A Fast, 3 MV Marx Generator

A FAST, 3 MV MARX GENERATOR FOR MEGAVOLT OIL SWITCH
TESTING AND INTEGRATED ABRAMYAN NETWORK DESIGN
________________________________________________________________________
A Thesis
Presented to
The Faculty of the Graduate School
University of Missouri-Columbia
________________________________________________________________________
In Partial Fulfillment
Of the Requirements for the Degree
Master of Science
________________________________________________________________________
by
LAURA K. HEFFERNAN
Dr. Randy Curry, Thesis Supervisor
DECEMBER 2005
ACKNOWLEDGEMENTS
I would like to thank my advisor, Dr. Curry, for his guidance throughout this research
and Dr. Kovaleski and Dr. Bryan for their collaboration. I would also like to thank all those that
assisted me in the many facets of this project including Dr. McDonald, Peter Norgard, Mark
Nichols, Josh Leckbee, Keith LeChien, Scott Castagno, and the many undergraduate assistants.
A special thanks goes to Mom, Dad, Mary, Brian, and Paul for their support and to Tim for his
constant encouragement.
ii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ..................................................................................................... ii
INDEX OF FIGURES.............................................................................................................. v
INDEX OF TABLES............................................................................................................... vi
CHAPTER 1. INTRODUCTION............................................................................................. 1
1.1. Background
1
1.2. Marx Generators
1
1.3. Abramyan Networks
4
References for Chapter 1
7
CHAPTER 2. DESCRIPTION OF SYSTEM ........................................................................... 9
2.1. Pulsed Power Research Laboratory
9
2.2. UMC Marx Generator
10
2.3. Charging
15
2.4. Triggering
20
2.5. Control Systems
22
2.6. Peaking Capacitor and Oil Switch
24
29
CHAPTER 3. DIAGNOSTICS............................................................................................... 31
3.1. Requirements
31
3.2. Voltage Divider Theory
32
3.3. Post-switch Divider
35
3.4. Pre-switch Divider
38
3.5. Calibration
40
45
CHAPTER 4. MODELING OF MARX................................................................................. 47
4.1. Marx Circuits
47
4.2. PSPICE Simulations
47
4.3. Breakdown Calculations
49
4.4. Electrostatic Simulations
53
56
CHAPTER 5. SWITCH EXPERIMENTS .............................................................................. 57
5.1. Experimental Setup
57
5.2. Experimental Results
60
65
CHAPTER 6. ABRAMYAN NETWORK DESIGN .............................................................. 66
6.1. PFNs, PFLs, and the Abramyan Network
66
6.2. Physical Implementation of the Abramyan
69
6.3. Circuit Simulations
72
82
CHAPTER 7. CONCLUSION................................................................................................ 84
APPENDIX A. SPARK GAP DIMENSIONS ...................................................................... 86
APPENDIX B. MARX SIMULATION CODE..................................................................... 87
iii
APPENDIX C. MARX OPERATION .................................................................................. 91

APPENDIX D. MARX SHOT LOG ..................................................................................... 93
APPENDIX E. MATLAB CODE ......................................................................................... 96
APPENDIX F. MATHEMATICA CODE .......................................................................... 100
APPENDIX G. ABRAMYAN SIMULATION CODE ....................................................... 103
iv
INDEX OF FIGURES
Figure 1.1. Marx stages.............................................................................................................. 2
Figure 1.2. Simple Marx equivalent circuit ................................................................................. 2
Figure 1.3. Typical output pulse of Marx generator into a purely resistive load.......................... 3
Figure 1.4. Simple Abramyan circuit.......................................................................................... 5
Figure 1.5. Abramyan circuit signal decomposition.................................................................... 5
Figure 2.1. UMC pulsed power lab ............................................................................................ 9
Figure 2.2. Marx circuit ........................................................................................................... 11
Figure 2.3. Hydraulic pumping system..................................................................................... 12
Figure 2.4. Indoor oil containment tank with capacitors in raised position .............................. 12
Figure 2.5. Secondary oil storage tank ..................................................................................... 13
Figure 2.6. Oil pumping station flow diagram.......................................................................... 14
Figure 2.7. Dual-polarity charging schematic ........................................................................... 16
Figure 2.8. Charging and triggering resistors............................................................................ 17
Figure 2.9. Ground reference resistors .................................................................................... 18
Figure 2.10. Spark gaps installed in Marx................................................................................. 19
Figure 2.11. Assembled spark gap ........................................................................................... 19
Figure 2.12. Trigger generator ................................................................................................. 20
Figure 2.13. Gas and trigger lines ............................................................................................ 21
Figure 2.14. Crowbar assembly with input charging and triggering lines .................................. 21
Figure 2.15. Control cabinet .................................................................................................... 22
Figure 2.16. Control system block diagram.............................................................................. 23
Figure 2.17. Peaking capacitor ................................................................................................. 25
Figure 2.18. Oil switch enhanced electrode ............................................................................. 26
Figure 2.19. Simplified Marx with peaking capacitor................................................................ 28
Figure 3.1. Voltage divider circuit............................................................................................ 32
Figure 3.2. Simplified voltage divider implemented in diagnostic circuit .................................. 33
Figure 3.3. Voltage divider middle electrode............................................................................ 34
Figure 3.4. Voltage divider stages ............................................................................................ 35
Figure 3.5. Post-switch voltage divider .................................................................................... 36
Figure 3.6. Pre-switch voltage divider...................................................................................... 40
Figure 3.7. Voltage divider calibration configuration ............................................................... 41
Figure 3.8. Digital calibration of post-switch divider ............................................................... 42
Figure 3.9. Analog calibration of post-switch divider............................................................... 42
Figure 3.10. Digital calibration of pre-switch divider ............................................................... 43
Figure 3.11. Diagnostic output connections and Faraday cage................................................. 44
Figure 4.1. Simplified 30 stage UMC Marx circuit.................................................................... 47
Figure 4.2. Simulated Marx output pulse with and without peaker........................................... 48
Figure 4.3. Two spheres with equal radii.................................................................................. 50
Figure 4.4. Sphere and plane ................................................................................................... 50
Figure 4.5. Electric field calculations with variable voltage and electrode separation................ 51
Figure 4.6. Field enhancement factor for sphere-sphere and sphere-plane geometries ............. 52
Figure 4.7. Electric field strength with varying electrode separation......................................... 53
Figure 4.8. Electrostatic simulation of output switch field strength before Marx fires.............. 54
Figure 4.9. Enlarged view of tip of enhanced electrode electric field strengths ........................ 54
Figure 4.10. Electrostatic simulation of output switch potentials before Marx fires ................. 55
Figure 5.1. Marx spark gap operating curve ............................................................................. 58
Figure 5.2. Trigger generator operating curve .......................................................................... 58
Figure 5.3. Oil peaking switch operating curve ........................................................................ 59
Figure 5.4. Marx output data ................................................................................................... 61
Figure 5.5. Marx simulated output........................................................................................... 61
Figure 5.6. Data from both pre and post-switch dividers (Shot 8) ........................................... 63
Figure 5.7. Marx shot with multi-channel arcs to ground (Shot 33) ......................................... 63
Figure 5.8. Oil breakdown with Hipotronics tester .................................................................. 64
Figure 6.1. Guillemin Type A PFN ......................................................................................... 67
Figure 6.2. Type A PFN for UMC Marx ................................................................................. 68
Figure 6.3. Two cylinders (approximating turns of copper pipe).............................................. 70
Figure 6.4. Equivalent Marx circuit with twenty-two stages ..................................................... 71
Figure 6.5. Basic Abramyan in twenty-two stage Marx............................................................. 71
Figure 6.6. Abramyan network with inductor values 1 H - 9 H............................................ 74
Figure 6.7. Varying Abramyan network inductor values .......................................................... 75
Figure 6.8. Signal decomposition of Abramyan with Marx ...................................................... 77
Figure 6.9. Conceptual design of UMC Marx with Abramyan network.................................... 77
Figure 6.10. Marx and Abramyan with 5 H inductor ............................................................. 78
Figure 6.11. Abramyan with damping resistor ......................................................................... 79
Figure 6.12. Full length output from Abramyan and Marx....................................................... 79
Figure 6.13. Output pulses with damping resistor in the 5 H Abramyan network .................. 80
INDEX OF TABLES
Table 2.1.
Table 2.2.
Table 4.1.
Table 4.2.
Table 5.1.
Table 6.1.
Table 6.2.
Table 6.3.
UMC Marx parameters ........................................................................................... 10

CuSO4 resistors....................................................................................................... 17
Pulse characteristics from Marx simulation without peaking circuit......................... 49
Pulse characteristics from Marx simulation with peaking circuit .............................. 49
Output pulse characteristics .................................................................................... 60
Expected pulse lengths from variable inductor values ............................................. 73
Abramyan network pulse characteristics, Ld = 5 H ............................................... 78
Pulse lengths for resistance values in Abramyan...................................................... 81
vi
CHAPTER 1. INTRODUCTION
1.1. Background
A fundamental component of pulsed power systems is energy storage.
The most
popular method of energy storage for many applications including high energy density physics,
particle accelerators and flash radiography, high power microwave generation, and directed
energy defense systems is the Marx generator. This broad application base requires a system that
is simple yet robust. In addition to the Marx, pulse-shaping circuits are often required to tune
the output for specific load parameters. This thesis will discuss the operation of a typical Marx
and include a design to generate a rectangular pulse using a simple circuit addition to the Marx.
1.2. Marx Generators

In 1923, Erwin Marx patented the circuit for a generator with the fundamental principle
of charging capacitors in parallel and switching the capacitors in series into a load [1]. The
capacitors were charged through resistors and switched using simple two-electrode spark gaps
triggered by overvoltage [2], [3]. More recent designs have presented variations, but the concept
is much the same [3], [4], [5], [6]. Figure 1.1 shows the first few stages of an ideal Marx
generator with a resistive load.
Figure 1.1. Marx stages
The Marx charges its capacitors through the charging resistors, Rc . After reaching the
desired voltage, the first spark gap breaks down. Twice the amount of voltage is then seen
across the second spark gap. The voltage is too high for that gap to hold off, so it also breaks
down. This is repeated for each stage of the Marx and is considered erecting the Marx. The
Marx can then be converted into an equivalent circuit including a single capacitor discharging, in
this example, into a resistive load. The series inductor in the equivalent circuit in Figure 1.2
represents the non-ideal case where [7]:
LMarx = Lswitches + Lcaps + Lconnections
Figure 1.2. Simple Marx equivalent circuit
(1.1)
The simplified erected Marx circuit allows all of the capacitive stages to appear as a single
capacitor, CMarx , where N is defined as the number of stages in the Marx.
CMarx =
Cstage
N
(1.2)
The output voltage, VMarx , of these combined stages, or the output of the Marx, is the product
of the number of Marx stages and the charge voltage per stage.
VMarx = N Vstage
(1.3)
The total energy, E Marx , stored in the Marx can also be calculated from the values per stage.
1
2
E Marx = C stage Vstage
N
2
(1.4)
1
2
E Marx = C Marx VMarx
2
(1.5)
Figure 1.3. Typical output pulse of Marx generator into a purely resistive load
The output of a Marx into a damped resistive load can be expected to appear similar to
the plot in Figure 1.3. A capacitive bank ideally switching into a resistive load will result in
Equation 1.6. Risetimes, fall times, and pulse widths will vary with parametric value changes, but
the result will maintain an exponential RC decay. For the same circuit containing an inductor L ,
the 10-90 % risetime can be expressed as 2.2
L
[8].
R
V = Vo e
RC
(1.6)
Although Marx generators can be relatively inexpensive and easy to build, a pulse of this
type may not be desirable for many applications. A flat-top pulse is especially valuable for
defense systems including high power microwave and directed energy applications. Two of the
most commonly used alternatives to increase pulse width include Pulse Forming Lines (PFLs)
and Pulse Forming Networks (PFNs). Both of these well-documented approaches have their
own sets of advantages and disadvantages [9], [10]. Plans for the University of MissouriColumbia (UMC) Marx utilize a different approach called an Abramyan network.
The
eponymous E.A. Abramyan developed this cost-efficient alternative to extending the pulse from
a Marx generator into a rectangularly shaped pulse with the addition of a single inductor.
1.3. Abramyan Networks

The basic Marx circuit in Figure 1.2 can be modified to produce a rectangular pulse with
the addition of a single component [11]. The Abramyan network reverse-charges a fraction of
Marx stages and requires placement of an inductor of an optimized value in parallel with those
stages [12]. The basic Abramyan network circuit is shown in Figure 1.4. Both C a and La are
components already in the Marx. They are simply stages of the Marx that are reverse-charged,
4
or oppositely charged from the other normally-charged stages. The inductance La is the
series inductance of those stages and not an actual physical inductor. The only addition to the
Marx circuit is Ld . This inductor is optimized to form a ringing circuit, or tank circuit, that
oscillates to extend and flatten the output pulse of the Marx. Taking the Laplace transform of
the circuit and plotting its signal components, as in Figure 1.5, emphasizes the importance of the
ringing portion of the circuit.
Figure 1.4. Simple Abramyan circuit
Voltage HMV L
2
1.5
1
0.5
0.5
1.5
2.5
0.5
1
1.5
Figure 1.5. Abramyan circuit signal decomposition
Time Hus L
In Figure 1.5, the green line is the main normally-charged capacitors decay. The red line
represents the reverse-charged capacitors decay. The blue line represents the ringing of the
Abramyan network section of the Marx. These three components combine to form the output
pulse of the Marx (dashed purple line). As can be easily seen, the pulse width is greatly increased
from its original form without the Abramyan network addition. The full Mathematica code for
this signal decomposition can be found in Appendix F. The design of this circuit for the UMC
Marx will be discussed in detail in Chapter 6.
[1]
E. Marx, Verfahren zur Schlagprfung von Isolatoren und anderen elektrischen

Vorrichtungen, Deutsches Reich Reichspatentamt Patentschrift 455933, Oct. 1923.
[2]
J. F. Francis, High voltage pulse techniques, Texas Tech University, 1976.
[3]
R. A. Fitch, Marx and Marx-like high-voltage generators, IEEE Transactions on

Nuclear Science, vol. 18, no. 4, pp. 190-198, Aug. 1974.
[4]
K. R. Prestwich and D. L. Johnson, Development of an 18-megavolt Marx

generator, IEEE Transactions on Nuclear Science, vol. 16, no. 3, pp. 64-69, Jun. 1969.
[5]
J. D. Graham, D. G. Gale, W. E. Sommars, and S. E. Calico, Compact 400 kV Marx

generator with common switch housing, Proceedings of 11th IEEE Pulsed Power Conference,
1997.
[6]
F. E. Peterkin, D. C. Stoudt, B. J. Hankla, K. A. Boulais, J. S. Bernardes, Performance

characteristics of a 1 MV miniature Marx bank, Proceedings of the 12th IEEE Pulsed Power
Conference, 1999.
[7]
R. J. Adler, Pulse power formulary, North Star Research Corporation, [Online].

