Transcript
TRIGGERING OF SURFACE
DISCHARGE SWITCHES
by
RANDY DALE CURRY, B.S.E.E.
A THESIS
IN
ELECTRICAL ENGINEERING
Submitted to the Graduate Faculty of Texas Tech University in
Partial Fulfillment of the Requirements for
the Degree of
MASTER OF SCIENCE
IN
ELECTRICAL ENGINEERING
Appr~
Accepted
August, 1985
ACKNOWLEDGEMENTS
I would like to express my gratitude for the guidance
and support given by Dr. Kristiansen and Dr. Hatfield
throughout the program and further to thank Dr. Guenther
and Dr. Agarwal for their advice and technical support
during the course of this study. I am also indebted to
George Voightlin at LLNL for the computer simulations of
the equipotential plots. Likewise, the support of the
graduate students in the Plasma and Switching Laboratories
was most appreciated. I am especially indebted to Mark
Fowler, Chris Young, Noreen, and my parents, without whose
patience and insistence this program might not have been
completed. I would also like to thank Gary Leiker and
Ellis Loree for their help and advice.
ii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS • . . . . . . . • • • • • • • • • • • • i i
ABSTRACT • • • . . . . . . . • • • . . . . . . . . • iv
LIST OF TABLES • • . . . . . . . . . . . . . . . . . . . v
LIST OF FIGURES • • • • • • • • • • • • • • • . • . vi
I. INTRODUCTION •••••.•••••••• • • 1
6
6
II.
III.
IV.
v.
VI.
VII.
OVERVIEW OF THE SYSTEM . . . . . . . Circuit Operation and Design . . . System Parameters • • • • • • • • • • • 9
Trigger Generators . . . . . . • • • • 14
DIAGNOSTICS . . . . . . . . . . . • • • 15
Voltage and Current Measurements ••• 16
Photographic Techniques • • • • • • • • 18
Surface Analysis . . . . . . . . • 21
TRIGGERING • • . . . . . . . . . . . . . . • 22
RESULTS . . . . . . . . . . . . . . . . • • 3 5
Polarity and Dielectric Effects . . • • 35
Surface Charging
Surface Analysis
. . . . . . . . • • • 50
. . . . . . . . . . • 57
GAS EFFECTS . . . . . . . . . . . . . • • • 6 0
CONCLUSION . . . . . . . . . . . . . . • • • 6 5
LIST OF REFERENCES • . . . . . . . . . . . . . . . • • • 6 7
APPENDIX • • • • • • • • • • • • • . . . • • • • • . . • 70
iii
ABSTRACT
The performance of a triggered, 45 kV, 3 kA surface
discharge switch was investigated under different operat
ing conditions. The performance of the switch was explor
ed, using G-10, Delrin and Lucite (polymer) discharge
substrates. Under optimum operating conditions, the
discharge characteristics of the switch using air, nitro
gen, SF6 , helium, argon, and various mixtures were eval
uated. The trigger performance was evaluated through
measurements of jitter, the number of channels per shot,
and the effects of charging voltage and trigger electrode
polarity. Similarly, surface charging and the character
istics of the dielectrics were studied using ESCA, SEM and
optical microscopy.
iv
LIST OF TABLES
TABLE
1. Switching Characteristics of G-10 as a Function of Trigger and Charging Voltage Polarities •.•.....•.
2. Switching Characteristics of Delrin as a Function of Trigger and Charging
37
Voltage Polarities • . . . • . • . . . . . . • . 43
3 •
4.
5.
Switching Characteristics of Lucite as a Function of Trigger and Charging Voltage Polarities •.•.•..
Comparison of Surface Resistances Before and After Exposure to Flashover
. . . . . .
List of Gases Employed in the Switching Experiments . . . . . . . . . . . .
v
48
59
61
LIST OF FIGURES
FIGURE
1.
2.
3.
4 .
5.
6.
7 .
8 .
9.
Block Diagram of the Surface Discharge Switch System and Diagnostics ...•.
Outline Drawing of the Surface Discharge Switch ....
Photographs of Single Channel and Multichannel Electrodes .
. . . . . . . . . .
Schematic of the Experimental Arrangement .
Photograph of the Surface Discharge Switch
Equivalent Circuit of the 6 ohm Load and a Schematic of the Trigger Generator
Schematic and Photograph of the Voltage Dividers and Feedthroughs .
Streak Camera Arrangement . . . . .
View of the Slit for Multichannel and Single Channel Development Measurements
10. Trial Locations of the Trigger Electrode
11. Jason Computer Code Plot of Equipotentials With and Without a Grounded Trigger Electrode Inserted . . . . . . . . . .
12. Jason Computer Code Plot of the Equipotentials With a 50 kV Trigger Pulse Applied to the Trigger Electrode and One
4
7
8
10
12
13
17
19
20
25
28
Resulting E-field Line . . . . . . . . . . . . . 29
13. Streak Photographs Showing Simultaneity of Multichannel Closure Near the Cathode and Single Channel Arc Development at the Surface . . . . . . . . . . . 31
vi
14.
15.
16.
17.
Open Shutter Photograph of Multichannels and Framing Photograph of Predischarge Luminosity ..•••••••.•.
Voltage and Current Characteristics of G-10 for a Negative 40 kV Charge Voltage and Positive 50 kV Trigger Voltage .•.
Overlay of 5 Shots Showing the Effect of Conditioning on the Risetime ...
Time Integrated Photographs of Multichannels of G-10 ..•••
18. Voltage and Current Characteristics of Delrin for a Positive 40 kV Charge Voltage and a Positive 50 kV Trigger Pulse
19. Multichanneling Characteristics of Delrin for a Negative 40 kV Charge Voltage and a
33
38
39
40
44
Positive 50 kV Trigger Pulse . . . . • • • . 45
20. Voltage and Current Characteristics of Delrin for a Negative 40 kV Charge Voltage and a Positive 50 kV Pulse ....
21.
22.
23.
24.
25.
26.
27.
Voltage and Current Characteristics of Lucite for a Negative 40 kV Charge Voltage and a Positive 50 kV Trigger
Charge Decay Curves of Lucite, G-10, and Delrin • • . • . • . . .
Dust Pattern of a G-10 Dielectric Showing the Charge Pattern Left on the Surface by 500 Shots (Multichannel) •......
Dust Patterns Showing Surface Charges Left by Single Channel Discharges Comparing High Rep-Rate and Low Rep-Rate Operation
Photograph of G-10 Exposed to 6 Shots in 80% N2 + 20% SF6 •..•.•. · · ·
Photograph of G-10 Exposed to 6 Shots in 80% SF
6 + 20% N2 • . . . . . . . •.
Photograph of Discharge in 5% air + 95% He Mixture, Negative 5 kV Charge Voltage, 50 kV Trigger ..•....
vii
47
49
53
55
56
62
63
64
28.
29.
30.
31.
Dimensions of the 6 n load •• . . . . . . . . . Dimensions of the gas chamber .••••••••
Top view of the SDS with dimensions •••••.
Dimensions of the SDS (Side View) . . . . . . .
viii
71
72
73
74
CHAPTER I
INTRODUCTION
High current, high voltage switches are used in many
pulsed power applications, such as lightning, EMP and
pulsed radiation simulation. Railgaps are frequently used
as the primary switch; however, these types of switches
suffer from inherent design limitations. Typically, a
transmission line provides the low inductance path to the
load, with a railgap incorporated into the transmission
line. The electrodes of the railgap are usually located
off the surface of the transmission line, thus adding
inductance and impedance mismatch to the line. The
inductance of the system increases proportionally to the
area added by the raised electrodes and inversely to the
number of channels which carry the current. The added
. d 1. . t th . di d 1. bl t th 1 d 1n uctance 1m1 s e max1mum dt e 1vera e o e oa .
Since a surface discharge switch (SDS) actually uses the
dielectric of the transmission line as the switching
surface, the surface discharge switch geometry does not
present any appreciable inductance to the system, if it
has a sufficient number of simultaneous discharge channels
sharing the current.
1
2
Historically, surface discharges have been used as
light sources, laser preionizers, and plasma cathodes [1,
2,3,4]. Only recently have these surface discharges been
widely employed as fast, high current switches [5]. When
used as the main switch, the SDS is usually pulse charged
and triggered before breakdown occurs [6]. The pulse
charging rapidly overvol tages the switch, thus enhancing
dense multichanneling and providing the low jitter
necessary for many high current, low inductance switching
applications.
