IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO.

10, OCTOBER 2010 2593

High Pulsed Power Sources for Broadband Radiation

Baptiste Cadilhon, Laurent Pécastaing, Thierry Reess, Antoine Silvestre de Ferron, Pascal Pignolet,
Stephane Vauchamp, Joël Andrieu, and Michèle Lalande, Member, IEEE

Abstract—This paper explains the design and production of in high pulsed power (HPP). In each case, the HPP generator
two autonomous ultrawideband (UWB) radiation sources. These is made up of a primary energy source and a load, separated
sources consist of a high-gain broadband antenna that is driven by
by a power-amplification system that forwards the energy from
one of two subnanosecond pulsed power sources. Each source is
made up of a Marx generator and a pulse-forming device based this source to the load. When the load is an antenna, it is
on the use of a gaseous spark gap. The first source combines possible to generate electromagnetic waves in the form of
a four-stage 200-kV/34-J Marx generator with a coaxial pulse- pulses. The frequency band of the wave that is radiated can
forming line. Its main characteristics are an output voltage of cover a very broad spectrum. In this case, the technique is
100 kV, a 250-ps rise time, a subnanosecond pulse duration, and
a repetition rate of about 40 Hz. The second pulsed source is of undoubted interest for a wide variety of civil and military
a ten-stage subnanosecond Marx generator that delivers pulses applications. These applications can include, for example, ultra-
in the 250-kV/1.5-J range, with a 300-ps rise time and a sub- wideband (UWB) pulse radars to detect buried mines [1] or to
nanosecond pulse duration at a pulse repetition rate of 350 Hz. rescue buried people, the production of nuclear electromagnetic
Probes were produced based on capacitive line dividers to measure
both the temporal characteristics and the high-voltage (HV) am- pulse (NEMP) simulators for electromagnetic compatibility and
plitude of the pulses delivered by the pulsed power sources. The vulnerability tests on electronic and IT equipment [2], and
antenna, combined with these two pulsed sources, is a traveling- UWB communications systems and electromagnetic jamming
wave antenna called the Valentine antenna. Some mechanical [3], the principle of which consists of focusing electromagnetic
modifications were made to the antenna to improve its dielec-
power waves on an identified target to compromise the target’s
tric strength. First, a 3-D model of the antenna was created on
time-domain electromagnetic software to study the influence of mission by disrupting or destroying the electronic components
these modifications on its main radiating characteristics. Its high that constitute it [4], [5]. Modern digital electronics are in-
gain and its capability to radiate short pulses without dispersion creasingly sensitive to electromagnetic radiation in the lower
allow us to achieve a high measured figure of merit (the maximum gigahertz range. The application of high-power microwave
value of far-field peak-to-peak electric field strength multiplied
by the distance). A new method called the Instantaneous Elec- (HPM) weapons to destroy such electronics generates a new
tromagnetic Field Measurement by Signature of a Neutral Object capability for defense, antiterror, and rapid-reaction forces.
(MICHELSON) method is used to measure the very intense elec- Due to the nonlethal and low-collateral-damage properties of
tromagnetic fields that are radiated. The incident field is diffracted HPM weapons, they will play an important role in future
by a special small-dimension target. The diffracted field is mea-
sured by a conventional low-power UWB antenna. The target that operations.
is used has small dimensions, and no cables are used in the field HPM technologies generate a significant threat for target
region; thus, the electromagnetic interference that is generated structures that are equipped with modern digital electronics.
and undergone by the measurement device is considerably limited. The aim of an HPM weapon system is to interact with the
The figure of merit that is measured is 436 kV.
increasing combination of software and hardware functions in
Index Terms—Compact pulsed power source, gaseous spark electronics. To affect any target that uses such radiation, the fre-
gap, high-bandwidth probe, high-voltage (HV) generator, Marx quency has to be matched to the target’s frequency absorption
structure.
spectrum. However, pulsed UWB radiation with pulse durations
in the subnanosecond range does not need to be matched to a
I. I NTRODUCTION /BACKGROUND
single target this way.

