provided by the main generator by opera- 491kV. For further tests with the Franklin rod the
tion of a delay circuit linking the two through fibre- peak impulse voltage was then increased to -725 kV, so
optic cables. The sequence is shown in Fig. 2. The that breakdown on the front of the wave could be
small generator was protected, in the event of sparko- assured, thus simulating the rise in field experienced by
ver to the rod, by a small parallel spark gap and series the termination as a downward leader approaches. The
resistor of 8kQ. degree of approximation is shown in Fig. 3 where com-
t=O
parison is made between the field strength at ground
level (disregarding the termination) under the labora-
+II +I tory impulse and that calculated according to Berger [4]
a,
I for the case of the leader of a 5kA lightning stroke.
0) I delay With this impulse, the mean time to breakdown of the
-
c
I 1
0
I
I
passive Franklin rod gap was 3 1 0 where,~ as Fig. 3
' impulse applied to rod shows, the laboratory and calculated leader fields are
1.2/50pswave)
reasonably close.
I
0
I
I f time

I
I
I

U b!
---___----
-I time to breakdown
0 40 80 120 160 200 240 280 320
Fig. 2 Sequence of voltage applications at sphere-rod gap
time, ps
Fi . 3 Comparison between calculatedfield due to an a proaching leader
The tests with the commercial ESE air termination andg$eld at ground produced by an impulse voltage reac&g peak value of
-72SkVat500p
were carried out in the same arrangement, (Fig. 1) but __ impulse -725 kV
were preceded by additional tests to demonstrate its -computed, I = 5kA
mode of operation. These consisted of the measure-
ment of the amplitude and frequency of the voltage Subsequent tests were carried out with the steady
pulses, generated by the device itself when the tip was voltage of -150kV at the sphere. Here, a composite
exposed to an electric field set up by the voltage on a voltage of -535kV (that is, a superimposed impulse of
separate sphere. Fuller details of these experiments are -385kV) was sufficient to cause breakdown on the
given in Section 5. wavefront; the mean time to breakdown was 1 1 7 ~ .
In all the experiments, initiation of corona at the ter-
minations was detected by photomultiplier observation 4.2 Active Franklin rod
of the light emission. Time to sparkover, between These tests were also carried out under conditions first,
sphere and termination, was found from the potential with zero steady field and second, with the steady field
divider observation of the collapse of voltage across the provided by a voltage of -150kV at the sphere.
gap, which was correlated with photomultiplier obser-
vations.