Available: http://www.northstar-research.com/PDF/Formweb1.pdf
[8]
G. A. Mesyats, Pulsed Power, Plenum, 2005.
[9]
G. N. Glasoe and J. V. Lebacqz, Eds., Pulse Generators, Massachusetts Institute of

Technology Radiation Laboratory Series, McGraw-Hill, 1948.
[10] P. W. Smith, Transient Electronics: Pulsed Circuit Technology, Wiley, 2002.

[11] R. Curry, K. Nielsen, I. Smith, Abramyan networks: a general purpose network for
linear induction accelerators, Pulse Sciences, Inc., San Leandro, CA, Apr. 1986
[12] E. A. Abramyan, E. N. Efimov, and G. D. Kuleshov, Correction of the pulse shape of

an Arkadev-Marks voltage generator, High-Temperature Institute of the Academy of
Sciences of the USSR, Moscow. Translated from Pribory i Tekhnika ksperimenta, No. 4,
pp.170-172, Jul.-Aug. 1979.
CHAPTER 2. DESCRIPTION OF SYSTEM
2.1. Pulsed Power Research Laboratory
Figure 2.1. UMC pulsed power lab
The UMC Marx and all its subsystems are housed in the Pulsed Power Research
Laboratory on campus. This lab enables students to design, build, and test large-scale pulsed
power projects. One of the goals of the lab was to develop a test facility to study oil breakdown
of enhanced and uniform gaps. The now operational test facility includes a thirty-stage, threemegavolt Marx generator. The pulser is designed to deliver a 3 MV output pulse with a risetime
(10-90%) of 10 ns with a peaking gap. The output polarity of the pulser can be easily reversed
for switch and dielectric testing. The configurations tested include large electrode gap spacings
with point-ball electrodes. This chapter will describe the system configuration and components
required for the test facility. Test results from these experiments will be reported in Chapter 5.
Chapter 6 will discuss a conceptual design for a simple modification which will allow a
rectangular pulse to be applied to the sample under test.
2.2. UMC Marx Generator

The UMC Marx generator has a total capacitance of 1.6 nF with a series inductance of
4.6 H [1]. At full charge, the Marx switches 7.2 kJ into the current 300 dummy load. The
Marx is modular in design with the capability for removal or addition of stages to increase or
decrease capacitance. Two capacitors can be added per half stage for a total erected capacitance
of 2.75 nF at 3 MV. The current Marx values are listed in Table 2.1 and the circuit is shown in
Figure 2.2.
Table 2.1. UMC Marx parameters
Number of stages
Stage charge voltage
Stage capacitance
Total capacitance
Series inductance
30
100 kV
48 nF
1.6 nF
4.6 H
10
Figure 2.2. Marx circuit
The Marx generators capacitors, resistors, and switches are supported by a nylon
backbone suspended below a steel ground plane. This frame can be raised to expose the
capacitors, resistors, and switches via a hydraulic lift system shown in Figure 2.3. The hydraulic
pumping system is positioned near the Marx and is responsible for pressurizing the four
cylinders that raise and lower the frame.
11
Figure 2.3. Hydraulic pumping system
Figure 2.4. Indoor oil containment tank with capacitors in raised position
12
During operation, the Marx is immersed in the indoor oil containment tank shown in
Figure 2.4. It is approximately 15 ft long, 3.5 ft wide, and 5 ft deep with a cylindrical oil-filled
extension at the high voltage end of the tank. The tank holds approximately 2,000 gallons of
Shell Diala AX transformer oil with the capacitors, resistors, switches, and support system
submerged [2]. The steel tank has been painted with International 850 Interline tank paint for
rust prevention [3].
Figure 2.5. Secondary oil storage tank
The secondary oil storage tank seen in Figure 2.5 is located outside of the building within
an additional retaining wall for environmental safety. The tank is used to store the oil during
periods of time in which the Marx is not in use. It has a storage capacity of 3,000 gallons and
remains unpainted on the inside to reduce impurities in the oil during long-term storage. To
13
prevent rust due to any outdoor moisture, the interior walls have been completely lubricated
with the oil. The outside of the tank has been painted and all openings have been resealed.
Figure 2.6. Oil pumping station flow diagram
The tank is connected to a pumping station also located within the retaining walls as
seen in the forefront of the image in Figure 2.5. The pumping station described in Figure 2.6
contains a Lincoln three-phase motor, 8300 Series Pall filter housing with a 42,326 cm2, 3 filter
[4], and refurbished control box. The control box consists of an Allen-Bradley AN16BN0AC
starter and C341BC transformer in a weather-resistant box attached to the pumping and filtering
cart frame. Three-phase power is supplied from the building to the control box and pumping
motor.
14
2.3. Charging
The Marx is dual-polarity charged and contains thirty capacitive stages. Each stage
consists of six 32 nF tubular capacitors. The six capacitors per stage are arranged three in
parallel to form a half-stage with two half-stages in series. The capacitors are members of the
Condenser Products KMOP Series [5]. The capacitors are housed in thermo-plastic tubing and
consist of a composite dielectric of kraft tissue and polyester film with a mineral oil impregnant.
The capacitor voltage maximum is 250 kV with a temperature range of -40C to +65C. The
capacitors are designed for 500,000 shots with no applied reverse voltage. With applied reverse
voltage, the number of shots in the lifetime of the capacitor will decrease according to the
formula described in Equation 2.1 given by Condenser Products [6].
1.57
number of shots =
ln (reverse voltage %)
2.2
500,000
(2.1)
According to this formula, for a 50% reversal, the capacitor lifetime is 82,000 shots. Similarly,
80% reversal yields 7,000 shots. The Marx is not initially configured for any voltage reversal but
possible alterations to the system discussed later in the Abramyan section will prove this
equation relevant.
The Marx is designed to be charged with two, 50 kV power supplies for a 3 MV
output pulse. Currently, the Marx is configured to charge with one +50 kV Glassman power
supply and one -40 kV, Kaiser power supply [7], [8]. The +50 kV, Glassman Model PS/WG50P6, is a 300 W power supply that can deliver 60 mA. The -40 kV Kaiser Series 1000 is a 500
W power supply that can deliver 12.5 mA. Because of this variation in power supply current, the
capacitors charge at separate rates and reach their charge voltage at different times. Although
not the ideal case, this poses no direct problems for operation of the Marx generator.
15
Figure 2.7. Dual-polarity charging schematic
With dual-polarity charging, the number of spark gaps required is reduced by half [9]. In
the simple single-polarity Marx charging configuration, a spark gap switch is required for each
capacitor. The switch is located between the high voltage end of one capacitor and the low
voltage end of the next capacitor. Figure 2.7 demonstrates how including both positive and
negative polarity charging in the circuit can eliminate half of the required spark gaps. This
charging configuration is utilized in the UMC Marx.
Although the Marx is largely a capacitive bank, many resistors are required for a variety
of applications within the pulser. For the UMC Marx, charging, triggering and ground reference
resistors were all essential. Values of resistances for these three types are noted in Table 2.2 [1].
Also described in the table are values necessary for constructing these resistors. Because the
energy requirements for resistors in the Marx were too high for traditional resistors, liquid
resistors were used. Copper sulfate (CuSO4) is the most commonly used solution for liquid
resistors and maintains a dielectric strength of > 200 kV/cm [10], [11]. Due to the ease and
cost-efficiency of design, liquid resistors were constructed for the UMC Marx with bead-blasted
copper electrodes, Tygon B-44-3 beverage tubing [12], 2-ear Oetiker clamps, deionized and
16
degassed water, and crystalline copper sulfate pentahydrate. Images of the charging, triggering,
and ground reference resistors are seen in Figure 2.8 and Figure 2.9.
Table 2.2. CuSO4 resistors [1]
Description
Quantity
Resistance
()
Charging
Trigger
GND Ref
62
30
30
2.5 k
1.5 k
10 k
Inner
Diameter
(in)
0.5
0.5
0.5
Length
(in)
13.75
11.75
9.25
Figure 2.8. Charging and triggering resistors
17
Volume
Resistivity
(/cm)
91
67
540
Solution
(gm/L)
38
63
4
Figure 2.9. Ground reference resistors
The spark gap switches installed in the UMC Marx (Figure 2.10) are similar in
configuration to the original simple two-electrode switches in early circuits [13]. They are very
similar in operation to the Series T-670 spark gap switches currently produced by Titan
Corporation Pulse Sciences Division [14]. Autocad drawings of the UMC Marx spark gap
dimensions are shown in Appendix A. The acrylic switch housing contains two brass electrodes
with a mid-plane for triggering. An image of a fully assembled switch can be seen in Figure 2.11
The switches are pressurized with sulfur hexafluoride (SF6) during operation of the
Marx. The use of SF6 enables the spark gaps to be operated at lower pressures than synthetic air
would allow. The switches will function with air, but at much higher pressures for comparable
voltage hold-off. The SF6 gas lines are attached at the top of the switches and are connected in
18
series to all 30 gaps. Swagelok fittings connect the segments of Dayco Nylo-Seal tubing between
the switches for filling and purging the SF6 between sets of shots.
Figure 2.10. Spark gaps installed in Marx
Figure 2.11. Assembled spark gap
19
2.4. Triggering
The Marx is triggered by the pulsed trigger system in Figure 2.12 consisting of a high
voltage power supply, a pressurized housing containing a capacitor and switching elements, an
intermediate trigger amplifier, and a crowbar switch assembly. A 75 nF Condenser Products
EC753-80M capacitor is DC-charged by a Glassman -80 kV power supply [15]. A pressurized
SF6 gas switch provides a low inductance short to ground at the high voltage electrode of the
capacitor [1]. The capacitor discharges a positive output pulse into a near open circuit formed
by the first stages of Marx generator spark gaps [16]. The trigger resistors described earlier form
trigger lines to each of the 30 spark gaps (Figure 2.13). Although it is not necessary to trigger
each switch, this method may decrease overall breakdown time [17]. The crowbar assembly in
Figure 2.14 is located in the low voltage end of the Marx tank. Spring-loaded pneumatic
cylinders keep the capacitors and all high voltage supplies shorted to ground until the charge
sequence is initiated.
Figure 2.12. Trigger generator
20
Figure 2.13. Gas and trigger lines
Figure 2.14. Crowbar assembly with input charging and triggering lines
21
2.5. Control Systems
Figure 2.15. Control cabinet
The master control system controls all major systems that are required for the operation
of the Marx. These systems include charging, triggering, gas, and safety. Each of these systems
will be discussed in detail.
The control system (Figure 2.16) monitors charging for all three power supplies. When
the trigger generators power supply reaches its preset charge voltage, the control system
22
indicates that the fire button can be pressed at any time, thus initiating the fire sequence. This
feature can be expanded to include the Marx charging power supplies, but is currently disabled
due to a non-compatible replacement supply. After the Marx has fired, the control system also
disables the high voltage from the supplies. The replacement supply presently prevents this for
the Marx supplies, but the trigger generator supply is still automatically disabled post-shot.
Figure 2.16. Control system block diagram
The triggering system is also monitored and operated by the control system. When the
trigger generator is fully charged and the fire button is pressed, the control system sends a 7 kV
trigger pulse to the trigger generator. The 7 kV pulse results in trigger switch breakdown and
initiates the Marx spark gap trigger pulse.
23
The gas line pressures, both air and SF6, are displayed above the control box. Venting of
the lines is achieved with electronically-controlled valves. Regulators are also located on the gas
control panel and allow for easy and controlled pressurization of the SF6 for the Marx spark gaps
and the trigger generator.
The safety system response consists of an automatic shutdown of the high voltage from
the supplies and a crowbar of the supplies and capacitors to ground after a shot or during an
abort of a shot. As discussed earlier, the automatic shut-off of the supplies only pertains to the
trigger generator supply. The Marx supplies are still crowbarred to ground, but require a manual
shutdown of high voltage. The crowbar (seen in Figure 2.14) shorts all high voltage supply
cables and trigger lines to ground. It is physically located in the low voltage end of the tank and
is pneumatically controlled by the control box.
2.6. Peaking Capacitor and Oil Switch