To date,
available on
D.C. charged
very little detailed information has been
D.C. charged surface discharge switches.
switches have been limited in application
because they did not afford desirable switching properties
(low jitter, low prefire). On the other hand, for repeti
tive operation, a high current, high voltage switch must
recover its voltage holdoff quickly, withstand the
reapplied charge voltage without pre firing, trigger
reliably on a repetitive basis, and maintain reproducible
voltage and current characteristics throughout the life of
the switch. The voltage holdoff characteristics of the
insulator in the surface discharge switch should remain
constant after the multichannel flashover of the switch,
which can alter the insulator behavior. During the
discharge, thermal shock, UV irradiation, high temperature
gradients, chemical degradation of the surface, carboni
zation, surface erosion and metal vapor deposition onto
3
the surface of the insulator can contribute to undesirable
insulator modifications. Therefore, a study was under-
taken to examine the operating characteristics of a D.C.
charged, triggered SDS under varying conditions.
The operating characteristics and the basic inter
action of the arc with different dielectrics were studied.
A 45 kV, 3 kA switch was built on a 1 m long, 6 ohm trans
mission line, thus allowing evaluation of the switch per
formance (Figure 1) . Under all possible trigger and
charge voltage combinations three radically different
dielectrics (G-10, Delrin, Lucite) were tested. G-10, an
epoxy laminate, is unlike Delrin, which is a polyacetal,
and Lucite, which is a polymethyl methacrylate. The best
voltage combination (negative charging voltage, positive
trigger) and the best dielectric (G-10) were tested in
various gases, including helium, nitrogen, argon, SF6 , and
air, as well as various mixtures of each. All tests were
conducted at 1 atm STP. The operation of the switch was
diagnosed by voltage, current and jitter measurements
under the aforementioned optimum voltage and current
arrangement in air. Streak photography was used to
examine the arc development at the dielectric interface
and to examine the simultaneity of closure times of the
multichannel arcs.
Finally, degradation of the surface of the dielec
tric was examined using ESCA (electron spectroscopy for
c ::pp
.~,.
~
POW
ER
SU
PP
LY
6 .n
~
TRIG
GE
RE
D
1 1
TRA
NS
MIS
SIO
N
SUR
FAC
E LI
NE
D
ISC
HA
RG
E
6ll
LOA
D
PULS
ER
25
kV-
80
kV
RE
VE
RS
IBLE
PO
LAR
ITY
FA
ST
RIS
ETI
ME
(I
OkV
/ns)
PU
LSE
R
50kV
FIX
ED
O
UTP
UT
SW
ITC
H
(!) z 2:
TRIG
GE
R
i=
LOW
V
OLT
AG
E
TIM
ING
AN
D
CO
UN
TIN
G C
IRC
UIT
S
(I)
0 ._ (I)
0 z ~
0
DIA
GN
OS
TIC
S
VO
LTA
GE
M
EA
SU
RE
ME
NTS
C
UR
RE
NT
ME
AS
UR
EM
EN
TS
STA
TIC
SU
RFA
CE
CH
AR
GE
ME
AS
UR
EM
EN
TS
TIM
E
INTE
GR
ATE
D P
HO
TOG
RAP
HY
SUR
FAC
E A
NA
LYS
IS
( S
EM
,ES
CA
,SU
RF
AC
E R
ES
ISTA
NC
E)
ST
RE
AK
P
HO
TOG
RA
PH
Y
Fig
ure
1
. B
lock
D
iag
ram
of
the
Su
rface
Dis
ch
arg
e
Sw
itch
S
yst
em
an
d D
iag
no
sti
cs.
,j::>
.
5
chemical analysis), SEM (scanning electron microscopy),
surface charge measurements and surface resistance
measurements. The results of the aforementioned program
are reported in the following sections.
CHAPTER II
OVERVIEW OF THE SYSTEM
Circuit Operation and Design
As previously mentioned, a 45 kV surface discharge
switch was constructed using a 1 m long, 20.3 em wide
stripline (Figure 2) (see the Appendix for dimensions).
This was actually a modified version of a SDS used in the
overvoltaged mode [7]. Both the single channel /and fi-
multichannel electrodes (Figure 3) were easily adjustable
to various spacings and were constructed of brass. A blue
nylon dielectric, 0.63 em thick and 30.5 em wide, provided
the insulation between the charged portion of the strip-
line and the ground return side. ,A slot, 33 em long and
0.16 em deep, was milled into the blue nylon line. A G-10
insert, placed in the slot, isolates the main trigger
electrode from the switch electrodes. This main trigger
electrode consisted of a 2.54 em wide strip of 2 mil thick
copper tape which was placed between the G-10 insert and
the blue nylon (Figure 2) line. During the main course of /.
the study, one edge of the main trigger electrode was
located 3 em from the charged electrode, and the other
edge 0. 66 em from the other electrode. The electrode
spacing was set at 6.2 em to give self-break voltages
6
CH
ARG
ED
SID
E
BA
KE
LITE
..,.
a
CLA
MP
S B
LUE
N
YLO
N
DIE
LEC
TRIC
S
AM
PLE
LUC
ITE
CH
AM
BE
R
MA
IN
ELE
CTR
OD
ES
GR
OU
ND
RE
TUR
N
G-1
0 IN
SE
RT
Fig
ure
2
. O
utl
ine
Dra
win
g o
f th
e
Su
rface
Dis
ch
arg
e
Sw
itch
. '-
.)
A. Single channel electrodes.
B. Multichannel electrodes.
Figure 3. Photographs of Single Channel and Multichannel Electrodes.
8
9
between 45 and 55 kV (dependent on the trigger and charg
ing polarities) [8]. For the 6.2 em electrode separation,
an insulator sample with dimensions 0.16 em x 30.5 em x
12 em was held in place by the electrodes. In addition to
the main trigger, an auxiliary electrode was placed on the
surface of the sample, at its edge. The auxiliary elect-
rode provided additional field distortion and UV preil-
lumination. The auxiliary electrode was a 0.63 em wide,
2.5 em long strip of 2 mil copper tape. This auxiliary
electrode was placed 5.2 em from the edge of the sample
and 2. 54 em from the grounded electrode. In this posi-
tion, an arc occurs between the auxiliary trigger elect-
rode and the main electrodes.
{-;(>j
/c<V ~
Jel . ..-J)"f System Parameters
The energy was stored in four parallel, 25 ohm cables
which are attached to the 6 ohm transmission line (Figure
4) • The four cables are D.C. charged through a 2 Mn
resistor. When discharged, the 20 m long lines provided a
200 ns wide pulse into a matched 6 ohm load. The dis-
tributed capacitance of the four cables is .0174 ~F,
producing a 13.9 Joule discharge, when charged to 40
kilovolts. When a trigger pulse was applied to the main
trigger electrode, the switch broke down, discharging the
stored energy into the 6 ohm load. The load consisted of
six each, 1 ohm ceramic disk resistors placed in series
(see the Appendix for dimensions). The SDS, the load
'
FO
UR
-25n
TR
AN
SMIS
SIO
N L
INES
2M
n-
POW
ER S
l.PPL
.Y
_0-1
00 k
V
--(7
/ /1
(. ~"<
..(__
..::
:,
lA
,',
j
CU
RR
ENT
/MO
NIT
OR
EL
ECTR
OD
ES
].)r.l
ckt.j<
-:;
?v /¥
~,
\ I
~____,
~. "
~
AU
XIL
IAR
Y.,/
".
TRIG
GER
2M
n
10000~~
40
00
pF
f/
30
kV
CO
NTR
OL
CIR
CU
ITS
CABL
E P
ll..S
ER
I
I
25
-IO
OkV
~
Gn
L~D
CAPA
CITI
VE
PRO
BES
6 M
ETER
S 50
0HM
CAB
LE
( RG
-19
)
Fig
ure
4
. S
ch
em
ati
c o
f th
e
Ex
peri
men
tal
Arr
an
gem
en
t.
1--'
0
11
and the connection of the four transmission lines to the
main transmission line are shown in Figure 5. The load was
designed for rep-rate operation and was capable of handling
4 kW of average power. The equivalent circuit of the load
is shown in Figure 6A. The time varying inductance and the
time varying resistance are shown in order to emphasize the
importance of the dimensions in pulsed power loads. The
effects of these were checked by considering the diffusion
width of the current into the load. The skin depth factor
(1/e penetration depth) is given by:
-a 0- -w~a
( 1 )
where w is 2n times the frequency, ~ is the permeability of
free space (4n x 10-7 H/m) and a is the conductivity of the
resistors. For the risetime of the system, the frequency
used in the calculation was approximated as f = 10 MHz and
the conductivity of the ceramic-carbon matrix as a =
.692 mhos/meter. This give a o = .19 m. Since the diameter
of the resistors is .11 m, which is smaller than o, L(t) and
R(t) are approximately constant. The capacitance CL across
the load results from the connecting plate and is approxi-
mately 200 pF. The inductance LL of the load was calculated
to be 200 nH. The inductance was measured at 1000 Hz and
found to be 220 nH.
