High Power Ignitrons

COMPREHENSIVE STUDY OF HIGH POWER IGNITRONS
by
DIANA LYNN LOREE, B.S.E.E., M.S.E.E.
A DISSERTATION
IN
ELECTRICAL ENGINEERING
Submitted to the Graduate Faculty
of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of
DOCTOR OF PHILOSOPHY
Approved
December, 1991
'.^ r -
Sol
/o//?/'
' ^ , ACKNOWLEDGMENTS
^j ^ *> This degree is a culmination of many years of public and

collegiate education and there are many people who helped me
along the way. I would like to express my appreciation to
Drs. Hagler, Hatfield, and Trost for their input and for
serving on my committee. I also thank the fine group of
people at the Texas Tech University Pulsed Power Laboratory
and especially: Lonnie Stephenson, Kim Zinsmeyer, Donna
Srader, Marie Byrd, Sandra Branch, Gene Igel, and Danny
Garcia. Special mention must go to Anton ^^Tony" Shulski at
Richardson Electronics, who has supported all my endeavors
with his professional advice and who began this line of
research by means of the construction of our main
experimental tube. Funding and technical support
acknowledgments are graciously extended to the capable group
of people at Lawrence Livermore National Laboratory including
Wayne Hofer, Dave Cummings, Ron Kihara, and Doug Larson.
Further appreciation is extended to the Texas Tech University
Research Foundation. My highest respect and appreciation
goes to Drs. Giesselmann, Krompholz, and Kristiansen. They
have taught me many things over the last five years, much of
it having to do with knowledge, responsibility, quality, and
interaction with people of all kinds and stations. I would
also like to thank some of my previous teachers without whom
I would have never made it this far: Mrs. Warner, Mr. Brown,
Mrs. Cooper, Ms. Brillhardt, Mr. Hittinger, Mrs. Jones, and
especially Mr. Laramore. I additionally thank my parents who
have put up with my many years of being a student. I also
thank Nella and Harry Loree, who have quickly become most
dear to me and who have seen me through the years of work
with love, support, and attention. Lastly, and most
important of all, I thank my husband Ellis who met me at the
beginning of my graduate career and has been understanding
through the tests, finals, preliminaries, projects, and
reports that have frequently taken up all of my attention.
11
CONTENTS
ACKNOWLEDGMENTS ii
ABSTRACT iv
LIST OF TABLES v
LIST OF FIGURES vi
CHAPTER
I. DESCRIPTION OF PRELIMINARY WORK 1
Ignitron Basics 1
Prototype Ignitrons 5
Axial Magnetic Field Effects 7
Other Work 13
II. EXPERIMENTAL APPARATUS 17
Major Test Circuits and Main Diagnostics .. 17
Sculptured Anodes and Arc Rotation 23
Mach-Zehnder Interferometers 28
Optical Interferometer 33
Microwave Interferometer 36
Spark Gap Igniter 38
III. TESTING AND RESULTS 41
Photographic Study of Anode Geometry 41
Optical Interferometry 49
Microwave Interferometry 62
Spark Gap Igniter Studies 77
IV. CONCLUSIONS/RECOMMENDATIONS 85
Results/Achievements 85
Anode Studies 85
Interferometric Studies 86
Spark Gap Igniter Studies 87
Overall Achievements 87
Recommendations 88
REFERENCES 94
APPENDICES
A. SCHEMATICS AND DEVICES 97
B. FRINGE SHIFT EVALUATIONS OF A
MACH-ZEHNDER INTERFEROMETER 108
C. PROJECT HERCULES 120
• I •
111
ABSTRACT
This dissertation describes research involving various

aspects of high power ignitron switches and their behavior
under high current and/or high coulomb conditions. Three
workshops held at Texas Tech University, Lawrence Livermore
National Laboratory, and Richardson Electronics, produced
suggestions for areas of research necessary to improve the
coulomb capability, peak current carrying capability, and
lifetime of ignitron switches. As a result, the effects of
axially applied magnetic fields have been studied, and plasma
diagnostic investigations, including optical and microwave
interferometry, have been conducted. Special ignitrons, some
of which had optical access to the discharge volume, have
been used to achieve this. Application of the magnetic field
caused a drastic reduction in the holdoff voltage capability
of the test switches, reducing it to just over 300 volts for
some tubes. Four anode geometries, flat, cup, slit-cup, and
spiral, were studied for their effects upon the discharge
plasma and its arc resistance. Photographic studies of anode
geometry reveal the sculptured anodes can induce discharge
motion, which can reduce material vaporization from the anode
thus increasing tube lifetime by limiting the contamination
of the mercury pool. Interferometry covering two density
ranges revealed the first measured plasma density for a full
size ignitron, which was on the order of 1.4*10^^ cm"^ at
115 kA with an average post-conduction diffusion rate
constant on the order of 2.7*10-^ s~^. Further research
comparing the performance of a spark gap igniter to a normal
igniter has also been performed with the jitter value for
both igniters shown to be on the order of 100 ns. Much of
this work has used a specially built, demountable ignitron,
and a 2.56 mF, 10 kV capacitor bank, charged with a 2 A
constant current power supply.
IV
LIST OF TABLES
Table Page
1. Average recombination values 76
2. Average diffusion rate constants 78
3. Conventional igniter trigger delay data .... 82
4. Spark gap igniter trigger delay data 84
5. Parameter comparison of NL-9000 to
previous high-power ignitrons 90
• ' ^
LIST OF FIGURES
Figure Page
1. Basic ignitron 3
2. Conventional igniter 4
3. The three prototype ignitrons used
at Texas Tech University 6
4. Holdoff voltage graphs of the three test
ignitrons 9
5. Diagram of Dynfax framing camera 12
6. Side view of three initial anode
geometries 15
7. Differential voltage versus peak current
for three anode geometries 15
8. The two basic testing circuits 18
9. Block diagram of the diagnostics and
triggering system for the crowbarred
circuit configuration 19
10. a) Cut-away view of test stand with
diagnostics and triggering boxes 21
10. b) Cut-away view of test stand with
diagnostics and triggering boxes 22
11. Diagram of the Megaframe camera 24
12. Ignitron with recommended modifications. ... 26
13. Machine drawing for cup anode 27
14. Machine drawing for slit-cup anode 29
15. Induced discharge motion in the cup
anode with slits 30
16. Machine drawing for spiral anode 31
17. Demountable ignitron with modified
anode section 32
18. Diagram of optical interferometer 34
19. Diagram of microwave interferometer 37
20. View of spark gap igniter 40
21. Critically damped waveforms for
Vbank = 2 kV and graphite anode 42
Vânk = 5 kV and graphite anode 42
23. Series and crowbar current waveforms
for V^,„, = 1 kV 43
VI
^
24. Voltage across and current through DIG
for crowbarred configuration 44
25. Photographic sequence for graphite anode. .. 46
26. Photographic sequence for cup anode 47
27. Photographic sequence for slit-cup anode. .'. 48
28. Exposure position for Megaf rame camera 50
29. Photographic sequences for spiral anode
with Vfcank = 5 kV 51
30. Photographic sequence centered at peak
current for spiral anode with Vj^^^k = 5 kV. . 53
^bank = 5 kV and spiral anode 55
32. Optical interferometer sequence of
coarse diamond-shaped fringes 57
33. Optical interferometer sequence of
fine diamond-shaped fringes 58
34. Adjusted optical interferometry sequence. .. 59
35. Fringe shift detection scheme 60
36. Microwave interferometer noise shot
for 1 kV level 64
37. Microwave interferometer data shot
for 1 kV level 65
38. Microwave interferometer shot for 1 kV
level with noise removed 66
43. Plasma density versus time for 1 kV level. . 71
48. Curve fitting of post conduction density
data to recombination and recovery curves. . 75
49. Traces showing current through and
voltage across conventional igniter 79
• I
Vll
50. Current through DIG and current through
conventional igniter 80
51. Voltage across spark gap igniter 83
52. Previous ignitron structure and new
ignitron structure 89
53. Comparison of detectable density ranges
for interferometers 91
54 . Plot of T^ and final peak current 93
A-1. Monocyclic power supply 99
A-2 . Basic monocyclic circuit 100
A-3. Bank charging waveform for V^harge = 2 kV. ... 100
A-4. Schematic for voltage level trigger
generator 103
A-5. Timing diagram for Megaf rame camera 105
A-6. Timing circuit for Megaf rame camera 106
B-1. Basic interferometry system 110
B-2 . Intensity versus phase difference plot 110
B-3. Generic microwave interferometer output
and associated density waveform 114
B-4. Angle factor in two-dimensional fringes. ... 115
B-5. Example of uniform phase shift
fringe pattern 118
B-6. Example of localized phase shift
fringe pattern 118
C-1.
Simulation circuit for Project Hercules. ... 123
C-2.
PSPICE input file for simulation circuit. .. 124
C-3.
Projected current for simulation circuit. .. 125
C-4.
Series capacitor set with resistors 127
C-5.
Differential voltage graph for
C-6. series capacitor sets 128
Generic metal oxide varistor
C-7. voltage-current characteristic 130
Series capacitor set with resistors
and MOVs 131
C-8. Diagram of electrolytic capacitor set
for Project Hercules 132
C-9. Basic floorplan for Project Hercules 134
C-10. Process flowchart for Project Hercules 135
C-11. Preliminary diagnostic and control plan
for Project Hercules 136
Vlll
IM«!,.J.I«v\
CHAPTER I
DESCRIPTION OF PRELIMINARY WORK
An ignitron is a triggerable enclosed switch that has a

liquid mercury cathode and a graphite or metal anode.
Mercury pool tubes have been used as rectifying devices for
more than fifty years. The first practical use was to supply
do for series-string street lights. Throughout the years,
ignitrons were used for rectification in welding power
supplies and similar applications at continuous current
levels of some 10's to some 100's of amps. However, this
domain has quickly been taken over by high power solid state
devices such as Silicon Controlled Rectifiers (SCRs).
Recent interest in ignitrons has focused on applications
with pulsed currents of up to 1 MA, charge transfer rates of
up to 1000 Coulombs, and multiple thousand shot lifetimes.
There are many reasons why ignitrons are considered one of
the best and only switches to potentially fulfill these
parameters. An ignitron is a vacuum switch that does not run
into some of the pressure shock problems associated with high
current, high coulomb discharges through switches operated at
atmospheric (or higher) pressures. Because of its liquid
mercury cathode providing a renewable electrode surface, an
ignitron has the highest simultaneous ratings on peak
current, coulomb capability, and lifetime. Envisioned
applications include the use of ignitrons as closing switches
for capacitively driven railguns, crowbar switches in energy
recovery circuits and closing switches for high-power laser
flash lamps.^
Ignitron Basics
The ignitron is a vacuum switch that operates on the
left side of the Paschen minimum in the insulating mode with
a pressure-distance product of 0.001 to 0.01 Torr-cm. The
tube is triggered by means of a small high voltage pulse
applied to an igniter which initiates the mercury vapor
stream from the cathode and closes the switch. The
1
positioning of the major components of an ignitron can be
seen in Fig. 1. The most prominent features that distinguish
ignitrons from other pulsed power switches are their high
current and coulomb capabilities. Most of this capability is
made possible by the mercury cathode which performs the dual
purpose of providing a source of electrons for the discharge
and a supply of vapor through which the discharge is
conducted. The anode material in most ignitrons is graphite,
although stainless steel is also used. When an arc or high
intensity discharge column reaches the surface of the wall,
anode, or an insulator, outgassing or vaporization from that
material can occur. The outgassed or vaporized material can
contaminate the vacuum of the tube and/or condense into the
mercury and form a sludge that can coat the igniter (or other
areas) and ruin the tube. The problems caused by evaporated
material have made research into ways of decreasing this
evaporation a priority.
The ignitor has turned out to be a major reason for
failures of ignitrons in pulsed power projects. Post-mortem
diagnoses on failed ignitrons have found shorted igniters
that cannot be triggered and igniters that have had their
tips broken off or have been broken off entirely. As seen in
Fig. 2, the conventional ignitor dips into the mercury
cathode and presses down the surface of the mercury to form a
negative meniscus. In a good tube the mercury does not wet
the ignitor, and a resistance of 10 to 100 12's usually exists
between the pool and the ignitor (this value can be as low as
1 ohm and still signify a good ignitor) . The ignitor is
usually a molybdenum support rod connected to a graphite
shank and tipped with a boron carbide compound. To initiate
an arc, a cathode spot is formed by passing a short, intense,
positive pulse of current through the ignitor into the
mercury pool. This pulse must be unidirectional because the
ignitor is easily damaged by reverse current flow. Plasma
caused by the ionization of the mercury by the cathode spot
then diffuses into the tube volume and a glow-to-arc
ANODE
TERMINAL
ANODE SEAL
(INSULATOR)
^ METAL
^ ENVELOPE
(WALL)
IGNITOR
MERCURY POOL
IGNITOR TERMINAL
Figure 1. Basic ignitron.

IGNITOR
CURRENT SUPPPORT
FLOW / ROD (MOLYBDENUM)
GRAPHITE SHANK
MENISC
IGNITOR TIP
MERCURY POOL
V77777777777j
BOTTOM OF TUB
* TIP IS OF A BORON
CARBIDE COMPOUND
Figure 2. Conventional ignitor.

transition takes place between the anode and the mercury pool
cathode, if there is enough potential difference across the
tube.2'3
Prior to the beginning of this research, ignitrons
designed for pulsed power service were simply stripped down
versions of welding ignitrons. This means that internal
components such as timing grids that surrounded the anode and
allowed for precise firing control, and a splash baffle,
whose purpose was to keep mercury from splashing up on the
grids or anode, were removed. Their maximum parameters at
this time were on the order of 350 kA and 300 C. However,
these ratings could not be used simultaneously with any
reasonable lifetime (> 100 shots) . When put into pulsed
power service, these tubes would display numerous failure
modes. The cable connection at the anode on some models
would pull out after a number of shots. After a tube became
untriggerable, internal examinations usually revealed a great
deal of contamination of the mercury pool. In one instance,
there was a semi-solid mass of metal granules and mercury in
the bottom of the ignitron rather than a liquid mercury pool.
Another failure mode encountered was the tube failing to
holdoff voltage after a few hundred high current, high
coulomb shots.^
Prototype Ignitrons
The work done at this institution involved three
different prototype ignitrons: a commercial NL-2909 ignitron,
a glass-walled ignitron, and a custom designed demountable
ignitron. Figure 3 shows cross-sectional drawings for all
three tubes. The NL-2909 was developed from an NL-5555 tube
used for welding applications by removing the splash baffle
and changing the cathode connection for high pulsed currents.
The glass-walled tube (of which two identical tubes were
tested) was of the basic ignitron design and was the largest
one of its kind ever tested at the current levels used in
this program. In this case, the absence of metal walls
c >i
o 4J
u •H
JJ
-H CO
M
c 0)
>
•H
-H
c
D
u
0)
4-> EH
c
a W
i CO
X
<D
O
E-«
4-)
(0
T3
0)
CO
CO
c c
4.1V 'I k V v ! ^
o o
-H 4-)
c •H
C7»
c
en
0)
•H
(0
» (D
I
0)
CO a
ID >i
crsT? t-i 4->
O
4->
O
<D
<D
M
.G
4->
0)
0)
M
o
CN
-H
S S ^ ^ [X4
2
turned out to be not only advantageous for optical
diagnostics, but also had an effect on tube performance in a
magnetic field. The demountable ignitron (nicknamed the DIG)
was designed at this laboratory and manufactured by National
Electronics/Division of Richardson Electronics. This tube
resembled a conventional steel-walled ignitron but had a
total of four viewports permitting visual access to the
discharge in the cathode and anode region. All components of
the DIG (i.e., igniters, anode, mercury filling, and
viewports) are easily removed, replaced, or modified for
research purposes. The DIG was designed with a small exhaust
tube on the top flange through which the ignitron could be
evacuated but pumping through one of the sideports was
sometimes required.
Axial Magnetic Field Effects