Positioned at the end of the thirty stages is a parallel plate oil dielectric peaking capacitor
as shown in the photograph of Figure 2.17. The high voltage electrode of the self-break singlesite oil switch is located in the center of the plate. The low voltage electrode is mounted inside a
cylindrical oil-filled extension of the tank. The Marx is capable of operating with both large and
small gap switch geometries. This fully-adjustable output switch allows for a variety of test
parameters.
24
Figure 2.17. Peaking capacitor
The self-break single-site oil switch functions as the output switch of the Marx. The
Marx generator requires an output switch for the capacitors to charge. If the capacitors were
permanently connected to the load, the capacitors would never completely charge as it would
dissipate into the load. The output switch of the UMC Marx is positioned at the end of the
stage capacitors in the center of the peaking capacitor. Triggered by overvoltage, the switch arcs
through the oil between the enhanced electrode (Figure 2.18) and the ball electrode.
25
Figure 2.18. Oil switch enhanced electrode
The ball electrode is connected to four parallel load resistors with a total resistance of
300 . Using Grovers equation [18] for self-inductance of a circular single conductor under
high frequency in Equation 2.2, the total inductance of the load (all four parallel resistors) is 35.6
nH. Also attached to the ball electrode is the post-switch voltage diagnostic described in
Chapter 3.
2l
L = 2 l ln
1
r
L
l
= inductance (nH)
= conductor radius (cm), 0.9525
= conductor length (cm), 24.3
26
(2.2)
Peaking capacitors are often placed at the end of Marx generators with the desire of
faster risetimes. The peaking capacitor in the UMC Marx has a value of 200 pF and is physically
a parallel plate oil dielectric capacitor.
With measurable dimensions and known dielectric
properties of the insulator, the capacitance can be approximated by Equation 2.3.
C peak =
C peak
k
k 0 A
d
(2.3)
= capacitance, F
= relative permittivity of oil, 2.2
= permittivity of space, 8.854 x 10-12 F/m
A
d
= area, 1.03 m2
= separation, 10 x 10-2 m
Applying J. C. Martins equation for breakdown [19], [20] to the peaker and the tank wall
in Equation 2.4, it can be shown that for the 10 cm spacing, 5.4 MV would be required to
achieve breakdown. With the Marx maximum operating voltage of 3 MV, the peaking capacitor
is unlikely to prematurely breakdown before the output switch has fired. Since the value of k
for transformer oil is independent of polarity, the above statement applies to both positive and
negative Marx operation.
Ft
1
10
=k
F
t
= breakdown field (MV/cm)
A
k
= electrode area (cm2), 1030
(2.4)
= time (s), 0.1

= 0.5 (for transformer oil)
In the simulation circuits described in detail in Chapter 4, the peaking capacitor is placed
between the switch approximating the 30 spark gaps and the output oil switch. In the case of
27
the UMC Marx, the entire circuit can be simplified to the circuit shown in Figure 2.19. The low
Marx inductance in combination with the peaking capacitor results in a theoretical UMC Marx
output pulse 10-90% risetime of < 10 ns [1].
Figure 2.19. Simplified Marx with peaking capacitor
It can be determined that the peaking capacitor in the UMC Marx is not matched to the
current load by using Equation 2.5 for an exponential decay though the load [21].
C peak
LMarx
2
R
load
=
L
1 + Marx 2
Rload CMarx
(2.5)
By that calculation, the peaking capacitor should be 50 pF. For the current peaking capacitor of
200 pF, a load resistance of 142 would produce the ideal exponential decay.
Minor
adjustments of the CuSO4 load resistor values for future experiments would optimize the circuit.
28
[1]
Spiral delay line pulser: operation and maintenance manual, Pulse Sciences, Inc., PSIO&M-2420, Dec. 1991.
[2]
Diala AX data sheet, Shell Lubricants, [Online]. Available: http://www.shelllubricants.com/products/pdf/DialaA&AX.pdf
[3]
Interline 850 data sheet, International Protective Coatings, [Online]. Available:

http://www.international-pc.com/pc/pds/850_us.pdf
[4]
8300/04/10/14 Series Filter Assemblies, Pall Corporation, [Online]. Available:

http://www.pall.com/variants/pdf/pdf/3112.pdf
[5]
Our capacitor specifications, Condenser Products, [Online]. Available:

http://www.condenser.com/specs.htm
[6]
Condenser Products, private communication, 2004.
[7]
Instruction manual: WG series, Glassman High Voltage Inc., PS/WG-50P6

M204250-01, July 1988.
[8]
Series 1000 high voltage power supply owners manual, Kaiser Systems, Inc.,Varian
IIS P/N E-81000065, 1991.
[9]
[10] R. Beverly III and R. Campbell, Aqueous-electrolyte resistors for pulsed power
applications, Review of Scientific Instruments, vol. 66, pp. 5625-5629, 1995.
[11] R. Beverly III, Application notes for aqueous-electrolyte resistors, [Online].
Available: http://www.reb3.com/pdf/r_appl.pdf
29
[12] Tygon beverage tubing, Saint-Gobain Performance Plastics Corporation, [Online].

Available:
http://www.tygon.com/media/documents/S0000000000000001013/tygb443.pdf
[13] R. A. Fitch, Marx and Marx-like high-voltage generators, IEEE Transactions on
Nuclear Science, vol. 18, no. 4, pp. 190-198, Aug. 1974.
[14] Series T-670 High Precision Spark Gap Switches, Titan Corporation Pulse Sciences
Division, [Online]. Available: http://www.titanpsd.com/PDFs/T-670.pdf
[15] Instruction manual: LG series, Glassman High Voltage Inc., PS/LG-80N-1.5
M204210-01, July 1988.
[16] Pulser to study the effects of EMP induced signals on analog models of large antenna
structures, Pulse Sciences, Inc. PSI-P-87-355, Sept. 1987.
[17] R. W. Morrison and A. M. Smith, Overvoltage and breakdown patterns of fast Marx
generators, IEEE Transactions on Nuclear Science, vol. 19, no. 4, pp. 20-27, Aug. 1972.
[18] F. W. Grover, Inductance Calculations, Dover, 1962.
[19] J. C. Martin, Nanosecond pulse techniques, Proceedings of the IEEE, vol. 80, no. 6, pp.
934-945, Jun. 1992.
[20] T. Martin, A. Guenther, and M. Kristiansen, Eds., J. C. Martin on Pulsed Power, New
York: Plenum Press, 1996.
[21] R. J. Adler, Pulse power formulary, North Star Research Corporation, [Online].
30
CHAPTER 3. DIAGNOSTICS
3.1. Requirements
The UMC Marx oil switch test facility required two voltage diagnostics: before and after
the output oil switch. It was necessary that these diagnostics minimally impact the system and
accurately measure the performance of the pulser. Since typical, off-the-shelf voltage probes
could not withstand the high peak voltages of the Marx output, other methods of determining
voltage waveforms needed to be developed. It was desired that the voltage probes attenuate the
Marx output pulse by a factor of 20,000 to safely acquire data with an oscilloscope. Liquid
resistive voltage dividers were selected for the diagnostics.
The two measurements desired were voltages in the pre-switch and post-switch
locations. The voltage waveform measured at the high voltage side of the output switch allows
for monitoring of the capacitor charge. The charge level of the capacitors at the time the trigger
pulse was applied can also be determined. The second voltage measurement required was on the
output side of the switch in parallel with the load resistors. This measurement is possibly the
more important of the two diagnostics since it monitors the operation of the oil switch and final
output of the Marx.
The first probe developed was the post-switch divider due to its less accessible location
at the end of the Marx in the oil-filled cylindrical extension of the tank. The output of the Marx
arcs through the oil from the enhanced electrode to the ball electrode and then dissipates
through the load resistors. The post-switch resistive voltage divider is positioned in parallel with
31
the load resistors. Minimizing the impact of the probe on the circuit, the resistance of the
divider was included in the load resistance calculations.
The pre-switch divider needed to be placed inside the tank with the capacitors and
attached to the top frame for ground. This divider had both extended space parameters and
considerably easier access than the post-switch divider.
3.2. Voltage Divider Theory

Liquid resistive voltage dividers are simple, effective, and easy-to-build probes that are a
cost-efficient solution to the need for a high voltage diagnostic. The circuit for the high voltage
resistive probe is based on the simple voltage divider circuit in Figure 3.1 [1]. The input voltage
V1 is related to the output voltage V2 by Equation 3.1. The ratio formed by the resistors R1
and R2 can be referred to as the division ratio [2]. Coaxial stray capacitance of the divider to the
tank wall is calculated in Equations 3.2 and Equation 3.3. The resultant value of 10.7 pF
distributed throughout the geometry can be assumed negligible for the UMC Marx
configuration.
V2 = V1
R2
R1 + R2
Figure 3.1. Voltage divider circuit
32
(3.1)
D
ln
d
= impedance, 149
= dielectric constant, 2.2
D
d
= outer conductor, inner diameter, 40 in
Z=
60
(3.2)
= inner conductor, outer diameter, 1 in
C=
Z
= capacitance, 10.7 pF
= transit time in oil, 1.5 ns/ft
= impedance, 149
(3.3)
An expanded version of the voltage divider circuit is shown in Figure 3.2. The additional divider
comprised of R3 and R4 is provided to match the line for reflection minimization.
Figure 3.2. Simplified voltage divider implemented in diagnostic circuit
Physically, the probe consists of two stages [3]. The first stage is the liquid resistor
portion which is responsible for the majority of the division ratio. It is built much like the liquid
resistors discussed in Chapter 2 [4], [5] with 1 inch inner diameter Tygon B-44-3 beverage tubing
33
[6]. The resistor is shaped much like a typical resistor with two electrodes separated in an
aqueous salt solution inside a sealed plastic tube.
The difference from those previously
described is the tap-off electrode. The tap-off electrode is positioned a small distance away from
the electrode on the low voltage end of the resistor. It is isolated from the low voltage electrode
through the middle by a section of RG-401 cable. Visually, the tap-off electrode can be noted in
Figure 3.3 at a spacing of 1.5 mm from the low voltage electrode.
The major advantage of using a liquid resistor is that both parts of the resistor will
maintain the same chemical composition regardless of the stresses applied. Knowing that the
resistors are linear in an electric field between 2 50 kV/cm when = 60 1200 cm
enables the assumption that the division ratio will also remain constant [4], [5].
Figure 3.3. Voltage divider middle electrode
The second stage of the divider is composed of discrete resistors in a small, electrically
shielded box. The RG-401 from the tap-off electrode is fed through one side of the box. A
BNC connection for an RG-58 cable is located on the opposing side. This can be seen in Figure
3.4.
34
Figure 3.4. Voltage divider stages
From Figure 3.4, it can be noted that the primary divider is comprised of R1 and R2 .
The tap-off electrode forms two resistors. The resistor R1 is the larger of the two and consists
of the region from the high voltage electrode to the tap-off electrode. Due to its large size, the
resistor value can be approximated to be the entire resistance value from high to low voltage
electrodes for measurement purposes. This resistor attenuates the input voltage down to a level
compatible with the input to an oscilloscope. The second of the two copper sulfate resistors is
physically located between the tap-off electrode and the low voltage electrode. This is a much
smaller value with minimal voltage drop between the electrodes. The voltage on the middle
electrode is transferred to the second stage of the voltage divider. The first of the secondary
divider resistors, R3 , can be used as a value to slightly adjust the output voltage. On the other
hand, R4 is used to match the output cable impedance so as to minimize reflections in the data.
Consequently, R4 is usually 50 .
3.3. Post-switch Divider

The post-switch divider seen in Figure 3.5 had a limited space requirement of 10 inches.
The resulting division ratio for the divider was 10,000:1. The CuSO4 resistor contributed a
35
division of 200 and the discrete dividers added an additional division of 50. The post-switch
divider is housed in the cylindrical oil-insulated extension at the end of the tank.
The total resistance of the liquid portion of the divider is 5 k as measured. In parallel
with the load resistors, it has the effect of minimally lowering the load resistance from 300 to
283 . With the inability to attain exact measurement values with CuSO4 resistors, the load can
be still approximated as 300 . One watt resistors were used for the discrete portion of the
divider. The resistor R3 has the value of 2.56 k and the resistor R4 has the value of 48.43 .
Figure 3.5. Post-switch voltage divider
The inductance of the probe can be calculated with Grovers equation for selfinductance of a circular single conductor under high frequency [7]. Given the parameters
described in Equation 3.4, the result is 101.3 nH.
36
2l
L = 2 l ln
1
r
L
l
= inductance (nH)
(3.4)
The capacitance of the probe can be calculated from both the high voltage electrode to
the tap-off electrode and from the tap-off electrode to the low voltage electrode. Equation 3.5
and Equation 3.6 can be used to determine these values. The 404 femtofarads calculated
between the high voltage electrode and the tap-off electrode are negligible. On the other hand,
the capacitance of 239.2 pF between the tap-off electrode and the high voltage electrode should
be factored into the frequency response of the probe.
Chv tap =
Chv tap
k
k 0 A
d
= capacitance, 404 fF
= relative permittivity of water, 80
A
d
= area, 506.7 x 10-6 m2

Ctap lv =
Ctap lv
k
(3.5)
k 0 A
d
A
d
= area, 506.7 x 10-6 m2

= separation, 1.5 x 10-3 m
37
(3.6)
Breakdown within the 10 inch (25.4 cm) probe is not a concern according to J. C.
Martins equation (Equation 3.7) [8], [9] which indicates that a positive voltage of 12 MV or a
negative voltage of 24 MV would be required for breakdown.
Ft 3 A
1
10
=k
F
t
A
k+
= electrode area (cm2), 20.3
= 0.6 (for negative electrodes in water)
(3.7)
= time (s), 0.1

= 0.3 (for positive electrodes in water)
3.4. Pre-switch Divider

The pre-switch divider seen in Figure 3.6 has a division ratio of 20,000:1. The CuSO4
resistor portion accounts for a division of 400. Additional discrete resistors contribute a division
of 50. The length of the pre-switch divider is approximately 35 inches.
The resistance of the liquid divider is also approximately 5 k. The divider is attached
to the shaper that forms the high voltage side of the peaking capacitor and to the top of the
grounded frame. One watt resistors were used for the discrete portion of the divider. The
resistor R3 has the value of 2.46 k and the resistor R4 has the value of 48 .
The inductance of the probe can be calculated with Grovers equation for selfinductance of a circular single conductor under high frequency [7]. Given the parameters
described in Equation 3.8, the result is 577.6 nH.
2l
L = 2 l ln
1
r
38
(3.8)
L
l
= inductance (nH)
The capacitance of the pre-switch divider can be calculated from both the high voltage
electrode to the tap-off electrode and from the tap-off electrode to the low voltage electrode
using Equation 3.9 and Equation 3.10. The capacitance of 1.41 pF between the high voltage
electrode and the tap-off electrode is slightly greater than the similar calculation for the postswitch divider. The value of Ctap lv is the same as the previous calculation since the electrode
spacing between the tap-off electrode and the low voltage electrode is identical in both cases.
Breakdown within the 35 inch (88.9 cm) probe is not a concern according to J. C. Martins
equation (Equation 3.11) [8], [9] which indicates that a voltage of 43 MV would be required for
breakdown.
Chv tap =
Chv tap
k
k 0 A
d
A
d
= area, 506.7 x 10-6 m2