Fig
ure
5
. P
ho
tog
rap
h o
f th
e
Su
rface
Dis
ch
arg
e
Sw
itch
.
1--'
"->
13
L(t) ~ 0 R(t) ~ 0
-----~
LL = 220 nH
CL = 200 pF
A. Equivalent circuit of the 6 Q load.
20 Mil 6 METERS
50n CABLE
o-4o kV REVER SIB
POLARITY
RG-19
3 CAPACITORS 2nF 40 kV
~--~~~HH~---~on
B. Schematic of the trigger generator.
Figure 6. Equivalent Circuit of the 6 ohm Load and a Schematic of the Trigger Generator.
PULSER 50kV
14
Trigger Generator
A fast cable pulser was constructed to produce the
fast dV/dt (10 kV/ns) pulse necessary for the multichannel
operation of the switch. The design of the pulser also
allowed the polarity and amplitude of the trigger pulse to
be varied for easy testing of the switch under different
conditions. The schematic of the pulser is shown in
Figure 6B. The 20 M~ charging resistor and the four 50
M ~ bias
flashover.
resistors were immersed in oil to prevent
Likewise, to meet the fast dV/dt requirements
of the spark gap, a commercially available, triggered gap
(Pulsar, model T-670) was mounted to minimize the induct
ance in the return current path. A 6 m long, 50 ohm
(RG-19) cable was charged to approximately 36 kV and then
discharged to ground to obtain the 10 kV/ns pulse risetime
needed to trigger the surface discharge switch. The spark
gap, a mid-plane, field distortion, triggered gap, is
triggered with a 50 kV pulse from a low energy pulser
(Atlantic-Pacific, model PT-55).
CHAPTER III
DIAGNOSTICS
Diagnostics, one of the most important aspects in the
design of an experiment, play a key role in understanding
and evaluating the physics and performance of an experi
ment. An interactive diagnostic program of on-line and
off-line diagnostics proved invaluable in the studies
(Figure 1). On-line (time resolved, active) diagnostics,
used during the operation of the experiment, measured
voltage and current risetimes, waveshapes, and the jitter
of the switch. Time integrated photography, along with
streak and framing photography, allowed streamer (channel
formation positions) and simultaneity of the closure times
to be assessed. Off-line (passive) diagnostics used under
static conditions (after exposure of the dielectric to a
selected number of shots) , evaluated the charging
characteristics of the dielectrics, surface degradation
(chemical and vapor deposition), and surface resistance
after exposure to flashover. Surface charge measurements
consisted of two types; first, a static measurement where
the insulator was dusted with an electrostatic powder
which attached to only positive charges; and second, a
15
16
field meter which measured the residual surface charge
decay over several hours. Surface chemical analysis
techniques, such as ESCA and SEM, analyzed the surface for
carbonization and chemical vapor deposits. Surface
resistance measurements were made with a deposited
metallic pattern. A voltage was applied to the deposited
pattern on the sample. The current was then measured and
the surface resistance calculated.
Voltage and Current Measurements
Voltage probes typically have resistive, capacitive,
or a combination of both resistive and capacitive
elements. Capacitive voltage probes, built into the
transmission line, were used to determine val tage rise
times and amplitudes on the trigger and across the load.
The capacitive probes consisted of thin brass sheets ( 1
mil thick) separated from the ground return by 3 mil thick
Mylar. The BNC feedthroughs incorporated into the line
isolated the output of the probes from the ground return
and allowed fast resistive dividers to be mounted under
the return side. The resistive feedthroughs (voltage
divider) (Figure 7) prevent the lower capacitor c2 from
discharging too fast and provide additional attenuation.
The voltage dividers were constructed using 2 Watt,
noninductive, carbon resistors and 3/4" copper tube
housings. The resistive dividers had a 2 ns (10% - 90%)
HIGH VOLTAGE ---T---
I I I I
_L c,
A. Schematic of the voltage probes.
B. Photograph of the SDS and the resistive feedthroughs (below the surface).
Figure 7. Schematic and Photograph of the Voltage Dividers and Feedthroughs.
17
18
risetime and the overall divider network had a 5 ns (10%-
90%) risetime (fastest in-situ calibration pulse avail
able). The capacitive probe monitoring the trigger pulse
was used to determine jitter. The output of the voltage
probe monitoring the trigger strip was used to trigger the
external input of an oscilloscope, allowing multiple
traces of the load voltage riset imes to be compared.
The current was monitored with a PEARSON coil mounted on
one of the 25 ohm transmission line return lines.
Photographic Techniques
Time integrated (open shutter) photography and streak
photography were used to study the spread of arc closure
times, arc development along the surface and the average
number of channels per shot. The framing and streak
measurements used a TRW lD camera with additional optics
(imaging lenses, mirrors, etc.). In the streak mode, a
slit selects a portion of the gap and views this area for
a selected period of time (Figure 8). The position of
these slits are shown on Figure 9. Framing photographs
taken from above for a 200 ns intervals show that luminous
predischarge currents were present. The result of the
measurements are discussed in more detail in subsequent
chapters.
IMAGING MIRROR LENS
#T-- --0----~-- ---: I : I SLIT
SURFACE DISCHARGE SWITCH
STREAK CAMERA
A. Top view for determining the multichannel closure times.
IMAGING LENS
~~---0-----~------SURFACE SLIT
DISCHARGE SWITCH
STREAK CAMERA
B. Side view for determining the arc development at the surface of the dielectric.
Figure 8. Streak Camera Arrangement.
19
CATHODE
CATHODE
I II 1 1 I
I'\L.JI
I ---'---
STREAK OF MULTICHANNELS
SURFACE DISCHARGE SWITCH
TIME
A. Top View
SLIT
B. Side View
(SINGLE CHANNEL ELECTRODE)
ANODE STREAK OF ARC
SURFACE DISCHARGE SWITCH
TIME
SURFACE
Figure 9. View of Slit for Multichannel and Single Channel Development Measurements.
20
21
Surface Analysis
Surface charge measurements, SEM, ESCA, and surface
resistance measurements were made to evaluate the inter
action of the arc and the insulator surface [ 8,10,11].
The samples were first cleaned with cyclohexane and then
inserted into the switch. After a predetermined number of
conditioning
the trigger
shots, the surface charge decay times over
and the uniformity of charge across the
surface were examined with a commercial fieldmeter (Monroe,
Model 171). The decay measurements were made over several
hours time (up to 24 hours) by means of a chart recorder.
During the measurements, the samples were left in the
switch and the field probe placed over them. Before the
measurements, the main electrodes were removed. Because
the probe lacked good spatial resolution, another means of
checking the charge patterns left on the insulator was
developed. An electrostatic powder (Xerox, yellow toner
powder #6500) was dusted onto the surface of the insulator
and the pat tern photographed. The sharp detai 1 and the
distinct contrast of the pattern revealed the distinctive
difference in the way the insulators retain charge. For
example, on G-10, the field meter detected a fairly
uniform charge distribution. Upon dusting with the
powder, a striated pattern was revealed. Unlike the other
two dielectrics (Lucite and Delrin), the striations were
quite dense and distinct.
CHAPTER IV
TRIGGERING
Triggering of SDS's is complicated by both the
(dielectric) as well as its geometry. S\vitch substrate
Various methods of triggering have been employed by other
investigators, including field distortion, and photo
induced triggering [12, 13, 14]. Field distortion, the
most popular method, is complicated primarily by
geometrical considerations and by surface charging of the
dielectric used as the switching substrate. The most
widely reported type of field distortion triggering
employs an electrode which is placed on the surface of the
sample, and biased at an equipotential level [ 14] • The
surface trigger electrode has several inherent drawbacks
which include erosion and the necessity for careful
alignment, as well as additional vapor deposition onto the
dielectric. The other reported form of field distortion
uses a trigger electrode located below the surface of the
sample, which is
switching surface
isolated from the main electrodes and
[ 15] . Several methods of triggering
were explored during the course of study. The results of
the study are reported herein.