The first series of tests on ignitrons run at Texas Tech
involved the effects of axial magnetic fields on the holdoff
voltage and on the conduction characteristics of the
prototype tubes. ^ It was the anticipation that an axial
magnetic field could prolong the life of an ignitron and/or
improve certain tube characteristics, like forward voltage
drop, that prompted this study. Past work on the effects of
axial magnetic fields on ignitrons is documented by Knight.^
He found that an axial magnetic field (even as low as 0.01 T)
reduced the holdoff voltage that the tube could withstand by
a factor of seven. Additionally, the application of an axial
field of 0.1 T caused the differential tube voltage trace
during current conduction to become significantly smoother
than without a field. Both of the preceding points were
verified and expanded upon by the work done here.
The vastly different dimensions of the three tubes
required the assembly of two different test stands. In both
stands, the test ignitron was placed inside a stack of donut-
shaped Magnion, water-cooled, high power electromagnets that
were energized by DC currents of up to 640 A. For the
8
holdoff tests, a 1.89 ^.F, 60 kV high energy capacitor was
used along with a series stack of four 0.2 O. carbon disk
resistors. A 120 kV, 10 mA supply was used to charge the
system until the test switch broke down and discharged the
capacitor. The effect of the axial magnetic field on the
holdoff voltage of each tube is plotted in Fig. 4. Each set
of points can be fitted to the same basic curve. In summary,
each ignitron holds off its full voltage until a certain
axial magnetic field is reached. At this point, there is a
steep drop-off in holdoff voltage until a minimum voltage is
reached; from then on, the holdoff voltage slowly rises. The
minimum field required to lower the holdoff voltage and the
minimum holdoff voltage varies from test specimen to test
specimen.
The NL-2909 displayed a drop in holdoff ability going
from over 55 kV to 800 V at an axial field value of around
0.005 Tesla. Data points of full holdoff potential (around
55 kV) are not shown in order to provide a higher resolution
graph of the other data points. The glass-walled tube
displayed its drop in holdoff voltage at an axial field value
of 0.022 Tesla. The holdoff voltage initially dropped from
an average of 4 5 kV down to 13 kV. The demountable ignitron
held off an average of 23 kV up to an axial field value of
0.014 Tesla at which point it dropped to 750 V. Again, the
plot of the data points for this tube does not include the
full holdoff points in order to improve the resolution of the
graph. Note that the minimum holdoff voltage encountered
with increasing magnetic field for the metal-walled tubes was
around 335 V, while the minimum for the glass-walled tube was
9.4 kV. The minimum holdoff value of several hundred volts
is consistent with other data gathered on the Paschen
characteristics of gases in crossed magnetic fields.^
Comparisons of the results from the three ignitrons
reveal the drastic effect that axial magnetic fields had upon
their holdoff abilities. The NL-2909 and the demountable
ignitron displayed large reductions in their respective
CO
C
o
u
^->
•H
C
O^
-H
4J
CO
Q)
4->
<D
O
M
x:
4->
[sjiOA] dSeiiOAJjopiOH
<D
4->
o
CO
a
(0
en
<D
Cn
4-)
o
d >
©
^-^
R o
T3
L\^] »8BIIOA JJopiOH o
d X
£
&
I 0)
M
I •H
b4
[sjiOA] aSeiiOAJjopiOH
10
holdoff voltages, dropping to less than 4% of their original
value at relatively low axial magnetic fields. Furthermore,
both of these tubes showed a very slow recovery toward full
holdoff potential as the axial magnetic field was increased.
However, the recovery of the NL-2909 was almost twice as fast
as that of the demountable ignitron. On the other hand, with
an applied field, the holdoff voltage of the glass-walled
tube only dropped to around 30% of its original value. The
magnetic field value at which this reduction took place was
almost twice that required by the DIG and over four times the
field required by the NL-2909. The glass-walled ignitron
also recovered to full holdoff potential more than ten times
faster with increased magnetic field than the two metal-
walled tubes. Tests with reversed magnetic field polarity
showed that these effects were independent of the polarity of
the magnetic field.
The reduction of the breakdown voltage in the presence
of an axial magnetic field was reasoned to be caused by the
deflection of primary electrons of suitable energy into a
cycloidal motion around the anode. For the ignitrons with a
metal wall (the 2909 and the DIG), the conditions for this
effect were ideal in the anode-wall gap, where electrons
accelerated by a radial electric field were exposed to a
perpendicularly oriented axial magnetic field. Due to the
smaller anode-wall spacing of the DIG, the electric field
strength in the gap was about 2 times higher than for the
2909 tube. In contrast to the tubes with metal walls, the
electric field (normal to the equipotential levels) in the
glass-walled tube was predominantly oriented in the axial
direction with the highest radial component near the lower
potential shield. The electron path is determined by the
superposition of the rotational motion in the magnetic field
and the radial acceleration in the electric field. The
cycloidal motion of the electrons in the magnetic field of
proper magnitude therefore led to an effective increase of
the crossed gap distance and a corresponding increase of the
11
pressure times distance product, with d being the actual path
length of the electrons, determining the breakdown voltage.®
The influence of axial magnetic fields during the
current conduction phase could be measured only for the glass
tube because the holdoff voltage for the other tubes was too
low to allow a reasonable operation. These tests were done
with a 2.56 mF, 128 kJ capacitor bank discharged by the
switch into a critically damped circuit. In terms of the
magnetic field's effect on the appearance of the voltage
trace, the differential voltage across the tube was markedly
smoother as the field increased. The data also indicated a
slightly lower arc resistance as the field was applied
indicating a more efficient conduction plasma.
In addition to electrical diagnostics during current
conduction with the magnetic field, optical diagnostics were
performed on the glass-walled tube. The camera utilized was
a Dynafax High Speed Continuous Writing Framing Camera, a
diagram of which can be seen in Fig. 5. The camera used
35 mm Tri-X Pan film which is located around the inside of
the camera drum. A separate rotating mirror flashed
alternating rows of exposures onto the rotating film. The
combined rotating-drum, rotating-mirror design of this camera
allowed framing rates of up to 26,000 pictures per second.
The top rotational speed combined with the 3 p.s exposure
stops (called the diamond stops) allowed around 40 pis between
frames. This camera was run in an open shutter mode for the
duration of each shot. Photographs showed that with no
applied field the plasma in the glass-walled tube after peak
current was very diffuse and filled the entire tube volume.
Early in shots of this kind, slightly irregular plasma
patterns were prevalent. There was also more luminosity near
the cathode and walls early on. Tests done with field values
of .01 to .02 Tesla showed patterns similar to the no field
shots. Again, there were plasma patterns and some vertical
striations in the volume. As the field approached 0.035 to
0.05 Tesla (which is the general amplitude where the holdoff
12
Figure 5. Diagram of Dynafax framing camera

13
voltage was at its minimum) , the Dynafax photos began to
lose much of their luminosity although the tube was
conducting the same peak current. The plasma at these field
levels displayed regions of varying, low, luminosities and
did not fill the volume as in lower field shots.
Furthermore, the plasma was significantly more confined to
the interelectrode region in the latter states. In studying
some of the exposures, the plasma column seemed to move from
side to side within the tube.
Other Work
Other tests on ignitrons run at Texas Tech University
involved the preliminary study of the effects of anode
geometry on ignitron behavior and spectral distribution of
luminosity.^ It is beneficial to the lifetime of an ignitron
to reduce all forms of erosion during current conduction
because contamination of the mercury affects ignitor
performance. Research with vacuum interrupters, railguns,
and other high current switching devices has shown that
stationary conducting plasmas (or arcs) cause heating of
contact faces which in turn causes the liberation of metal
vapor from the contact surface. To combat this source of
erosion it seems reasonable to either find ways of externally
affecting the conduction plasma to create a totally diffuse
plasma that does not develop any stationary high intensity
columns or design contacts which force the the conducting
plasma to move over the surface of the contacts rather than
remain stationary. However, at higher currents even a
totally diffuse discharge plasma can vaporize metal from the
electrode. The first suggestion has been addressed in the
previous section of this chapter. The second suggestion of
sculptured contacts is addressed in this section.
The design of practical vacuum interrupter contacts has
evolved proven contact designs which enhance arc motion
during current conduction. •^^ The designs include spiral
and/or slit features which use the interaction between the
14
current flowing in the arc and a transverse magnetic field
resulting from the current flowing in the sculptured contacts
to continually force the arc to move in a transverse
direction. The effect of anode geometry on the arc impedance
was determined utilizing the demountable ignitron and three
different anodes: a flat graphite anode, a stainless steel
cup-shaped anode, and a stainless steel cup-shaped anode with
slits. The three anodes are illustrated in Fig. 6. Numerous
test shots were fired at different current levels for each
anode and the corresponding differential (i.e., anode-
cathode) voltage values at peak current were taken. Figure 7
is a plot of the differential voltage values versus peak
currents for all three anodes. As can be seen in Fig. 7, the
slit-cup anode produced the smallest voltage spread and the
lowest arc voltages overall. The average arc impedance for
the flat anode, the cup-shaped anode, and the slit cup anode
respectively were 5.998 mQ, 5.246 mQ, and 3.764 mCl. Some
high speed photography was performed in conjunction with the
arc impedance measurements but these results will be detailed
and expanded upon in a later section of this dissertation.
In further research by Burke, spectroscopic analysis was
performed on the discharge plasma in the DIG using optical
bandpass filters for mercury in the visible light spectrum
and two bandpass filters centered away from any
characteristic mercury lines. In this case, photographs of
the conduction plasma were taken with the Dynafax camera
through the sideport of the DIG through the various line
filters. The optical filters were centered at the following
wavelengths: 405 nm, 435.8 nm, 546.1 nm, 577 nm, 580 nm,
690 nm, 488 nm, and 514.5 nm. The last two filters listed
are the ones not centered on characteristic mercury lines.
The luminosity as a function of the following parameters was
compared: the anode geometry (flat anode versus slit cup
anode), the optical filter, and the peak current level. In
varying the current level, it took 100 kA of peak current
before any images began to appear on the film. At this
15
Graphite Anode Cup Anode Slit-cup Anode

(Cross-section) (Stainless Steel) (Stainless Steel)
Figure 6 Side view of three initial anode geometries.
«
D GRAPHITE ANODE
800- O CUP ANODE O D
03
A SLIT-CUP ANODE O D
700-
o
600- n
< ID D^
500- 6b
O • o
> 400- ^ 4^ AOA
"" ^ A
300-
a
200-
QA
100-
' 1 » 1 • 1 ' 1 • 1 • 1 •
—r-' 8 01 • 1 • 1 ' 1 • 1 ' 1 • 1
0 10 20 30 40 50 60 9 0 100 1 1 0 1 2 0 1 3 0 1 4 0
70
PEAK CURRENT (kA)
Figure 7. Differential voltage versus peak current for

three anode geometries.
16
level, the flat anode would show 3 or more frames fully
exposed with the mercury filters and 1 to 2 frames half
exposed with the off-the-line filters. At the same current
level, the slit cup anode showed 1 frame fully exposed with
the mercury filters and 1 frame half exposed with the off
line filters. Tests at higher current levels displayed
similar results. The photographic data was then
qualitatively related to electron density with a resulting
generalized statement that the electron density in the
discharge of the flat anode was 1.4 times greater than in the
discharge of the slit cup anode.^
CHAPTER II
EXPERIMENTAL APPARATUS
Manor Test Circuits and Main Diagnostics

In the pulsed power industry, high current and/or high
coulomb capability in a switch is a prime concern.
Therefore, for the research to be relevant, tests were run
with a system capable of hundreds of kiloamps and/or tens of
coulombs. All testing involved one set of capacitors; four
cabinets of ten 64 JIF, 10 kV capacitors each, resulting in a
total bank capacity and energy content of 2.56 mF, and
125 kJ, respectively. During the early part of this
research, a 10 kV, 1 A constant voltage supply and a stack of
current limiting resistors was used to charge the bank.
However, this provided charging times of over 5 minutes.
Therefore, a ±10 kV, 2 A DC power supply (Hipotronics Model
810-1A/5356) was modified to 10 kV, 2 A constant current.
This supply, frecjuently called a monocyclic supply, provided
charging times of less than 1 minute. A schematic and
description of this supply and its control circuitry can be
found in Appendix A.
The research involved two basic circuits, both of which
can be seen in Fig. 8. The critically damped configuration
was employed for the spark gap jitter testing and some of the
anode geometry studies. This configuration was used in order
to obtain higher peak currents and a better means of
measuring the ignitor jitter in the DIG. The crowbarred
circuit configuration was used for all of the interferometric
testing and some of the anode geometry studies. The benefit
of this configuration was the longer current pulse in the DIG
and the corresponding increase in coulombs carried. A block
diagram of the diagnostics and triggering system for the
crowbarred configuration can be seen in Fig. 9. As can be
seen in the figure, the triggering signal for the crowbar
switch, which must be provided at some time near the peak
current, is initiated by a voltage level sensing instrument
and a HV probe. A schematic and working explanation for the
17
Test 18
Ignitron
(DIG)
MAr
©
2 A
Critically Damped Configuration
Ri
C^ Series
VAr
Ignitron
Test
Ignitron
(DIG)
Crowbarred Configuration
Capacitor bank, 2.56 mF

L = Test stand and system inductance, ~1 ÎH
Ri = Power supply s a f e t y r e s i s t o r , 200 Q
R2 = Lumped system r e s i s t a n c e , 1 0 ' s of mfl
R, = Damping r e s i s t a n c e , 25 mfl
Figure 8. The two b a s i c t e s t i n g c i r c u i t s .