Ctap lv =
Ctap lv
k
(3.9)
k 0 A
d
A
d
= area, 506.7 x 10-6 m2

= separation, 1.5 x 10-3 m
39
(3.10)
Ft 3 A
1
10
=k
F
t
A
k+
= 0.6 (for negative electrodes in water)
(3.11)
= time (s), 0.1

= 0.3 (for positive electrodes in water)
Figure 3.6. Pre-switch voltage divider
3.5. Calibration
Calibration of the resistive voltage dividers was done using a high voltage self-break and
triggered-break switch with a Tektronix P6015A high voltage probe [10]. Both the analog
Tektronix 2467B and digital Tektronix TDS3034 [11] oscilloscopes were used to calibrate the
40
dividers. The +50 kV Glassman high voltage power supply [12] from the charging circuits
discussed earlier was utilized for the tests. The configuration used for this calibration is seen in
Figure 3.7. Although, the ideal calibration scenario is to test within the actual circuit parameters,
this was not possible for the UMC Marx due to physical constraints. Data from the calibration
of both the post-switch and pre-switch dividers can be seen in the plots following.
It can be noted from the calibration plots for the post-switch voltage divider (Figure 3.8
and Figure 3.9) that the risetimes and fall times are identical.
This accuracy allows for
exceptional data collection from the output of the Marx. The pre-switch divider was originally
constructed without using a RG-401 cable connection to the tap-off electrode. Instead, a
conductor and insulator without a matching 50 impedance was used. The result is a series of
reflections in the output pulse. This can be seen in the pre-switch calibration data in Figure 3.10.
Figure 3.7. Voltage divider calibration configuration
41
Figure 3.8. Digital calibration of post-switch divider
Voltage
Divider
200 mV/div
Voltage
Divider
200 mV/div
Tektronix
Probe
2 V/div
Tektronix
Probe
2 V/div
50 ns/div
20 ns/div
Figure 3.9. Analog calibration of post-switch divider
42
Reflections
Figure 3.10. Digital calibration of pre-switch divider
During operation of the Marx, the voltage dividers are located in their respective
locations and the oscilloscope is located in a copper-mesh Faraday cage seen on the right-hand
side of Figure 3.11. The copper sheet in the figure lies underneath the cables as a path to the
shielded cage. Despite the many opportunities for signal interference in a pulsed power system
[13], the output pulses recorded and discussed in Chapter 5 provide great clarity for all necessary
tests.
43
Figure 3.11. Diagnostic output connections and Faraday cage
44
[1]
[2]
A. J. Schwab, High-Voltage Measurement Techniques, MIT Press, 1972.
[3]
D. G. Pellinen and I. Smith, A reliable multimegavolt voltage divider, Review of

Scientific Instruments, vol. 43, pp. 299-301, 1972.
[4]
R. Beverly III and R. Campbell, Aqueous-electrolyte resistors for pulsed power

applications, Review of Scientific Instruments, vol. 66, pp. 5625-5629, 1995.
[5]
R. Beverly III, Application notes for aqueous-electrolyte resistors, [Online].

Available: http://www.reb3.com/pdf/r_appl.pdf
[6]
Tygon beverage tubing, Saint-Gobain Performance Plastics Corporation, [Online].

Available:
http://www.tygon.com/media/documents/S0000000000000001013/tygb443.pdf
[7]
F. W. Grover, Inductance Calculations, Dover, 1962.
[8]
J. C. Martin, Nanosecond pulse techniques, Proceedings of the IEEE, vol. 80, no. 6, pp.
934-945, Jun. 1992.
[9]
T. Martin, A. Guenther, and M. Kristiansen, Eds., J. C. Martin on Pulsed Power, New

[10] Passive High Voltage Probes, Tektronix, [Online]. Available:

http://www.tek.com/site/ps/56-10262/pdfs/56W_10262.pdf
[11] Digital phosphor oscilloscopes data sheet Tektronix, [Online]. Available:
http://www.tek.com/site/ps/3G-12482/pdfs/3GW_12482.pdf
45
[12] Instruction manual: WG series, Glassman High Voltage Inc., PS/WG-50P6

M204250-01, July 1988.
[13] E. Thornton, Electrical Interference and Protection, Ellis Horwood, 1991.
46
CHAPTER 4. MODELING OF MARX
4.1. Marx Circuits

The basic Marx circuits described in Chapter 1 were enhanced to show the detailed
parameters of the UMC Marx in Chapter 2.
For modeling and simulation purposes, the
expanded Marx circuit drawing in Figure 2.2 was simplified to the circuit shown in Figure 4.1.
This equivalent circuit is appropriate for simulations of the erected Marx since the firing of all
stages occurs nearly simultaneously and appears to the load as a single capacitive energy store [1].
The Marx capacitance C Marx is the equivalent total capacitance of all the stages of the Marx.
The Marx inductance LMarx is the total series inductance of the Marx and all its connections.
Figure 4.1. Simplified 30 stage UMC Marx circuit
4.2. PSPICE Simulations

Simulations were run on the Marx parameters for several different reasons. Determining
the expected output of the Marx is important for both operational and safety concerns. By
47
determining the output pulse specification details before a shot is taken, the diagnostics can be
optimized to accurately verify pulse shape. The PSPICE [2] simulations of the Marx both with
and without the peaking circuit are shown in Figure 4.2. The PSPICE code for these results is
documented in Appendix B. The ripple in the pulse from the simulation of the Marx with the
peaking circuit is a result of the peaking capacitor and load producing a non-ideal exponential
decay as described earlier in Chapter 2. The characteristics of the two pulses are outlined in
Table 4.1 and Table 4.2. The risetime of 30 ns from the simulation without the peaking circuit
can be attributed to the low inductance of the Marx and the triggering of all spark gaps
simultaneously. The introduction of the peaking circuit to the Marx improves the risetime by an
additional 20 ns.
Figure 4.2. Simulated Marx output pulse with and without peaker
48
Table 4.1. Pulse characteristics from Marx simulation without peaking circuit
Risetime (10-90%)
Pulse width (90-90%)
Fall time (10-90%)
30 ns
90 ns
1.6 s
Table 4.2. Pulse characteristics from Marx simulation with peaking circuit
Risetime (10-90%)
Fall time (10-90%)
10 ns
50 ns
1.6 s
4.3. Breakdown Calculations

To verify the functionality of the Marx at specific output switch spacings, voltage
breakdown calculations were computed. The Bouwers and Cath [3], [4] electric field strength
configurations (described in Figure 4.3 and Figure 4.4) with variable separation produced the
results given in Figure 4.5. For sphere-sphere calculations, the electrodes were approximated as
two spheres with diameters of 3 cm. For sphere-plane calculations, the enhanced electrode was
approximated as a sphere with the same diameter (0.28 cm) as the tip of the electrode. Since
these are both approximations, the actual result lies between the two.
For comparison, the J. C. Martin [5], [6] liquid breakdown equation (Equation 4.3) was
used to compute the oil breakdown of a 300 ns pulse with the electrode geometry used in the
sphere-sphere calculation. The result in Figure 4.5 indicates that the output switch can be
successfully operated at all voltages (500 kV 3 MV) required for future high pressure switch
tests. Given that the actual maximum electric field of the output switch is between the two
calculations described earlier, the breakdown fields from the J. C. Martin equation occur on the
low side of the two. It appears that the only area in which breakdown has a slightly greater
49
chance of not occurring is at voltages under 1 MV and gap spacings under 4 cm. Decreasing the
gap further with electrode geometry changes would rectify this issue. During fabrication of
future electrodes, the length could be easily extended if operation at lower voltages is required.
Overall, the comparison of the maximum electric fields to the Martin equation suggests that the
Marx will work appropriately within the given parameters.
Figure 4.3. Two spheres with equal radii
Emax
Emax
a
V r+ 2
= 0 .9
a
r
(4.1)
= maximum electric field, MV/cm
= potential difference between electrodes, MV
= electrode spacing, 3-7 cm
= electrode radius (approx. sphere), 1.5 cm
Figure 4.4. Sphere and plane
Emax = 0.9
Emax
V r+a
r
a
= maximum electric field, MV/cm
= electrode radius (approx. point), 0.14 cm
50
(4.2)
Ft 3 A
1
10
=k
F
t
A
k
= time (s), 0.3

= 0.5 (for transformer oil)
Figure 4.5. Electric field calculations with variable voltage and electrode separation
51
(4.3)
Additionally, the field enhancement factor (FEF) for the two electrode approximations
can be calculated using Equation 4.4.
FEF =
Emax
Emean
where
Emean =
V
a
(4.4)
The result from this equation with both electrode approximations is shown in Figure 4.6.
Figure 4.6. Field enhancement factor for sphere-sphere and sphere-plane geometries
The FEF for the sphere-plane geometry is significantly higher than the FEF for the spheresphere geometry. The electric field strengths with the same electrode separation distances are
seen in Figure 4.7. Since the electrodes resemble the sphere-sphere approximation with slight
enhancement, it appears likely that an electric field of at least 0.5 MV/cm will occur with a 5 cm
52
separation. The red line representing the mean voltage across the gap can be interpreted as the
minimum due to the field enhancement design present in the pointed high-voltage electrode.
This can be further verified with the electrostatic simulations in the next section.
Figure 4.7. Electric field strength with varying electrode separation
4.4. Electrostatic Simulations

Electrostatic simulations were conducted on the geometry of the output switch to verify
proper breakdown. The cross-sectional simulation results seen in the following Quickfield [7]
color maps were a product of a 1 MV charge with a 5 cm gap spacing. Figure 4.8 is a mapping
of the field strengths before the Marx has fired. Vectors of field strengths were also included on
the plots.
53
Strength
E (108V/m)
1.330
1.197
1.064
0.931
0.798
0.665
0.532
0.399
0.266
0.133
0.000
Figure 4.8. Electrostatic simulation of output switch field strength before Marx fires
Strength
E (108V/m)
1.330
1.197
1.064
0.931
0.798
0.665
0.532
0.399
0.266
0.133
0.000
Figure 4.9. Enlarged view of tip of enhanced electrode electric field strengths
54
The field strength values from the Quickfield simulations coincide with the electric field
calculations in the previous section. From the enlarged view of the enhanced electrode in Figure
4.9, the maximum electric field strength attains a value of 1.33 MV/cm. This result is greater
than the 0.48 MV/cm sphere-sphere geometry approximation. As mentioned earlier, this was
expected due to the enhanced tip on the high-voltage electrode. The 1.33 MV/cm field strength
from Quickfield results in a FEF of 6.65 with a 5 cm electrode separation. Comparing this result
to the plot in Figure 4.6, the UMC Marx configuration is more closely approximated by the
sphere-sphere geometry approximation than the sphere-plane geometry approximation.
It is interesting to note that the electric fields on the field shaper ring in the Quickfield
simulations appear to have a noticeable effect on the output switch. The experimental result
from this is discussed in Chapter 5. The possibility of moving the low voltage electrode from its
current, recessed position is an option for future experiments. The gradient of potentials
surrounding the switch is seen in Figure 4.10 with expected results.
Voltage
U (106V)
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Figure 4.10. Electrostatic simulation of output switch potentials before Marx fires
55
[1]
[2]
Orcad 9.1 [Online]. Available: www.orcad.com
[3]
A. Bouwers and P. G. Cath, The maximum electric field strength for several simple
electrode configurations, Philips Tech. Review, vol. 6, no. 9, pp. 270-278, 1941.
[4]
W. D. Greason, Electrostatic Discharge in Electronics, Research Studies Press Ltd., 1992.
[5]
934-945, Jun. 1992.
[6]

[7]
Quickfield [Online]. Available: www.quickfield.com
56
CHAPTER 5. SWITCH EXPERIMENTS
5.1. Experimental Setup

After the results from the previous chapter, it was determined that a 1 MV output pulse
from the Marx would be an appropriate test for its functionality without overstressing its
components. The 1 MV output from the Marx will be described specifically in this section.
Parameters to be discussed include charging voltages, triggering voltages, gas pressures, and
switch gap spacings. Testing was conducted by varying these parameters to check both system
versatility and simulation verification.
For a 1 MV output with 30 stages, each stage was charged to 33.3 kV. Since the
charging schematic utilizes dual-polarity charging, half of each stage was charged to +16.7 kV
and half was charged to -16.7 kV. Applying these charge voltages to Figure 5.1 [1], it was
appropriate to pressurize the spark gaps with SF6 to 4 psi. Using the original operation manual
guidelines to charge the trigger generator to approximately twice the half-stage voltage, the
trigger generator was typically charged to 40 kV. Figure 5.2 [1] indicated the trigger generator
should be pressurized to 22 psi. These values combined with a gap spacing number of 180 from
Figure 5.3 [1] proved reliable for the Marx to fire 1 MV shots consistently. All three of the
following original operation manual plots were verified by observing the parameter guidelines
with the testing of the Marx. The test points from the current Marx operation fall along the
same plot lines with the exception of the oil switch operating curve where changes in operation
have been noted.
57
Figure 5.1. Marx spark gap operating curve [1]
Figure 5.2. Trigger generator operating curve [1]
58
Figure 5.3. Oil peaking switch operating curve [1]
The quality of the shots was greatly dependent on oil filtration. As mentioned in
Chapter 2, the output switch of the Marx is an oil dielectric switch. The quality of the Diala AX
[2] used to insulate the Marx directly affected the arc formation across the output switch. It was
visually noted that multiple arc channels formed after sets of shots without filtering between
them. Large multi-channel arcs produced very large carbon clouds and bubbles that appeared to
release smoke (or carbon) when they reached the surface. These multi-channel arcs began at the
enhanced, high-voltage electrode and reached both the low-voltage electrode and the grounded
field shaper (seen in the electrostatic simulations in the previous chapter). From the simulations
conducted before the Marx was fired, it was known that the field shaper had higher electric field
strengths than the recessed low-voltage electrode. As mentioned earlier, the possibility of
moving the low voltage electrode closer to the high-voltage electrode is an option for future
testing. Oil impurities appeared to increase the probability of the multi-channels occurring.
59
Pumping the oil through the filter located in the side of the tank for 20-30 minutes between
shots proved sufficient to eliminate breakdown. Preparation for Marx operation and detailed
records of individual shots are given in Appendix C and D respectively.
5.2. Experimental Results