4 Tests with the Franklin rod T

4. I Passive Franklin rod
For these tests, a sphere-termination spacing of 0.75m
was adopted. The tip of the termination was 0.75m
above the laboratory floor. The end of the rod was
machined to a 20" cone.
= 100
Since Franklin rods are commonly mounted on
prominences on buildings, rather than on a flat plane,
steady corona normally precedes the augmented 0
0 50 100 150 200 250 300
streamer emission resulting from the approach of a delay, ps
downward leader. Related work in the same high volt- Fig. 4 Times to breakdown of sphere/active Franklin rod gap under sim-
age arrangement [7] has also shown that where the ple impulse voltage asfunction of time delay to application of 1.2/S0p aux-
steady voltage is insufficient to initiate corona at an iliary impulse at rod
Steady voltage = zero. Time to breakdown of sphereipassive rod shown as
electrode, it has no significant effect on the corona and horizontal dashed line. Limits of dispersion about the mean are shown.
sparkover occurring on application of a superimposed
impulse voltage. Therefore, although producing a gen-
eral field near the ground plane which is greater than With zero steady field, the -7251tV impulse was
that normally encountered in lightning storms, it was applied to the sphere and an auxiliary +40kV peak
202 IEE Proc.-Sei. Meas. Technol, Vol 145, No. 5, September 1998
lightning impulse (1.2/50ps) applied to the rod at vari- over between the sphere and the Franklin rod by
ous times after the start of the main impulse. The between 4 0 p and 7 0 p .
sequence is shown in Fig. 2. Times to breakdown were
measured as a function of the time delay to application 5 Tests w i t h a commercial ESE device (the
of the lightning impulse; the results are shown in Fig. 4 ’Pulsar’)
where comparison is made with the passive rod. It is
clear that the auxiliary impulse had no significant effect 5. I Functioning of the device
upon the average time to sparkover of the main gap, It is known that most commercial ESE devices operate
that is, about 310p, until it was delayed by more than by application of a pulsed voltage to a pointed termi-
2 0 0 p from the start of the main impulse. There was nation. For meaningful tests, therefore, it was neces-
then a sharp reduction of about 7 0 p in time to break- sary to characterise the device being tested, known as a
down at -240p delay accompanied by a reduction to ‘Pulsar 7’. This device consisted of a rod, 0.75m long,
an insignificant level of the dispersion in breakdown tapering to a tip of radius lmm approximately. The
times. For longer delays, the times to breakdown then rod surmounted a cylindrical ‘can’, containing a power
approached the value in the absence of the auxiliary unit.
impulse. This was sealed, self-contained and integral with the
Using the -150kV steady voltage at the sphere, two device. It was evident, therefore, that the energy
sets of tests were carried out, with +10kV and +20kV required for the high voltage pulses must be derived
auxiliary impulses at the rod. The peak value of the from the corona set up at the tip of the rod in the
main impulse in these tests was -385kV which, when enhanced electric field existing throughout a lightning
added to the steady voltage, gave a composite voltage storm. The energy is stored until the power unit dis-
at the sphere of -535kV. Times to breakdown were charges to produce a pulse of high voltage which then,
measured; the results are shown in Figs. 5a and b. presumably, produces a more energetic corona.
In these experiments, the tip of the device was placed
160
r a few centimetres below a sphere, of diameter 0.25m,
which was connected to a variable DC supply up to
30kV (Fig. 6). Voltages and sphere-rod gap were
arranged so that corona was set up at the tip. The
resulting voltage developed on the rod itself, by the
functioning of its power supply, wa:; monitored via a
2 t - 1
1OOO:l Tektronix divider probe of input resistance
1OOMQ and input capacitance 3pF and a further 1O:l
divider. The tip of the rod was viewed by a photomulti-
0 20 40 60 80 100 120 plier. The voltage and photomultiplier signals were dis-
delay, ps played simultaneously on an oscilloscope.
a
160

to voltage
-
h
n

n
n
n
generator

._
c photo-multiplier
n
q-
01 +--rJ== _A. gap spacing
0 20 40 60
delay, ps
b
80 100 120 to oscilliscope

I
+ETand
Fig.5 Times to breakdown obtained as in Fig. 4, but with a steady volt-
age of -150kVat the sphere
a +lOkV peak impulse at rod oscilloscope
b +20kV peak impulse at rod

In the absence of the auxiliary impulses, the mean Fig.6 Preliminary experiment: arrangement oJ sphere, rod and associ-
time to breakdown was 1 1 7 ~When . +10kV auxiliary
ated measurements of rod voltage and corona light output
impulse was applied with delays less than S o p , there
was a reduction in time to breakdown which was nev-
ertheless small and hardly significant. At 80 p delay, An example of the voltage oscillogram is shown in
the time to breakdown was reduced from that for the Fig. 7a. Here, the tip of the rod was placed 4cm below
passive rod by about 4 0 ~ to, be compared with 7 0 p the sphere, to which a steady voltage of -29kV was
in the previous case described where the steady field in applied. The oscillogram shows that repeated negative
the gap was zero. When the +20kV auxiliary impulse charging of the rod occurred, between large, fast volt-
was applied, the result was generally similar. As in the age excursions; the cycle was approximately 25 ms, that
case with zero steady field, the dispersion in the results is, a frequency of 40 Hertz for the sphere voltage used.
was reduced to insignificance at the time delay corre- Use of a faster sweep showed that the voltage excur-
sponding to the maximum reduction in the time to sions were oscillatory as in Fig. Sa; here the oscillation
sparkover. frequency was approximately 40 kHz and the peak to
Thus, with or without the steady field, the effect of peak amplitude was found to be constant, from shot to
the auxiliary impulse voltage was to advance the spark- shot at approximately 7kV. (The variation in ampli-
IEE Pvoc.-Sei. Meas. Technol.,Vol. 145, No. 5, September 1998 203
tudes apparent in Fig. 7a was a function of the sam- Table 3: ESE device characteristics at constant voltage
pling process of the oscilloscope on this very slow of 25 kV
sweep).
Gap Charging Oscill. Oscill.
distance, cm frequency, Hz frequency, kHz amplitude, kV
4.4 2.6 35.7 6.2
4.1 6.3 36.4 6.3
3.5 13.2 35.8 6.4
3.0 18.8 35.9 6.8