22
23
Laser triggering was tried with single channel
(shaped) electrodes, placed 4 em apart in one atmosphere
of ambient air. A one megawatt pulse from a nitrogen
laser was focused onto the surface of the switch from
above. Even at 90% of selfbreak voltage, no triggering
was achieved. When the laser beam was focused onto the
electrodes (either cathode or anode), visible sparks off
the electrode surfaces also failed to trigger the switch.
These results may indicate that higher energy levels would
be needed for successful laser triggering or that laser
triggering is not practical for surface discharge
switches. It is also likely that the length of the laser
pulse was too short (20 ns). Longer laser pulses, which
are on the order of the discharge growth time, are prob
ably needed to trigger surface discharges. Likewise, a
selected laser wave length corresponding to a resonant
frequency of the dielectric and the surrounding gas medium
may be required to trigger a switch of this type success
fully.
Two types of field distortion triggering were also
investigated. The first method employed various geome
tries of trigger electrodes placed on the surface of the
switch. Included among the electrode shapes were rods and
metal strips placed at various positions (mid plane, 2/3
point, etc., and biased) on the surface. Although the
multichannel electrodes (Figure 3B) were employed,
24
multichannel ing did not occur for any of the electrode
trigger polarity combinations. The final and most
successful trigger scheme also utilized field distortion.
In this case, the main trigger is isolated from the
discharge and the main electrode by the G-10 insert, as
previously described (Figure 2).
Jitter, as well as channel formation, is dependent on
the risetime of the trigger pulse, the energy coupled into
the discharge by the trigger and the polarity of the
trigger. These factors are dramatically affected by the
switching medium below which the trigger is located, the
positioning of the trigger, and the width of the trigger
strip. Various locations for the isolated trigger were
tried (Figure 10).
Three dielectrics were tried during the tests (G-10,
Delrin, Lucite). The optimum dielectric (in terms of
multichanneling performance and jitter) along with the
optimum polarity of charging voltage and trigger pulse
polarity (G-10, negative charging, and positive trigger
voltages) were utilized to determine the optimum trigger
location. With the optimum conditions fixed, the trigger
width and positions were varied. As the unbiased trigger
strip (2.54 em width) was moved toward the charged elec
trode, the hold-off voltage of the gap decreased quite
rapidly. Likewise, the multichanneling of the switch
~
T ~
A.
Tri
gg
er
ele
ctr
od
e,
.64
em
w
ide
was
lo
cate
d at
the
mid
po
int.
~ .~
C.
Tri
gg
er
ele
ctr
od
e,
2.5
4
em
wid
e
was
lo
cate
d
3 em
fr
om
th
e
ch
arg
ed
ele
ctr
od
e
(mo
st
su
ccessfu
l).
~
-,-
.--.
a
B.
Tri
gg
er
ele
ctr
od
e,
2.5
4
em
wid
e
was
lo
cate
d at
the
mid
po
int .
~
.,......
. D
. T
rig
ger
ele
ctr
od
e,
2.5
4
em
wid
e
was
lo
cate
d
un
der
the
gro
un
ded
ele
ctr
od
e.
Fig
ure
1
0.
Tri
al
Lo
cati
on
s o
f th
e T
rig
ger
Ele
ctr
od
e.
1\.
)
Ul.
26
decreased and triggering became erratic (Figure lOB). At
the midpoint, the width of the strip was narrowed to
0.63 em. The multichanneling of the switch also decreased
and triggering became unreliable (Figure lOA). The width
of the trigger strip was once again increased to 2.54 em
and the strip placed 3 em from the charged electrode in
the 6.2 em gap (Figure lOC). This seemed to be the most
optimum position in terms of jitter, triggering and
multichanneling. One other position was tried. The 2.54
em wide strip was moved under the uncharged electrode
(Figure lOD). In this case, triggering was not success
ful. During the tests, the selfbreak voltage varied with
the trigger electrode position. At the best position, the
selfbreak voltage was approximately 55 kV. With the
charge voltage held at -40 kV, the best position of the
trigger seemed to be 3 em from the charged electrode
(Figure lOC). At this position and charging voltage
level, the best results in terms of multichanneling and
jitter were obtained. The jitter (5 shots) in this posi
tion was 5 ns and the average number of channels was 8-9
per shot (with an auxiliary trigger added) when employing
G-10. The material used as the switching substrate, as
well as the voltage polarities, had dramatic effects on
the voltage and current characteristics of the switch.
These results, along with the surface charging character
istics, are reported in later chapters. The charging of
27
the dielectric has been hypothesized to be the principal
mechanism which contributes to the dense multichanneling.
However, no theory has been developed to account for the
dramatic differences seen in materials which have the same
charge polarity, but have dramatically different multi
channeling characteristics. The charging of the dielec
tric surfaces probably provides localized field enhance
ment due to anisotropic charging.
Once the optimum trigger position, and trigger and
charging voltage polarities, were determined, modeling of
the equipotentials in the switch were done on the JASON
code at LLNL. All computer simulations were done assuming
ideal conditions; no surface charging, no degradation of
the trigger pulse and no surface irregularities. The
first computer output (Figure llA) is a simulation of the
equipotentials in a surface discharge switch without a
trigger strip added. Figure llB is the same switch
geometry, but with a trigger electrode added. All these
simulations assumed that a -40 kV charging voltage is
applied to the left hand electrode and that the trigger is
held at ground potential until a +50 kV pulse is applied
to the trigger. Figure 12A shows the equipotential lines
with a trigger pulse applied. When a trigger is applied,
the fields at the surface of the insulator are increased
at three different points. The fields at the edges of the
8.00
e.oo
>- ... oo
2.00
o.oo
8.00
e.oo
>- ... 00
2.00
o.oo
8 8 8 8 8 8 8 8 8 8 0 ('f "" 00 10 0 ('I "" 10 co .- - - -
X
A. Equipotential plot without the trigger added.
8 8 8 8 8 8 8 8 8 0 ('I "" 00 co 0 ('I "" 10 .- -
X
B. Equipotential plot with a grounded triggger electrode added.
8 co -
Figure 11. Jason Computer Code Plot of Equipotentials With and Without a Grounded Trigger Electrode Inserted.
28
8 0 l:'t
8 0 ('f
8.00
8.00
>- -4 .oo
2.00
o.oo
8 0
6.00
4.00
2.00
0.00 0 C! 0
29
8 8 8 8 8 8 8 8 8 8 N .. "'
., 0 <'I .. ~ ., 0
~ ~ .... ~ .... N
X
A. Equipotential plot with a 50 kV pulse applied to the trigger electrode, with a -40 kV charge voltage applied.
-40 kV
0 0 0 0 0 0 0 C! 0 C! C! C! C! C! N ~ CQ CQ 0 N "'d' - - -
B. Sample field line derived from the above equipotential plots.
0 0 0 C! C! C! CQ CQ 0 - - N
Figure 12. Jason Computer Code Plot of the Equipotentials With a 50 kV Trigger Pulse Applied to the Trigger Electrode and One Resulting E-field Line.
30
main electrodes and at the edge of the trigger are
increased. Note that this plot does not actually show the
real fields present during the operation of the switch,
but does give an approximation of the initial conditions.
The surface charging, which is anisotropic in nature,
radically changes the tangential and normal components of
the E-fields. Figure 128 shows one sample field line
derived from the equipotential plot of Figure 12A. For
the optimum polarities (negative charging voltage and
positive trigger), the electrons are pulled toward the
surface of the insulator. When the polarities of the
charging and trigger voltages were reversed, erratic
behavior of the switch resulted, indicating that this may
be an important consideration when designing a triggered
SDS's (reported further under polarity and dielectric
effects).
The simultaneity in closure times of the channels
after triggering were determined with streak photography,
as previously mentioned in Ch. III. Figure l3A shows the
simultaneity in closure of the channels near the cathode.
Due to the nonlinearity (H-D curve) of the film, an exact
time spread can not be determined. However, the record on
most of the streak photographs indicated that the spread
in the development times is between 30-60 ns. This may
indicate that the arcs are resistive phase dominated and
that the spread in development times may be due to the low
A. Streak (top view) of the multichannel arc closure near the cathode, 500 ns = 5 em.
t ~
B. Side view of a single arc developing at the surface of a G-10 sample, 500 ns = 5 ern.
Figure 13. Streak Photographs Showing Sirnultaniety of the Multichannel Closure Near the Cathode and Single Channel Arc Development at the Surface.