Initial testing of the Marx was conducted using the point-ball electrode geometries
discussed in the previous chapter. The tests have shown expected and repeatable results.
Output pulse characteristics are listed in Table 5.1. The risetime of 18 ns represents a slight
increase from the simulated results which could be attributed to bandwidth filtering within the
oscilloscope. Waveforms from the post-switch divider at an output voltage of 1 MV are shown
in Figure 5.4. The repeatability of the pulser can be noted from these waveforms. Comparing
the experimental waveforms to the simulated waveform in Figure 5.5, the results are similar.
The ringing in the experimental data appears to have a slightly larger influence on the pulse than
in simulated results. Figure 5.6 shows output from both the pre-switch and post-switch dividers
on the same plot. As indicated in Chapter 3, the pre-switch divider contains reflections due to
impedance differences between the tap-off electrode and the 50 cable.
Table 5.1. Output pulse characteristics
Risetime (10-90%)
Fall time (10-90%)
18 ns
50 ns
1.6 s
60
Figure 5.4. Marx output data
Figure 5.5. Marx simulated output
From the early results, the necessity for proper oil filtration was apparent. As discussed,
the presence of certain levels of both carbon particles and bubbles occasionally facilitated arcs to
61
ground during the switching of the pulse.
When multiple plasma channels contacted the
grounded field shaper during a shot, at least one of the channels always made contact with the
low-voltage electrode. An example of data from a shot when multi-channeling occurred can be
seen in Figure 5.7. This data can be compared to the J. C. Martin equation [3],[4]. It can be
observed from the plot that the arc initially contacted the low voltage electrode and later broke
off to ground resulting in an approximate pulse width of 0.5 s. With an electrode area A of
7.07 cm2 (radius 1.5 cm) and k equaling 0.5 for transformer oil, the result of the Martin equation
in seen in Equation 5.1.
F =
k
1
t 3A
1
10
= 518 kV/cm
(5.1)
This result indicates a likelihood of breakdown at 518 kV/cm across the gap. In
actuality, with a charge voltage of 1.2 MV and a gap spacing of 5.4 cm, there was a mean electric
field of 222 kV/cm across the gap. With the FEF of 6.65 determined in the previous chapter,
the maximum electric fields were closer to 1.48 MV/cm. This exceeds the field strength
necessary for breakdown according to the Martin equation. The impurities present in the
unfiltered oil near the switch area assist in the formation of the extra channels. Additionally, a
small misalignment was noted after the shot indicating the high-voltage electrode was closer to
ground than originally calibrated.
Because the insulating oil quality was so important to the breakdown of the output
switch, additional breakdown tests were conducted with a Hipotronics OC60D [5] ASTM D877
dielectric tester on a sample of the Diala AX from the switch area of the tank. Results are
shown in Figure 5.8 and appear to slightly diverge from the manufacturers specifications of
breakdown voltages greater than 35 kV RMS [2].
62
Figure 5.6. Data from both pre and post-switch dividers (Shot 8)
Figure 5.7. Marx shot with multi-channel arcs to ground (Shot 33)
63
Figure 5.8. Oil breakdown with Hipotronics tester
It appears from the results of the ASTM D877 standard [6] breakdown test that the oil
may be minimally saturated with water. The oil in the Marx tank is exposed to the atmosphere
while the Marx is in operation. After operation, the oil is pumped out to the sealed, secondary
storage tank, but the opportunity for water to infiltrate the oil exists when the Marx is in use.
Water extraction can be attained though increased filtering.
64
[1]
Spiral delay line pulser: operation and maintenance manual, Pulse Sciences, Inc., PSIO&M-2420, Dec. 1991.
[2]
Diala AX data sheet, Shell Lubricants, [Online]. Available: http://www.shelllubricants.com/products/pdf/DialaA&AX.pdf
[3]
934-945, Jun. 1992.
[4]

[5]
OC60D Liquid dielectric test set, Hipotronics, [Online]. Available:

http://www.hipotronics.com/pdf/OC6090D-DS0803.pdf
[6]
D877-02e1 Standard test method for dielectric breakdown voltage of insulating liquids
using disk electrodes, ASTM International, [Online]. Available:
http://www.astm.org/cgibin/SoftCart.exe/DATABASE.CART/REDLINE_PAGES/D877.htm?L+mystore+x
tyx3815+1125378065
65
CHAPTER 6. ABRAMYAN NETWORK DESIGN
6.1. PFNs, PFLs, and the Abramyan Network

Desiring a rectangular pulse for a potential directed energy technology testbed with the
UMC Marx in the most economical and space-conscious manner required a review of typical
pulse-forming techniques.
As introduced in Chapter 1, a pulse-forming line (PFL), or
transmission line, lends itself to physical impracticalities due to excessive length requirements.
In contrast, a pulse-forming network (PFN) may be smaller in size but more expensive due to
specialty components. The third option is the implementation of an Abramyan network [1], [2],
[3] into the current Marx configuration. Minimal addition of components is required thereby
maximizing cost-efficiency and ease of installation.
For the required output parameters of approximately 1 MV for 1s, PFLs, PFNs, and
Abramyan networks all are reasonable methods of pulse shaping [4], [5], [6], [7]. As mentioned
earlier, the sheer size of a PFL addition to the current configuration is a major concern. It is
known that the duration of the output pulse is equal to twice the one-way transit time of the
transmission line.
p =
2l
= 2
vp
vp =
p
l
= duration of the pulse, s

= length of line, m
66
(6.1)
(6.2)
vp
= velocity of propagation, m/s
= one-way transit time, s
= velocity of light, m/s
= relative permittivity of dielectric, F/m
As Smith discusses in Transient Electronics [4], if a polymer plastic, like polypropylene, with
a low dielectric constant ( r = 2 3 ) is used as the dielectric, then the velocity of the pulse is
approximately 2 108 m/s. This results in a transit time of about 20 cm/ns. To produce a 1 s
pulse from this configuration would require 100 m of transmission line. This obviously violates
the space constraints and is impractical for use with the UMC Marx. Additional disadvantages
include the limited availability of impedances other than 50 with the alternative of building a
stripline of a different impedance both expensive and difficult.
PFNs are essentially lumped parameter approximations of PFLs and can require less
space, but more specialized components. Using a Guillemin Type A PFN [4], [5] as an example,
the required additional components are shown in Figure 6.1.
Figure 6.1. Guillemin Type A PFN [4], [5]
The values of the inductances for the network are calculated by multiplying the number
supplied by the network impedance, Z N , and the pulsewidth desired, . The values of the
67
capacitors are calculated by multiplying the number supplied by the pulsewidth and dividing by
the impedance. As with any transmission line, or variant of, a matched load is always ideal.
Therefore, Z N = 300 for the example. These parameters result in the PFN described in
Figure 6.2.
Figure 6.2. Type A PFN for UMC Marx
The values of the capacitors and inductors seem to encompass a large range. Although
inductors can be wound differently for different inductances, capacitors are not nearly as
adaptable. Compared to the pulse-forming line, this pulse shaping method results in the need
for many additional components.
In contrast, the Abramyan network utilizes components readily available, and only
requires the addition of a single, easy-to-construct inductor. The Abramyan network is simple to
adjust for multiple output pulse requirements and can operate under the conditions of a
collapsing load. Inductors of several values can be constructed for alternative parameters and
can be interchanged in the system for varying pulse-forming needs. The Michigan Electron
Long Beam Accelerator (MELBA) employed an Abramyan network to utilize the voltage
compensation for a 1 s collapsing diode load impedance [8], [9], [10], [11], [12], [13].
68
6.2. Physical Implementation of the Abramyan

Developing an Abramyan network for implementation in the UMC Marx began with an
analysis of physical requirements. Ideally, all components should fit in the existing oil tank. To
accomplish this, a few of the Marx stages must be removed. Testing of the Marx at 1 MV
proved successful and therefore subtraction of stages from the 3 MV configuration is not a
concern. Given that each stage is approximately 5 inches wide with supports, the removal of 5
stages provides approximately 2 more feet of length in the tank to utilize for other purposes. It
is assumed this is enough space to fit an inductor in the tank given the calculations following
(Equation 6.3 and Equation 6.4).
Applying a reasonable inductor value of 5 H to the Wheeler formulas [14], [15] for
inductance, basic geometry can be specified to determine the corresponding number of turns
required. With the use of 1.27cm (0.5 in) bendable copper pipe, a coil diameter of 30.48 cm (12
in) and coil length of 30.48 cm (12 in) appears to be reasonable.
The Wheeler formula
determines that 5 turns are necessary to reach a 5 H inductance.
L=
d 2 n2
l + 0.45 d
L
d
inductance (H), 5.26
coil diameter (m), 0.3048
coil length (m), 0.3048
number of turns, 5
(6.3)
The next step in the inductor design is to verify that the electric field strength between
the coils is not greater than the breakdown strength of the transformer oil. The coil radius,
spacing, and voltage between coils can be used with the Bouwers and Cath maximum field
strength formula for two cylinders [16]. Observing a conservative transformer oil breakdown
69
strength of 40 kV/cm [17], the resultant electric field strength of 26 kV/cm is well within the
limits of safe operation. Therefore, the space allocated for the inductor is sufficient.
Figure 6.3. Two cylinders (approximating turns of copper pipe)
Emax = 0.9
2
r +a
2
2.3 r log
r
Emax
maximum electric field, 26 kV/cm
potential difference between coils, 90 kV
coil spacing, 6.096 cm (12in / 5turns = 2.4 in)
coil radius, 1.27 cm (0.5 in)
(6.4)
It was also desired to have a small amount of room in the tank to slightly separate the
reverse-charging stages from the forward-charging stages. An additional 3 stages were allocated
for removal for this purpose. A total of 8 stages are then designated removable. The remaining
22 stages will need to be optimized in the Abramyan network configuration. The original 3 MV
Marx configuration from Figure 2.2 can be adjusted with the removal of stages and simplified to
the circuit in Figure 6.4.
70
Figure 6.4. Equivalent Marx circuit with twenty-two stages
Implementing the Abramyan network into the above circuit results in the schematic
shown in Figure 6.5. Values are not listed for C Marx and LMarx due to recalculations needed
based on the reallocation of stages for C a and La . These calculations will occur in the
following section.
Figure 6.5. Basic Abramyan in twenty-two stage Marx
71
This circuit can be implemented in the UMC Marx with minimal cost. Simulations were
conducted with the removal of eight stages for a total of twenty-two functioning stages.
Optimizing this circuit for a 1 MV output with a 1 s pulsewidth, a total of five stages were
reverse-charged in the simulations. Determining the number of stages to be reversed is based
on the desired output voltage. The output of the Marx with the Abramyan is the result of
subtracting the voltage on reverse-charged stages from the voltage on the forward-charged
stages. All capacitors were charged to 90 kV (+50 kV, -40 kV) per stage to match the current
power supplies in the system. The equivalent capacitance of the reversed stages is 9.6 nF. At 90
kV per stage charge, the reverse-stage capacitors will charge to a total voltage of -450 kV. The
series inductance of these capacitors is 766.7 nH. The remaining stages constitute an equivalent
capacitance of 2.8 nF with a series inductance of 2.6 H. The charge on these stages at 90 kV
per stage is 1.53 MV. The peak expected output from the Marx and Abramyan with the reversal
of 5 stages is 1.08 MV (forward 1.53 MV reverse 450 kV).
6.3. Circuit Simulations

The challenge in Abramyan network design is the optimization of the additional
inductor, Ld . By its nature, the design of an Abramyan network is an iterative process of
determining possible solutions and realizable components [18]. One method of analysis is to
approach the Abramyan as the addition of a simple oscillating tank circuit. A tank circuit
consists of a capacitor and inductor in parallel with a resonant frequency f r .
fr =
1
2 LC
72
(6.5)
With the selection of five reverse-charged stages, the capacitance Ca and inductance La
of the reverse-charged stages maintain fixed values of 9.6 nF and 766.7 nH respectively. The
only remaining variable is the additional inductor, Ld . Because La and Ld can be assumed to
be in series, the sum of the two will serve as the total parallel inductance of the tank circuit.
Choosing Ld = 5H results in the following calculations.
fr =
1
= 676kHz
2 5.7667 H 9.6nF
(6.6)
1
1
=
= 1.5s
f r 676kHz
(6.7)
Tr =
The period calculated is 150% of the pulse length desired from the output of the Marx
and Abramyan circuit. This result can be used for a variety of inductor values to determine
approximate pulse length (Table 6.1). These values can be verified by the simulation results in
Figure 6.6 and Figure 6.7. The slight increase or decrease in inductor values as described above
facilitate changes in both the pulse width and amplitude to suit many output specifications.
Table 6.1. Expected pulse lengths from variable inductor values
Inductor value
1 H
3 H
5 H
7 H
9 H
Expected pulse length

546 ns
798 ns
986 ns
1.14 s
1.28 s
73
Figure 6.6. Abramyan network with inductor values 1 H - 9 H
74
Figure 6.7. Varying Abramyan network inductor values
Additionally, previous work at Pulse Sciences, Inc. [18] indicates a correlation between
the frequency of the reversal stage and the rate of change of voltage on the reversal stage VR (t )
at the zero crossing point. This relationship is given by [18]:
dVR (t )
=
dt
Va
(La + Ld ) Ca
(6.8)
assuming
(Ra + Rd ) << (La + Ld )

Ca
(6.9)
The load voltage droop is given by
dVL I L
VL
=
=
dt
Ce RL Ce
75
(6.10)
where Ce = Ca || Cm [18]. Following the derivation in this reference, Equation 6.10 can be set to
equal Equation 6.8 to solve for Ce . Since Ce is a known variable in our situation, this equation
can be used to solve for Ld . The previous equations are combined and give Equation 6.11.
Ld =
Va2 RL2 Ce2