19
X
J o,
m
CO'
H
o 0)
> H:1 EH
M
O
^ M
W CO £
H 4-)
o CO
>i
CO
c
JJCJ -H
EH u
2 M 0)
o cn
> cn
•H
t^ u
M +J
2^ -o
c
(tJ
•
CO C
o o
-H -H
+J 4-)
CO fO
O M
(x: u M _ /^ o
o c n
H M H cn tj>
CO Ou (d -H
m EH u -H m
EH
2:
T3 c
pa o
o: 0) u
ex: 4-> 4->
D •H
CJ
o o
I e-H
CO
M
o
CO
(0 Q)
< •H M
C!)
•O U
M
CO
U 5
O O
.-H M
OQ U
U
04
O CM
0^
2 o
M CO (D
tt)
M u
O o
t^
2 M cn
H O U
H CO .H -H
CO o
20
voltage level device, which was designed and constructed
specifically for this project, can be found in Appendix A.
Other devices that were also designed for this project
include the ignitor trigger pulsers, the differential probe
amplifier, and the fiber optic transmitter/receiver.^ The
critically damped circuit configuration used much of the same
equipment as the crowbarred configuration with the deletion
of the voltage level sensor and the series switch.
In order to implement all of the planned diagnostic
techniques, a test stand had to be designed which would fit
into the existing optical interferometer, allowing for the
demountable tube and another ignitron to be mounted in the
proper configuration, and have an inductance of approximately
1 p.H. Figure 10.a shows a three-dimensional cut-away view of
the test stand taking into special consideration the
placement of the two ignitrons and the main current sensors.
In Fig. 10.b, another view of the stand is shown this time
emphasizing the positioning of the voltage sensors and
ignitor triggers. As can be seen, holes in the stand are
positioned in line with the throughports of the DIG. This
feature allows for the use of interferometry and the high
speed photography of the discharge plasma. The stand was
made out of .5" thick aluminum with a 1" aluminum rod as a
central current conductor. The holding braces for the series
ignitron had to be modified slightly as two different sizes
of switches were used during the past years of testing.
Initially, size D NL-5555 tubes were used but these tubes
always failed after a relatively low number of shots.
Therefore, a larger size E tube, the NL-1057, was used and
has performed well with no failure modes displayed as of yet.
Besides the Dynafax framing camera described in the
previous chapter, another high speed camera was used for this
project. This camera was a Model 189B Synchronous Framing
Camera loaned to us by Lawrence Livermore National
Laboratory. To shorten all references to this camera, it
will be called the Megaframe camera in reference to its
21
connection to outer
conductor of bank cables
Coaxial Gabies
Coaxial Cables
from Banlcs
from Banlcs
Crowt>ar Currant
Rogowski Coll
Series Ignitron
Crowt>ar Ignitron
Figure 10. a) Cut-away view of test stand with

diagnostics and triggering boxes.
Voltage probe and
22
Compensation Box
Voltage Level
for Crowbar
Timing
Anode
Voltage
Signal
Cathode
Voltage •
Signal
1:1
Pulse
Transformer
Trigger
Generator Generator
Series Ignitron
Trigger Signal
Crowbar Ignitron^
Trigger Signal
Figure 10. b) Cut-away view of test stand with

diagnostics and triggering boxes.
23
highest framing speed of 1 |ls per frame. The camera is
operated by compressed gas, turning a turbine, which rotates
a mirror. The camera, which can be seen in Fig. 11, was
operated from the compressed air lines of the laboratory, but
needs helium to reach its maximum framing speed. The air was
filtered by a Matheson Gas Purifier which was connected to a
gas pressure regulator and a 120 V air control valve. This
camera contains 120 sets of relay lenses which take the
images from the rotating mirror and relay them to the
stationary film. The minimum framing time allowed by the
pressure in the laboratory's air lines was around 1.875 p.s
per frame. Because of its limited recording arc and high
framing speed, the Megaframe camera could not be run in the
same open shutter mode that the Dynafax camera was. A system
had to be designed which would fire the experiment only when
the rotating mirror was in the proper position and also
protect the camera from overwriting on the film. Previously,
the camera had been modified and contained an internal lens
and light system which fed a signal to a photomultiplier.
The output of this system gives a small voltage pulse just
before the mirror face arrives at the first lens. A circuit
was therefore designed which would synchronize the Megaframe
timing signal, the triggering of the ignitron, and the
shutter pulse to a ferroelectric light valve. The valve will
be described in a later section of this (inapter. A schematic
and description of this circuit can be found in Appendix A.
Sculptured Anodes and Arc Rotation

As part of the research involving the improvement of the
ignitron for pulsed power service, four different anode
geometries were tested in the DIG. A conventional flat
graphite anode was supplied by the manufacturer. The other
three anodes were designed and machined in our laboratory out
of stainless steel. The basic dimensions of the anodes,
including the main radius and the overall length, remained
the same throughout all designs.
24
Polaroid film pack mount
35 mm film
Relay lenses
Image Path
adjustable
main lens
assembly
Ragrulator
Figure 11. Diagram of the Megaframe camera

25
As previously stated, stationary conducting plasmas (or
arcs) cause heating of contact faces, in the case of an
ignitron mainly the anode face, and there can be a high
amount of evaporation of metal vapor. In an ignitron, the
unwanted metal vapor can coat the walls of the tube, the
anode insulator, or the ignitor, and promote tube
malfunctions. Evaporated material can also mix with the
liquid mercury cathode and inhibit proper tube triggering and
mercury plasma formation. For all of these reasons,
alternate sculptured anode designs which would in theory
promote discharge motion and/or rotation were tested in the
DIG.
The first alternate anode design did not initially
concentrate on promoting arc motion but came about from the
need to increase ignitor lifetime. Through multiple
discussions with industry personnel, it was planned to move
the ignitor out of the way of the main discharge.-^^ ^^ The
ignitor was moved to the center of the tube and the main
current discharge relegated to the outer edge of the mercury
pool. One way to concentrate the discharge to an outer
circumference of the tube was to sculpture the anode in the
form of an inverted cup. Additionally, the spacing between
the anode and mercury pool would be lowered in order to
promote anode-to-cathode current conduction over anode-to-
wall conduction. Figure 12 depicts an ignitron during
conduction with all of the feature changes previously
mentioned. The research into sculptured anodes began with
this cup design; however, the anode to cathode spacing within
the DIG was not modified in order to be able to use our
optical diagnostics and in order to be able to make
comparative studies between the conventional anode and the
sculptured anodes. Figure 13 shows the machine drawing for
the cup shaped anode.
The goal of the next alternate anode designs was to take
the cup anode and modify it to try and induce discharge
motion or induce the arc motion by a spiral design. The rim
26
CO
C
o
-H
-M
m
o
-H
'4H
-H
-d
o
e
-o
<D
T3
c
B
o
o
(1)
JH
-H
o
M
-P
-H
C
Cn
Csl
Q)
u
p
cn
•H
all around 1
c C a,
<D - H o p
cn -H o
JH - d Cn 0) \
<XJ Q) Q) «M TJ N.
,c; U M O o ^
o Q) c:
CO 4-> CO (U m
-H C -H cn
-d Q) ^ -d
O -P <u
27
0 375 R
094 Dia. hole along central axis

(dtmenstons In inches)
Figure 13. Machine drawing for cup anode

28
of the cup anode was cut with sixteen angled, 0.125" slits.
The slits were designed so that the top of a slit vertically
overlapped the bottom of the next slit. The machine drawing
for the resulting slit-cup anode can be seen in Fig. 14. It
can be seen that the slanted slits provide a transverse
component to the magnetic field which may induce the
discharge plasma within the tube to rotate as depicted in
Fig. 15. The final anode design was based on the standard
spiral contact anode used in vacuum interrupters. The
earliest design for this type of electrode was patented by
Schneider in 1960.^^ In this type of anode, the self magnetic
field generated by the current flowing in the spiral arms
interacts with the current flowing in the arc to force the
arc to move between the spirals.^^ The machine drawing for
this anode can be seen in Fig. 16.
After initial testing of the cup anode, a problem was
encountered. Due to the high current tests, some spot
welding had occurred between the threads of the anode rod and
the cup anode. Therefore, the cup anode could not be removed
from the original anode section of the DIG. In order to
continue the testing, a new anode section was designed for
the DIG. A 6" diameter ConFlat ceramic top insulator and a
6" ConFlat blank flange were used in the new anode
connections. Extended length anode rods (one for each
unattached anode) were designed which could be threaded into
the existing anodes and screwed through a hole threaded in
the blank. This entire assembly is shown in Fig. 17. The
rest of the testing was done by alternating the entire anode
rod-anode assemblies.
Mach-Zehnder Interferometers
A Mach-Zehnder interferometer is a field visualization
interferometer which detects the phase shift in a beam
introduced by an object such as a plasma. The phase shift
introduced by the plasma is dependent upon the plasma charged
particle density and the plasma dimensions. In the
29
0 375R
3312
.094 Dia hole along central axis

16 silts spactd
22.S dsgPMt apart
slit dirrenslons
(dimensions in inches)
Figure 14. Machine drawing for s l i t - c u p anode

30
- ^ ^ o
o -d
- H -H
o
4J QJ
cu
1
C fe4 M
tJ^ p
(0 u
M
s (U
CO
4-)
S -H
.H
CO
.c
-H
<U
T3
O
C
"O m A ftJ
o
o 0)
CO -H c
o M a
- H "M • p "S.• H p
cd
« « O O
4J A u
u CJ E n
o M CD
O n
M C 0)
o
C7< E •P 3 c
C C •H
c
o
© © ©^ , •H
4-J
O
6 6 d' :i B
<D
u
CO
-H
T3
T5
0)
u
p
-d
c
in
0)
M
p
cn
-H
31
0375 R
0.094 Dia. hole along centnl axis

Taper on bottom ends here
(dimensions in inches)
Bottom View
Figure 16. Machine drawing for spiral anode

32
Figure 17. Demountable ignitron with modified anode

section.
33
interferometry system, two beams (a reference beam and a
plasma beam) are derived from a common source. The plasma
beam passes through the plasma while the reference beam does
not, thus the phases of the two beams will now differ. The
beams are then recombined with the angle between the beams
creating a fringe pattern. Changes in the plasma density as
a function of time will cause phase differences which create
a shifting in the fringe pattern, all of which can be
photographed or recorded over time and analyzed.^^ Part of
the investigation of the discharge plasma parameters in the
DIG involved the use of two Mach-Zehnder interferometers.
Both interferometers use the classical Mach-Zehnder method of
comparing a reference beam to a beam passing through the test
plasma. One interferometer utilized a visible laser and a
system of beam splitters and mirrors. The other
interferometer was designed for a microwave frequency with a
proper source, waveguides, amplitude attenuator, and phase
shifter.
Optical Interfergmeter
A basic layout of the entire system can be seen in
Fig. 18. The interferometer was a large prebuilt model
mounted on casters. It totally enclosed the optical
components and had the pitch and yaw adjustments mounted
externally to make interferometer alignment easier. A CW
Argon Laser (Continental 3020 AS) with 2 watts maximum output
power (multi-line, multi-mode) provided a continuous beam,
which was pulsed for the measurements. A small helium neon
laser, was helpful during the alignment procedure. It is
important to check and make sure that the difference between
the path lengths of the two beams is smaller than the
coherence length of the laser. Otherwise, interpretable
fringe patterns will not result. This laser has a
longitudinal mode spacing specification of 160 MHz. This is
equal to the reciprocal of the coherence time, which turns
out to be 6.25*10"^ seconds. The coherence length of a laser
34
SH
0)
-p
o
u
Q)
4H
t5 Q)
•P
o •H
u
o
-H
4-)
O.
o
14H
o
B
cn
(d
-H
Q
oo
M
P
-H
35
is the speed of light times the coherence time. Therefore,
the coherence length of this laser is approximately 1.875 m.
Thus, the path lengths can differ by almost 2 meters without
losing fringe pattern information. With the rigid, symmetric
construction of the interferometer, it is not anticipated
that the path lengths will vary by more than a few
centimeters. As a check of the longitudinal mode spacing
value, the coherence time of a cavity is equal to 2 times its
length divided by the speed of light. The length of this
laser's cavity is .94 m, which corresponds to a coherence
time of 6.27*10"^ s. Another way to get a rough approximation
for the coherence length of a laser is to simply take twice
its length. In this case, the approximate coherence length
would have been 1.88 m.^^
A narrow bandpass laser line filter, of which there were
two (one for each of the main Argon laser wavelengths) , was
inserted into the beam path of the interferometer just before
the camera to avoid any illumination of the film by the
ignitron discharge plasma. The filter for the blue line was
centered at 488.9 nm with a 6.7 nm half transmission
bandwidth. The filter for the green line was centered at
516.4 nm with a 7.8 nm half transmission bandwidth. This
interferometer creates a two-dimensional output of shifting
fringe patterns which could then be photographed. The
Dynafax rotating camera, which was referred to earlier, was
run in an open shutter mode at full speed. For some
experiments, the laser line filter was removed from the
optical path. This allowed both the fringes and the
discharge light to be photographed, thus allowing the
checking of the laser window timing.
For testing, the laser beam was pulsed through the
system with either a normally-open, normally-closed cascaded
shutter combination or a TTL controlled ferroelectric
shutter. The cascaded shutter combination consisted of two
Uniblitz Model 23X shutters, one normally open and one
normally closed, with a special reflective coating on the
36
.125" aperture shutter. When triggered simultaneously, they
provided a 400 Us window through which the beam could pass
before traveling on to the beam expander. However, these
mechanical shutters turned out to have a great deal of
inherent jitter due to their 0.6 ms delay time and 0.6 ms
opening time. Additionally, the coil of this shutter needed
a special trigger pulse of 60 V followed by 5 V of holding
power. Many test runs were lost due to the improper timing
of the laser pulse. To correct this predicament, a
ferroelectric light valve Model LVIOOAC, made by Displaytech,
Inc., with a 25 mm clear aperture and the corresponding
control box were used. This light valve contains two
adjustable polarizers and an active liquid crystal element in
the center. The operation is based on the rotation of the
plane of polarization in the center element upon application
of a voltage of ±15 V. This shutter had approximately an 800
to 1 contrast ratio, a 40 \Ls opening or closing time, and a
50-70 \ls delay time. For use with this shutter, a small
monostable circuit and inverter was designed. The circuit
had a potentiometer to control the pulsewidth of the shutter
pulse, and was triggered prior to the DIG.
Microwave Interferometer
A Mach-Zehnder interferometer functioning in the
microwave range was used in order to test for a different
range of plasma densities. A basic layout of the entire
system can be seen in Fig. 19. As can be seen, the reference
path in this interferometer is not physically parallel to the
plasma path as in the previous interferometer. Instead, the
reference path uses a variable attenuator and variable phase
shifter to match the two signal paths. Another difference
between this interferometer and the previous one is that the
output signal is not two dimensional but a time-varying
voltage signal provided by a detection diode mounted into a
tunable crystal mount. The source for this system is a
reflex klystron, VRE-2101AA, which has a maximum output power
37
% c
rH X O
O Q) u
lH r-\
CM MH CO
u
1 0) >
w fr; •H
0)
4-J
(U
e
o
c «4H
4-) 0)
CO 4-»
M 0)
0) O CO c
-H 4-) fO
^ 03 x: -H
p Oi
C >
Shift
M 0 0)
 4-> M
O
< •H
M B
MH
(0 O
>
(d
u
cn
-H
Q
(T\
0)
p
-H
[L4
t.
§= #
J
38
of 450-600 mW, depending upon the operating frequency, into a
matched load. A reflex klystron is a single-cavity klystron
oscillator in which an electron beam passes through a cavity
gap into a region where the negative voltage on a repeller
electrode causes the electrons to reverse direction. The
combination of a signal across the cavity gap and the bunched
electrons returning to the modulating gap at the correct time
provide the feedback loop which causes the device to
oscillate.^^ Before the system was utilized, the output of
the interferometer system was maximized by tuning the crystal
mount. Prior to each test shot, the output voltage of the
interferometer system is adjusted to be as close to zero as
possible by balancing the interferometer utilizing the
attenuator and phase shifter.
Spark Gap Ignitor

Additional research has concentrated on replacing the
conventional ignitor with a spark gap style ignitor. It is
thought that this will remove a major mechanism of ignitron
failure at high currents. Conventional igniters break off or
become shorted and/or untriggerable in the high current, high
coulomb regimes of interest to the pulsed power community.
The main test circuit used the existing ignitor trigger
generator with the addition of only a step-up transformer.
The transformer was built in the laboratory. It had a
toroidal ferrite core on which the primary and secondary
coils were wound. The coils were made from insulated high
voltage wire and were wound with an approximate 1:4 turns
ratio. The entire assembly was attached to external high
voltage BNC connections and then placed in a sealed oil-
filled container.
Two styles of spark gap igniters were designed and built
to be mounted on the same 2.75" ConFlat high voltage
feedthrough, which was rated at pulsed voltages of up to
40 kV. As a special note of interest, the center conductor
of this feedthrough was nickel plated to provide a metallic
39
surface that was compatible with mercury. This feedthrough
was then bolted to the angled port of the DIG. The first
ignitor was a straight tipped design used to perform a proof
of principle test. After this model proved successful in
triggering the ignitron, another model was developed that was
angled perpendicular to the mercury pool and whose
longitudinal dimension was adjustable. A drawing depicting
this ignitor and its positioning within the ignitron can be
seen in Fig. 20. The main purposes of the testing was to
show that this type of ignitor could indeed trigger the
ignitron and make a comparison of the jitter and delay
statistics of this ignitor versus the conventional ignitor.
These measurements were made by comparing the delay between
the beginning of the current going through the ignitor to the
beginning of current flowing through the DIG.
40
o
o
HIGH VOLTAGE FEED-THROUGH

ARRANGEMENT
Initial spark gap tip
Figure 20. View of spark gap ignitor.