31
32
E-fields or low current per arc channel present in the
switch. Near the anode, however, the spread in
development time is about 5-20 ns. This may indicate that
the arc is cathode directed,thus placing the trigger near
the anode may facilitate triggering which is stabler in
terms of jitter.
Another aspect of the arc development was explored
using streak photography. With the single channel elec
trodes in place, the arc development at the surface of the
dielectric was checked by a sideview (Figure 13B) . The
arc is seen to move up off the surface during its develop
ment. Figure 13B shows the arc movement off the surface
(lower half of image) with respect to its reflection from
the surface. When measured at 2. 8 kA to 3 kA, the arc
moves about .3 rnrn off the dielectric. At high currents,
this may present a problem if the current is not divided
evenly between many channels. The arcs, if they push too
far off the surface, may extinguish themselves, resulting
in higher resistive and inductive losses in the switch,
but perhaps lower dielectric erosion and higher repro
ducibility.
The predischarge current is quite important in a
switch of this type. If predischarge current was present
before triggering, reproducibility and hold-off character
istics would be affected. Using long exposure, 200 ns in
length, triggered before the main rise of the current,
predischarge luminosity was (Figure 14) observed. This is
A
N
0 D
E
c A
T
H
0 D
E
c A
T
H
0 D
E
A.
Pre
dis
ch
arg
e
lum
ino
sity
, 2
00
n
s B
. O
pen
sh
utt
er
ph
oto
gra
ph
o
f ex
po
sure
, te
rmin
ati
ng
20
n
s m
ain
d
isch
arg
e.
befo
re m
ain
cu
rren
t ri
se.
Fig
ure
1
4.
Op
en S
hu
tter
Ph
oto
gra
ph
o
f M
ult
ich
an
nels
an
d
Fra
min
g
Ph
oto
gra
ph
o
f P
red
isch
arg
e
Lu
min
osi
ty.
w
w
34
not surprising when compared with the long formative time
delays present in SDS's (100-200 ns). The development of
the luminosity near the cathode was compared against the
development of the luminosity near the anode. The lumi
nosity near the anode tended to appear 100-200 ns earlier
than the luminosity near the cathode. This indicates that
the discharge is cathode directed.
CHAPTER V
RESULTS
Polarity and Dielectric Effects
Three dielectrics, G-10, Delrin and Lucite, were
tested in dry air under all possible charging and trigger
polarities. Along with these dielectrics, teflon, poly-
ethylene, mica paper and alumina bonded to G-10 were
tested for multichanneling capabilities. Results for the
three principal dielectrics are given in terms of multi-
channeling characteristics, di (at 50% of peak) and dV (at dt dt
50% of peak and jitter measurements). Initially the rise-
time of the voltage and current waveforms were compared
[8]. After reevaluation of the data, the risetimes (10% -
90%) were found not to be a good comparison due to the
differences in wave shapes in the different cases. The
differences in peak voltage, peak current, and varying wave
shapes all resulted from the varying number of channels in
each case. The varying number of channels in multichannel
switches also affect the inductive and resistive phase
time evaluation of the arcs. In switches of the type
investigated and at the level of currents used, the arc may
or may not leave the resistive phase. Other investigators
have seen this effect and use comparisons other
35
36
than the risetime of the voltage and current [9]. Through
out the tests, a 50 kV ( 10% - 90%) trigger voltage was
used while the charging voltage was kept constant at 40
kV. The isolated (main) trigger electrode edge was
positioned 3 em from the charged electrode. The repeti-
tion rate was l-2 pps throughout the tests.
G-10, an epoxy laminate, had the best characteristics
(in terms of jitter and multichanneling) of the three rna-
terials tested (Table 1). The first case tested was with
a negative charging voltage and positive trigger. The
voltage falltime was 70 ns and the current falltime 70 ns
(Figure 15). The jitter of an initial five shots was 16
ns, however, after 500 conditioning shots, the jitter
decreased to 5 ns. The characteristic of the voltage
falltime also changed during the first 500 shots (Figure
16). The change in the wave shape may be due to the
additional channels. The additional number of channels
would decrease the discharge inductance, and add more high
frequency components to the pulse (wave shape). However,
di dV the effect of this is not clear either from the dt or dt
values .in Table l. Multichanneling characteristics also
changed as the G-10 eroded away. Initially, about 5
channels per shot were visible; after 500 shots, 8-9
channels were typical (Figure 17). All of the other
electrode polarity combinations gave considerably poorer
performance. For the case of positive charging and
I 2 3 4
CH
AR
GE
VO
LTA
GE
TA
BL
E
1.
Sw
itch
ing
C
hara
cte
risti
cs o
f G
-10
as
a F
un
cti
on
o
f T
rig
ger
an
d
Ch
arg
ing
V
olt
ag
e P
ola
rit
ies
TR
IGG
ER
(5
0% o
f P
eak
) (5
0% o
f P
eak
) JI
TT
ER
P
ULS
E
d!
X
10
10
dV
X
1
01
1
dt
TIM
ES
N
O.
OF
dt
PO
LA
RIT
Y
PO
LA
RIT
Y
(A/s
ec)
(vo
lts/
sec)
(5 S
hots
) C
HA
NN
EL
S
+
2.1
2
.5
Sn
s 8
+
+
2.7
3
.0
30
ns
2
2.2
2
.3
2 --
+
2.3
2
.8
lOO
ns
3
.
CO
MM
EN
TS
Pre
fir
es
Pre
fir
es
an
d
~1isfircs
' ' I w
-.
..]
A. Voltage across the 6 ~ load. 3340 V/div., 50 ns/div.
B. Current through the 6 ~ load. 1000 A/div., 50 ns/div.
Figure 15. Voltage and Current Characteristics of G-10 for a Negative 40 kV Charge Voltage and Positive 50 kV Trigger Voltage.
38
A. Overlay of 1st five shots. 3340 V/div. 1 20 ns/div.
B. Overlay of 1st five shots after 500 shots.3340 V/div. 1 20 ns/div.
Figure 16. Overlay of 5 Shots Showing the Effect of Conditioning on the Risetirne.
39
A. Open shutter photograph of the 1st shot.
B. Open shutter photograph of the 1st shot after 500 shots.
Figure 17. Time Integrated Photographs of Multichannels of G-10.
Anode {-
Electrode Position
t
Cathode
Anode
Electrode Position
t
Cathode
40
trigger voltages, the di dt
dV and dt were
41
higher than the
aforementioned case. In this case multichanneling was
poor, resulting in only 2 channels. . dV di The h1gher dt and dt
may result from the smaller number of arcs thus increasing
the switch's inductance. Since fewer arcs are carrying
current, the current per arc is higher. Other possible
conclusions may also be drawn from these data. In the
multichannel case where there is a large number of arcs,
the arcs may be resistive phase dominated, due to the
lower current per arc. In this case if the arcs were
resistive phase dominated the lower dV dt and di
dt could
result. However, reviewing the equivalent circuit for the
6 Ohm load, the reason for the differences are not clear
since the load is complex (R + jX). The jitter (5 shots)
for the positive charging and trigger voltage case dropped
from 140 ns to approximately 30 ns after the first 500
shots. Similiarly, mul tichanneling was nonexistent for
the negative charging, negative trigger voltage case. The
jitter was too great to measure with the sweep speeds used
on the oscilloscope. Characteristics of this case includ-
ed a large number of pre fires and lower selfbreak volt-
ages. The selfbreak voltage decreased from 45 to 36 kV
after 500 shots. In the last case of positive charging and
negative trigger voltage, the voltage and the current wave
shapes varied considerably from shot to shot. The jitter
was approximately 100 ns. Multichanneling decreased from
42
5 channels per shot to an average of 3 channels per shot
after the initial 500 shots. Prefires and misfires were
common for this case.
The performance of Delrin, a polyacetal, was similiar
to that of G-10 (Table 2). For the case of negative
charging and positive trigger voltage (Figure 18) the ~~
and the ~~ were slightly different from the similiar case
of G-10 (Figure 15). This may result from the fewer
number of channels or from a slight error introduced by
the interpretation of the oscilloscope trace. The jitter
was 15 ns in this case, slightly higher than the similiar
case of G-10 (Figure 15). Multichanneling remained stable
at 5 channels per shot for the entire run (Figure 19).