La
VL2 Ca
(6.11)
This equation was utilized in a MATLAB program to quickly calculate values for a
variety of parameters. The code is documented with sample output in Appendix E. The output
of the program gives a range of values for Ld that are optimized for output voltages between 1 1.5 MV and apply to the selected input parameters. The ranges supplied can be simulated in
PSPICE to determine the optimum value for the desired pulse shape.
The previous sets of calculations also show correlation with the plot from the Laplace
transform developed in Mathematica in Figure 6.8 [19], [20]. This graph represents the signal
components (red, blue, and green) that together compose the output pulse (dashed purple). The
green line represents the forward-charged capacitors, the red line represents the reverse-charged
capacitors, and the blue line represents the ringing of the Abramyan circuit. With an oscillation
period of 1.5 s, the highest voltage peak of the ringing circuit occurs at 0.5Tr shown by the
blue line. Therefore, the peak voltage from the Abramyan or tank circuit would occur at 750 ns.
The voltage remaining on the forward-charged capacitors shown in green is approximately 50%
at this time. This appears to be the ideal combination for the parameters and output required.
Full documentation of the Mathematica code developed to calculate this plot is provided in
Appendix F.
76
Voltage HMV L
2
1.5
1
0.5
0.5
1.5
2.5
Time Hus L
0.5
1
1.5
Figure 6.8. Signal decomposition of Abramyan with Marx
Figure 6.9. Conceptual design of UMC Marx with Abramyan network
For the pulse specification proposed, it appears that the selection of a 5 H inductor is
appropriate.
Further PSPICE simulations were conducted on the circuit in Figure 6.9.
77
Simulation code is documented in Appendix G. The resultant pulse is seen in Figure 6.10 with
the characteristics listed in Table 6.2. The risetime of the pulse increased minimally from the
original Marx configuration while the pulse width increased dramatically.
Table 6.2. Abramyan network pulse characteristics, Ld = 5 H
Risetime (10-90%)
Fall time (10-90%)
Pulse-top ripple
24 ns
0.8 s
0.5 s
8%
Figure 6.10. Marx and Abramyan with 5 H inductor
Simulations were also conducted to test the possibility of triggering the spark gap
switches at different times and the results indicated no advantage. Triggering the switches at the
same time, or without modification, appeared to optimize the pulse.
78
Figure 6.11. Abramyan with damping resistor
The drawback to the Abramyan network is the voltage reversal on the reverse-charged
stages. Damping resistors can be placed in the Abramyan section of the circuit in Figure 6.11 to
reduce the ringing seen in Figure 6.12 but accompany a decrease in pulse width in Figure 6.13
and Table 6.3. The ringing circuit that reverses the charge on the capacitors is also responsible
for the pulse extension.
Figure 6.12. Full length output from Abramyan and Marx
79
Figure 6.13. Output pulses with damping resistor in the 5 H Abramyan network
80
Table 6.3. Pulse lengths for resistance values in Abramyan
Resistance value
1
5
10
15
Pulse length (90-90%)

800 ns
720 ns
580 ns
410 ns
The decrease in pulse width is attributed to the same resistor that damps the late-time
ringing of the pulse. From the plots in Figure 6.13, the optimum resistance value to supplement
the UMC Abramyan appears to be between 1 and 5 . The late-time ringing is damped in
approximately three periods while decreasing pulse width by only 80 ns. This is a significant
portion of the 800 ns pulse, but the long-term benefit of increasing capacitor lifetime may
outweigh the alternative.
Another possible solution was employed on MELBA [8], [9], [10], [11], [12], [13]. The
circuit contained a high pressure crowbar switch that was triggered to remove 90% of the
generator voltage in approximately 40 ns [8]. With this option, smaller resistance values can be
used if the crowbar terminates the pulse before the late-time ringing can occur. Additionally, the
introduction of time-variable resistance into the Abramyan stage presents many potential
possibilities.
Keeping in mind that the Abramyan network utilizes components already available and only
requires the addition of a single component, the drawbacks discussed can be seen as a minor
compromise to its ease of construction. While some capacitor reversal is inherent in a circuit of
this type, the simplicity of the Abramyan network addition to operational systems provides a
cost-effective solution to the requirement of a rectangular pulse.
81
[1]
E. A. Abramyan, E. N. Efimov, and G. D. Kuleshov, Correction of the pulse shape of

an Arkadev-Marks voltage generator, High-Temperature Institute of the Academy of
Sciences of the USSR, Moscow. Translated from Pribory i Tekhnika ksperimenta, No. 4,
pp.170-172, Jul.-Aug. 1979.
[2]
E. A. Abramyan, B. A. Altercorp, and G. D. Kuleshov, Microsecond intensive ebeams, Proceedings of the 2nd International Topical Conference on Power Electron and Ion Beam
Research and Technology, pp. 743-754, Oct. 1977.
[3]
E. A. Abramyan, E. N. Efimov, and G. D. Kuleshov, Energy recovery and power

stabilization of pulsed electron beams in Marx generator circuits, Proceedings of the 2nd
International Topical Conference on Power Electron and Ion Beam Research and Technology, pp.
755-760, Oct. 1977.
[4]
P. W. Smith, Transient Electronics: Pulsed Circuit Technology, Wiley, 2002.
[5]
G. N. Glasoe and J. V. Lebacqz, Eds., Pulse Generators, Massachusetts Institute of

Technology Radiation Laboratory Series, McGraw-Hill, 1948.
[6]
S. T. Pai and Q. Zhang, Introduction to High Power Pulse Technology, World Scientific, 1995.
[7]
I. Smith, Pulsed power for 0.3-0.5 s durations, Lawrence Livermore National Laboratory
Report, 18318, Oct. 1979.
[8]
R. M. Gilgenbach, L. D. Horton, R. F. Lucey, Jr, S. Bidwell, M. Cuneo, J. Miller, and L.

Smutek, Microsecond electron beam diode closure experiments, Proceedings of the 5th
IEEE Pulsed Power Conference, pp. 126-132, Jun. 1985.
82
[9]
J. I. Rintamaki, Effects of RF plasma processing on the impedance and electron

emission characteristics of a MV beam diode, Doctoral dissertation, University of
Michigan, 1999.
[10] J. M. Hochman, Microwave emission of large and small orbit rectangular gyrotron
devices, Doctoral dissertation, University of Michigan, 1998.
[11] M. T. Walter, Effects of tapering on high current, long pulse gyrotron backward wave
oscillator experiments, Doctoral dissertation, University of Michigan, 1995.
[12] M. E. Cuneo, Characterization of the time-evolution of a microsecond electron beam
diode with anode effects, Doctoral dissertation, University of Michigan, 1989.
[13] M. E. Cuneo, Spectroscopic study of anode plasmas in a microsecond electron beam
diode, IEEE Transactions on Plasma Science, vol. 15, no. 4, pp. 375-379, Aug. 1987.
[14] H. A. Wheeler, Formulas for the skin effect, Proceedings of the I.R.E., pp 412-424,
Sept. 1942.
[15] H. A. Wheeler, Simple inductance formulas for radio coils, Proceedings of the I.R.E.,
pp 1398-1400, Oct. 1928.
[16] A. Bouwers and P. G. Cath, The maximum electric field strength for several simple
electrode configurations, Philips Tech. Review, vol. 6, no. 9, pp. 270-278, 1941.
[17] R. J. Adler, Pulse power formulary, North Star Research Corporation, [Online].
[18] R. Curry, K. Nielsen, I. Smith, Abramyan networks: a general purpose network for
linear induction accelerators, Pulse Sciences, Inc., San Leandro, CA, Apr. 1986.
[19] G. V. Lago and D. L. Waidelich, Transients in Electrical Circuits, New York: Ronald Press,
1958.
[20] Mathematica 4.1 for Students [Online]. Available: www.mathematica.com
83
CHAPTER 7. CONCLUSION
This thesis has presented the development of the large-gap oil dielectric switch test
facility at the University of Missouri-Columbia.
Designs for future Abramyan network
implementation have been included in this work. A description of the system for all future
research on the UMC Marx is documented and thoroughly explained. The simulation circuits
and code have been developed and accurately predicted experimental results. Early test data has
proven the Marx to reliably operate at the desired 1 MV output with capability for increased
output voltages up to 3 MV. Abramyan network designs for multiple output ranges are available
for implementation.
Chapter 1 provided an introduction to Marx generators and the need for pulse shaping
systems.
Basic Marx circuits were developed and analyzed.
Abramyan networks were
introduced with the possibility of implementation into the UMC Marx generator.
The UMC Marx facility was described in detail in Chapter 2. Enhanced circuits with
detailed parameters were determined. The subsystems of the Marx including the charging
system, the triggering system, and the peaking switch were discussed relative to the control
system.
Diagnostics for the facility were developed in Chapter 3. Voltage divider theory was
discussed and applied to the UMC Marx. Construction and calibration of the dividers was also
mentioned. Testing of the dividers proved successful and resulted in implementation into the
Marx system.
The UMC Marx was modeled and simulated in Chapter 4. Simulations of the Marx
circuits produced expected results. Breakdown of the output switch was calculated for use with
variable gap spacing and charge voltage in experimental work. Electrostatic simulations were
84
conducted to analyze the electric fields present with the geometry of the output switch. The
electric field calculations and electrostatic simulations provided similar results.
From the
maximum field strengths determined, the field enhancement factor was calculated and proved
reasonable in the following chapter.
Chapter 5 provided experimental parameters and results.
The Marx performed as
expected under 1 MV operation. Operation at higher voltages may require modifications to the
output switch geometry. The increased electric fields on the field shaper appeared to affect the
pulse after successive shots with minimal carbon filtration. Industry standard oil breakdown
tests indicated the possibility of water saturation. Both carbon and water accumulation in the oil
can be addressed with enhanced filtering capabilities.
An Abramyan network was designed for the UMC Marx in Chapter 6. Background
information on the alternative pulse-forming lines and pulse-forming networks was presented.
Designs of the physical implementation into the Marx were discussed and simulations of
expected output followed.
Future work includes further studies in oil breakdown and specific levels of impurities in
the oil. It may also be determined that a threshold level of carbon exists allowing the Marx to
maintain a repeatable result such as in Figure 5.7. Additionally, the output polarity of the Marx
can be reversed for further oil switch testing.
At the conclusion of this project, the UMC Marx is now operational and oil switch
testing is currently being conducted. Conceptual Abramyan network designs for the UMC Marx
have been optimized to deliver a 1 s pulse at 1 MV and are available for implementation.
85
APPENDIX A. SPARK GAP DIMENSIONS
86
APPENDIX B. MARX SIMULATION CODE

This code was developed from PSPICE schematics to simulate the Marx parameters
both with and without the peaking circuit. Results from these simulations are discussed in
Chapter 4. The first section describes the Marx circuit without the peaking capacitor. The
second section includes the peaking capacitor in the analysis of the circuit.
*Libraries:
* Local Libraries :
* From [PSPICE NETLIST] section of pspice91.ini file:
.lib "nom.lib"
*Analysis directives:
.TRAN 0 2u 0 1ns
.PROBE
.INC "simple marx-SCHEMATIC1.net"
**** INCLUDING "simple marx-SCHEMATIC1.net" ****

* source SIMPLE MARX
C_C1
0 N00016 1.6n IC=-3MEG
L_L1
N00116 N00022 4.6u
R_R1
0 N00022 300
X_U1
N00016 N00116 Sw_tClose PARAMS: tClose=.399u
ttran=1n Rclosed=0.01
+ Ropen=1Meg
**** INCLUDING "simple marx-SCHEMATIC1.als" ****
.ALIASES
C_C1
C1(1=0 2=N00016 )
L_L1
L1(1=N00116 2=N00022 )
R_R1
R1(1=0 2=N00022 )
X_U1
U1(1=N00016 2=N00116 )
.ENDALIASES
.END
87
****
Voltage Controlled Switch MODEL PARAMETERS
**************************************************************
RON
ROFF
VON
VOFF
X_U1.Smod
.01
1.000000E+06
1
0
****
INITIAL TRANSIENT SOLUTION
TEMPERATURE =
27.000 DEG C
**************************************************************
NODE
VOLTAGE
VOLTAGE
NODE
VOLTAGE
(N00016) 3.000E+06 (N00022)

(X_U1.3)
0.0000
NODE
VOLTAGE
899.7300 (N00116)
899.7300
VOLTAGE SOURCE CURRENTS

NAME
CURRENT
X_U1.V1
0.000E+00
TOTAL POWER DISSIPATION
0.00E+00
WATTS
JOB CONCLUDED
TOTAL JOB TIME
.19
The following is from the Marx schematic including the peaking switch.
*Libraries:
* Local Libraries :
.lib "nom.lib"
.TRAN 0 2u 0
.PROBE
.INC "marx-SCHEMATIC1.net"
88
NODE
**** INCLUDING marx-SCHEMATIC1.net ****

* source MARX
C_Cm
N00009 0 1.6nF IC=1.02MEGV
L_Lm
N03740 N00015 4.6uH IC=0
R_R
0 N00017 300
C_C3
0 N00015 200p IC=0
X_U4
N00015 N00017 Sw_tClose PARAMS: tClose=399ns
ttran=0 Rclosed=0.01
+ Ropen=1Meg
X_U2
N00009 N03740 Sw_tClose PARAMS: tClose=350ns
ttran=0 Rclosed=0.01
+ Ropen=1Meg
**** RESUMING marx-schematic1-marxsim.sim.cir ****
.INC "marx-SCHEMATIC1.als"
**** INCLUDING marx-SCHEMATIC1.als ****
.ALIASES
C_Cm
Cm(1=N00009 2=0 )
L_Lm
Lm(1=N03740 2=N00015 )
R_R
R(1=0 2=N00017 )
C_C3
C3(1=0 2=N00015 )
X_U4
U4(1=N00015 2=N00017 )
X_U2
U2(1=N00009 2=N03740 )
.ENDALIASES
**** RESUMING marx-schematic1-marxsim.sim.cir ****
.END
****
**************************************************************
RON
ROFF
VON
VOFF
X_U4.Smod
.01
1.000000E+06
1
0
X_U2.Smod
.01
1.000000E+06
1
0
****
TEMPERATURE =
27.000 DEG C
**************************************************************
NODE
VOLTAGE
VOLTAGE
NODE
VOLTAGE
89
NODE
VOLTAGE
NODE
(N00009) 1.020E+06 (N00015) 2.038E-06 (N00017) 611.2E-12

(N03740) 1.019E+06
(X_U2.3)
0.0000 (X_U4.3)
0.0000

NAME
CURRENT
X_U4.V1
X_U2.V1
0.000E+00
0.000E+00
0.00E+00
WATTS
JOB CONCLUDED
TOTAL JOB TIME
.98
90
APPENDIX C. MARX OPERATION
Preparation for Marx Operation

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Determine Marx charge voltage, _______ kV per power supply.