CHAPTER III
TESTING AND- RESULTS
Now that all of the test circuits, diagnostics, and

research goals have been introduced, the actual data can be
presented and discussed. As displayed previously,
experiments were done on the DIG in two different circuit
configurations: critically damped and crowbarred. Figures 21
and 22 show typical traces for the voltage across and current
through the DIG for two bank charging levels with the circuit
in the critically damped configuration. Interpreting these
levels over many shots shows that there is approximately 18.5
kiloamps of peak current in the DIG for each kilovolt of bank
charging level. Factors change for the crowbarred
configuration. With the damping resistors removed, there is
more peak current in the main current loop. However, not all
of the current is crowbarred into the DIG loop when it is
triggered. Figure 23 shows standard traces of the currents
through the series ignitron and the crowbar ignitron (the
DIG) for the same shot. In this particular case, the peak
current of the DIG was 60% of the peak current in the series
ignitron. Overall, the peak current in the DIG averaged out
to be approximately 56.2% of the peak current in the series
ignitron and turned out to be higher at greater current
levels. The heavy oscillations in the current waveforms are
due to too much inductance in the banks and the series switch
not turning off immediately after the crowbar switch has
fired. For this configuration, approximately 20 kiloamps of
peak current was observed in the DIG for each kilovolt of
bank charging level. A typical trace for the voltage across
and current through the DIG in the crowbarred configuration
can be seen in Fig. 24.
Photographic Study of Anode Geometrv

The optical study of anode geometry on the discharge
behavior in the DIG was done on four anodes with two
different cameras. Although some of this work has been
41
42
I river t e d
i qnit ron
r n r r ont.
27 kA/V
Ignitron
voltage
1 3 3 3 . 3 3 V/V
Figure 21 CT i t i c a l l y damped w a v e f o r m s f o r Vjâ^^j^. = 2 kV

and g r a p h i t e a n o d e .
Inverted
ignitron
current
27 kA/V
Ignitron
voltage
1 3 3 3 . 33 V/V
Fignie 22 C r i t i c a l l y damped w a v e f o r m s f o r Vj^g^^ 5 kV

and g r a p h i t e a n o d e .
43
>
u
o
CO
M
PI O
IM
(1)
>
.s >* 4J
s& c

P
o
M
4->
4-)
C

(A^) a5e:noA ' (O^/V^l) :û9Jano

Csl
0)
u
p
-H
1X4
45
previously discussed, the photographic interpretations will
be reexamined and expanded upon in this section. ^'^^'^^
Exposures for all photographic data were made on Tri-X Pan
film and processed in the laboratory darkroom using HC-110
developer. The negatives were then made into photographic
prints using medium and/or high contrast photographic print
paper and Dektol developer. During the making of some
prints, transparent masks depicting the viewport dimensions
were superimposed along with the actual plasma exposures to
make viewing the prints easier. The individual frames are
numbered in secjuential order beginning with frame #1 with the
time between frames dependent upon the camera. Some of the
data in this section will be scanned images rather than
actual photographs, to make the images more reproducible.
The images are scanned at 360 dpi spatial resolution with a
256 level gray scale scanner and then printed out with a 300
dpi laser printer.
Numerous shots were taken at various bank voltage levels
for each anode. Example images made for the graphite anode,
the cup anode, and the slit-cup anode can be seen in
Figs. 25, 26, and 27. All of these images were made with the
Dynafax camera and with the circuit in the crowbarred
configuration. Therefore, the data can be seen to brighten
as the DIG progresses through the main current peak, dim as
the tube approaches zero current conduction, and grow
luminous again as the tube goes on to the secondary current
peak. Note that the data shots made with the slit-cup and
cup anodes showed the most plasma motion and/or distortion
while the shots with the graphite anode showed a uniform
diffuse plasma. Detection of this motion feature was the
stated intention of the sculptured anode testing, so it can
be said that the tests are successful. Due to the large
interframe time, no valuable judgments on the velocity of any
plasma motion could be made.
Images of the conduction discharge with the spiral anode
in the DIG were made with the Megaframe camera and with the
46
Bank level = 5 kV, 45 |ls interframe time
Figure 25. Photographic sequence for graphite anode

47
Figure 26. Photographic sequence for cup anode.

48
Figure 27. Photographic sequence for slit-cup anode

49
circuit in the critically damped configuration, which is
easier because of the timing problems with the Megaframe
camera. Figure 28 shows the relative placement of the
recorded camera image in reference with the anode, cathode,
and sideport of the DIG. Two examples of the Megaframe
camera's data can be seen in Figs. 29 and 30. Each example
only shows a few frames at four time intervals throughout a
shot. Again, the frames are numbered consecutively with
frame #1 being near the beginning of current conduction. The
small, oblong bright area in the corner of each frame is a
reflection of the light bulb used for the photomultiplier
timing. Investigation of the exposures over multiple shots
taken at a bank charging level of 5 kV revealed some
discernible plasma motion. As an example, frames 48 through
51 of Fig. 29 show a bright region of plasma moving
diagonally from the cathode region of the tube to the anode
region. As a rough approximation, the left edge of the
plasma seems to move approximately 2.3 cm (1.27 cm
horizontally and 1.9 cm vertically) over two frames, which is
3.75 p,s in time. In terms of velocity, the plasma turns out
to be moving at approximately 6.1*10^ cm/s. This is on the
same order of magnitude as the electron drift velocity of
4.8*10^ cm/s that is calculated in the next section.
Additionally, the differential voltage waveform for shots
made with this anode, an example of which can be seen in
Fig. 31, were markedly different from the voltage waveforms
for the graphite anode, an example of which was shown in
Fig. 22. The arc resistance at peak current for this anode
averaged out to be 13.76 mil at an average peak current of
87.83 kA.
Optioal Interferometry
The optical interferometer system introduced previously
was used on the DIG in the crowbarred configuration. The
method used to interpret the data is derived in Appendix B.
All shots were done with the flat graphite anode in the DIG
50
E x p o s u r e P o s i t i o n Of_
Megaframe Camera
Figure 28. Exposure p o s i t i o n for Megaframe camera

51
Bank l e v e l = 5 kV, 1.8 îs i n t e r f r a m e t i me
32
33
34
35
Figure 29 Photographic sequence for spiral anode

with Vbank = 5 kV.
52
48
49
50
51
Figure 29. (continued)

53
Bank level = 5 kV, 1.8 îs interframe time
30 34
31 35
32 36
37
Figure 30 Photographic sequence centered at peak current

for spiral anode with Vj^^^k = 5 kV.
54
38 42
39 43
40 44
41 45
Figure 30 (continued)
55
(D
-o
O
c
(0
(d
-H
a
CO
-d
c
(d
>
IT)
_ •" "O
o
<+-4
CO
a B
u
o
>
(d
-o
0)
(d
-o
(d
u
-H
4-J
•H
U
o
CO
0)
(OC/A) »IPA •(•1) >xx»ajt>D P
[I4
56
and photographed by the Dynafax camera. Figures 32, 33, and
34 display interferometric sequences for three different
shots. As can be seen, different views of the fringe pattern
were observed by changing the focal length of the camera in
order to find the best option. The diamond shaped patterns
occurred when an external extension to the lens of the camera
was added on. Consequently, the interferometric beams are
routed through the camera optics in such a way that the
diamond shaped exit pupil defines the beam shape and is
projected in the sharp outline onto the film. By adjusting
the focusing control of the lens, more dense fringes could be
obtained as in Fig. 33. By removing the extension and proper
camera positioning, the more circular patterns of Fig. 34
could be obtained. The fringes in Fig. 34 are not perfectly
straight because of imperfect surfaces on the optics and
viewports of the interferometer system.
Initial analysis of the interferometric images was
inconclusive as very little shifting of the fringe patterns
could be visually detected. Two main methods were arrived at
to aid in the detection of fringe shifting. By enlarging the
individual frames of a shot by as much as 400%, minute shifts
in the fringe pattern from frame to frame could be seen. In
addition, an image processing program. Image 1.29, was used
to plot out the gray levels along a straight line taken
through the center of an individual fringe exposure frame.'^^
The plots could then be aligned and the shifting evaluated,
as seen in Fig. 35. Although there may have been some
localized fringe shifting, only the maximum fringe shift
observed on an exposure was taken into account. Careful
evaluation of several shots at various peak current levels
centered around 115 kA revealed an average fringe shift which
corresponded to plasma densities in the DIG on the order of
1.445*10^^ cm~^. Data before the peak level, at a current in
the area of 50 kA, indicated plasma densities on the order of
7.225*10l^l9'20
57
Laser line filter removed, 45 |ls interframe time
Figure 32. Optical interferometer sequence of coarse

diamond-shaped fringes.
58
Figure 33 Optical interferometer sequence of fine

diamond-shaped fringes.
59
Bank level = 5 kV, 45 ^s interframe ti me
Figure 34. Adjusted optical interferometry sequence

60
Alignment Alignment
Line #1 Line #2
numbers refer to
exposure number
on Fig. 34
Figure 35. Fringe shift detection scheme

61
Some of this data can now be compiled to calculate some
approximate parameters about the discharge plasma. The
current density of the discharge, J, can be related to the
electric field, E, by
J = e(n ^l^ + p ^lp)E, (1)
where
e = charge of electron,
n = electron density [cm"-^],
p = ion density [cm~^],
\i^ = mobility of species type i [cm^/Vs],
and the current density and electric field are assumed to be
one-dimensional and parallel. Using the dimension of the
anode and the average peak current value, it can be shown
that
J = — - — = 115000 ^ 1120.77 [A/cm^]. (2)
area 7C(5.715)
The measured electric field at 115 kA is approximately

voltage 632.5 >,T oc TTT/ i /•:.x
E = « - _ - — — « 47.96 [V/cm] , (3)
distance 13.19
where we are neglecting the fact that part of the voltage
drop is in the anode and external connections rather than in
the discharge. Since we are looking at a highly ionized
discharge plasma, it can be assumed that the electron
mobility is much greater than the ion mobility (i.e., \ij^»\ip) .
Therefore, Eq. (1) can be used to solve for JI^ ^Y
N„ = i = 10,107.63 [cm^/Vs] . (4)
^" e n E
Therefore, the measured electron drift velocity is ecjual to
v^ = \lÊ = 4.85*10^ [cm/s]. (5)
Referring t o l i t e r a t u r e , t h i s d r i f t v e l o c i t y for
electrons in mercury vapor can be related to a value for E/p^,
with po being the reduced gas pressure in Torr at O^C.^^ The
62
value for drift velocity given in Eq. (5) is related to a
value of E/Po « 15 [V cm"^ Torr"^] . This means that the
reduced gas pressure in the discharge at a peak current of
115 kA is approximately 3.2 Torr. By using the standard
relationship between gas pressure and density (po = nKT) and
an assumed temperature of 0°C, an approximation for the
neutral gas density can be calculated to be 1.13*10^"^ [cm"^] .
The previously calculated values can be used to compare
the index of refraction for the electron plasma and the index
of refraction for the mercury ions. The magnitude of the
index of refraction, \ij., for a species is given by the
following formula
2 .2
CO A
<^^r-l) • =—V^ ' <^)
^ species ^ 2 2
B n C
where COp is the plasma frequency and is related to the plasma
density of the species. Next, we will use the approximation
that the mass of a mercury ion is its atomic number (i.e.,
number of protons), which in this case is 80, times the mass
of a proton. Using the electron density from the previous
data and the neutral plasma density previously calculated,
the values for Eq. (6) become
(^l^-l) = 1.714*10'^ and (^lr-l) = 9.13*10~^\ (7)
electron ion
Notice, the index of refraction of the mercury ions is much

smaller than the electron index of refraction, we therefore
made an appropriate assumption that the data from the optical
inteferometer is due to the electron density only.^^
Mir-.rowave Intf>rferometrv
The microwave interferometer system previously
introduced was also used on the DIG in the crowbarred
configuration. The method used to interpret the data is
derived in Appendix B. The output voltage of the
63
interferometer was recorded with a Nicolet 2090 digital
oscilloscope along with the corresponding current in the DIG.
Because of the close proximity of the interferometer to the
discharge and the lack of shielding, there was some discharge
noise present in the interferometer's output. In order to
remove the system noise from the output waveforms, data were
taken with the receiving horn of the interferometer blocked
and the resulting signal subtracted from the normal data
waveform. A typical noise shot can be seen in Fig. 36. A
normal interferometer shot for the same bank charging level
can be seen in Fig. 37. It can be seen that the noise
waveform has none of the oscillations describing the plasma
density that were present in normal data waveforms. In any
case, no noise occurs when data is taken after the current
equals zero. Additionally, notice that the data shot has no
voltage oscillations at the beginning of or during current
conduction. This is because N^ is reached so cjuickly by the
plasma that the interferometry system did not have time to
respond and the density of the conduction plasma stays well
above N_ for the entire current conduction period.
Therefore, we were only able to look at the plasma decay
after conduction.
Tests were done at five different bank charging levels,
and a single data shot with the noise removed for each
charging level can be seen in Figs. 38 through 42. For
multiple shots at each voltage level, the timing of the
voltage oscillations was related to the associated
differential phase change and then to the respective density
by the method detailed in Appendix B. Figures 43 through 47
show the density versus time plots for the shots presented in
Figs. 38 through 42. Note that the lowest density detected
by the interferometer was 0.364*10^^ [cm"^] .
Since the density data occur after conduction, some new
principles must be introduced so that the data can be fully
interpreted. If we assume that the plasma present in the DIG
well after conduction has ceased is a distribution of ions
64
Q)
>
>
U
O
»M
4J
O
x:
CO
Q)
CO
-H
O
c
M
0)
a -M
(U
B
o
u
0)
»M
U
Q)
J-)
C
•H
0)
>
(d
?
o
u
o
•H
CO
0)
M
P
&>
•H
(OI/VM) :iuajjno U4
65
>
>
U
O
<4-l
4J
O
Xi
CO
(d
4-)
(d
-O
M
Q)
4->
a 0)
e
o
M
0)
U-4
M
<D
4-»
C
-H

(d
?
o
M
O
-H
CO
Q)
M
3 3 3 P
cn
-H
[14
(OI/V^) ûajjno
66
>
o
g
(U
M
to (U
cvi CO
•H
O
c
4->
-a •H
5
>
>
M
to
O
U-i
-a +J
a "^ O
Xi
CO
M
Q)
4-J
Q)
to o
M
(U
U-4
0)
4-)
<4 C
•H
>
(d
^
o o
u
I u
CVI ^ - <4
3 3 3 3 3 o
I
oo
(0I/Y5() :ûaaano CO
0)
M
P
Cr>
-H
67
>
o
(D
U
Q)
CO
-H
O
C
x:
4-)
-H
>
CU
O ^ >
CO Csl
M
O
*4-l
> to
oa .—. oi 4-)
o
o CO
M
" H O , 0)
CJ 4-)
0)
.^1 O
0)