When the voltage and current waveforms, using Delrin, are
compared to the voltage and current characteristics of
G-10 (Figure 15), under the same conditions, a dramatic
difference is seen in the voltage wave shapes and rise-
times. However, di dV when the dt and dt values (at the 50%
point) are compared, they are within a few percent of each
other. The difference is mainly in the 70%-90% slope,
which also may indicate that in several cases the arc may
not be leaving the resistive phase as previously mention
ed. Although the current dependence has been mentioned,
one other factor also controls this. The electric field
present at the time of triggering also affects the way the
arc develops. The effect of different trigger charging
I 2 3 4
TA
BL
E
2.
Sw
itch
ing
C
hara
cte
risti
cs o
f D
elr
in as
a F
un
cti
on
o
f T
rig
ger
an
d
Ch
arg
ing
V
olt
ag
e P
ola
rit
ies
CH
AR
GE
T
RIG
GE
R
(50%
of
Pea
k)
(50%
of
Pea
k)
JIT
TE
R
NO
. V
OLT
AG
E
PU
LSE
d
! X
1
01
0
dV
X
lOll
T
IME
S
OF
PO
LA
RIT
Y
PO
LA
RIT
Y
dt
dt
(5 S
hots
) C
HA
NN
EL
S
(A/s
ec)
(vo
lts/
sec)
+
2.7
2
.4
15
ns
5
+
+
2.6
2
.1
1 --
2.8
2
.5
1 --
-
+
2.9
2
.2
1 --
~--~-
--
--
----
---
CO
MM
EN
TS
Pre
fir
es
Som
e P
re fir
es
"" w
A. Voltage across the 6 n load 3340 V/div., 50 ns/div.
B. Current through the 6 n load. 1000 A/div., 50 ns/div.
Figure 18. Voltage and Current Characteristics of Delrin for a Negative 40 kV Charge Voltage and a Positive 50 kV Trigger Pulse.
44
' Anode
+
Electrode Position
t
Cathode
Figure 19. Multichanneling Characteristics of Delrin for a Negative 40 kV Charge Voltage and a Positive 50 kV Trigger Pulse. (Arc channel to the far right is the arc formed between the main electrodes and the auxiliary electrode.)
45
46
polarity combinations is quite apparent when the case of
positive trigger and positive charging voltage (Figure 20)
are compared with the most successful case of negat~ve
charging voltage and positive trigger (Figure 18). The most
significant change seen
the dV dt
di and dt values
is in the wave shapes; however,
are relatively unchanged. One
explanation for the drastic wave shape changes is the final
arc phase present after triggering. If the surface charge
of Delrin differed drastically in magnitude or polarity, the
electric field needed to trigger the surface discharge would
differ from dielectric to dielectric. As discussed earlier,
if the electric field at triggering and during the discharge
is not sufficient to drive the arc out of the resistive
phase, the discharge remains resistive, which changes the
wave shape. The following chapter discusses the differences
in charge polarities and magnitudes which could account for
the different electrical characteristics of G-10, Delrin and
Lucite.
Results using Lucite, a polymethylmethacrylate
(PMMA), differed considerably from those using G-10 or
De 1 r i n ( Tab 1 e 3 ) • Once again, a negative charging voltage
and a positive trigger pulse gave the best overall re-
sults. The waveform of this case (Figure 21) compared
favorably with the simular cases for G-10 and De1rin
(Figure 15 and Figure 18, respectively). However, multi-
channeling decreased to only two channels for the Lucite
A.
B.
Figure 20.
Voltage across the 6 n load. 3340 V/div., 50 ns/div.
Current through the load. 1000 A/div., 50 ns/div.
Voltage and Current Characteristics of Delrin for a Positive 40 kV Charge Voltage and a Positive 50 kV Trigger Pulse.
47
I 2 3 4
TA
BL
E
3.
Sw
itch
ing
C
hara
cte
risti
cs o
f L
ucit
e as
a F
un
cti
on
o
f T
rig
ger
an
d
Ch
arg
ing
V
olt
ag
e P
ola
rit
ies
CH
AR
GE
TR
IGG
ER
(5
0% o
f P
eak
) (5
0% o
f P
eak
) JI
TT
ER
N
O.
VOLT
AG
E P
ULS
E
d!
X
101
0
dV
X lO
ll
TIM
ES
O
F P
OLA
RIT
Y
PO
LA
RIT
Y
dt
dt
(5 S
ho
ta)
CH
AN
NE
LS
(A/s
ec)
(vo
lts/
sec)
+
2.3
2
.1
40
ns
2
+
+
2.0
2
.0
1 -
2.2
2
.1
1 -
-
+
2.4
2
.0
1 -
CO
MM
EN
TS
No
Pre
fir
es
No
Pre
fire
s
~
00
A.
B.
Figure 21.
Voltage across the load. 3340 V/div., 50 ns/div.
Current through the load. 1000 A/div., 50 ns/div.
Voltage and Current Characteristics of Lucite for a Negative 40 kV Charge Voltage and a Positive 50 kV Trigger.
49
50
dielectric. Unlike the G-10 results, the jitter was too
great to measure (with the oscilloscope sweep speeds used)
for three of the charging trigger voltage combinations,
however, no prefires or misfires were observed.
In addition to the dielectrics mentioned above, tef
lon, bonded mica paper, polyethylene and alumina bonded to
a G-10 substrate were tested under the optimum conditions.
Teflon failed to multichannel and the triggering proved
erratic. Likewise, the mica paper had about 2 channels
per shot and triggering was unreliable. Polyethylene dif
fered from the other two materials. Polyethylene had
about 5-7 channels per shot initially; however, multichan
neling decreased after several hundred shots and the trig
ger reliability became quite poor. An alumina sheet, 0.076
em thick, glued on to G-10, multichanneled quite well with
an average of 5-7 channels per shot. However, the
alumina, being brittle, cracked after only 50 shots.
Surface Charging
Using techniques described in Chapter II, Diagnostics
(field meter measurements and dusting), the surface
charging of G-10, Delrin and Lucite were assessed.
Although previously reported for some surface discharge
switches, surface charging had never been examined in
detail [ 17] . The large residual fields present in the
switch differed from insulator to insulator. These dif
ferences in surface charging (e.g., polarity of charge,
51
magnitudes, and different patterns) probably account for
the differences in mul tichanneling characteristics pre
viously mentioned. The amount of charge present on the
surface modifies the discharge characteristics and the
delay times needed for the arc to develop at the surface
of the dielectric. The polarity of the charge and the
amount of charge present on the surface alter the time
needed for the arcs to form (space charge effects). The
charge pattern also determines, to a large extent, the
number of arcs which form and the position of formation.
If a continuous sheet of charge was present, it is likely
that only one arc would form. Subsequent arc formation
would be prevented due to rapid electric field collapse
throughout the entire length of the gap. Multichanneling
would then be prevented from the rapid voltage drop across
the one arc. However, if striated charging was present as
in the case of G-10, transient time isolation between arcs
would be provided by the localized field enhancement
regions. This transient time isolation would also allow a
spread in the formative development times as seen
previously in the streak photographs. Likewise, mottled
(nondescript) charging would inhibit arc formation due to
unenhanced field regions on oppositely charged regions
across the discharge gap. For example, the striated
charging may help to guide the path of the arc, whereas
the mottled charging may tend to also guide it, but in a
52
direction that does not enhance subsequent closure of the
arc channel. The different charge patterns, striated
versus mottled, would also change the field enhancement
factors (FEF) present at the surface during the operation
of the switch. FEF' s as high as fifty percent of the
local fields have been noted on insulator surfaces [18].
The fields measured on the insulators indicate that the
average residual fields are as high as 47 percent of the
applied field (after 500 shots in the SDS). The localized
field enhancements are probably much higher.
On G-10, the residual electric field was found to be
fairly uniform when the field probe was moved parallel to
line A-A in Figure A-3, but non-uniform along line B-B.
The field probe gives the average macroscopic field due to
charges spread over several square centimeters (probe is
approximately 1 ern above the surface) . Three different
samples subjected to 500 shots, each using the parallel
electrodes (multichannel), exhibited positive charging
with the residual field near the cathode ranging from 3.8
kV/crn to 5.7 kV/crn. The residual field above the trigger
electrode varied from 7.2 kV/cm to 14 kV/crn and the 1/e
decay times for these charges ranged from 3. 85 hours to
7.4 hours. Although the decay times differed for differ
ent samples, the general shape of the decay curve remained
approximately the same for each material type. Typical
decay curves are shown in Figure 22. Notice that after
kV/ /Cm
4
2
-4
-6
-8
-14
-16
LUCITE
2 3 4 5HOURS
DELRIN
Figure 22. Charge Decay Curves ofLucite, G-10, and Delrin.