From Figure 4.2 in Spiral Delay Line Pulser Manual, determine Marx spark gap operating pressure, _______ psig.
From Figure 4.3, determine oil switch gap spacing, _______ cm.
Determine trigger generator charge voltage, _______ kV.
From Figure 4.1, determine trigger generator operating pressure, _______ psig.
Verify intended charge polarity is consistent with Marx input charge lines.
Positive output = Positive charge to first capacitor, negative to second.
Negative output = Negative charge to first capacitor, positive to second.
Slightly pressurize spark gaps (5-10 psi) to discourage oil leaking into them.
Verify all resistors and gas lines are correctly installed.
Check all connections at low voltage end of Marx are secure. Pay special attention to trigger lines into resistor connection balls.
Plug in both power cables to hydraulic pump.
Control box cable
Pump cable
Turn on pump and raise Marx slightly to remove safety pins.
Remove all 4 safety pins.
Lower Marx into tank.
Check Marx top frame is sitting on current contacts on both low and high voltage ends of tank.
Turn hydraulic pump off.
Unplug both pump power cables (control and pump).
Set oil switch spacing to distance determined in Step 3.
Clear area and move scaffolding away from tank.
Open valves to oil wash recirculator pump. Turn on pump.
Verify safety ground connections from trigger generator, control/charging cabinet, and tank to wall.
Open window and secure both SF6 and air vents outside.
Pressurize pneumatic cylinders with air to 95 psig. (Do not exceed 100 psi - Versa valve maximum.)
Plug isolation transformer into wall.
Turn transformer on.
Check all 4 plugs in control/charging cabinet power strip.
Turn on control/charging cabinet power strip.
Pressurize Marx spark gaps with SF6 to pressure determined in Step 2.
Pressurize trigger generator with SF6 to pressure determined in Step 5. (Purging need only be done one every 15 shots.)
Verify output circuit configuration with associated monitor connections.
Set up barriers around test area.
Operating Procedure
**Verify scope is ready for output data**
Turn off oil wash recirculating pump.
31
Turn CONTROL POWER ENABLE KEY to turn control panel on.
32
Manually depress the DUMP button to clear controller of any abnormal start-up logic states.
33
Check the trigger charge rate (pot on back of cabinet). It should be set to track slightly ahead of desired Marx charge level.
34
Set the pot on each Marx power supply to zero.
35
Verify all interlocks are satisfied.
36
Evacuate test area.
37
Announce start Pre-charge Sequence.
38
Push CHARGE button.
39
Slowly increase voltage on both Marx power supplies to set the voltage to the correct charge level for firing.
40
Turn off Marx power supplies and hit DUMP. The Marx is now ready for the firing sequence.
41
Announce start Firing Sequence.
42
Follow the numbered-button firing procedure on the control cabinet.
43
Charge button
80kV Glassman power on
50kV Glassman power on
-40kV Kaiser power on
Monitor charge sequence on control panel. Sequence can be terminated at any time by pushing DUMP button.
44
When FIRE light turns RED the system can be triggered by pushing the FIRE button.
45
91
46
47
Follow the numbered-button Turn-off Procedure on the control cabinet. **Be sure to begin this immediately after Marx fires**
-40kV Kaiser power off
50kV Glassman power off
80kV Glassman power off
Control power enable off
Manually depress DUMP switch after each shot as an additional safety precaution.
Shutdown Procedure
Short-term:
Set the pressure in the spark gaps to 10 psig.
48
Set the pressure in the trigger vessel to 10 psig.
49
Turn off the CONTOL POWER ENABLE KEY.
50
Close valves in the compressed air supply line.
51
Close valves on SF6 cylinder and in supply line.
52
Close valves on oil recirculating pump.
53
Remove barriers around test area.
54
Long-term:
For extended shutdown periods, the system should be drained of oil. Prior to startup the Marx should be carefully inspected.
55
92
Shot #
Marx Charge
Voltage (kV)
Spark Gap
Pressure (psi)
Trigger Charge
Voltage (kV)
Trigger Gen
Pressure (psi)
Gap Spacing (dial

reading)
Ch.1 File #
Ch. 2 File #
3/17/2005
17
40
15
180
trigger generator pre-breaking, need

to increase TG pressure
3/17/2005
17
40
20
180
arcing to ground at output, data

appears as rectangular pulse
3/17/2005
20
40
20
160
decreasing gap spacing
3/17/2005
20
40
20
140
3/17/2005
20
40
20
190
3/17/2005
20
40
20
190
3/17/2005
20
40
20
190
10
11
ran pump briefly, picture
3/17/2005
17
40
20
180
12
13
waited approx. 45min
3/17/2005
17
40
20
180
14
15
no wait
3/17/2005
10
17
40
20
180
16
17
no wait
3/17/2005
11
17
40
20
180
18
19
no wait
21
ran pump 10 min, saw arc begin to

form off to side and corrected to
electrode, bubbles
3/17/2005
5/11/2005
12
13
17
20
4
4
40
40
20
20
93
180
90
20
Notes
Date
APPENDIX D. MARX SHOT LOG
flushed carbon away from electrode
pumped 20 min before, change to

bandwidth limited, Ch.1 20 dB driven
by 50
5/11/2005
14
20
40
20
190
5/11/2005
15
20
40
20
190
shortened Ch. 2 cable after seeing

reflections
5/12/2005
16
20
40
20
190
post1
pre1
5/12/2005
17
20
40
20
190
post2
pre2
pump ran day before at sat all night,

multichanneling to electrode and gnd,
20 dB, 50 on both Ch.1,2
pump 37 minutes, stopped scope
from averaging, change to 200ns,
tried to take digital picture but
blurred
5/12/2005
18
20
40
20
190
post3
pre3
no pump, bad pic, change to 400ns
5/12/2005
19
20
40
20
190
post4
pre4
pump 13 minutes, TG pre-fired,

change scope to 20 MHz bandwidth
limited
5/12/2005
20
20
40
20
190
post5
pre5
pump 20 min, increase SF6 in TG to

22 psi, no file
5/12/2005
21
20
40
20
190
post6
pre6
no pump, picture taken
5/12/2005
22
17
40
22
180
post7
pre7
5/12/2005
23
17
40
22
180
post8
pre8
pump 40 min, no fire, possibly no 40kV supply HV, one SG fire, purge
SG SF
6
5/12/2005
24
17
40
22
180
post9
pre9
no pump, no data, scope triggered to

early, sounded, looked like good shot
5/12/2005
25
17
40
22
180
post10
pre10
pump 5 min, no data, scope triggered

to early, sounded, looked like bad
shot
5/12/2005
26
17
40
22
180
post11
pre11
pump 10 min, sounded looked good,

arc had legs but didn't touch gnd
5/12/2005
27
17
40
22
180
post12
pre12
pump 40 min, big time difference

between charge and fire pulse on
scope
5/12/2005
28
17
40
22
180
post13
pre13
no pump, wait 15 min, good shot,

data
5/12/2005
29
20
40
22
180
post14
pre14
no pump, wait 20 min, bad shot
5/16/2005
30
20
40
22
180
post1
pre1
realigned electrodes by adjusting

supports, pumped H20 out of oil 3
days, multichannel to electrode, gnd
5/16/2005
31
20
40
22
180
post2
pre2
pump 20 min, multichannel to

electrode, gnd
94
5/16/2005
32
20
40
22
180
post3
pre3

electrode, gnd
5/16/2005
33
17
40
22
180
post4
pre4

electrode, gnd, pulse looked better
5/17/2005
34
32
15
60
40
90
post1
pre1
5/17/2005
35
32
15
60
40
90
post2
pre2
no pump, took bandwidth filtering

off, overshot scope so changed to 30
dB, observed billowy carbon clouds
pump 20 min, realigned with
supports, pump 20 min, supports fell
during shot, square pulse, lots carbon
again
5/17/2005
36
32
15
60
40
90
post3
pre3
pump 30 min, fixed diagnostics with

copper, realigned with cable
5/19/2005
37
32
15
60
40
190
post1
pre1

electrode, gnd
5/19/2005
38
32
15
60
22
240
post2
pre2

electrode, gnd, large carbon cloud
5/19/2005
39
17
40
22
180
post3
pre3
no fire, no data
5/19/2005
40
17
40
22
180
post4
pre4
no fire, no data, SG pressure too high
5/19/2005
41
17
40
22
180
post5
pre5
reduced SG pressure, good shot
5/19/2005
42
17
40
22
180
post6
pre6
no pump, multichannel to electrode

only
5/19/2005
43
17
40
22
180
post7
pre7
thick arc to electrode only
95
APPENDIX E. MATLAB CODE

This MATLAB program accepts inputs for parameters for Abramyan network design. The
program is based on the UMC Marx and prompts for the number of stages to reverse-charge from
of the total number of stages in the Marx. The charge per stage is assumed to be 90 kV (+50 kV, 40 kV) based on the current configuration of the Marx, but the program prompts for changes. The
load impedance is assumed to be 300 but the program again prompts for alterations. From these
inputs, capacitance and inductance values are calculated for both the forward and reverse-charged
stages. The inductor to be added to the Abramyan section is computed using the following
equation.
Ld =
Va2 RL2 Ce2

La
VL2 Ca
Prior work used to derive the equation above is referenced in Chapter 6. Outputs from the
program can be used for circuit simulation purposes.
clear;
format('short');
total_stages = input('Total number of stages in Marx: ');
reverse_stages = input('Total number of stages to reverse:
forward_stages = total_stages - reverse_stages';
charge_per_stage = input('Charge per stage: [90kV] ');
if isempty(charge_per_stage)
charge_per_stage = 90000;
end
load = input('Load impedance: [300ohm] ');
if isempty(load)
load = 300;
end
forward_voltage = charge_per_stage * forward_stages;
reverse_voltage = charge_per_stage * reverse_stages;
96
');
total_capacitance = inv(1/(48e-9)*(forward_stages +
reverse_stages));
load_voltage_beginning = 1.5e6;
load_voltage_end = 1e6;
Cm = inv(1/(48e-9)*(forward_stages))
V_Cm = forward_voltage
Lm = (4.6e-6/30 * forward_stages)
Ca = inv(1/(48e-9)*(reverse_stages))
V_Ca = -reverse_voltage
La = (4.6e-6/30 * reverse_stages)
Ld_sweep_beginning = (reverse_voltage * load * total_capacitance
/ load_voltage_beginning)^2 * (1/Ca) - La
Ld_sweep_end = (reverse_voltage * load * total_capacitance /
load_voltage_end)^2 * (1/Ca) - La
ApproxMarxOutput = V_Cm + V_Ca
Example of results:
Total number of stages in Marx: 22
Total number of stages to reverse: 5
Charge per stage: [90kV]
Load impedance: [300ohm]
Cm =
2.8235e-009
V_Cm =
1530000
Lm =
2.6067e-006
Ca =
9.6000e-009
V_Ca =
-450000
La =
7.6667e-007
Ld_sweep_beginning =
3.2499e-006
Ld_sweep_end =
8.2705e-006
ApproxMarxOutput =
1080000
97

Cm =
2.4000e-009
V_Cm =
1800000
Lm =
3.0667e-006
Ca =
4.8000e-009
V_Ca =
-900000
La =
1.5333e-006
1.5747e-005
Ld_sweep_end =
3.7347e-005
ApproxMarxOutput =
900000

Cm =
2.8235e-009
V_Cm =
1530000
Lm =
2.6067e-006
Ca =
1.6000e-008
V_Ca =
-270000
La =
4.6000e-007
5.8976e-007
Ld_sweep_end =
1.9020e-006
ApproxMarxOutput =
1260000
98

Cm =
2.6667e-009
V_Cm =
1620000
Lm =
2.7600e-006
Ca =
6.8571e-009
V_Ca =
-630000
La =
1.0733e-006
7.4616e-006
Ld_sweep_end =
1.8130e-005
ApproxMarxOutput =
990000
99
APPENDIX F. MATHEMATICA CODE

This code was developed to solve the Laplace transform of the Marx and Abramyan circuit.
Two current loops in the Marx-Abramyan circuit were developed and solved accordingly. Both the
output voltage and ringing Abramyan voltage are plotted. Also, all signal components that combine
to form the output pulse are shown in an additional plot. Values were used in place of variables due
to extended computation time.
Clear@I1, I2, Ca, Va, La, Ld, Cm, Vm, Lm, R, s, V6, num, den, f6D;
1
Va
I1N J
N + Hs La I1L + Hs Ld HI1 I2LL 0,
SolveA9J
s Ca
s
1
Vm
J
I2N J
N + Hs Lm I2L + HR I2L + Hs Ld HI2 I1LL 0=, 8I1, I2<E
s Cm
s
J
+Las+LdsN Vm y
i
IR + 1 + Ld s + Lm sM Ld Va + Cas
Cms
s
Vm
{
k
::I1
,
Ld s2 Ld s ILd2 s2 I 1 + La s + Ld sM IR + 1 + Ld s + Lm sMM
Cas
J 1 +Las+LdsN Vm
s
>>
1 + La s + Ld sM IR + 1 + Ld s + Lm sM
Cas
Cms
I 1 + Las+ LdsM Vm
Cas
Ld Va +
s
;
1
2
2
Ld s I Cas + Las + LdsM IR + Cm1s + Lds + Lm sM
I2
I2 =
Cms
Ld2 s2 I
Ld Va +
Cas
Vm = 1.53 10^6;
Cm = 2.8 10^9;
Lm = 2.6 10^6;
Va = 450 10^3;
Ca = 9.6 10^9;
La = 766.7 10^9;
Ld = 5 10^6;
R = 300;
V6 = I2 R
num = Collect@Numerator@V6D, sD
numroots = Solve@num == 0, sD
den = Collect@Denominator@V6D, sD
denroots = Solve@den 0, sD
100
i
k
8
i
y
1.53106 1.0416710 +5.7667106 s y
s
k
{
s
{
8
8
I 1.0416710 + 5.7667 106 sM I300 + 3.5714310 + 7.6 106 sM +
300 9 +
4
1971.92
s2
40000000000
4.78125 1016
s2
88s 0. 4.9241 106 <, 8s 0. + 4.9241 106 <<
3.72024 1016 3.125 1010