all conditions, the oscillatory frequency and amplitudes

remained about constant. The results implied that the
oscillation was a characteristic of the circuit in the
power supply and that the charging time and, therefore
the repetition rate of the oscillations, depended upon
the field at the tip of the device. In Tables 2 and 3 the
increase in voltage and the decrease in gap length were
respectively taken to limits beyond which breakdown
b would have ensued.
Fig.7 Oscillograms: tip of rod lcmfrom sphere ut -29kV Figs. 7b and 8b show photomultiplier signals corre-
Horizontal scale ?Oms/div
a Voltage at rod, recorded by probe arrangement, arbitrary units sponding to the voltage oscillograms. Each downward
b Photomultiplier record of light emitted by corona pulses peak represents light from a corona pulse. Amplitudes
in Fig. 7b are variable; this, again was a function of the
sampling process. Some light peaks clearly coincided,
in this event, with peaks of oscillatory voltage, shown
on an expanded time scale in Fig. 86. However, corona
did not always occur at voltage peak. Such a case is
shown in Fig. 9b where, for the same voltage excursion
as in Fig. 7 , a corona pulse occurred near the end of
the charging period, but not during the oscillation. It is
assumed that this pulse, with others observed in Fig.
7b, was due to coronas occurring during the charging
period which thus contributed to the subsequent high
voltage oscillation.

b
Fig.8 Oscillograms: conditions as for Fig. 7
Horizontal scale: 20p/div
a Voltage at rod, recorded by probe arrangement, arbitrary units
b Photomultiplier record of light emitted by corona pulses

The repetition rate of charging, the frequency and

the amplitude of the fast voltage oscillations were
measured in two ways:
(a) by varying the sphere voltage at constant sphere-tip
distance; results are given in Table 2.

Table 2: ESE device characteristics at constant sphere-

b
tip distance of 3.2cm
Fig.9 Oscillogrums: conditions us for Figs. 7 and 8
Horizontal scale: 20pidiv
Sphere Charging Oscill. Oscill. a Voltage at rod, recorded by probe arrangement, arbitrary units
voltage, kV frequency, Hz frequency, kHz amplitude, kV b Photomultiplier record of light emitted by corona pulses