53
54
all external fields were removed, the G-10 sample con
tinued to show an increasing residual field for the first
10-15 minutes. Figure 23 shows the "dust" pattern of a
G-10 sample subjected to 500 shots. The pattern exhibits
a great deal of structure (anisotropic charging) which
was, of course, not detected with the low resolution field
probe. Single channel discharges on G-10 produced the
dust patterns shown in Figure 24. The surface shown in
Figure 24A was subjected to 850 shots at a repetition rate
of 0.12 pps, while the one shown in Figure 24B was sub
jected to 1,000 shots at 1.3 pps. Experiments indicated
that the difference in the number of shots had no visible
effect, but samples run at the lower repetition rate
always displayed the more clearly defined pattern.
In the case of Delrin, the residual field along the
length of the dielectric varied only slightly and was
uniform across the width (line B-B, Figure A-3). The
measured field near the cathode ranged from -8.6 kV/cm to
-13.5 kV/cm for three samples, indicating negative charg
ing of the dielectric surface. For the three different
samples, the field above the trigger electrode varied from
-6.6 kV/cm to about -8 kV/cm. A typical decay curve for
Delrin, is also shown in Figure 22. It exhibits an in
creasing field for the first 30-40 minutes before charge
decay occurs. The 1/ e decay time in this case ranged
between 9.7 and 20 hours, significantly longer than in the
-+
An
od
e
+ +
Cath
od
e
Fig
ure
2
3.
Du
st P
att
ern
o
f a
G-1
0 D
iele
ctr
ic
Sh
ow
ing
th
e
Ch
arg
e
Patt
ern
Left
o
n
the
Su
rface
by
50
0
Sh
ots
(M
ult
ich
an
nel)
.
Ul
Ul
A. Single channel charge pattern left by 1000 shots at 1.3 pps.
B. Single channel charge pattern left by 850 shots at .12 pps.
Figure 24. Dust Patterns Showing Surface Charges Left by Single Channel Discharges Comparing High Rep-Rate and Low Rep-Rate Operation.
56
57
case of G-10. The "dusting" charge patterns on Delrin
subjected to single channel discharges were not as well
defined as the single channel discharge dust patterns on
G-10. In the multichannel case, the striated charge
regions were more widely spaced compared to the charges
left on G-10 which would account for the longer decay
times in the case of Delrin.
Lucite showed results which were quite different from
the G-10 and the Delrin. The residual field along the
length (line A-A) varied by relatively large amounts
(>> 1 kVIcm). The field near the cathode ranged from -3.8
kV I em to +1. 9 kV I em, and above the trigger electrode it
varied from -3.2 kVIcm to +2.4 kVIcm when the probe was
moved parallel to line A-A (Figure A-3) for three dif
ferent samples. Unlike Delrin and G-10, this dielectric
showed a long charge build-up time with no detectable
decay after fifteen hours (Figure 22). Several attempts
to study the spatial charge distribution by dusting did
not reveal any definite charge pattern.
Teflon, another dielectric which failed to multi
channel, was checked for charging patterns. Just as for
Lucite, only a mottled, nondescript pattern was observed.
Surface Analysis
surface analysis performed \'lith ESCA, SEl-l, surface
resistance and optical microscopy on G-10, Lucite and
Delrin provided an interesting insight into the processes
58
involved during the operation of the switch. Optical
microscopy revealed that the G-10 typically has a cracked,
rough surface compared to the other two polymers. After
several hundred shots, the G-10 surface was in fact
rougher than the virgin surface. The other polymer
surfaces tend to be smooth even at 450 x magnification,
except for the scratches present from handling. After
several hundred shots, the other polymer surfaces tended
to become polished.
Using ESCA, the flashed-over surfaces showed no
carbonization. The Lucite and G-10 showed no sign of
chemical degradation; however, Delrin showed signs of
permanent surface damage. No signs of vapor deposition
from evaporated electrode material was seen in the dis
charge region of the gap. However, toward the ends of the
gap, where no discharge was present, signs of metal
deposition were noted. SEM confirmed that no surface
irregularities were present on the surface of the flashed
over samples. However, surface resistance measurements
did reveal that the surface resistance of all the dielect
rics did drop by several orders of magnitude (Table 4).
Material
G-10
G-10
Delrin
Delrin
Lucite
Lucite
Lucite
TABLE 4
Comparison of Surface Resistances Before and After Exposure to Flashover
Surface No. of Shots Resistance (ohms)
Virgin 5.0 X 1012
4000 shots 7.5 X 10 7
Virgin 2.4 X 1015
500 shots 2.3 X 1013
Virgin 3.4 X 1013
500 shots 4.3 X 1010
4000 shots 4.3 X 1010
59
CHAPTER VI
GAS EFFECTS
Besides laboratory air
V), helium, nitrogen, SF6
,
these were tried under the
(results reported in Chapter
argon, co2 and mixtures of
"optimum" conditions ( G-1 0,
negative charging voltage and positive trigger). The
results of these tests are reported in Table 5.
Unlike the results when air was used with G-10, which
showed no visible signs of surface carbonization after
4,000 shots, most of the other gas mixtures caused
catastrophic damage after 5-6 shots. One hundred shots in
nitrogen showed light carbon tracks left on the surface;
however, rnultichanneling was still occurring at the end of
the run. The 80% N 2 + 20% SF6 mixture rnultichanneled as
indicated; however, after 5-6 shots, heavy carbonization
led to tracking and failure of the insulator (Figure 25) .
On epoxy insulators used in SF6 , this phenomenon had been
previously noted [19]. Similarly, the 20% N2 + 80% SF6
mixture formed a light carbon path on the surface and the
arc usually tracked along this path during the rest of the
run (Figure 26). The pure SF6 atmosphere also caused
tracking and failure of the insulator after only a few
shots. The argon + SF 6 mixtures behaved like the nitrogen
60
61
TABLE 5
List of Gases Employed in the Switching Experiment
Gas Type Vselfbreak No. of Channels Per Shot
Air 55 kV 8-9 channels
N2 32 kV 3-4 channels
SF 6 * 1 channel
20% SF6 + 80% N2 * 3-4 channels
20% N2 + 80% SF6 * 1 channel
Argon 6 kV 1 channel
He 6 kV diffuse discharge
95% He + 5% Air 6 kV diffuse discharge
95% He + 5% N2 6 kV diffuse discharge
80% Ar + 20% SF6 46 kV 2-3 channels
20% Ar + 80% SF 6 * 1 channel
10% Ar + 10% SF6 + 80% N2 * 1 channel
C02 * 1 channel
* The system's maximum charge voltage was limited to 55 kV therefore the selfbreak voltages were not measured. However in the cases shown, the switch did trigger. The pressure was kept constant at 1 atmosphere during the tests.
Anode
Cathode
Figure 25. Photograph of G-10 Exposed to 6 Shots in 80% N2 + 20% SF 6 .
62"
A N 0 D E
Figure 26. Photograph of G-10 Exposed to 6 Shots in 80% SF 6 + 20% N2 .
c A T H 0 D E
63
64
and SF6 mixtures. When the argon, co 2 and helium gases were
used, no damage was seen on the surface even after 100
shots. The helium mixtures all gave diffuse discharges when
run at 5 kV with a 50 kV trigger applied. These discharges
appeared to be diffuse and seemed to have a high impedance
(Figure 27). Voltage and current measurements showed no
voltaqe drop across the load and no current flow through the
load at measurable levels. These diffuse type discharges
have been reported before in the 1 atmosphere range near
dielectric surfaces [20].
Figure 27. Photograph of Discharge in 5% Air + 95% He Mixture, Negative 5 kV Charge Voltage, 50 kV Trigger.
CHAPTER VII
CONCLUSIONS
In reviewing the data obtained, it is clear that many
mechanisms interact to affect the performance of a surface
discharge switch. The polarity of the charging and
trigger voltages seemed extremely important, as seen from
the results of G-10, Delrin, and Lucite. The striated
charging (dust) patterns of the most successful dielectric
candidates indicate that materials which display these
surface charging patterns are the ones to employ in an
actual switch. It seems likely that the striated surface
·charge patterns actually aid in forming multichannels,
since this does modify the normal and tangential fields of
both the applied electric field and the applied trigger
fields. One experiment which would verify this hypothesis
is to apply a substance to the surface of dielectric which
would cause it to charge in this anisotropic manner. When
compared to the dielectric without the applied substance,
it would clarify whether the charging pattern is a
symptomatic behavior or the cause of the multichanneling
behavior in these switches.
Further tests which should be run include the perfor
mance of the switch under high currents and higher
65
voltages.
66
Further exploration of the erosion of the
surface at these high currents should also be investi
gated. Although some erosion data have been reported, the
pulse wave shape and currents have measurable effects on
the amount of material actually eroded away [21, 22].
Also, the tests previously reported used relatively low
energy pulses. The jitter, voltage, and current charac
teristics and multichannel behavior may change measurably
at higher energies. Longer current discharge times may
also affect the multichannel behavior. If the arcs do not
carry the same currents on millisecond tirnescales, some of
the arcs in a multichannel discharge may extinguish.
However, the experiments performed with different
gases, polarities, and dielectrics, indicate that the
discharge, voltage, and current characteristics and the
jitter may be tailored to some degree. This unusual
property of the surface discharge switch promises to
provide a switch sui table for many of the high current,
high voltage applications of interest in pulsed power
technology.
1.
2.
3.
4.
5.
LIST OF REFERENCES
S. I: ~ndreev, "The Use of a Spark Di~9harge_Gor Obta1n1ng Intense Light Flashes of 10 - 10 Second Duration," Soviet Physics-Technical Physics, Vol. 7, No. 8, pp. 703-708, February 1963.
V. Yu Baranov, v. M. Borisov, A. M. Davidovskii, and 0. B. Khristoforov, "Use of a Discharge Over a Dielectric Surface for Preionization in Excimer Lasers," Sov. J. Quantum Electron, 11(1), pp. 42-45, January 1981.
V. Yu Baranov, v. M. Borisov, 0. B. Khristoforov, "Excimer Electric-Discharge Laser with Plasma Electrodes," Sov. J. Quantum Electron, 11(1), pp. 93-94, January 1981.
W. J. Sarjeant, "High-Pressure Surface Discharge Plasma Switches," IEEE Trans on Plasma Science, Vol. PS-8, No. 3, pp. 216-226, September 1980.
G. I. Belyaev, P. w. Dashuk, S. A. Kozak et al., "Switching of Megampere Impulse Currents by a Creeping Discharge Arrestor," Allerton Press, Inc., pp. 137-142, 1981.
6. H. M. Von Bergmann, "Triggered Multichannel Surface Spark Gaps," Journal of Physics, E, Vol. 15, No. 2, v
pp. 243-247, February 1982.
7. D. Johnson, M. Kristiansen, L. Hatfield, "Multichannel Surface Discharge Switch," Conference on Electrical Insulation and Dielectric Phenomena, 1982, p. 573, Amherst, MA.
8. R. Curry, M. Kristiansen, L. Hatfield, and V. K. Agarwal, "Surface Charging of Insulators in a Surface ~ Discharge Switch," Conference on Electrical Insula-tion and Dielectric Phenomena, 1983, Buck Hill Fall, PA.
9. Dr. Nunnally, private communication.
67
68
10. G. L. Jackson, "Pulsed Flashover of Solid Dielectrics in Vacuum and Gases," Ph.D. Thesis, Texas Tech University, pp. 138-141, December 1983.
11. M. Borger, "Surface Resistivity," Senior Project, Physics Dept. Texas Tech University, 1983.
12. A. G. Bedrin, v. E. Lavrentyuk, I. v. Podmoshenskii, and P. N. Rogovtsev, "Photoinduced surface discharge," Sov. Phys.-Tech Phys., Vol. 24, No. 10, pp. 1181-1185, October 1979.
13. S. T. Pai and J. P. Marton, "A Preliminary Study of the Breakdown Mechanism of Surface Discharge Switches," IEEE Fifteenth Modulator Symposium, pp. 153-159, Baltimore, HD., 1982.
14. Y. Saito, Taro Hino, and M. Suzuki, "A Dry Tracking Resistance Test Method Using a Trigger Discharge," IEEE Transactions on Electrical Insulation, Vol. EI-3, No. 1, pp. 18-23, February 1968.
15. S. I. Andreev, E. A. Zobov, and A. W. Sidorov, "Investigation of Sliding Spark in Air," Plenum Publishing Corporation, pp. 309-314, 1978.
16. E. L. Neau, "A Surface-Breakdown, l-lultichannel Air Gap Switch For Megampere Currents," Sandia National ~ Laboratory, Contract No. AT-(29-1)-789, March 1971.
17. J. c. ~lartin, private communication.
18. A. S. Pillai and R. Hackarn, "Modification of Electric Field at the Solid Insulator-Vacuum Interface Arising From Surface Charges on the Insulator," J. Appl. Phys., Vol. 54, No. 3, pp. 1302-1313, March 1983.
19. T. Suzuki, S. Nakayama, and T. Yoshirnitsu, "Degradation of Insulating Materials Including Sio 2 due to sF
6 Gas Dissociation Products," IEEE Trans. on
Electrical Insulation, Vol. EI-15, No. 1, pp. 53-58, February 1980.
20. Y. B. Golubovskii and R. Zonnenburg, "Discharge Contraction in Inert Gases. IV Results of Studies in Helium," sov. Phys. Tech. Phys., Vol. 25, No. 10,
21.
pp. 1220-1222, October 1980.
P. N. Dashuk, A. K. Zinchenko and M. D. Yarysheva, "Erosion of Dielectrics in the Switching of High-Pulsed Currents by a Grazing Discharge," Sov.
69
Phys. Tech. Phys., Vol. 26, No. 2, pp. 196-201, February 1981.
22. Y. Saito, F. Noto, "Test of Surface Electrical Failure of Organic Insulators by Means of Intermittent Discharges from a Capacitor," IEEE Trans. on Electrical Insulation, Vol. 1/2, pp. 2-8, March 1965.
APPENDIX
The dimensions of the switch's gas chamber, the SDS
and the load are shown in Figures A-1, A-2, A-3 and A-4.
The dimensions of the chamber were selected to fit the
switch dimensions (clamps not shown in Figure A-2) and to
insure that flashover on the walls did not occur. Like
wise, the load was designed for voltage hold off and a
rep-rate power rating of 4 kW. The resistors used in the
load were 40 kV hockey puck type, made by Stackpole
(Figure A-1).
The dimensions of the transmission line were selected
to allow operation up to 55 kV. All dimensions are in
centimeters. The Figures A-3 and A-4 show the switch
without clamps, thus allowing easier viewing. Clamps (not
shown) were made of bakelite, which allowed for easy
changing and maintenance.
70
~m.dia
CO
PP
ER
IN
SE
RT
S
\ L
OA
D
rR
ES
IST
OR
, Il
l ea
. -
_ ........
.... _,
._
r r
1---· f.
---~--· ~-
-··-~
~---_ __.
-·
~---
--. 1
---·
...
~ L
1 , i
l I.
Ocm
. 14
. Ocm
.
L :
-: I
0.6
35
em
. ~
SL
9T
S
r----
---
1----
_ ~-
~ ~ --
----
-· -·-
-
_..
. ...
..J
Fig
ure
2
8.
Dim
en
sio
ns
of
the
6 ~
load
. -..
.J t-'
~
---
----
----
----,
I I L -+- r I I
-------------
_j
~~---------
----
~3.8
7cu,
.~
20
.32
cm
• ~3
.87c
rn~
r-·-
----
----
----
--1
-_
r- 1 I I I I I I I
-, 1---
+-·
r I I I I I I I I
I 1.
27cm
l !
I
4
-.- 2.54
cm
.LT
em
6.
03cm
li
~
12.0
7cm
t---------i
I iJ
I
I I
I ...!
. PT
I ~
7.62
cm
I 4
I I
j I
1 •
I r-
--1
1.2
.7cm
I
L-.
....
J __
I
I +
I I
13.9
7cm
I I
~
.L
Fig
ure
2
9.
Dim
en
sio
ns
of
the
gas
ch
am
ber.
-...
.1 l'o
J
-Q)
-*~--- --- ----------------1 -r-eI I 1 T
~---~-------------- ___ J (0 ~ ~ ~ 0
5 Q) ---------------~--~~ ~ 0 rt)
(\J+
~(\J
f(O
73
. Ul c 0
·.-I Ul c Q)
E ·.-I 't1
..c +J ·.-I ~
Cf.l Q Cf.l
Q)
..c +J
4-1 0
~ (1)
·.-I :.> 0.. 0 8
. 0 C"")
(1)
"-! !j tJ'l
·.-I ~
I 0 N .. .... ,~~---!!
@
74
T
top related