11 2
0.00173001 s 1.88269 10 s
s2
s
2851.2
88s 9.04156 107<, 8s 1.22694 106<, 8s 123828. 4.21867 106 <, 8s 123828. + 4.21867 106 <<
f6 = Simplify@InverseLaplaceTransform@V6, s, tDD
Expand@LaplaceTransform@InverseLaplaceTransform@V6, s, tD, t, sDD
Plot@f6, 8t, 0, 3 10^6<, PlotStyle 8RGBColor@1, 0, 0D<D
7
6
H1.1785 106 + 0. L 9.0415610 t + H1.59052 106 + 0. L 1.2269410 t
H206009. 10070.5 L
I123828.4.21867106 M t
H206009. + 10070.5 L
I123828.+4.21867106 M t
206009. + 10070.5
206009. 10070.5
1.59052 106 + 0.
+
H123828. 4.21867 106 L + s
H123828. + 4.21867 106 L + s
1.22694 106 + s
110
800000
600000
400000
200000
510
-7
-6
-6
-6
-6
110 1.5 10 210 2.5 10 310
-6
-200000
Graphics
a1 = 9.041563404309268`*^7;
a2 = 1.2269389967064043`*^6;
aratio = a1 a2
C1 = 1.1784976473604566`*^6;
C2 = 1.5905157613165756`*^6;
Cratio = C1 C2
73.692
0.740953
101
1.1785 106 + 0.
9.04156 107 + s
sinfunc = ExpToTrigAH206009.05697805953` 10070.534283799661` L

H206009.05697805953` + 10070.534283799661` L
H123828.2413040823`4.218669197291801`*^6 L t
H123828.2413040823`+4.218669197291801`*^6 L t
Plot@sinfunc, 8t, 0, 5 10^6<, PlotStyle 8RGBColor@1, 0, 0D<D
H206009. 10070.5 L Cos@H4.21867 106 + 123828. L tD H206009. 10070.5 L Cosh@H123828. + 4.21867 106 L tD +
H10070.5 206009. L Sin@H4.21867 106 + 123828. L tD + H206009. 10070.5 L Sinh@H123828. + 4.21867 106 L tD
200000
110 -6
210 -6
310 -6
410 -6
510 -6
-200000
-400000
Graphics
ftime1 = C1 Exp@a1 tD;
ftime2 = C2 Exp@a2 tD;
ftime3 = ftime1+ ftime2 + sinfunc
Plot@8ftime1, ftime2, sinfunc, ftime3<, 8t, 0, 3 10^6<,
PlotStyle 8RGBColor@1, 0, 0D, RGBColor@0, 1, 0D, RGBColor@0, 0, 1D, RGBColor@1, 0, 1D<, AspectRatio 1,
PlotRange 82 10^6, 2 10^6<D
7
6
1.1785 106 9.0415610 t + 1.59052 106 1.2269410 t
H206009. + 10070.5 L Cos@H4.21867 106 + 123828. L tD H206009. 10070.5 L Cosh@H123828. + 4.21867 106 L tD +
H10070.5 206009. L Sin@H4.21867 10 + 123828. L tD + H206009. 10070.5 L Sinh@H123828. + 4.21867 10
6
210 6
1.5 10 6
110 6
500000
510 -7 110 -61.5 10 -6210 -62.5 10 -6310 -6

-500000
-1 10 6
-1.5 10 6
-2 10 6
Graphics
102
L tD
APPENDIX G. ABRAMYAN SIMULATION CODE

This code was developed from PSPICE schematics to simulate the Marx parameters with
the inclusion of the Abramyan network designed by the programs discussed in Appendix E and
Appendix F. The program was instructed to compute a range of multi-variable parameters with
varying inductor values. Results from these simulations are discussed in Chapter 6.
*Libraries:
* Local Libraries :
.lib "nom.lib"
.TRAN 0 3us 0
.STEP LIN PARAM L 1uH 9uH 2uH
.PROBE N([N00467])
.INC "abramyan0601-SCHEMATIC1.net"
**** INCLUDING abramyan0601-SCHEMATIC1.net ****
* source ABRAMYAN0601
C_Cm
N00465 N00180 2.8nF IC=1.53MEGV
R_Rload
N00467 0 300ohm
L_Lm
N01367 N00467 2.6uH
R_Rd
N03766 0 1ohm
C_Ca
N05437 0 9.6nF IC=-450kV
X_U1
N05344 N00180 Sw_tClose PARAMS: tClose=0 ttran=1n
Rclosed=0.01
+ Ropen=1Meg
R_R1
N05424 N05344 1ohm
X_U2
N00465 N01367 Sw_tClose PARAMS: tClose=0 ttran=1n
Rclosed=0.01
+ Ropen=1Meg
L_La
N05437 N05424 766.7nH
L_Ld
N00180 N03766 {L}
.PARAM L=1uH
**** RESUMING abramyan0601-schematic1-abramyan22stage.sim.cir
****
103
.INC "abramyan0601-SCHEMATIC1.als"
**** INCLUDING abramyan0601-SCHEMATIC1.als ****
.ALIASES
C_Cm
Cm(1=N00465 2=N00180 )
R_Rload
Rload(1=N00467 2=0 )
L_Lm
Lm(1=N01367 2=N00467 )
R_Rd
Rd(1=N03766 2=0 )
C_Ca
Ca(1=N05437 2=0 )
X_U1
U1(1=N05344 2=N00180 )
R_R1
R1(1=N05424 2=N05344 )
X_U2
U2(1=N00465 2=N01367 )
L_La
La(1=N05437 2=N05424 )
L_Ld
Ld(1=N00180 2=N03766 )
.ENDALIASES
**** RESUMING abramyan0601-schematic1-abramyan22stage.sim.cir
****
.END
****
RON
ROFF
VON
VOFF
X_U1.Smod
.01
1.000000E+06
1
0
X_U2.Smod
.01
1.000000E+06
1
0
****
TEMPERATURE =
27.000 DEG C
****
CURRENT STEP
PARAM L =
1.0000E-06
****************************************************************
NODE
VOLTAGE
VOLTAGE
(N00180)
(N01367)
NODE
VOLTAGE
NODE
VOLTAGE
-1.9795 (N00465) 1.530E+06 (N00467)

458.8600
458.8600
(N03766)
-1.9795 (N05344)-450.0E+03 (N05424)-450.0E+03
(N05437)-450.0E+03
(X_U1.3)
0.0000 (X_U2.3)
0.0000
104
NODE

NAME
CURRENT
X_U1.V1
X_U2.V1
0.000E+00
0.000E+00
0.00E+00
WATTS
****
TEMPERATURE =
27.000 DEG C
****
CURRENT STEP
PARAM L =
3.0000E-06
****************************************************************
NODE
VOLTAGE
VOLTAGE
(N00180)
(N01367)
NODE
VOLTAGE
NODE
VOLTAGE
-1.9795 (N00465) 1.530E+06 (N00467)

458.8600
NODE
458.8600
(N03766)
-1.9795 (N05344)-450.0E+03 (N05424)-450.0E+03
(N05437)-450.0E+03
(X_U1.3)
0.0000 (X_U2.3)
0.0000

NAME
CURRENT
X_U1.V1
X_U2.V1
0.000E+00
0.000E+00
0.00E+00
WATTS
****
TEMPERATURE =
27.000 DEG C
****
CURRENT STEP
PARAM L =
5.0000E-06
****************************************************************
NODE
VOLTAGE
VOLTAGE
NODE
VOLTAGE
NODE
VOLTAGE
(N00180)
-1.9795 (N00465) 1.530E+06 (N00467)
(N01367) 458.8600
105
458.8600
NODE
(N03766)
-1.9795 (N05344)-450.0E+03 (N05424)-450.0E+03
(N05437)-450.0E+03
(X_U1.3)
0.0000 (X_U2.3)
0.0000

NAME
CURRENT
X_U1.V1
X_U2.V1
0.000E+00
0.000E+00
0.00E+00
WATTS
****
TEMPERATURE =
27.000 DEG C
****
CURRENT STEP
PARAM L =
7.0000E-06
****************************************************************
NODE
VOLTAGE
VOLTAGE
(N00180)
(N01367)
NODE
VOLTAGE
NODE
VOLTAGE
-1.9795 (N00465) 1.530E+06 (N00467)

458.8600
NODE
458.8600
(N03766)
-1.9795 (N05344)-450.0E+03 (N05424)-450.0E+03
(N05437)-450.0E+03
(X_U1.3)
0.0000 (X_U2.3)
0.0000

NAME
CURRENT
X_U1.V1
X_U2.V1
0.000E+00
0.000E+00
0.00E+00
WATTS
****
TEMPERATURE =
27.000 DEG C
****
CURRENT STEP
PARAM L =
9.0000E-06
****************************************************************
106
NODE
VOLTAGE
VOLTAGE
(N00180)
(N01367)
NODE
VOLTAGE
NODE
VOLTAGE
-1.9795 (N00465) 1.530E+06 (N00467)

458.8600
458.8600
(N03766)
-1.9795 (N05344)-450.0E+03 (N05424)-450.0E+03
(N05437)-450.0E+03
(X_U1.3)
0.0000 (X_U2.3)
0.0000

NAME
CURRENT
X_U1.V1
X_U2.V1
0.000E+00
0.000E+00
0.00E+00
WATTS
JOB CONCLUDED
TOTAL JOB TIME
.11
107
NODE

A Fast, 3 MV Marx Generator

Uploaded by

Copyright:

Available Formats

A Fast, 3 MV Marx Generator

Uploaded by

Document Information

Original Description:

Original Title

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

A Fast, 3 MV Marx Generator

Uploaded by

Copyright:

Available Formats

A FAST, 3 MV MARX GENERATOR FOR MEGAVOLT OIL SWITCH

TESTING AND INTEGRATED ABRAMYAN NETWORK DESIGN

APPENDIX C. MARX OPERATION .................................................................................. 91

UMC Marx parameters ........................................................................................... 10

1.2. Marx Generators

Figure 1.1. Marx stages

LMarx = Lswitches + Lcaps + Lconnections

Figure 1.2. Simple Marx equivalent circuit

1.3. Abramyan Networks

Figure 1.4. Simple Abramyan circuit

References for Chapter 1

E. Marx, Verfahren zur Schlagprfung von Isolatoren und anderen elektrischen

J. F. Francis, High voltage pulse techniques, Texas Tech University, 1976.

R. A. Fitch, Marx and Marx-like high-voltage generators, IEEE Transactions on

K. R. Prestwich and D. L. Johnson, Development of an 18-megavolt Marx

J. D. Graham, D. G. Gale, W. E. Sommars, and S. E. Calico, Compact 400 kV Marx

F. E. Peterkin, D. C. Stoudt, B. J. Hankla, K. A. Boulais, J. S. Bernardes, Performance

R. J. Adler, Pulse power formulary, North Star Research Corporation, [Online].

G. A. Mesyats, Pulsed Power, Plenum, 2005.

G. N. Glasoe and J. V. Lebacqz, Eds., Pulse Generators, Massachusetts Institute of

[10] P. W. Smith, Transient Electronics: Pulsed Circuit Technology, Wiley, 2002.

[12] E. A. Abramyan, E. N. Efimov, and G. D. Kuleshov, Correction of the pulse shape of

CHAPTER 2. DESCRIPTION OF SYSTEM

2.1. Pulsed Power Research Laboratory

Figure 2.1. UMC pulsed power lab

2.2. UMC Marx Generator

Table 2.1. UMC Marx parameters

Figure 2.2. Marx circuit

Figure 2.3. Hydraulic pumping system

Figure 2.5. Secondary oil storage tank

Figure 2.6. Oil pumping station flow diagram

Figure 2.7. Dual-polarity charging schematic

Table 2.2. CuSO4 resistors [1]

Figure 2.8. Charging and triggering resistors

Figure 2.9. Ground reference resistors

Figure 2.10. Spark gaps installed in Marx

Figure 2.11. Assembled spark gap

Figure 2.12. Trigger generator

Figure 2.13. Gas and trigger lines

2.5. Control Systems

Figure 2.15. Control cabinet

Figure 2.16. Control system block diagram

2.6. Peaking Capacitor and Oil Switch

Figure 2.17. Peaking capacitor

Figure 2.18. Oil switch enhanced electrode

= conductor radius (cm), 0.9525

= conductor length (cm), 24.3

With measurable dimensions and known dielectric

properties of the insulator, the capacitance can be approximated by Equation 2.3.

= permittivity of space, 8.854 x 10-12 F/m

= breakdown field (MV/cm)

= electrode area (cm2), 1030

= time (s), 0.1

Figure 2.19. Simplified Marx with peaking capacitor

References for Chapter 2

Diala AX data sheet, Shell Lubricants, [Online]. Available: http://www.shelllubricants.com/products/pdf/DialaA&AX.pdf

Interline 850 data sheet, International Protective Coatings, [Online]. Available:

8300/04/10/14 Series Filter Assemblies, Pall Corporation, [Online]. Available: