Content uploaded by Ron Focia
Author content
All content in this area was uploaded by Ron Focia on Nov 18, 2015
Content may be subject to copyright.
* This work was supported AFRL, Directed Energy Directorate, under contract number FA9451-05-C-0021.
ε email: rfocia@pulsedpwr.com
A COMPACT, LOW JITTER, FAST RISE TIME, GAS-SWITCHED PULSE
GENERATOR SYSTEM WITH HIGH PULSE REPETITION RATE
CAPABILITY*
Ronald J. Fociaε
RF Engineering/Pulsed Power Laboratories, Inc., PO Box 429
Edgewood, NM 87015 USA
Charles A. Frost
Pulse Power Physics, Inc.
Albuquerque, NM 87122 USA
Abstract
We present the experimental results of an ongoing
research effort focused on the development and
refinement of a compact, low jitter, fast rise time,
command triggered, high peak power, high pulse
repetition rate (PRR), gas-switched pulse generator
system. The main component of the system is a gas-
switched Marx-like pulse generator module designed for
applications including UWB radar, microwave sources,
and triggering large scale multi-module pulsed power
systems of all types. The pulse generator system,
comprised of a single or multiple Marx modules, is
command triggered by a single or multiple TTL level
pulses generated by a timing and control system
implemented using LabVIEW software and a PXI-based
hardware system. The TTL trigger pulses fire all solid-
state high voltage trigger pulsers that close the first stage
switches in the Marx modules using a novel method to
reduce jitter. The control system also accepts user input
to set the desired output conditions, adjusts the charge
voltage of a high voltage capacitor charging power
supply, inhibits capacitor charging during firing of the
pulse generators, and can control the system in a closed-
loop fashion to maintain relative timing and output
characteristics during timing drifts and changing
environmental conditions. The individual Marx stages
are compact and stackable and utilize field enhanced
spark gap switches. The stage capacitors are charged in
parallel through mutually coupled inductors in series with
resistors. This charging scheme allows for high PRR
operation limited only by the stage switch recovery time
and the power of the available capacitor charging power
supply. The stage switches are optically coupled to aid in
Marx output voltage formation and to minimize system
jitter. The Marx generator is housed in a lightweight
aluminum pressure vessel and is operated in a low
pressure dry air environment. The design exhibits a low
inductance which varies depending on the number of
stages used. Using a five stage prototype, we have
generated output voltages of ~100 kV with a rise time of
<4 ns. The output pulse width is variable and is
dependent on the value of the Marx stage capacitors used
and the load resistance. The pulse generator system has
been operated in a burst mode at a PRR in excess of 1
kHz with good output voltage regulation. The total jitter
of the Marx generator system, i.e. from the application of
the trigger pulse to arrival of the output pulse, was
measured and found to be <1 ns.
I. INTRODUCTION
This paper summarizes the current status of a Phase II
Small Business Innovation Research (SBIR) effort
funded by the Air Force Research Laboratory, Directed
Energy Directorate. The solicitation for the Phase I topic
(SBIR 04.1, topic A04-009) called in general for high
pulse repetition rate (PRR) pulsed power generators. The
choice of technology for a solution to the problem and the
operating design goals were left open to suggestion.
When formulating our Phase I plan of action, we studied
the literature available on the subject and devised a
solution that combined the best ideas from multiple
independent experiments and our own new ideas in order
to provide a unique solution to the problem. Our efforts
were successful and led to award of a Phase II project.
The Phase II results to date are summarized in the
following paragraphs.
We feel that the work presented here is unique. In
Phase I we designed a gas-switched Marx generator that
can be operated at a PRR in excess of 1 kHz in burst
mode. In Phase II the system design was revised to allow
for a configurable output polarity from a single charging
voltage polarity, a lower overall inductance, and low total
system jitter. A dual-triggered trigatron scheme was
devised that allows for sub-nanosecond trigger to output
jitter and synchronization of multiple pulse generators.
II. SYSTEM OVERVIEW
The Phase I research effort focused on using a gas-
switched Marx generator design to provide the desired
output voltage levels. In Phase II, the design was refined
to allow for better overall performance. The key features
of the Phase II design are:
• A compact, lightweight, and portable design that
is housed in a tubular containment pressure
vessel,
• A configurable output polarity using a single
polarity charging source,
• A modular construction that allows setting the
output voltage and pulse width as desired,
• Utilization of corona-stabilized (or field
enhanced) spark gap switches for high PRR
operation [1],
• Optical coupling of spark gap switches to reduce
system jitter,
• A triggering scheme that allows for a sub-
nanosecond command trigger to output jitter,
and
• Parallel resonant charging using a series
inductance and resistance in each stage to
support high PRR operation.
The tubular containment pressure vessel is shown in
Figure 1. The outer diameter at each end flange is 127
mm. Each flange has a removable, O-ring sealed cover
plate installed which is held in place by six Allen head
cap screws. The main tube inner diameter is 102 mm.
The inner length of the containment vessel is 406 mm
which allows for housing a design maximum of ten Marx
stages with some room left over for additional
components, such as a peaking switch, if desired. The
overall length of the vessel is 460 mm. If loaded with all
ten stages, the total weight of the Marx generator is 5.3
kg. The physical attributes of the pressure vessel make it
a highly compact and portable system.
Figure 1. Tubular containment pressure vessel.
All connections to the Marx generator are made
through one flange cover. In the standard configuration,
the protrusions through the flange cover are for
pressurized gas inlet, high voltage for capacitor charging,
high voltage output, trigger inputs, and B-dot probe
outputs. The metal pressure vessel acts as the ground
reference. A Teflon liner is installed on the inside of the
pressure vessel to mitigate high voltage arcing. The
output of the Marx generator folds back through the
center of the stages in order to minimize inductance.
The Marx generator is designed to work using dry air
as the pressurizing gas. Other gasses have been
evaluated. However, dry air provided the best
performance at high PRR operation. Portable operation is
easily accomplished using a scuba tank as the source of
pressurized air.
Each stage of the Marx generator is designed for a 20
kV maximum charge voltage. The design uses a single
positive polarity charging voltage and the stages can be
configured to provide a positive or negative polarity
output voltage. The design accommodates a maximum of
ten stackable stages. Each stage can have various types
and sizes of capacitors installed to achieve the desired
output characteristics. Each stage capacitor is charged in
parallel through a series resistance and inductance with
mutual coupling as elaborated on by Baum and
O’Loughlin.[2,3] For high PRR operation, a high power
charging supply must be used. For low PRR or portable
operation, a high voltage DC-DC converter can be used
to charge the stage capacitors.
Each stage of the Marx generator can be stacked
together and the stack plugs into a base plate that is
mounted to the flange cover. Once the stack is assembled
on the flange cover, it slides into the Teflon lined
pressure vessel. The capacitance in each stage and
number of stages is variable and can be set to achieve the
desired pulse width and output voltage magnitude.
The output connector is a pressure sealed re-entrant
type. Two types of coaxial cable can be accommodated
using different size re-entrant connectors depending on
the maximum output voltage that the Marx generator is
configured for.
Low jitter triggering of the Marx generator requires a
bipolar trigger input. The low jitter triggering scheme is
described in the following section. For low jitter
triggering, the first two stage switches are needed and the
minimum number of stages is three. If low jitter
triggering is not required, only the first stage switch is
triggered and the minimum number of stages can be set at
two. The solid-state bipolar trigger generators can be
powered from 220 VAC or from 12 VDC for low PRR or
portable operation.
The Marx generator has two built-in B-dot probes
installed to monitor current in the Marx ground return
path. The B-dot probes are configured to provide
opposite polarity outputs so that common mode noise can
be rejected if necessary.
III. LOW JITTER TRIGGERING
The Phase I Marx generator design used a trigatron
switch in the first stage and only the first stage switch
was triggered. Although the compact, UV coupled, and
corona stabilized spark gap design resulted in a relatively
low jitter for the Marx generator itself (σRMS ≈ 636 ps),
the jitter from command trigger to output was on the
order of several nanoseconds.
One goal of the Phase II effort was to reduce the
command trigger to output pulse jitter to <1 ns so that
several Marx generators could be synchronized to form a
high power array if desired. To achieve this goal, several
switch geometries were evaluated, including standard
trigatrons with various trigger pin geometries, trigatrons
using high-K ceramic materials surrounding the trigger
pin, mid-plane triggered field distortion gaps, overvolted
two-electrode triggered gaps, and a new type of triggered
spark gap that we call the dual-pulsed trigatron design. In
the end, the dual-pulsed trigatron design allowed for
achieving the desired low trigger jitter goals. This
triggering scheme is discussed in the following
paragraphs.
Figure 2 shows schematically the new dual-pulsed
trigatron switch configuration as it would be employed in
the first two stages of a Marx generator. Key to
implementation of this method is the use of a solid-state,
low jitter, high voltage trigger pulser with a bi-polar
output described in a subsequent section.
Figure 2. Schematic of the new dual-pulsed trigatron
switch configuration.
The circuit in Figure 2 is quite similar to a standard
trigatron circuit. However, the positive going output of
trigger transformer T1 is coupled through trigger
capacitor Ct to the anode of the first stage switch. When
the trigger pulse is first applied, almost the full potential
drop occurs across the trigger gap T-K which rapidly
breaks down. This is because the capacity of the trigger
gap is much smaller than the series combination of
capacitors Ct and Cp. When the trigger gap closes, the
full transformer potential is rapidly applied to the switch
anode and heater capacitor Cp. This circuit achieves a
simultaneous UV illumination of gas molecules in the
gap and over-voltage of the gap potential. The heater
capacitor Cp rapidly heats the discharge channel and
switch closure causes the large potential to appear on the
cathode of the following stage.
Figure 3 shows the trigger pulse and voltage
waveforms across the anode-cathode gap of the dual
pulsed trigatron circuit of Figure 2. Separate waveforms
show charge voltage on the gap of 0%, 50%, and 80% of
the 11.4 kV self-breakdown (SB) voltage. The main gap
spacing is 0.05 inches, and operating pressure is 35 psia
dry air. The only difference between the physical
configuration of this gap and the standard trigatron gap is
that the trigger pin is slightly recessed behind the flat
surface of the cathode, rather than being flush with it.
The data of Figure 3 show that the voltage on the anode
starts to rise immediately after the firing of the trigatron
gap. For the case with zero initial voltage on the anode,
the voltage rises to 50% of the breakdown level in
approximately 50 nanoseconds without firing. With 50%
of the self-breakdown voltage charging the gap,
breakdown occurs near the center of the pulse charging
ramp and at a voltage level that is almost 100% of the
self-breakdown level. With higher charging voltage the
voltage across the anode-cathode gap can actually exceed
the self-breakdown before the switch fires. This mode of
operation leads to more intense ionization of the channel
and reduced resistive phase loss.
-6
-4
-2
0
2
4
6
8
10
12
0 50 100 150 200
Figure 3. Voltage waveforms across anode-cathode gap
of dual-pulsed trigatron for 0%, 50%, and 80% of the
11.4 kV self-breakdown (SB) voltage.
Figure 4 shows the voltage waveforms which would be
applied to the cathode of the second stage of the Marx
generator circuit in actual operation. This waveform is
measured across monitor resistor R2. The risetime here is
substantially faster than for a standard trigatron operating
at the same gap spacing and pressure. The over-shoot on
the front of the waveform indicates that the resistive
phase of switch closure occurs more rapidly, and some
inductive resonance ring-up on the waveform is observed.
If this switch was driving the second stage, breakdown of
the second stage switch would occur in less than 1
nanosecond and with extremely low jitter. This is the
desired situation for the low jitter triggering of a Marx
generator.
t (ns)
kV
Trigger
0% SB
50% SB
80% SB
-12
-7
-2
3
8
0 50 100 150 200
Figure 4. Voltage waveforms at the cathode of the
second stage switch when the dual-pulsed trigatron fires.
Figure 5 shows the trigger delay measured as a function
of charging voltage for three different methods of
triggering a trigatron switch. For all three cases the basic
geometry and dimensions of the switches are identical
with a 0.05 inch anode-cathode gap operating at a
pressure of 35 psia. Identical trigger levels were applied
to the switches for all cases. For all three cases the
switch was operated at a 200 Hz repetition frequency.
Data for the new dual-pulsed trigatron is shown by the
triangle symbols. The circle symbols show data for an
over-volted two-electrode gap without the trigger pin.
Comparing the curves it can be seen that the dual-pulsed
trigatron design is superior in giving a lower trigger delay
and allowing triggering down to 20% of the self-
breakdown voltage. The standard trigatron performs well
above 90% of self-breakdown but the delay drops rapidly
with charge voltage. The over-volted two-electrode gap
without UV illumination from the trigger pin gives a
greater delay and ceases to trigger below 80% of self-
breakdown voltage. The data for the over-volted gap,
however, is indicative of what would occur for the non-
triggered second stage gap in the Marx generator without
the presence of UV illumination from the first stage gap.
This data indicates that it is necessary to use UV coupling
in the Marx to achieve a low jitter.
Figure 6 shows the standard deviation of the trigger
delay corresponding to the data points of Figure 5.
Again, each data point represents the average of 64
individual measurements of the time delay, and the RMS
jitter value corresponds with one standard deviation from
the mean. The data indicate that the new dual-pulsed
trigatron design allows triggering with sub-nanosecond
jitter when operated at 80% of the self-breakdown
voltage. This result meets our requirements for use in the
Phase II Marx generator application. The other gap
designs show a substantially higher jitter and cannot be
triggered at reduced charge voltages.
0
20
40
60
80
100
120
0 20406080100
Percentage of Self-breakdown Voltage
Trigger Delay Nanoseconds
Dual Trigatron Standard Trigatron Overvolted Gap
Figure 5. Switch trigger delay measured for three
different methods of triggering a trigatron switch with
11.4 kV hold-off.
0
2
4
6
8
10
0 20406080100
Percentage of Self-breakdown Voltage
RMS Jitter Nanoseconds
Dual Trigatron Standard Trigatron Overvolted Gap
Figure 6. Standard deviation of the trigger delay data
shown in Figure 5.
The effects of changing the gap spacing and operating
pressure were investigated in order to optimize
performance of the dual-pulsed trigatron design. As
anticipated, operation at higher pressures with shorter gap
lengths gives the lowest switching jitter. A gap spacing
of 40 mils provides a reasonable compromise between
performance and switch lifetime.
In order to explore the dual-pulsed trigatron lifetime,
we operated the switch for a total of 3.6x105 pulses by
firing the triggered switch at a constant pulse repetition
frequency of 100 Hz for a period of one hour. No
noticeable change in the switch risetime or triggering
jitter was observed. After the test run the switch was
disassembled and inspected. Neither the trigger pin nor
the main electrodes showed significant signs of wear. It
seems reasonable that this switch design could operate for
~106 discharges between maintenance intervals. This
should be more than adequate for most applications. The
Marx discharge capacitor Cm employed for these tests had
a value of 150 pF. If a larger value of capacitor is used,
the lifetime will be reduced proportionate to the total
charge passing through the switch.
t (ns)
kV
Trigger
0% SB
50% SB
80% SB
IV. TRIGGER PULSE GENERATOR
Key to achieving sub-nanosecond trigger jitter in the
Phase II Marx generator was development of a bipolar,
solid-state, high voltage, and low jitter trigger pulse
generator. Several designs were evaluated and
prototyped. The final design uses a four-stage
semiconductor Marx generator as the primary pulse
generator with pulse sharpening provided by a ferrite core
magnetic switch.
A simplified schematic of the bipolar trigger pulser is
shown in Figure 7. The solid-state Marx generator uses
inexpensive consumer-grade IGBT (Insulated Gate
Bipolar Transistor) type closing switches. A high
resistivity Ni-Zn iron ferrite material was used in the fast
magnetic switch. In this application we employed a total
of twelve of the IGBT devices in the four-stage Marx
generator where three parallel devices were employed in
each stage. When combined with a single ferrite core
magnetic pulse-sharpening switch, a pulse rise time of
approximately 5 nanoseconds was applied to the primary
of the pulse transformer. This was transformed to the
high voltage level required to match the high impedance
capacitive load provided by the triggered switch
electrodes.
Figure 7. Simplified schematic of the trigger pulser.
The trigger pulser requires +/- 600 VDC input voltage.
For high PRR operation, the input voltage is provided by
220 VAC mains converted to DC using a pair of standard
half-wave voltage doublers operated in opposite polarity
modes. For portable or low PRR operation, a DC-DC
converter can be used to provide the required input
voltage.
The output parameters of the complete spark gap
trigger system were characterized by feeding the
differential output voltage from the positive and negative
pulser output leads directly to a high-voltage nanosecond-
rise time resistive voltage divider probe. The probe
output was attenuated with a Barth high-voltage
attenuator followed by a standard 50 Ω attenuator. The
waveforms were recorded on a Tektronix model
TDS3044B digitizer.
Figure 8 shows the measured trigger pulser output
voltage waveform. For this waveform the pulser was
operated at a sustained repetition rate of 300 Hz and the
AC line input voltage was set at 220 VAC. Measured
output voltage was 28.1 kV and rise time was 7.3 ns.
Time delay through the unit was <20 ns with a statistical
triggering jitter of 130 ps.
-5
0
5
10
15
20
25
30
35
010203040
Figure 8. Trigger pulser output voltage waveform.
V. TIMING AND CONTROL SYSTEM
Triggering the Phase II gas-switched Marx generator
with sub-nanosecond jitter allows for synchronizing
multiple pulse generators to form a high power array.
The closed loop timing and synchronization control
system is described in this section.
A block diagram of the timing and synchronization
control system is shown in Figure 9. Although only two
pulse generator modules are shown in the block diagram,
the control system could be expanded to a larger number
of units. The control system is based on using a National
Instruments (NI) PXI-1042 chassis which houses a PXI-
8196 embedded controller, a PXI-6653 timing and multi-
chassis synchronization module, a PXI-7831R
reconfigurable input/output (RIO) card, an Agilent
(formerly Acqiris) DC140 PXI digitizer card, and has a
resistive touch panel human machine interface (HMI).
A Highland Technology P400 digital delay generator is
used to drive each trigger pulser. One channel of the
P400 is set as the reference and the relative delay of the
other channel is adjusted to synchronize the outputs.
The output of each main Marx pulse generator is
monitored using D-dot probes attached to the output
coaxial cables. Analog to digital conversion (ADC) of
the D-dot probe outputs is accomplished using the DC140
PXI digitizer card. A random interleaved sampling (RIS)
scheme allows for over-sampling of the D-dot probe
output at a rate of >20 Gs/s. This over-sampling rate
allows for time resolution of the analog signals to <50 ps
which is more that adequate for synchronization of
multiple Phase II Marx generator modules.
t (ns)
kV
Figure 9. Timing and synchronization control system.
The outputs of the digitizer card are analyzed in
LabVIEW code to find the peak in the D-dot probe
outputs and determine the appropriate delays for the
trigger pulsers. After the delays are calculated, the
system controller communicates with the P400 digital
delay generator via Ethernet to set the appropriate delays
between channels. The P400 is also used as the rate
generator and sets the pulse repetition rate.
If desired, the control system can monitor atmospheric
pressure and adjust the Marx generator pressure and input
voltage to maintain the desired output voltage magnitude.
This mode of operation could be used on an air-mobile
platform or in those cases where the unit is transported
through various elevations. The electronic pressure
regulator communicates with the control system via
Ethernet and the output of the high voltage capacitor
charging power supply is controlled using an analog
output of the RIO board.
VI. TEM HORN ANTENNA
A TEM horn antenna, shown in Figure 10, was
fabricated to be used with the Phase II Marx generator in
order to provide a complete high power impulse source.
The TEM horn is an unbalanced design having a tapered
horn section cantilevered above a ground plane. Other
basic design constraints were that the antenna should be
compact and transportable, be easily assembled and
mounted, have the largest radiating aperture practicable,
and be designed for operation at high voltages and high
peak power levels.
The ground plane size was set at 4x10 ft and was
fabricated by laminating aluminum sheet to a composite
substrate. The height of the aperture was set at 32 inches
above the ground plane giving a width to height ratio of
1.5. Choosing a desired characteristic impedance of Zc =
200 Ω, design guidance [4] was referenced to determine
the angular width of the triangular section (α) and the
angular separation between the triangular section and
ground plane (β/2) of 25o and 16.5o, respectively. TDR
was used to characterize the horn impedance which
measured close to that predicted by theory.
Figure 10. TEM horn antenna.
VII. REFERENCES
[1] S. J. MacGregor, et al., “Factors Affecting and
Methods of Improving the Pulse Repetition Frequency of
Pulse-Charged and DC-Charged High-Pressure Gas
Switches,” IEEE Trans. Plasma Science, 25, 110-117
(1997).
[2] C. E. Baum and J. M. Lehr, “Charging of Marx
Generators,” CESDN 43, AFRL/DEHP (2000).
[3] J. O’Loughlin, et al., “High Repetition Rate Charging
a Marx Type Generator,” in Conference Record of the
Pulsed Power and Plasma Science Conference (2001).
[4] R. Todd Lee, and Glenn S. Smith, “On the
Characteristic Impedance of the TEM Horn Antenna,”
IEEE Trans. Antennas and Propagation, 52, 315-318.
P400 Digital
Delay Generator
Tri
gg
er PG1
Tri
gg
er PG2
HV
PG1
PG2
Load
PXI-7831R
RIO Board
At
t
At
t
D1 D2
PS
HV
Source
PR
Air
Aout
ENET
PXI Chassis
PXI-8196
Controller
Acqiris DC140
Digitizer