(d
o
I M
O
s
I
CO
(03/V^) :;uajjnD
0)
P
Cn
-H
[14
68
•o
>
o
1 B
<D
U
- fM 0)
CO
- -H
O
c
Xi
4J
•H
u — to
(U
4J
0) (1)
>
e
o _ 3
u
0) "•
<4-( ~
>-l >
(U M
4J
c PO
t-l
U
- O
„2 «M
4->
O
-
xi
CO
CJ S
-3
— 0)
4->
CD
•*-• i /'^'''^ B
O
^
«u 1
1 v^ " •
u
M I ^^^ <D
^^^-^^ _
^
^ fi U-l
u 1 - ^
^4
ro (U
4-)
C
-H
0)
>
(d
o
I u
o
1 -H
1 1 1 1 1 1 1 1 1 1 1
a tq ^ N
cvi <;} ^ r^ ^ 4 3 S
I I
O
(OG/V^) tûeajno
Q)
U
P
-H
[lL4
69
TJ
>
O
B
0)
to <D
rJ CO
•H
O
c
x:
-3 4-J
-rH
>
- 3 <1>
>
-3 CM
M
O
fl
5
t*.
M-l
a 1
4J
O
"a m
m
CO
U
<D
4->
(U
to B
o
u
0)
«M
U
(D
4J
- ?i
c
>
(d
CM
e o
I M
O
•H
a CM • { t q ^ c v i - ^ a Q t q ^ c s
S S 3 S
I I I
(0^/Y^) :iuajjno
M
p
-H
[14
70
>
o
B
Q)
M
0)
CO
•H
O
C
4J
•H
0)
>
O
m >
w CM
in
> H
CM
u
o
«4-4
o 4J
I o
n xi,
to
M
Q)
4-J
Q)
o
M
0)
tM
in V4
0)
4-J
C
•H
(U
>
(d
o
u
I _ I o
-H
3 a *" 3 2 :s !:i -^ 3 3 5 ?! 3
I
sI sI sI
(0S/V5() ûejjno
eg
0)
M
P
•H
[L4
71
0.5 MWD15R7
1 0.4
B
^û
^
r-t
0.3
o
iH
0.2
4-)
-H
CO
C 0.1
<D
a
0
500 1000 1500
Time (p,s)
Figure 43. Plasma density versus time for 1 kV level
0.5 MWD13R1
I 0.4
B
u
0.3
o
f-H
•K
>i
0.2 I
4J
-H \
(0 0.1 •H
c
(U
Q 0
500 1000 1500
Time i\is)

72
0..5 MWD13R7
1 0..4
B
^m^ u
'9'
iH
0,.3
o
TH
*
>1 0..2
4-)
-H
CO
c
0.. 1
(U
Q
0
500 1000 1500 1750
Time (îs)
0.5 MWD14R7
fO
0.4
B
u 0.3
t. 0.2
o
4->
1 0.1
Q
0
500 1000 1500 2000
Time (îs)

73
0. 5 MWD12R2
^.^
ro
1 0,.4
B
^^
o
'a*
iH
o
tH
0,.3
v
4-J
-H
C
0)
Q
> 1
CO
0,.2
0,.1
0
X ^
0 500 1000 1500 2000

Time (|is)

74
and electrons in a dense background of neutrals, the plasma
density will decay by two processes; recombination and
diffusion. When the density is high, recombination is
dominant. After the density has decreased to a low value,
diffusion becomes dominant. The general equation of
continuity for a decaying plasma is
n = q n^ + Cj n , (8)
where Ci and C2 are constants and n is the density function.

When recombination is dominant in plasma decay, the n^ term of
Eq. (8) is dominant and the solution for n is proportional to
1/t. As time passes further, the density versus time curves
should approach a pure exponential decay as diffusion becomes
dominant and the second term of Eq. (8) takes more control.^^
Applying this reasoning to the density versus time curves
obtained in this experiment involved matching the first part
of the curves to a linear, reciprocal of time function and
matching the tail end of the data to an exponential function.
An example of this type of curve fitting can be seen in
Fig. 48.
If we take the earliest density decay data to be of the
form
n = ^ , (9)
with A and B being constants, then a plot of t versus 1/N can

be used to find A and B for each data set. The average
values of A and B for the five voltage levels are given in
Table 1. For recombination, the continuity equation can be
approximated by
n = -ari^ . (10)
The constant of proportionality, a, is called the

recombination coefficient. If Eq. (9) is substituted into
Eq. (10) and a solved for, then a turns out to be the
75
onential Fit
400 1500
Time (|is)
Figure 48. Curve fitting of post conduction density

data to recombination and recovery curves
76
Voltage [kV] A B a
1 7.562e+15 -267.9 1.322e-16
2 2.011e+15 -457.8 4.973e-16
3 7.389e+15 -809.6 1.353e-16
4 6.545e+15 -989.6 1.530e-16
5 6.187e+15 -950.8 1.620e-16
* equation: t[ns]= ^T ~ ^
n [ cm ]
Table 1. Average recombination values

77
reciprocal of A. As the table shows, the recombination
coefficient for all charge levels feel within the same order
of magnitude. However, the value does seemed to slowly
increase as the charge voltage (and its associated peak
current) is increased except for a drastic jump at the 2 kV
level. One note about this level was that the current was at
a value where the tube just barely turned off and back on
again as the current oscillated for a few half cycles.
Therefore, the interferometry output was observed to
oscillate for a few cycles, cease due to cutoff, and then
return to oscillations after current had ceased.
Next, in fitting the diffusion part of the decay data,
the tail section of each density versus time set was fitted
to an exponential (n proportional to exp[-Ddft]). A listing
of the average values for D^f is given in Table 2. Again, all
values in question fell within the same order of magnitude.
The average value of Ddf did seem to increase with the charge
voltage level except at the 4 kV level. Table 2 also lists
the average time between the end of current conduction and
the end of the interferometer's output, T^^j^, for the five
charging voltage levels. These values will be used in the
recommendations section of this dissertation.^^
Spark Gap Ignitor Studies

The jitter comparison studies between the spark gap
ignitor and the conventional ignitor were done with the
graphite anode in the DIG and the test circuit in the
critically damped configuration. For the conventional
ignitor, ten shots were taken for each of two ignitor
triggering levels and two bank charging voltages. A typical
trace for the current through and voltage across the ignitor
can be seen in Fig. 49. Figure 50 displays a typical data
shot showing the current through the ignitor and the current
through the ignitron. The time delay between the beginning
of the ignitor current and the beginning of the current
through the DIG is the quantity of interest in the gathering
78
Voltage [kV] Ddf Tdk [^^s]
1 1.823e-3 1944.0
2 2.480e-3 947.0
3 3.004e-3 1392.7
4 2.204e-3 1503.6
5 4.192e-3 972.7
* e q u a t i o n n=const*exp (-%f *t [|is])
Table 2 Average diffusion rate constants

79
M
O
4-)
•H
C
CP
-H
(d
c
o
a •H
CM 4-)
C
0)
>
c
o
u
_ » •
CO
CO
o
u
o
(d
CM CU
cn
(d
4-J
.-H
o
TI >
d
s• •O
c
e (d
u
o x:
a cn
p
o
VH
0 Xi
4-)
4->
c
CU
u
M
p
o
cn
c
-H
:*
o
CO
CO

C

c
o
o
£
p
o
M
4-)
4-)
c
<u
V4
p
u
-o
a c
(d
C!)
M
Q
x:
p
o
u
Xi
4J
4J
c
0)
iH
M
p
o
o
(OOI/V) JO:^TU6TI ' (0T/V5{) fixpi in
Q)
P
Cn
•H
(X4
81
of jitter statistics. The best resolution of the digital
oscilloscope used to take these data was 50 ns. Table 3
lists the delay times for all test shots with the
conventional ignitor along with the mean and standard
deviation values for each data set. Since the goal of this
testing was to do an analysis of the jitter, the value of
interest is the standard deviation of the delay values. As
can be seen, the smallest deviation occurred when the bank
was at 2 kV and the ignitor trigger generator was set to
3.5 kV. The largest standard deviation occurred when the
bank was at 2 kV and the ignitor trigger generator was set to
2 kV.
The same type of measurements were done with the spark
gap ignitor. Ten shots were done with the capacitor bank
charged to 2 kV and then to 5 kV. However, the trigger
voltage on the ignitron trigger generator was set to a single
level that would break down the spark gap ignitor for all
shots. This level was 2.5 kV on the output of the trigger
generator but measurements of the spark gap voltage, such as
that seen in Fig. 51, showed that after the transformer the
peak voltage was only 4.4 kV or less. Inspection of the
spark gap ignitor revealed a gap between the ignitor tip and
the mercury pool of less than 0.125". Tempering of the
ignitor was also revealed in inspections made after its use.
Table 4 displays the delay values for all shots done with the
spark gap ignitor. As can be seen, the results are very
similar to the results for the conventional ignitor.
Therefore, this spark gap ignitor successfully triggered the
DIG without a drastic change in the jitter range.
82
Vcharge = 2 kV Vcharge = 5 kV
Trigger
Level V=2 kV V=3.5 kV V=2 kV V=3.5 kV
0.60 0.65 0.45 0.65
0.45 0.70 0.65 0.60
0.80 0.65 0.40 0.80
0.75 0.60 0.50 0.45
Delay (|is) 0.60 0.75 0.60 0.60
0.50 0.65 0.60 0.75
0.45 0.75 0.60 0.70
0.80 0.75 0.55 0.60
0.75 0.80 0.65 0.55
0.70 0.65 0.70 0.60
Mean 0.64 0.695 0.57 0.63
Standard 0.139 0.0643 0.0949 0.101

Deviation
Table 3. Conventional ignitor trigger delay data

83
1333.33 V/V
Figure 51. Voltage across spark gap ignitor

84
Vcharge = 2 kV Vcharge = 5 kV
Trigger
Level V=2.5 kV V=2.5 kV
0.80 0.60
0.70 0.60
0.65 0.50
0.55 0.50
Delay (jls) 0.60 0.70
0.65 0.75
0.85 0.60
0.70 0.45
0.60 0.40
0.65 0.35
Mean 0.675 0.545
Standard 0.092 0.128

Deviation
Table 4. Spark gap ignitor trigger delay data

CHAPTER IV
CONCLUSIONS/RECOMMENDATIONS
In this work, different features of ignitron performance

have been experimentally investigated. The research areas
included time-lapse plasma density during and after
conduction, triggering mechanisms, and anode geometry.
Additionally, results of feature modifications have been
reported. All of this work was geared toward an improved
general understanding of ignitron behavior and an
investigation of possible avenues of improvement for
ignitrons geared toward use in the pulsed power industry. In
the following section, each area of study will be addressed
individually in terms of a summary of experimental results
and overall in terms of general contribution to the ignitron
field. This section will include a mention of a new
commercial ignitron which has resulted from some of the work
done here and the improvement in switch parameters that has
resulted. The final section will cover overall
recommendations on how the research could have been improved.
Results/Achievements
Anode Studies
The primary goal of the photographic anode studies was
to determine whether anode sculpting could be used to improve
ignitron performance by concentrating the main discharge away
from the ignitor and/or inducing discharge motion across the
surface of the anode. The photographic studies of the plasma
behavior close to the anode, and especially the comparison of
standard anodes to sculpted ones, revealed that sculpted
anodes induce plasma motion, which greatly reduces anode
erosion and contamination of the mercury pool. A further
discharge control using a cup-shaped anode reduces the plasma
density at the ignitor position. Again, this feature will
increase tube lifetime by protecting the ignitor from the
main discharge. It is believed that the addition of these
anode features to commercial ignitrons will improve ignitron
85
86
lifetime by decreasing the vaporization of material from the
anode and/or protecting the ignitor.
Examinations of the anodes after use did reveal
differences. The flat graphite anode, although showing wear
and discoloration overall, displayed a very dark circular
area at one side of the rim. This area was also deeper,
indicating more material had vaporized from this spot. This
was an example of the effect that a stationary high intensity
discharge can have on a flat graphite anode. During some
shot, a discharge column anchored at this spot and vaporized
a great deal of material. The discharge from successive
shots tended to migrate toward the same area and released
more and more material. None of the sculptured anodes
displayed any anchor spots. Part of this may be due to the
fact that they were sculptured and part due to the fact that
the sculptured anodes were made of stainless steel, which
makes them able to conduct more of the heat from the
discharge into the anode rather than vaporize material from
the surface.
Interferometric Studies
For the first time, the electron density during
conduction and post conduction phase has been measured
interferometrically. These data are essential for the
understanding of the basic conduction and recovery mechanisms
of ignitrons. The optical interferometry provided an
estimate of the plasma density with values on the order of
10^^ cm"-^, at a peak current of approximately 115 kA. The
microwave interferometry provided information on the recovery
phase of ignitrons. The recombination phase had a
recombination coefficient of approximately 2.17*10"^^ cm^/s,
and the diffusion phase had a time constant of approximately
2.74*10-^ s"-'-. Both of these parameters depended only slightly
on the charging voltage. Some of the data concerning the
general shape of the recombination curve has been pointed out
87
in previous literature. In a paper by Turner and Butler,
indirect measurements of the electron density in the
deionization phase of an ignitron showed a two-stage
exponential decay.^^
Spark Gap Janitor Studies

This dissertation has demonstrated the successful use of
an experimental spark gap ignitor. The spark gap ignitor is
expected to increase ignitron lifetime by at least one order
of magnitude because of its lack of wetting potential and
shock wave insensitivity. Comparative studies showed that
this ignitor had similar delay and jitter characteristics
when compared to the commercial ignitor. Special care,
however, should be taken in the designing of a spark gap
ignitor for commercial ignitrons regarding the orientation
dependent gap width. Orientation dependence of ignitrons is
already a problem, and an improper spark gap ignitor design
could only worsen it. The breakdown potential between two
electrodes depends upon the^spacing, thus the shifting of the
mercury pool level by any means (including tilting the tube),
will affect the spark gap trigger voltage, which may or may
not be a problem in terms of the available trigger voltage.
If the spark gap trigger is properly overvoltaged, then it
could make the ignitron less orientation dependent than with
the conventional ignitor. On another point, a spark gap
ignitor operated at negative polarity would be even easier to
trigger because of enhanced electron emission at the tip of
the spark gap.
Ovprall Anhifivfiments
Some of the first results and recommendations resulting
from the previously described research have already been used
in the manufacturing industry. In particular, Richardson
Electronics has produced a next generation pulsed power
ignitron, the NL-9000. The change in general ignitron
geometry that has resulted from the years of discussion and
88
research can be seen in Fig. 52. As can be seen, the anode-
cathode spacing of the new ignitron is smaller and the cup
anode has been adopted to confine the main discharge away
from the centrally spaced ignitor. Additionally, two
separate water cooling coils allow for the cathode area to be
at a lower temperature than the tube wall. This feature aids
in keeping mercury droplets from adhering to the walls of the
tube, an occurrence which can harm tube lifetime. For better
performance in the rigorous pulsed power industry, the anode
connection and insulating feedthrough of the new pulsed power
ignitron has been strengthened. Table 5 shows a comparison
of some of the general ratings for the NL-9000 and previous
high power ignitrons.^^ As can be seen, there has been a vast
improvement towards some of the ratings that the pulsed power
industry has been needing.
Rf^ commendations
There are many ways in which these lines of research
could have been improved. Overall, a larger viewport window
would have been a great diagnostic improvement. Another
lower viewport through which to view the discharge would have
allowed for viewing a shorter anode-cathode spacing and a
better investigation of the anode effects.
Although the interferometers performed properly and
information was obtained from their data, this area of study
could have been improved in three major areas. Figure 53
displays the approximate density ranges which were detectable
by the two interferometry systems. Although the optical
interferometer with the argon laser was useful at the higher
current levels, too much of the plasma density seemed to fall
below the optical interferometer's density range and above
the microwave interferometer's range. More detailed data in
this area could have generated density versus time data that
could have been used to compare the anode geometries.
Additionally, a better method needed to be determined which
could accurately map out and reference the two-dimensional
89
U
P
4-J
O
P
M
4-)
CO
C
eo O
u
t. 4->
-H
c C
Cn
•rH
0^
c
O
C
•a (d
0)
"3 M
P
4->
O
P
M
4-)
CO
o
M
4->
-H
C
Cn
-H
CO
p
c o
o -H
>
0)
M
04
C CN
in
Q)
>H
p
cn
-H
[L4
90
Parameter NL-9000 Previously
Holdoff >10 kV >15 kV