21 6.4 35.0 6.8 Table 4: Probability of corona pulse during voltage oscil-
23 11.4 36.4 6.7 lation
25 16.7 36.4 6.8 Sphere Probability of
26 19.5 37.7 6.9 voltage (kV) corona pulse
27 22.7 37.7 7.1 -2 0 0.37
29 28.4 37.6 7.0 -22 0.40
29.7 31.2 37.7 7.1 -24 0.60
-2 5 0.83
(b) by varying the sphere-tip distance at constant Gap = 3.5 cm
sphere voltage; these results are shown in Table 3.
A general consistency is evident between the charging It was found that the probability of corona occurring
frequencies obtained by the two procedures but under at the peak of the oscillation increased with the applied
204 IEE Proc -Sei Meas Technol, Vol 145, No 5, September 1998
voltage at the sphere, despite the constancy of ampli- 6 Discussion and concluding remarks
tude of this oscillation, Table 4. The reason for this
behaviour is not apparent. It may be associated with In these experiments, the passive and active Franklin
changes in the ion space charge set up around the tip, rods and the Pulsar have been compared under simu-
since the rate of change of tip voltage changes with lated lightning storm conditions. The average stresses
sphere voltage. applied to each sphere-rod gap were similar, but in the
The highest sphere voltages used in these experiments active Franklin rod case, it constituted an overvoltage,
yielded an average stress in the gap of the order whereas in the Pulsar case, it was a threshold voltage
900kVm-', which approaches values encountered in a for sparkover. Absolute rates of risle of voltage were
lightning storm. No corona could be detected below similar in each case, and a comparison between the two
-200 kVm-'. Thus, the experiments covered much of devices can therefore be made. With the active Frank-
the range of conditions under which the device could lin rod, working at approximately 40% overvoltage for
function. the gap, it was shown that independent energising of
the rod reduced the time to sparkover by between 40ps
5.2 Performance in tests under impulse and 7 0 p , compared with that for the passive Franklin
voltage rod. This reduction was obtained, however, only over a
The sphere-rod test arrangement was used again, as narrow range of times of application of the auxiliary
with the passive and active Franklin rod. The tip of the 1.2150ps impulse at the rod, which themselves extended
pulsar was 1m above the ground plane. Comparison over 7 0 p prior to the breakdown time of the passive
was again made with the characteristics of a passive rod. These results were obtained whether or not the
Franklin rod of similar length, with a hemispherical tip applied impulse, to the sphere, was preceded by a
of diameter 10". Two sets of experiments were car- steady electric field. Thus, the time of energisation of
ried out, using sphere-rod gaps of 1.0m and 1.4m. the active rod was critical in achieving an advance of
Prior to testing with the combined circuit of Fig. 1, the breakdown time. Only one time-to-peak was used
the DC voltage alone was raised from zero to deter- here; it was chosen to simulate the rate of rise of aver-
mine the condition at which the Pulsar would start to age field indicated in Fig. 3. It must be expected that a
function. It was found that it started to charge at different advance would be obtained with a different
sphere-voltages of -50 kV and -60 kV respectively for time-to-peak and it follows that othcr average stresses,
the 1.0 and 1.4m gaps and voltage pulses commenced over a range that might be expected from the leaders
at -204kV and -206kV. The voltage measuring circuit occurring in nature, would also affect the results.
was removed from the Pulsar during these tests to Comparison of the Pulsar rod with, the passive Fran-
avoid the risk of damage at sparkover. klin rod showed little difference in sparkover character-
Sparkover voltages were measured, first in the 1.0m istics, except at the longer gap used, where a possible
gap using simple negative impulse only and then in time advantage to the Pulsar of 401s was noted. The
both gaps, using pre-stress negative voltages of -250kV nature of the tests precluded observation of the timing
(I .Om gap) and -300 kV (1.4m gap) to assure repeated of the Pulsar's oscillatory voltage pulse, however, and
operation of the device. Simultaneous measurements direct comparison in this respect with the active Frank-
were made, over 50 shots in each case, of times to (a) lin rod is not possible. The preliminary tests showed an
the first corona after the start of the impulse, in the 1.0 increase in the repetition rate of this pulse with applied
m gap, (b) the breakdown in the 1.0m and 1.4m gaps. stress, but this was always less than ;!5 Hertz and abso-
The results are shown in Table 5. lute values may have been affected 'by the presence of
Table 5: Sparkover voltages, average times T, to first the measurement circuit. The pre-breakdown corona
corona and average times TBto breakdown showed a higher frequency of corona pulses than the
Franklin rod, probably due to the sharpness of the
ESE rod Gap = 1 m Franklin rod Gap = I m point. There was no clear evidence of an effect due to
VDC = 0 -250 kV 0 -250 kV the oscillatory voltage pulse on this (corona, which was
V50 = -712kV -745 -709 -724 assumed to be due to the main impulse voltage on the
T/=14p~ 9 15.5 11 sphere.
T g = 2 4 7 ~ ~ 219 219 221 Thus, the maximum reduction in time to breakdown,
TB was from 310ps to 240ps, achieved with the active
ESE rod Gap = 1.4m Franklin rod Gap = 1.4m
Franklin rod. Reference to the measured impulse wave-
VDC = 0 -250 kV 0 -250 kV form shows this to correspond to a reduction in stress
v50 = - -992 - -985 from 86% of that at peak voltage to 76%, that is, a
T, = - - - - reduction to 88% of the value for the passive rod.
TB= - 195 - 233 Direct comparison with the work of Berger [4], is not
Impulse voltage = 200/1400ys approximately possible, because only times to leader inception were
presented there.
For both rods, sparkover voltages in the 1.0m gap Differences in time to sparkover must depend on gap
were significantly increased by application of the nega- length, since the growth of the upward leader must
tive prestress voltages; times to the first corona were, determine the breakdown time. Since the maximum
on average, reduced. Times to breakdown with both time advance found in the present vvork was 70ps and
passive Franklin rod and Pulsar did not significantly since the velocity of the leader with both active and
differ from each other, either under simple impulse or passive rods can be assumed to be the same, that is,
with the prestress voltage. In the 1.4m gap, however, about 2 x IO4ms-' [ 5 ] ,it is clear that the maximum time
the average time to sparkover was 38ps less, for the advantage for leader formation found in this work is
Pulsar than for the Franklin rod, though this difference 70ps. This is to be compared with -1OOps in [5] for a
was within the respective standard deviations. gap of 9.5m.
IEE Proc -Sei Meas Technol, Vol 145, No 5, September 1998 20s
 These results, and the comparison with [4, 51 indicate 8 References
that the time advantage may not depend strongly upon
the gap, and Table 5 indicates that the variation with 1 UMAN, M.A.: ‘The lightning discharge’ (Academic Press, 1987),
Chapter 3
stress is not strong. The most important result is the 2 BOUQUEGNEAU, C.: ‘Laboratory tests on some radioactive
critical nature of the timing of energisation of the rod; and corona lightning rods’. Proceedings 18th international confer-
in nature, therefore, it would be essential that a self- ence on Lightning protection, Munich, 1985, pp. 37-45
3 GRZYBOWSKIj S., LIBBY, A.L., GUMLEY, J.R., and GUM-
powered device such as the Pulsar must energise the LEY, S.J.: ‘Comparative testing of ionizing and non-ionizing air
rod at an optimum time in relation to the rate at which terminals’. Proceedings 10th international symposium on High
voltage engineering, Montreal, 1997, Vol. 5, pp. 331-334
the field due to the downward leader is increasing. This 4 BERGER, G.: ‘Determination of the inception electric field of
rate of increase is variable due to the great variation in the lightning upward leader’. Proceedings 8th international sym-
measured velocities of the approaching downward posium on High voltage engineering, Yokohama, 1993, Paper
70.02
leader [l]; however, the present work shows that the 5 ALEKSANDROV, G.N., BERGER, G., and GARY, C.: ‘New
pulsed oscillations occur at only low repetition rates investigations in the lightning protection of substations’. CIGRE,
over a wide range of fields. Further work would be Paris, 1994, Paper 23/13-14/
6 CRISTESCU, D., and GARY, C.: ‘Laboratory simulation of the
needed to show that such a device is able to energise lightning impact to the ground’. Proceedings symposium Light-
itself at a time which provides a significant time advan- ning and mountains, Chamonix, 1994
tage in leader initiation. 7 MACKERRAS, D., DARVENIZA, M., and LIEW, A.C.: ‘Criti-
cal review of claimed lightning protection of buildings by early
streamer emission air terminals’, IEE Proc Sci. Meas. TechnoL,
7 Acknowledgment 1997, 144, pp. 1-10
8 ALLEN, N.L., HUANG, C.F., CORNICK, K.J., and
GREAVES, D.A.: ‘The sparkover of air gaps under composite/
The authors thank English Heritage for their financial direct voltages’. Proceedings 10th international symposium on
support for this work. High voltage engineering, 1997, Vol. 3, pp. 157-160

206 IEE Proc.-Sei. Meas. Technol., Val 145, No. 5, September 1998