Voltage
Peak Current 1 MA 200-400 kA
Peak Coulombs 500 C 300-400 C
Repetition 1 shot/minute >4 s h o t / m i n u t e

Rate at maximum a t maximum
(mostly a ratings ratings
thermal limit)
can do 700 kA, c a n n o t do b o t h

Overall 250 C ratings
capabilities simultaneously s i m u l t a n e o u s l y
arc resistance
has been found
General arc to be much s e v e r a l mQ
resistance lower than
previous tubes
(<1 mQ)
Dissipates
less power at
Comments higher current
than previous
models
Table 5. Parameter comparison of NL-9000 to previous

high-power ignitrons
91
Problem optical
range Interferometer
xoeter < Argon)
I i i i i i i i f I I I
1 0^2 1 0^' 1 0^^ 1 0^^ 1 0^^ 1 0^^

-3
Density (cm )
Figure 53. Comparison of detectable density ranges for

interferometers.
92
fringe pattern exposures so that two dimensional density data
could be easily obtained. An improvement in these data could
have been made by using an optical interferometry with a CO2
laser. The calculated density range for this interferometer
is also shown in Fig. 53. Assembling this type of
interferometer was considered. It would have required
special optics and special viewports. The detector would
have also been a problem as the output of a CO2 interferometer
would not have been as easily photographed as the optical
interferometer discussed in this report.
A second area of improvement would have been the ability
to use the Megaframe camera for the optical interferometer
data. The long interval between the exposures of the Dynafax
camera no doubt obscured some of the gradual fringe shifting.
Unfortunately, the camera was not donated until the latter
half of this research and the designing of the timing
circuitry delayed the camera's utilization even further. A
last area which could have improved the data from the
interferometers would have been to use the critically damped
circuit configuration rather than the crowbarred
configuration. The optical interferometry was done first,
and in order to get the highest peak currents and longest
conduction periods, the crowbarred configuration was used.
The microwave interferometry was therefore done in the same
circuit configuration. This turned out to cause a problem in
the data. It is now believed that the oscillatory current
waveform influenced the plasma density decay data to a higher
degree than the overall peak current value. This effect can
be seen in Fig. 54. This plot shows the similarity of the
trend for T^k and the value of the final peak current for the
various bank charging levels. It is thought that more
information could have been gleaned from the microwave
interferometer data if the circuit had been run in the
critically damped configuration.
93
10000
CO
Avg. Tdk
10001
M
O
1001
4-»
Avg, Peak of F i n a l
C
(V
101 Conciuction Perioci
M
M
U
2 3 4 5
Charge V o l t a g e (kV)
Figure 54. Plot of T^ and f i n a l peak c u r r e n t

REFERENCES
1. Proceedings of the first ''Workshop on High-Power,
High-Coulomb Ignitrons," edited by M. Kristiansen and
E. Honig, Texas Tech University, Lubbock, Texas,
April 21, 1987.
2. National Electronics, Industrial Tnh^s by National:

Operation and Maintenance Hints for Ignitrons.
Thvratrons. and ReotififtrR^ pp. 1-14, n.d.
3. General Electric Tube Products Department, ignitrons:
Capacitor Discharge and Crowbar Serviof^^ publication
PT 57B, October 1974.
4. C. R. Spikings, C. M. Oxley, and J. R. Beacham, ''High
Current Evaluation of the New Size E Ignitrons With a
Ringing Waveform," Proceedings of the 8th Pulsed Power
Conferencer June 1991.
5. D.L. Adkins, AxJal Magnetic Field Effects on Ignitron

Switchesr Masters Thesis, Texas Tech University, 1988.
6. H. de B. Knight, L. Herbert, and R.C. Maddison, "The
Ignitron as a Switch in High-Voltage Heavy-Current
Pulsing Circuits," Proc. of TEE. vol. 106, pt. A,
supplement n2, pp. 131-137, April 1959.
7. A. E. D. Heylen, "Paschen Characteristics of Gases in a
Crossed Magnetic Field," University of Leeds, LS2 9JT,
1980.
8. D.L. Adkins-Loree, M. Giesselmann, and M. Kristiansen,
"Dielectric Strength of High Power Ignitrons in Axial
Magnetic Fields," Proceedings of Lh£ 1MB ILEEL
conferenr-e on Electrical insulatjon and Dielectric
Phenomena. October 1988.
9. J. Burke, Plasma Diagnostios on High Current Ignitrons,
Masters Thesis, Texas Tech University, 1989.
10. P. Slade, "The Vacuum Interrupter Contact," IEEE Trans.
on Components. Hybrids, and Manufacturing Technology,
vol. CHMT-7, no. 1, pp. 25-32, March 1984.
11. Proceedings of the second "Workshop on High-Power,
High-Coulomb Ignitrons," edited by M. Kristiansen and
E. Honig, Lawrence Livermore National Laboratory,
Livermore, California, April 21, 1988.
12. H.N. Schneider, "Contact Structure for an Electric

Circuit Interrupter," U.S. Patent 2,94 9,520, Aug. 1960.
94
95
13. J. Thompson and M. Kristiansen, Plasma Diagnostic

Seminar, 2nd ed., Texas Tech University, pp. 105-132,
June 1976.
14 Orazio Svelto, Principles of Lasers, Plenum Press, New

York, 1976.
15. A. S. Gilmour, Jr., Microwave Tubes, Artech House, Inc.,

Norwood, Maine, 1986.
16. J. E. Burke, D. L. Loree, M. Giesselmann, and M.

Kristiansen, "Optical Study of the Effects of Anode
Geometry on the Performance of an Ignitron," Proceedings
Of the 7th IEEE Pulsed Power Conferenoe, June 1989.
17. D. L. Loree, M. Giesselmann, M. Kristiansen, and A.

Shulski, "Plasma Diagnostics for High Power Ignitron
Development," IEEE Transactions on Magnetics, vol. 27,
no. 1, pp. 181-184, January 1991.
18. Wayne Rasband, IMAGE: Image Processing and Analysis,

ver. 1.2 9, National Institute of Health, Research
Services Branch, NIMH, 1990.
19. D. L. Loree, J. E. Burke, M. Giesselmann, and M.

Kristiansen, "Preliminary Interferometric Investigations
of a Demountable Ignitron," Proceedings of the 7th
Pulsed Power Conference, June 1989.
20. D. L. Loree, M. Giesselmann, M. Kristiansen, A.P.

Shulski, and R. Kihara, "Recent Advances in High-Power
Ignitron Development," IEEE Transactions on Electron
Devices, vol. 38, no. 4, pp. 720-725, April 1991.
21. A. A. Garamoon and A. S. Abdelhaleem, "Properties of

Electron Swarms in Mercury Vapour, " J. Phys. D: Appi.
Phys.. vol. 12, 1979.
22. Francis F. Chen, Introduction to Plasma Physics j^^^

Controlled Fusion. Volume 1; Plasma Physics, 2nd ed..
Plenum Press, New York, 1984.
23. D.L. Loree, M. Giesselmann, E. Loree, M. Kristiansen,

and A. Shulski, "Recombination Studies of a High Current
Ignitron," Proceedings of the fith Pulsed Power
Conference. June 1991.
24. T, F. Turner and H. S. Butler, "Performance of Ignitrons

in Pulse Service," Proceedings of the 7th SymposJi^Tn on
pydrogen Thyratrons and Modulators, May 1962.
96
25. National Electronics, A' Division of Richardson
Electronics, Ltd., NL-9QQQ Data Sheet, LaFox, Illinois,
n.d.
26. M. Kristiansen and J. Francis, Some Basic Concepts and

Fundamentals of H . V . Generators, U.S. Air Force Pulsed
Power Lecture No. 2, Texas Tech University, n.d.
27. Stanley Marshall and Gabriel G. Skitek, Electromagnetic

Concepts and Applications, 2nd ed., Prentice-Hall,
Inc., Englewood Cliffs, New Jersey, 1987.
28. Max Born and Emil Wolf, Principles of Optics, 5th ed.,
Pergamon Press, Inc., New York, 1975.
29. Lawrence Livermore National Laboratory, LLNL ICE Program

Executive Summary and Technical Contracts. Nova Upgrade
Campaign. Book 1, UCRL-TB-104287, August 29, 1990.
30. Lawrence Livermore National Laboratory, LLNL ICE Program

Precision Nova and Nova Upgrade Laser, Nova Upgrade
Campaign, Book 2, UCRL-TB-104288, August 29, 1990.
31. Louis L. Reginato and Bob H. Smith, "Construction and

Testing of the 2XIIB Flat-Topping Power Supply,"
Proceedings of the 6th Symposium of Engineering Problems
of Fusion Research, November 1975.
32. Paul Horowitz and Winfield Hill, The Art of Electronics,
2nd ed.. University of Cambridge Press, New York, 1989.
33. Phasetronics, Specifications and Installation Sheet,

PTR6000 Three Phase SCR Gate Drives, n.d.
APPENDIX A
SCHEMATICS AND DESCRIPTIONS
97
98
Monocyclic Power Supply
A monocyclic power supply was used to charge the 2.5 6 mF
capacitor bank used for most of the work detailed in this
dissertation. The schematic for the supply can be seen in
Fig. A-1. This supply was designed with a special control
switch that can be used after the bank is charged, to prevent
the supply from recharging the bank after the shot.
Otherwise, as soon as the voltage on the bank is less than
the set point on the voltmeter, the power supply will turn on
and automatically begin charging the bank again.
Additionally, the control for the safety discharge system for
the banks (the dumps) was built into the supply.
A monocyclic power supply provides the shortest charging
time for a given maximum current capability. This is because
the supply is designed to deliver a constant current level,
independent of the load. This feature can best be derived by
doing a single phase example. The following derivation
refers to variables seen in Fig. A-2. In the figure, the
load impedance, ZL, represents the primary of the main
transformer of the power supply. The capacitor in parallel
with the load impedance can be said to be
^" . . 1 1 + jcocz, • ^^^^

^L^-^coc"
The voltage across the load is therefore

zjc
V, = jCOL
. . . . \+ Z
„ j..l ^C Vo , (A2)
which can be reduced to
\ = .. . . J A V ^ U...T "Ô • <A3)

Z L (1-0)^ L C ) + jCOL
If we now assume resonance, which is

99
k k k k ig ig
ap
CO
u
(D
o
u
-H
.-I
o
>1
u
o
c
o
I
<
0)
M
P
Cn
-H
100
^wm
+ +
V. V,
Figure A-2. Basic monocyclic circuit.
Figure A-3. Bank charging waveform for V^^^L^^^ = 2 kV.

101
COL = - A — , (A4)
ox:
then Eq. (A3) can be reduced to
V = -h^ V . {A5)
^ jCOL °
The current flowing through the load is therefore

VT V
I = - ^ = "TT;" ' <A6)
ZL DWL
which is independent of the load impedance.^^ This type of

supply will therefore attempt to drive a certain current
through whatever load is attached. Theoretically, if there
is an open circuit at the load terminals, an infinite voltage
will be produced as the supply attempts to drive current
through the open circuit. In practical terms however, some
component of the supply will be overvoltaged and destroyed.
To ensure that no overvoltages occurred, the supply used for
this project was designed with an overvoltage protection
circuit. This circuit senses the voltage on the primary and
can shut off the main power to the supply if the primary
voltage exceeds a preset level. Figure A-3 is a typical
charging waveform displaying the 2.5 6 mF capacitor bank being
charged to 2 kV by the 2 A monocyclic supply. At this
voltage level, the bank has a calculated value of 5.12 C of
stored charge. As can be seen in Fig. A-3, it takes about
2.6 seconds for the supply to charge the bank, which
corresponds to 5.2 C of charge, which is very close to the
calculated value.
Voltage T.pvel Trigger Generator

The crowbarred circuit configuration requires some means
for the proper timing of the trigger signal to the crowbar
switch. Some systems have simply used a set time delay
between the firing of the series switch and the firing of the
102
crowbar switch. This is a good method, but it takes some
initial adjusting of the timing to get it correct.
Additionally, the gradual changing of the circuit parameters
as components age can require the constant adjustment of the
timing delay value. For this project, an alternate scheme
was decided upon. Instead of using a time delay, the crowbar
switch would be triggered by a device which could sense the
bank voltage and transmit an output signal when the bank
voltage was near zero. A circuit was designed and built
which could sense the bank voltage with a 1000:1 voltage
divider probe, compare this value to an adjustable preset
value, and give an output signal when the voltages were
equal. In addition, the input of the device was designed to
be slope dependent, otherwise the device would trigger during
the charging of the bank. A schematic for the voltage level
trigger generator can be seen in Fig. A-4.
This circuit can be broken into four basic building
blocks: input conditioning, precision reference,, amplifier/
comparator, and pulse and reset logic. The input
conditioning block allows the user to AC or DC couple the
input signal into an input impedance of 50 Q or 1 Mil. The
variable reference level is generated by the precision
reference block. The base for this part of the circuitry is
a temperature controlled, precision 6.99 V Zener diode. This
reference voltage is used, along with an external 20 kQ
trigger level potentiometer, to create the reference voltage
with the aid of an operational amplifier (LM308) . The
voltage that the reference level has been set to is shown on
an external digital display. Both of the previously
discussed signals, the input signal and the reference
voltage, feed into the amplifier/comparator block. An
operational amplifier, LF357, both amplifies the input signal
and subtracts the reference level voltage from it. A high
speed comparator, NE5105, then takes this signal and compares
it to ground to determine if the input signal is above or
below the reference level. Two diodes at the inverting input
103
o
4-J
fd
M
c
cn
<D
Cn
-H
M
-P
(D
>
0)

M
O
<4-l
U
•H
4->
CO
Q)
£!
U
CO
I
<;
Q)
M
P
cn
-H
u4
UJ
u
UJ
s
104
of the comparator prevent the comparator from saturating and
losing speed. The TTL compatible signal from the comparator
then flows into the pulse and reset logic block of the
circuit. The slope select switch determines which voltage
slope the system should trigger off by setting one of the
inputs of an exclusive-OR gate (XOR gate). The rising edge
of a signal from the XOR gate, which occurs when the input
signal has the correct slope and is close to the reference
voltage level, triggers a monostable multivibrator (74LS123).
An output pulse will be generated by the entire circuit if a
flip flop (74LS74) is set in the ready mode and if the
monostable multivibrator is triggered as previously
mentioned. There are two ways for the flip flop to be set.
A reset pushbutton can be used to clear the flip flop and/or
the flip flop can be automatically reset by a second
monostable multivibrator which is triggered by the first
monostable multivibrator.
Megaframe Camera Timing Circuit

In order to synchronize the Megaframe camera, the light
valve, and the DIG, a special timing circuit had to be
designed and built. All signals of interest, .with their
proper timing sequence, are depicted in Fig. A-5. The goal
of the circuit is to enable when the "fire" signal is
received, wait until the camera's mirror is in the correct
position, which is when the photomultiplier emits a signal,
and then provide a variable width pulse for the light valve
and an alternate output pulse for the ignitor trigger
generator. The circuit designed for this purpose can be seen
in Fig. A-6. As can be seen, the photomultiplier signal from
the Megaframe camera is amplified, and then used as an input
to a comparator (LM311) which converts it to TTL compatible
square pulses. The "fire" signal, which is initiated to
begin a shot, is fed into the clock input of a flip flop
(74LS74), which enables another flip flop. The second flip
flop waits for the photomultiplier pulse before triggering
105
Photomultiplier _A_ J^-.

Signal
Comparator
Signal
"FIRE" Signal i
Shutter Pulse
Alternate Output
Pulse
TIME
Figure A-5. Timing diagram for Megaframe camera

106
u
ôwH? <D
B
to
o
B
(0
u
(0
cn
0)
M
O
•P
-H
P
O
-H
O
cn
c
-H
B
-H
EH
I
<
M
P
-H
107
both monostable multivibrators (74123) which provide the

alternate output and the shutter output. One component
accidentally left out of the circuit was involved with the
fact that the light valve is closed when the input voltage is
high and open when the input voltage is low. Therefore an
inverter circuit had to be inserted between the shutter
output of the Megaframe circuit and the light valve control
box.
APPENDIX B
FRINGE SHIFT EVALUATIONS OF A
MACH-ZEHNDER INTERFEROMETER
108
109
The theory behind interference phenomena in waves can be

used as a base from which the fringe pattern interpretation
for both of the. interferometers can be derived. To begin
with. Fig. B-1 shows a basic Mach-Zehnder interferometer with
various parameters which will be used in the following
equations. The basic motivation behind the theory is that
two beams produced by a common source are sent through
different paths and dielectrics and recombined at a detector
plane. The two beams can be represented mathematically in
plane wave form as
E,= C,e^"^^-'*' =C,e^*^ (Bl)

and
In this development, k^^ is equal to the phase constant of the

medium it represents. It has units of radians per meter and
multiplication by k^ converts a length into a phase angle. ^"^
The intensity recorded at the detector plane is then
I = |Ei+ E2 I = ICie"^"^^ +C2e-^'^^| . (B3)
This equation can be evaluated using Euler's Relation and

algebra to form
I = C^ + C2 + 2 CiC2COs((t)i - <)2), (B4)
which upon proper substitutions is then

I = C,^ + C" + 2 C,C2COS(k^(L,- L2) + (k^- k^)D). (B5)
Adjusting I to be zero without the plasma present results in

k^(Li-L2)=7C and Ci=C2. Being that the data for this
dissertation were all relative, meaning simply being able to
detect voltage oscillations or light and dark areas, the
constants Ci and Cj can simply be set equal to one.
Therefore, Eq. (B5) reduces to
110
Source-
V \
' k.
T
plasma
k,
\ D V-
Detector Plane
Figure B-1. Basic interferometry system.
Figure B-2. Intensity versus phase difference plot.

Ill
1=2 (1 - cos(k^- k^)D)), (B6)
which can be further reduced to
I = 4 cos^ ('^Y>'") . (B7)
In future discussions, the differential phase shift caused by

the plasma will be defined as A<J) = (k^ - ki)D. So, the
intensity recorded at the detector plane is a function, which
can be see in Fig. B-2, that depends on the phase difference
between the two paths caused by the plasma and varies between
a relative maximum value and a relative minimum value. •'••^'^®
Fringe Shift Evaluations Applied

to Microwave Interferometer
The interpretation of the data from the microwave
interferometer follows directly from the previous derivation.
Microwave interferometry is often used to measure the
electron density in laboratory plasmas. The phase velocity
of the high frequency wave propagating through the plasma is
a function of the electron density. This plasma wave is
combined with a reference wave and the output of the
interaction converted to a voltage waveform whose time
history can be related back to a plasma density versus time.
The phase constants for normal air and for a plasma can be
written as
k„= co^Vii;e7 = -^ and k i = co„V^l„eê,= —^-^—^ , ( B 8 )

with
c = speed of light, 3*10^0 [cm/s],
\i^ = free space permeability = 1.26*10"^ [H/m] ,
e^ = free space permittivity = 8.854*10-^2 [F/m],
COQ = radian frequency of microwaves.
and
112
COp= plasma frequency = S ne^ /mEo . (B9)
In Eq. (B9), n corresponds to the instantaneous average

plasma density, e is the charge of an electron, m is the mass
of the species. Another relation that can be derived from
Eq. (B8) is that the plasma can be considered to be a
dielectric with a relative dielectric constant of
er = (1 - (•%•) ) . (BIO)
CO
For this equation, a fully ionized plasma is assumed.

Additionally, notice that if COp > CO^ t h e n k^^ i s not a real
number and t h e microwaves w i l l n o t p r o p a g a t e t h r o u g h t h e
plasma. The c u t o f f d e n s i t y , N^., i s d e f i n e d as t h e d e n s i t y
where COp = CO^ and from Eq. (B9) can be w r i t t e n as
N^ = m e^(COye)^ [m"^] . (Bll)
The c u t o f f density is approximately 6 * lO-^-^ cm"-^ f o r the

70 GHz microwave i n t e r f e r o m e t e r used h e r e . Previous
e q u a t i o n s can be used t o d e r i v e a r e l a t i o n s h i p between t h e
d e n s i t i e s and t h e r a d i a n f r e q u e n c i e s , which i s
COp _ n
(B12)
CO? N^
Using all previous equations, the differential phase

shift, A<|), that the wave experiences as it propagates through
the plasma can be related to the cutoff density, N^, and the
instantaneous plasma density, n, by
A(t) = - ^ D [ l - ^1 - ^ ] . (B13)
In this equation, n is assumed to be constant over the entire

plasma (i.e. a uniform plasma density profile). If a
nonuniform density, n=n(z), is assumed then the phase shift
must be found by integrating the term in the
113
A(|) equation with the plasma phase constant, k^, over the
plasma length. Solving Eq. (B13) for n yields an equation in
a form which can relate phase shift data to the density,
n(t) = N e - (NC (l - ^ f ) . (B14)
Figure B-3 is an example of a generic output waveform and the

associated plasma density data which could be associated with
it. The change in voltage from point a to point b in
Fig. B-3 represents a phase change of n radians introduced
into the transmitted wave by the plasma by the time t^. The
density at this time is then calculated by inserting A<j)=7C
into Eq. (B14) . Two characteristics to point out about the
displayed example are that when the plasma density exceeds
the critical density, the interferometry waveform goes into
cutoff. Additionally, the changing plasma density
encountered as time passes attenuates the microwave signal to
varying degrees. The higher the density, the more thermal
and/or collisional effects attenuate the transmitted
microwave signal and unbalance the C^=C2 requirement which was
made when deriving Eq. (B6) .^^ The interpretation procedure
described here was the one used to interpret the plasma
density in this experiment.
Fringe Shift Evaluations Applied

to Optical Interfereometer
The interpretation of two dimensional fringe shift
patterns created by the optical interferometer involves the
same development as before; however, one new consideration
must be taken into account. In the previous derivations, the
detector plane detected the overall intensity present over
the entire beam interference pattern. For a two dimensional
interpretation, the angle between the two incident beams must
be taken into account. Figure B-4 displays the incident
beams and some necessary variables that will be used in the
114
TIME
Figure B-3. Generic microwave interferometer output

and associated density waveform.
115
reference
beam
plasma
detector
plane
-X
Figure B-4. Angle factor in two-dimensional fringes

116
following equations in addition to the terms displayed in
Fig. B-1. The addition of a angle into the plane wave
representations .of the beams, as in Eqs. (Bl) and (B2),
results in ^^ and ^2 being changed to
<|)i = k^x cos(0) + k^y sin (6) + k^L ^ - cot (B15)
and
^2 = ô^ -^ ô<L2" ^) + îD - cot . (B16)
The intensity recorded at the detector plane, x=0, follows

the same procedure as before and results in
I = C^+ C2+ 2C^C2COs(k^y sin (8) +
ko (Li- L2) + (ko- ki)D). (B17)
This function has been shown to exhibit maxima and minima.

For this case, the minima define dark fringes and are located
at positions defined by
k^y sinO) + k^ (L^- L2) + (k^- ki)D = (2d + 1) 7C , (B18)
with d = 0, ±1, ±2, .... Solving this equation for y will

shows minima can occur along the y-direction of the detector
plane at positions defined by
(2d + 1)TC - k^(L^- L2) - (k^- k^)D
y = . (B19)
koSin(e)
This type of interferometer can operate in two modes, with

6=0, which corresponds to an infinite fringe pattern mode and
with QÔ, which corresponds to a finite fringe pattern mode.
For the infinite fringe case, y='» and the minimum detectable
phase difference in 2% radians. The finite fringe mode
allows for smaller phase differences, up to .01*27C depending
upon the detection method, and the sign of the phase
difference to be detectable by means of detecting which
direction the fringes shift.
117
Figure B-5 is a generic example of a uniform shift
fringe pattern with some of parameters labeled for use in
further derivations. Figure B-6 is an example which shows
the two dimensional data that can result from this type of
interferometer if the change in plasma density is high enough
and localized within the limits of the beams. Equation (B19)
shows that with no plasma present, the minima are separated
by a distance, 5y, called the vacuum fringe spacing that is
equal to
5y = —^^^—7T- = — — , (B20)
k^sin(e) sin (9)
where we used t h e f a c t t h a t k^ = CO^/c = 27Cf/c = 2K/X. AS can

be s e e n , a s m a l l e r vacuum f r i n g e spacing r e s u l t s from 6 b e i n g
c l o s e t o 90°. The e q u a t i o n f o r t h e change i n minima
p o s i t i o n . Ay, c a u s e d by plasma can be o b t a i n e d by t a k i n g
d i f f e r e n c e between Eq. (B19) w i t h plasma and Eq. (B19)
without plasma. This e q u a t i o n can be w r i t t e n as
( k . - k )D A* 8y
Ay = —i 2-— = _1_JL . (B21)
^ koSin(e) 27C
with A<j) equal to the phase shift caused by the plasma that
can be detected by measuring the fringe shift data. Ay and
6y. Compared to the phase shift caused by the charged
particle density, the magnitude of any additional phase shift
by the neutral gas density in the ignitron is negligible.
The same equations for k^, COp, n, and k^ as previously
introduced hold. However, to be able to easily relate A(|) to
n, ki must be approximated by
,^=_^(l -4- --^(1 - -4-)- (B22)

' c ^ COo ^ 2(ol
From this, Acj) as defined previously becomes

118
6y = vacuum fringe spacing
Figure B-5. Example of uniform phase shift

fringe pattern.
.y
:sLZ- 5y,,
Figure B-6. Example of localized phase shift

fringe pattern.
119
2
A(h = ^ D _ n e^ D
^ 2cC0o ~ 2c m e^cOo ' ^^^^^
where Eq. (B9) was used. If we set the A(j) of Eq. (B23) equal
to the A<t) of Eq. (B21) , set CO^=27Cc/>., and solve for nD we get
n D = M g m E , ^ ^ (B24)
During all experiments, the optical interferometer was

aligned to produce finite fringes rather then infinite
fringes. Introducing the two possible main wavelengths of
the argon laser, 488 nm (blue) and 515 nm (green) we get:
n D [cm"^] = 4.570*10^^ -~- (for 488 nm) (B25)

and
n D [cm^] = 4.335*10^^ - ^ (for 515 nm) . (B26)
Using an estimated plasma length in the DIG of 15 cm we get:

dy
n [cm'^] = 3.050*10^^ -~- (for 488 nm) (B27)
Sy
and
-T^T _ o ortA + i r?-^
n r[cm^] = 2.890*10^^-5—- (for 515 nm) . (B28)
The frequency of the laser light will determine the minimum

as well as the maximum detectable plasma density. The lower
detection limit will be given by the smallest fringe shift
that can be evaluated in relation to the vacuum fringe
spacing whereas the high limit is reached when the light
absorption within the plasma becomes excessive. The high
limit is usually given when the plasma frequency reaches 10%
of the light frequency. Higher wavelengths mean lower
detectable plasma densities as well as lower maximum
densities.^^
APPENDIX C
PROJECT HERCULES
120
121
Late in the phases of the previously discussed research,
a contract was awarded to us to conduct testing involving
prototype ignitrons. The main objective of the contract was
to lifetime test some prototype switches for Lawrence
Livermore National Laboratory's (LLNL's) Nova laser upgrade
for their Internal Confinement Fusion (ICF) program.^9/30 rphe
laser will be upgraded from its current 100 kJ level up to
roughly 2 MJ over the next five to ten years. Our work is an
eighteen month project which began on July 15, 1991.
Detailed in this chapter will be the design goals and
preliminary diagnostic and control system designs for what we
have named Project Hercules.
Project Hercules will test prototype switches with the
required parameters of 300 kA peak current, 200 C per shot,
and a 25 kV voltage holdoff ability. The ultimate goal is to
recommend a switch which can do all the required parameters
for 10,000 shots. None of the first generation prototypes
are expected to fulfill all the parameters for the full
lifetime. Therefore, the^switches will be examined after
their failure, design changes recommended, and second
generation switches retested. The switches will come from a
variety of competing sources including Richardson Electronics
and English Electric Valve (EEV).
Basic Circuit Design

The first work done was the modeling of the test circuit
in order to ensure that the parameters could indeed be met
with existing or future laboratory supplies (i.e.,
capacitors, power supplies, and diagnostics). For a test run
of 10,000 shots to be reasonably monitored, it must not take
too long. It was then decided to try and do one shot per
minute so that a 10,000 shot test would take one week. To be
even more time saving, it was decided that two switches
would be tested at the same time. This feature would also
allow two switches to be run under an identical test
situation. All of these decisions affected the power
122
supplies and capacitor bank which would be required. Since a
simple critically damped system was to be designed, the bank
size, the charging voltage, and the needed coulombs could be
related by
Q = C * V (CI)
where
Q = charge = 200 C
C = capacitance of bank
V = charging voltage.
Through several designs, including the possible use of a
transformer to step up the current, it was apparent that only
one set of capacitors was available which could fulfill all
requirements. The capacitors in question were from a system
which in the past formed LLNL's 2XIIB Flat-Topping Modules
for their 2XIIB coils.^^ These capacitors were 1500 ÎF, 450 V
electrolytic capacitors made by Mallory and we had over
11,000 of them for our use. With these capacitors in mind, a
test circuit could then be designed.
The schematic of the test circuit can be seen in
Fig. C-1. The model for the ignitron switches came from a
model proposed in the first ignitron workshop. •'• The value
for the bank was arrived at by relating the necessary charge
voltage required to produce 300 kA to the needed coulombs
using Eq. (CI), keeping in mind the capacitors which would
have to be used. The circuit was simulated on PSPICE to help
optimize all parameters and to ensure that physically
reasonable system values could be developed. The PSPICE
input file can be found in Fig C-2 while the resulting
current through the prototype tubes can be found in Fig. C-3.
One note to make here is that the 25 kV holdoff parameter is
not being met by the main test circuit which works at only
5 kV. Therefore, a high voltage testing (i.e., hipotting)
circuit had to be added. This circuit, which was placed in
parallel with the second test switch, will test the holdoff
voltage for each tube (one with positive polarity, one with
123
o >
-H ,i^
4->
>iLr)
o •P
0)
M
4-) o
U B
CM
CO
Hh
-H CO
Xi 0)
u TJ rH
c p
fO -o o
M P U
fO X) <-H 0)
CO U
M G -H
<D
O 4-)
-H Q)
CJ > 4-)
O
0) O (U
U C •(—(
Q Q) (0
4-)
4-)
o
O M
CO
-H
CO U
0) O
M-4
^ 4->
U -H
4-> P
-H U
^ M
CO -H
•K O
c
Cn o
> -H
X 4-J
o fO
-^WyYYYyll- (0
-H
P
-H •
e
Q -H
CO
I
O
u
^ H -^ . 0)
M
P
V Cn
-H
[14
124
2 SWnCH TEST CKT. NO XFRMR. ELECTROLYTIC BANK 39.46mF (13 x 342)

.OPTIONS ITL5=0 ITL4=10000
CB 1 0 39.46mF IC=5070 ;Bank 342 X 13 n series(390V per,4446 tot)

RB2 3 6mOhm ;Resist of Bank and circuitry
LB 1 2 25nHenry ;Inductance of Bank
D1 0 31 DREC ;Power Diode to prevent Bank Reversal
RC 31 3 0.2Ohm ;Resistance in series with Power Diode
LS2 3 7 325nHenry ; Circuit Inductance
LIG2 7 8 125nHenry ;Inductance of 2nd Test Ignitron
S2 8 9 12 0 S2MOD ;Switch for 2nd Test Ignitron
.MODEL S2MOD VSWIT(^(RON=.004 ROFF=10E+7 VON=l VOFF=0)
VPLS 12 0 PULSE (0 2 5u 50n 50n 5m 8m)
RDM? 9 13 1.5460hm ;Damping Resistor for H.V. Cap
LHVC 13 10 lu ;Inductance in series with H.V. Cap
CHVC 10 0 1.88uF IC=25000 ;Capacitance of H.V. Cap
RIGl 9 11 3mOhm ;Resistance of 1st Test Ignitron
LIGl 110 125nHenry ;Inductance of 1st Test Ignitron
.MODEL DREC D aS=lE-9) ;Diode Model
.TRAN/OP lOu 3m 0 lOu UIC ;Normal Mode Tran. Anal., 3ms nin time
.PRINT TRAN I(LIG2)
.PROBE
.END
Figure C-2. PSPICE i n p u t f i l e for simulation circuit.

125
300000-
i 200000 -
4J
c
o
u
u
100000-
0.000 0.001 0.002 0.003
Time (seconds)
Figure C-3. Projected current for simulation circuit

126
reverse polarity) prior to the main discharge. The polarity
of the hipotting circuit will have to be periodically
reversed to maintain an identical test situation for the two
test switches.
Project Construction
With the main and highpotting circuits conceptually
formulated, the physical system could begin to be designed
and tested. The design of the capacitor bank was dealt with
first. The main circuit provided the requirements for
voltage capability and total capacitance. For a charging
voltage of 5.07 kV, thirteen capacitors in series would be
required so that each capacitor had only 390 volts across it
and 377 sets of thirteen 1500 \1F capacitors in series would
be required to make the calculated bank capacitance of
39.46 mF. This made a total of 4446 capacitors needed to
build the bank.
Placing electrolytic capacitors in series posed some
design problems. Electrolytic capacitors are generally known
for their nonuniform leakage resistances. Therefore, when
these capacitors are placed in series and a potential placed
across the group, the voltage does not necessarily divide
evenly across each capacitor even with the addition of a
voltage division network in parallel with the capacitors. To
demonstrate this point, a series set of ten electrolytic
capacitors with parallel resistors was built and tested.
Figure C-4 shows a schematic for the system and Fig. C-5
shows the resulting differential voltage across each
capacitor for a total charge voltage of approximately 2.4 kV.
Note how one of the capacitors is dangerously close to its
maximum voltage rating of 450 V while another capacitor has
only 120 V across it. If perfect voltage division was going
on then each capacitor would have had 240 V across it.
Similar results were obtained for other test sets. For a few
of the capacitors on the existing 2XIIB modules this was
taken care of by adding two, series connected, 200 V Zener
127
(^ U
PC u CO
o
4-)
CO
-H
CO
PC o Q)
U
x:
4-)
-H
a
^
h4
PC u 4J
(U
^
CO
o
o o ^4
o
00 LO 4->
<X) rH PC o -H
U
(0
II II fO
u
PC u CO
PC o 0)
-H
>^
0)
CO
PC u
I
o
0)
PC o P
-H
U4
PC u
(0
0)
PC u a>
<1> o
128
cn
(0
+->
o
>
Capacitor Position
No MOV
With MOV
Figure C-5. Differential voltage graph for series

capacitor sets.
129
diodes (1N3015A and 1N3015RA) in parallel with each
capacitor. This feature clamps the maximum voltage allowed
across each capacitor to 400 V. In looking into using the
same scheme for the design of the main bank, it was found to
be too expensive. The total cost of the diodes would have
been over $20 per capacitor coming to a bank total of over
$89,000. Therefore, another way to make the voltages across
the series capacitors divide evenly had to be found.
A device called a metal oxide varistor (MOV) was found
to be much less expensive and yet fulfill our needs. This
device resembles a disc ceramic capacitor and behaves
electrically like a bi-directional zener diode.^^ The
voltage-current characteristics of this device resemble those
of Fig. C-6. When an MOV is added in parallel to the
capacitor and resistor set, as seen in Fig. C-7, it will help
the voltages divide more evenly. Ten MOVs were added to the
series set of electrolytic capacitors and resistors which
were previously tested and the differential voltages measured
again. The results can be seen in Fig. C-5. As can be seen,
there is a vast improvement in the uniformity of the voltage
division among the set. As a result, almost 5000 MOVs will
be used for the capacitor bank. The type to be used are from
Allied Electronics, model S20K250, and have a 390 V knee
voltage that should properly protect the 450 V electrolytic
capacitors. For this quantity, the cost per MOV was only
40<:, which corresponds to a cost of only $2000 for the
entire bank.
With the basic capacitor bank components designed, a
preliminary floorplan for the entire system could be
developed. The cabinets housing the 2XIIB system of
capacitors could also be used to house the Hercules bank.
Six of the cabinets will be modified in order to be able to
hold up to 63 series capacitor sets each. The design for a
standard capacitor set can be seen in Fig. C-8. The entire
project will need the following components: a main power
supply/ a highpotting power supply, a capacitor bank, a
130
Current
Figure C-6. Generic metal oxide varistor voltage-

current characteristic.
131
o
LO
>^r p:;< a^S
r" ^
S20
^"l o
oo CO
t
CO LO >
P^< p— o II o
II II > •c
ors an
o
MOV
P^ CJ S
PC< Z (J
4-)
CO
-H
CO
> ^
o ^ PC< o M
s^ ^ ^ ^ M
.C
4-)
-H
U;
> ^ / ^ ^
set
PC< Z o
^**-- >-i
o
4-)
-H
> ^ U
(^< Z (J (0
s-* (0
ri
Series (
MOV
E ^ < Z CJ
MOV
E
; C-7.
C^< Z CJ
U
Figi
^ <
Z u
age
>5
23t PC < • ^ —
^T^ •0
132
CO

VH O
(L> (U
-o
^ • i ^ •r—i
O
O
c«
O
*-> u
Idered
O
piece
4-1
4->
<D
CO
c S
«3
VH
O M
o
•a V-t
r> -4—>
"-1
O
4-J
o
<D o cd
•»-» o -H
O
a c« Cd (0
O D O a
o
VH ^
i:Ô CO
O
.1:^ cd
o
CI. -H
O TN C/5 4-)
O ^ O >i
i—t VH rH
- Cd o O
in par
Ited a
cn u
_x 4->
O
• 0)
r* rH
> ^ 0)
VO O TS 4H
r—• O
1-H s § e
(0
cn
(0
•c •H
Q
o
.'-I
oo
CH ;-i I
r/1 O U
4—>
paci
lina
<U
u
cd cn
(U
^-t
o -H
<D
o bO
b^ C
H 03
133
highpotting capacitor, a tes^t stand, a cooling water
circulating system, various diagnostic and control devices,
and a screenroom. A 12 kV, 10 A power supply was obtained to
charge the main bank of this project. However, in the
existing constant voltage configuration, the supply would
require a large current limiting resistor and it would take
too long to charge the bank. It is therefore planned to
convert this supply to 8 A constant current by means of phase
controlling the primary of the power supply's transformer
with an SCR controller.^^ This supply will then charge the
bank in only 25 seconds. Using existing component
dimensions, the basic floorplan worked out for Hercules can
be seen in Fig. C-9.
Diagnostic and Control Design

Discussions with the contract monitor at LLNL revealed
that certain data would be required to fully appreciate the
switch performance for the project goals. The peak current
value and corresponding anode-cathode voltage value at peak
current would be required for every test switch and every
test shot. The complete waveforms for current and voltage
would be required every 25 shots. The ignitor resistance
values would also be needed approximately every 25 shots. In
addition to the data related monitoring, safety and control
monitoring will also be required. It was decided to make the
entire test system as automatic as possible with the use of
computer control via a GPIB interface. A concept flowchart
for a single test shot can be seen in Fig. C-10. The actual
diagnostic plan was arrived at by considering what equipment
was available. The preliminary diagnostic and control plan
arrived at for Hercules can be seen in Fig. C-11. A 100 MHz
Yokogawa digital oscilloscope will be used to acquire the
entire waveforms when needed. A GPIB compatible system
voltmeter, HP3437A, will be used to acquire the peak levels
needed every shot. Additional circuitry to control various
solid state relays with the GPIB interface and to switch the
134
M il
Double lines of
Low Inductance
Coax
17'
Probe Amplifier
33'
Trans Feedthrough 4'

former Panel
•H BNC h"
t^le
I tif [ i i n t t M i I tri ,tt**mmu*m,4ntnmnii
Table
I r 111111 m 11M
Screenroom 10'
Desk
-\ IXX?R h
Figure C-9. Basic floorplan for Project Hercules

135
Is everything ready??
safety, power, diagnostics,
cooling, automation, shot #
r Dump the HV cap

îve Breakdown erro Dump the HV capacltoi
Dump main bank I

Give charge erroij
G Dump main bank

Give ignitor erro)
Try again
Wait one second, then dump main bank

Give current erro
3
Save peak current and corresponding
voltage for both switches; save
entire waveforms every 25 shots
Yes
(
Return to I
^
Yes
k Reset diagnostics
Give ignitor check signa
3
Figure C-10. Process flowchart for Project Hercules
•WP«
136
Supply
25 kV
Power CO
0)
o
n
ac
Relay
O
(I)
•t—I
L.-.
O
u
M
O
o
-p
c
o
o
n
c
nj
o
•H
c 4-)
OT
O
3:: s &
C
cn
m
•H
c «« "C
c^ c/5 a^
-o
« 13 & >1
(0
c
O C/3 •H
> e
•H
rH
0)
M
I
U
Q)
U
3
•H
xMMMWMMMM
^*MMMMM*MM.M*
137
peak data into the voltmeter need to be designed. There is a
possibility that the automatic functions of the digital
oscilloscope could perform the peak acquisition function.
Status Rfipnrt
At the time of the writing of this dissertation,
construction on the project had been going on for some four
months. A 10' by 10' screenroom had been assembled, wired
with proper power and filters, and fitted with wheels. Over
one half of the capacitor sets had been assembled and placed
into the cabinets. Prior to the sets being inserted, each
cabinet was totally stripped, four 900 pound casters mounted
on the bottom, and holes drilled for the proper shelf
spacing. Parts to modify the power supply had been gathered.
The SCR controller board had been chosen and ordered. The
necessary six SCR's and their heatsinks had been acquired.
Design of the test stand was in progress. Additionally, some
programming for the computer control system had begun.

High Power Ignitrons

Uploaded by

Copyright:

Available Formats

High Power Ignitrons

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

High Power Ignitrons

Uploaded by

Copyright:

Available Formats

COMPREHENSIVE STUDY OF HIGH POWER IGNITRONS

^j ^ *> This degree is a culmination of many years of public and

This dissertation describes research involving various

An ignitron is a triggerable enclosed switch that has a

Figure 1. Basic ignitron.

Figure 2. Conventional ignitor.

Axial Magnetic Field Effects

Figure 5. Diagram of Dynafax framing camera

Graphite Anode Cup Anode Slit-cup Anode

Figure 6 Side view of three initial anode geometries.

Figure 7. Differential voltage versus peak current for

Manor Test Circuits and Main Diagnostics

Critically Damped Configuration

Capacitor bank, 2.56 mF

Figure 8. The two b a s i c t e s t i n g c i r c u i t s .

Figure 10. a) Cut-away view of test stand with

Figure 10. b) Cut-away view of test stand with

Sculptured Anodes and Arc Rotation

Polaroid film pack mount

Figure 11. Diagram of the Megaframe camera

094 Dia. hole along central axis

Figure 13. Machine drawing for cup anode

.094 Dia hole along central axis

Figure 14. Machine drawing for s l i t - c u p anode

0.094 Dia. hole along centnl axis

Figure 16. Machine drawing for spiral anode

Figure 17. Demountable ignitron with modified anode

Spark Gap Ignitor

HIGH VOLTAGE FEED-THROUGH

Initial spark gap tip

Figure 20. View of spark gap ignitor.

Now that all of the test circuits, diagnostics, and

Photographic Study of Anode Geometrv

Figure 21 CT i t i c a l l y damped w a v e f o r m s f o r Vj^a^^j^. = 2 kV

Fignie 22 C r i t i c a l l y damped w a v e f o r m s f o r Vj^g^^ 5 kV

(A^) a5e:noA ' (O^/V^l) :^u9Jano

Figure 25. Photographic sequence for graphite anode

Bank level = 4 kV, 45 |ls interframe time

Figure 26. Photographic sequence for cup anode.

Figure 27. Photographic sequence for slit-cup anode

Figure 28. Exposure p o s i t i o n for Megaframe camera

Bank l e v e l = 5 kV, 1.8 ^is i n t e r f r a m e t i me

Figure 29 Photographic sequence for spiral anode

Figure 29. (continued)

Bank level = 5 kV, 1.8 ^is interframe time

Figure 30 Photographic sequence centered at peak current

Laser line filter removed, 45 |ls interframe time

Figure 32. Optical interferometer sequence of coarse

Figure 33 Optical interferometer sequence of fine

Bank level = 5 kV, 45 ^s interframe ti me

Figure 34. Adjusted optical interferometry sequence

Figure 35. Fringe shift detection scheme

The measured electric field at 115 kA is approximately

Notice, the index of refraction of the mercury ions is much

Figure 43. Plasma density versus time for 1 kV level

Figure 44. Plasma density versus time for 2 kV level

Figure 45. Plasma density versus time for 3 kV level

Figure 46. Plasma density versus time for 4 kV level

0 500 1000 1500 2000

Figure 47. Plasma density versus time for 5 kV level

E,= C,e^"^^-'' =C,e^^ (Bl)