F OR the last few years, the generation of high-power elec-

tromagnetic waves has been one of the major applications
It is in this context that this paper has been carried out,
which deals with the study, production, and characterization
of an “all-electric” device that can radiate intense transient
Manuscript received September 1, 2009; revised November 27, 2009; ac- electric fields in a repetitive mode of operation. Our device is
cepted January 25, 2010. Date of publication March 22, 2010; date of current
version October 8, 2010.
composed of a dc/dc (300 V/60 kV) autonomous power supply,
B. Cadilhon is with the Atomic Energy Commission (CEA), 33114 Le Barp, a subnanosecond Marx generator (a few hundred kilovolts) and
France. a traveling-wave high-gain antenna that has been improved to
L. Pécastaing, T. Reess, A. Silvestre de Ferron, and P. Pignolet are with the
Laboratoire de Génie Electrique, Université de Pau et des Pays de l’Adour, withstand high-voltage (HV) pulses in a wide frequency band.
64053 Pau, France (e-mail: [email protected]). An extremely short pulse is generated by this source and is
S. Vauchamp, J. Andrieu, and M. Lalande are with the Department of transmitted through the transmission antenna. Our device must
Waves and Associated Systems (OSA), XLIM Research Institute, University
of Limoges (National Center for Scientific Research (CNRS), 19100 Brive, transmit waveforms in a frequency bandwidth of 300 MHz–
France. 2.5 GHz.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. The basic structure of our proposed repetitive HPM system
Digital Object Identifier 10.1109/TPS.2010.2042732 is shown in Fig. 1.

0093-3813/$26.00 © 2010 IEEE

2594 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO. 10, OCTOBER 2010

Fig. 1. Basic structure of our repetitive HPM system.

II. D ESIGN OF T WO S UBNANOSECOND HPP S OURCES

BASED ON M ARX G ENERATORS
A. Pulsed Power Source of 100 kV-40 Hz: Leonardo
This source is made up of a 200-kV nanosecond-rise-time
Marx generator (described in [6] and [7]) that supplies energy
to two gaseous sharpening spark gaps in a pulse-forming line
(PFL) with an impedance of 50 Ω. The stability of the UWB
source’s parameters is closely linked to the stability of the HV
pulsed source’s pulse characteristics. It seems that the primary
driver has to deliver a HV pulse with a fast rise time and good
reproducibility to lead to the stable erection of the switches in
the PFL [8].
The mechanical and electrical characteristics of the Marx
generator on its own are described in [6]. The generator is
Fig. 2. Three-dimensional schematic of the pulse-forming set.
then associated with a pulse-forming device. The energy that is
released by the Marx generator thus supplies a forming line that
allows a voltage pulse to be created, the temporal characteristics to be the design in which the crowbar spark gap has a smaller
of which entirely depend on the operation of the two forming interelectrode distance (1 mm) to obtain the shortest possible
spark gaps. fall times. The peaking spark gap’s interelectrode distance must
The line is created in an aluminum cylinder with a length of be large enough to reach voltages of 100 kV but small enough
240 mm and an outer diameter of 130 mm. It is made up of a to switch as quickly as possible. The peaking spark gap’s
section of a coaxial line with an impedance (Z0 = 50 Ω) and distance must also be greater than that of the crowbar spark
of two pressurized spark gaps (see Fig. 2): 1) the peaking spark gap to favor short pulse durations. Furthermore, the center-to-
gap and 2) the crowbar spark gap. The first spark gap sharpens center distance between the two spark gaps must be as small as
the rising edge of the incident pulse. The second spark gap (the possible to reduce the duration of the pulses. This distance is
“crowbar” spark gap) short-circuits the line by forming a falling set at 8 mm. For these last two reasons, the peaking spark gap
edge. The section of the line allows the signal to be formed was set at 1.5 mm.
according to the principles of a line discharge, i.e., A series of experimental tests was carried out to determine
the performance of the complete device in this configuration
• a rectangular pulse whose duration is equal to twice the
according to the air pressure (Marx) and hydrogen pressure
electric length of the line (only if the crowbar spark gap
(PFL) involved. The rise time of a gas switch is found to be
switches after this propagation time; otherwise, it is the
strongly dependent on the electric field value [9]. Moreover,
setting of this spark gap that establishes the duration);
for a given electric field, the rise time decreases with the gas
• an amplitude that is equal to half of the Marx generator
density. Hydrogen is therefore chosen in the PFL to minimize
when the load is equal to Z0 .
this rise time.
The second section of the line downstream these two spark All the tests were performed on a resistive load of 50 Ω.
gaps allows the transverse electromagnetic (TEM) mode to First, we present the evolution of the device’s output voltage
be reestablished to forward the signal that is generated to the according to the hydrogen pressure in the line’s spark gaps
transmitting antenna. The insulating material that constitutes (see Fig. 3). This paper is carried out at different pressures in
the line is identical to the material for the construction of the the Marx generator and, therefore, at different charging voltage
generator output to maintain the diameters and impedances levels. Results are presented for only five significant pressures.
(50 Ω) of these two structures. The line’s imperviousness and When there is a constant pressure in the line, the value of the
mechanical strength allow the spark gaps to be pressurized up output voltage does not vary very much but slightly increases
to approximately 100 bars. as the amplitude of the voltage that is delivered by the Marx
An initial study allowed the most appropriate interelectrode generator increases. This result is due to an overvoltage on
and inter (spark gap distances to be determined. In light of the the forming peaking spark gap when the pressure in the Marx
results of this paper, the most satisfactory configuration proves generator increases. For a pressure in the line of 50 bars and a
CADILHON et al.: HIGH PULSED POWER SOURCES FOR BROADBAND RADIATION 2595

Fig. 3. Output voltage in relation to the pressure in the PFL, parameterized by Fig. 5. Fall time in relation to the pressure in the PFL, parameterized by the
the voltage in the Marx generator. pressure in the Marx generator.

Fig. 4. Rise time in relation to the pressure in the PFL, parameterized by the
pressure in the Marx generator.
Fig. 6. Setting the amplitude of the output voltage by adjusting the pressure
in the line.
pressure in the generator of at least 18 bars, the amplitude of
the voltage reaches 100 kV. than 200 ps (an average of five shots) are obtained in this
Fig. 4 shows the results of the experimental study on the rise configuration with good reproducibility at the highest hydrogen
time of the output pulse in the same experimental configura- pressures.
tions. When there is constant pressure in the line, the rise time Fig. 6 superimposes three pulses that are achieved on a
decreases as the pressure in the Marx generator increases. The 50-Ω broadband load. These signals are generated with three
variation in this pressure modifies the amplitude of the output hydrogen pressures (i.e., 20, 35, and 50 bars) in the line at a
voltage of the Marx generator and, therefore, the value of the constant pressure of 17 bars in the Marx generator. θ represents
electric field that is applied to the peaking spark gap [8]. For the midheight pulse width. The amplitude of the voltage can be
the highest voltage that is applied to the line (the pressure in set at 55, 75, and 98 kV, respectively, with an uncertainty of
the Marx generator is up to 18 bars), the rise time seems to be 5%–10%.
only almost independent of the pressure in the line, whereas
it is fully independent for low-amplitude pulses (the pressure
B. Pulsed Power Source of 250 kV-350 Hz: George
in the Marx generator is down to 10 bars). When the Marx
generator is pressurized under 18 bars, the average rise time of The second pulsed power source called “George” was de-
the output pulse is relatively constant at approximately 340 ps, veloped to reach higher output voltages and a higher repetition
whereas when the Marx generator is pressurized under 10 bars, rate at a more suitable volume (power ratio. The original idea
the output rise time decreases from 550 ps to 450 ps as the of this source was to directly integrate the pulse-forming device
pressure in the PFL increases from 15 bars to 35 bars. (a peaking stage and a crowbar switch) within the last stage of
The fall of the pulse is governed by the crowbar spark gap. a Marx structure.
The graph in Fig. 5 clearly shows that the pressure in the This generator is made up of a stack of ten vertical stages.
line plays an important role in optimizing the switching time Each stage consists of a capacitor, a set of stainless steel
of the spark gap. Similar to the previous study of the rise electrodes, and a polyvinyl chloride (PVC) disc. The energy
time, the application of the higher amplitude pulse to the spark is stored in HV 500-pF/100-kV dc capacitors through 33-kΩ
gap results in the shortest switching times. Fall times of less axial resistors. These ten capacitors are equally divided into two
2596 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO. 10, OCTOBER 2010

Fig. 8. Diagram of the insertion of the PFL in the last stage of the Marx
generator.

Fig. 7. Photograph of the generator, equipped with its load and referencing
resistors.

columns of five (see Fig. 7), which allows the loading time to
approximately be halved in relation to a conventional structure
(when the DC power supply is sufficient). Thus, the capacitors’
charging time (to reach 99% of the maximum load voltage [1])
is 2.3 ms. In this configuration, the repetition rate limitation is
435 Hz, with no loss of output voltage (compared to the single-
shot mode of operation).
Each spark gap is formed by two stainless steel electrodes.
These ten horizontal gaps are adjusted by increasing the gap
distance from the first gap to the last gap so that the generator
switch is regular. Thus, the first stage has a 0.8-mm gap, and
the last stage (largest) is set at 1.25 mm. Their 5-mm curvature Fig. 9. Output voltage in relation to pressure.
radius means that we obtain a relatively homogeneous field
in the insulating gap. The complete generator with the vessel
represents a cylindrical volume of 20 L (diameter = 20 cm and to the metal enclosure. This second electrode is a 5-mm-thick
length = 66 cm). stainless steel disc. After the crowbar, a coaxial line structure is
This generator’s main original feature lies in the design of its directly integrated to avoid dispersion of the pulses’ temporal
last stage, as represented in Fig. 8. characteristics.
The last stage was designed different from the other stages The first set of experiments was carried out without the
to insert a peaking capacitor CP , which is necessary to obtain crowbar switch to determine the nature of the gas and the
very fast rise times, and to match the output to a 50-Ω coaxial related pressure that would make it possible to reach output
line structure. This peaking capacitor is made up of a PVC voltages with short rise times. The results of these different
component that is maintained between the electrode and the tests are set out in Fig. 9. They were carried out in air and in
vessel enclosure. This component’s geometrical dimensions an air-SF6 mixture (75%–25%).
are defined in such a way that it withstands the considerable For a set number of stages, the Marx generator output voltage
dielectric constraints on this last stage and to establish a low- increases according to the product pressure multiplied by the
value peaking capacitor (around 40 pF). gap distance for a given gas. In this case, voltage amplitude can
Moreover, with a crowbar spark gap on the output electrode be adjusted to between 50 kV and 150 kV on a 50-Ω load in
of the generator, it is possible to generate a subnanosecond air for pressures that range between 2 and 20 bars. A small per-
pulse width based on the Marx-generated pulse. This crowbar centage of SF6 (25% in this case) makes it possible to increase
switch allows the pulse’s exponential decrease to be cut off. It the amplitude of pulses by approximately 1.7 times in relation
is made up of two electrodes: the first electrode is the generator to those obtained with pure air. A 250-kV voltage amplitude
output (an HV electrode due to the switch in the last stage), is obtained for a dc charging voltage of approximately 50 kV
and the second electrode (the ground electrode) is connected with 20 bars of pressure in this air-SF6 mixture. For a maximum
CADILHON et al.: HIGH PULSED POWER SOURCES FOR BROADBAND RADIATION 2597

III. I MPLEMENTATION OF T RANSIENT HV S ENSORS

To record fast transient phenomena, high-bandwidth diag-
nostics are required. Conventional measurement techniques do
not allow an estimation of both the rise time and the voltage
amplitude of the output waveform. Thus, a home-made probe
was manufactured. This probe is based on the principle of a
capacitive line divider that is fitted on the coaxial output of
the generator. It must have, on the one hand, a high bandwidth
to measure fast pulses and, on the other hand, an attenuation
rate that is constant within the bandwidth and is sufficient
to not damage the acquisition devices. The probe’s transient
characterization, the experimental setup for which is shown
in Fig. 12, is carried out using a Kentech APG1 generator.
Fig. 10. Rise time of the output pulses in relation to pressure. This generator produces 220-V amplitude pulses. The spectral
range extends up to 5.1 GHz at −20 dB. The signal is recorded
using two PSPL attenuators of −20 dB/18 GHz and a Tektronix
DPO70804 scope (8 GHz/25 Gs.s−1 ). All coaxial cables and
attenuators were characterized up to 40 GHz to withdraw their
responses from the following results. A probe at the Kentech
generator output is used to give a reference pulse measurement.
The probe for this measurement is a commercially available
probe with a cutoff frequency of 5 GHz and a voltage atten-
uation of −56 dB.
In Fig. 13, the signal that is measured by our probe is com-
pared with the reference measurement (signals are standardized
for easier understanding). The perfect correspondence between
these two curves allows us to conclude that our probe can
measure pulses with rise and fall times of 100 ps without
distortion. A Fourier transform of the measured signal also
Fig. 11. Marx generator output pulse shape (P = 17 bar, 75% air (25% SF6 , allows us to find the probe frequency response and to compare it
τ = 360 ps, VOU T = 225 kV).
with the frequency analysis result by using a network analyzer.
In this numerical stage, all the frequency responses of the
cables, attenuators, and 50-Ω matching cones were taken into
dc charging voltage of 100 kV per stage, a 500-kV amplitude
account to leave only the probe’s response (see Fig. 14).
pulse can be reached if the generator can withstand a pressure of
The frequency response that is determined by this method
44 bars.
(red curve) leads to the conclusion that the home-made probe
A similar study was carried out on the evolution of rise time
has a high cutoff frequency of 2.15 GHz at −3 dB. The attenu-
in relation to pressure for these two gases. The results are shown
ation varies between −61.4 dB and −64.4 dB. The results that
in Fig. 10. For each gas configuration, the shortest rise times
are obtained by both frequency and transient characterization
are observed at the highest pressures. At a given pressure in
show a very satisfactory correlation.
the generator, air induces rise times that are shorter than in
the air-SF6 mixture [10]. The measured rise times lie between
IV. I MPROVEMENT OF A UWB T RANSMITTING A NTENNA
250 ps and 350 ps in the air and between 350 ps and 500 ps in
the air-SF6 mixture. The two pulsed power sources that are previously presented
Fig. 11 shows a typical pulse of the George generator. It will be combined to a UWB transmitting antenna to radiate high
is connected to a 50-Ω coaxial line and to the Valentine HV electromagnetic fields. Due to its high gain and its capability
antenna [11]. First, the crowbar switch gap distance setting to radiate short pulses without dispersion in a frequency band
was tested to determine the solution that leads to a break- of 300 MHz–2.5 GHz, the Valentine antenna, which was first
down in the rising edge of the Marx-generated pulse. Thus, a used as a transmission antenna for a former radar application
1.5-mm distance seems to be the best configuration to obtain a [11], was chosen to be the radiating element of our high-
subnanosecond pulse with minimized loss. power UWB sources. The antenna was mechanically matched
The pulse amplitude is 226 kV, the rise time is 362 ps, and using the CST Microwave Studio time-domain electromagnetic
the midheight width is 410 ps. Tests were also carried out in a software [12] to improve its dielectric strength. The antenna
repetitive mode of operation. The dc HV charger for these tests was originally composed of two metal strips that were curved to
allowed us to test the generator up to a maximum repetition rate form a specific profile and a 50-Ω coaxial-to-stripline transition
of 350 Hz for an output voltage of 170 kV on 50 Ω. The tests at the input. This structure is supported by two wooden side
were carried out by bursts of pulses that lasted a few seconds, pieces. Thus, this whole unit is 1200 mm high, 1000 mm long,
separated by a few minutes of downtime. and 400 mm wide.
2598 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO. 10, OCTOBER 2010

Fig. 12. Experimental setup of the transient analysis.

According to Foster [13], the length of such a transition must

be one quarter of the lowest frequency wavelength. This length
is fixed at 300 mm. In addition, Farr [14] gives the following
simple formula for the rise time, which can be preserved in such
a transition as a function of several transition dimensions:
√  2
π 2 εr D2
tr = (1)
2 lc
where εr is the relative dielectric permittivity of the insulating
material, D is the inner diameter of the output conductor on
the coaxial transition part, l is the transition length, and c is the
light velocity. Based on (1), the selected dimensions and the
use of Univolt oil with a relative permittivity of 2.3 gives a
Fig. 13. Transient response of the capacitive divider. 105-ps rise time, which is in accordance with the power
sources’ time-dependent parameters.
A PVC part was then inserted between the two strips and
was extended from the new transition to the end of the antenna
to ensure good dielectric insulation for the HV pulses. The CST
Microwave Studio software was used as a design-support tool.
The design of this modification, its shape, and its influence on
the antenna’s electromagnetic characteristics were studied. The
complete design of the HV antenna is shown in Fig. 15.
The HV antenna was studied in the 0–3 GHz frequency band.
A Gaussian pulse with a frequency spectrum that is larger than
the spectrum of the antenna, is used as the input signal for the
time-domain simulation, which seems to be the best solution
to study the calculation of the S11 parameter. The input port is
defined on the coaxial part of the transition.
Fig. 16 shows the measured S11 reflection parameter of the
Fig. 14. Comparison of the two characterization methods. modified antenna. This result is compared to the calculated S11
parameter in a frequency range of 300 kHz–3 GHz. Results are
We first focused on the development of the new coaxial- quite similar. Except at 600 MHz, the S11 parameter is less than
to-stripline transition. This transition was initially meant to −10 dB between 300 MHz and 3 GHz.
transform (in air) a 50-Ω coaxial transmission line geometry The new HV antenna’s measured gain up to a frequency of
into a 50-Ω stripline geometry at the antenna input. Simulation 3 GHz is shown in Fig. 17. This response is compared to the
allowed us to modify this transition for our application with calculated response, as well as to the measured gain of the
a very high voltage and very fast transients. When designing original low-voltage antenna. Both results are realized gains,
the transition, it is essential to take the bandwidth into account. which take account of the matching of the antenna.
CADILHON et al.: HIGH PULSED POWER SOURCES FOR BROADBAND RADIATION 2599

and stabilize the gain at around 10 dB within this frequency

range. Totally avoiding using a wooden structure would make
it possible to obtain a more satisfactory result.
The measured and calculated radiation patterns in the H
and E planes at 500 MHz, 1 GHz, and 2 GHz are shown in
Fig. 18(a)–(f).
Good harmony is obtained between the theoretical results
and the experimental results. In each plane, the radiation pat-
terns become narrower as the frequency increases. The back-
ward radiation is greater at lower frequencies. This result is due
to the low-frequency parts of the current pulse that are reflected
at the end of the antenna and propagated toward the end of
the metal strips. On the contrary, high frequencies are radiated
before reaching the end of the antenna. The backward radiation
is lower than −20 dB for frequencies greater than 1.5 GHz.
Moreover, the main lobe is narrower in the H plane than in the
E plane. The directivity of the antenna in this plane depends on
the width of the metal strips. The wider the strips, the greater
the antenna’s directivity in the H plane.
Fig. 15. Three-dimensional modeling of the Valentine HV antenna and the
Tulipe transition.
V. E XPERIMENTAL T ESTS OF THE T WO P ULSED
R ADIATION S OURCES
In this section, we present different experiments that aim at
determining the amplitude and temporal characteristics of the
electric field that is radiated along the axis of the Valentine
antenna when it is associated with each of the aforementioned
two pulsed HV sources. Two methods of measuring the field
are studied: 1) a direct measurement by using a derivative
field sensor and 2) an indirect measurement of the field that
is diffracted by a sphere, i.e., the Instantaneous Electromag-
netic Field Measurement by Signature of a Neutral Object
(MICHELSON) method [16], [17].

A. First Set: Leonardo Pulsed Power Source and

Fig. 16. S11 parameter of the new Valentine HV antenna.
Valentine Antenna
The first whole emitting system (see Fig. 19) is composed
of the autonomous dc charger, the Leonardo Marx generator, a
70-cm-long 50-Ω coaxial line on which the capacitive probe is
inserted, and the Valentine HV antenna.
The pressure in the Marx generator is maintained at 17 bars
in an air-SF6 mixture throughout the tests. In this case, the
amplitude of its output voltage, which attacks the forming line,
is close to 200 kV. This line is set according to the most effective
spark gap configuration. Varying the pressure in these spark
gaps to between 20 and 55 bars allows the amplitude of the
voltage at the input of the antenna to be increased to between
55 kV and 100 kV. The rise times are between 300 and 400 ps,
and the width of the pulses remains close to 500 ps.
Fig. 17. Realized gain of the Valentine HV antenna in the axial direction. In Fig. 20, we compare the rate of the pulse that was obtained
in the configuration where the pressure in the forming line is
It has been shown in [15] that the wooden structure of the 55 bars, with an estimation that was calculated based on the
antenna is responsible for the gain drop to below 10 dB between transfer function of the antenna and the Fourier transform of the
800 MHz and 1.6 GHz. A parametric study was conducted excitation pulse. The fields are measured at a distance of 10 m
on the variation of the wood’s relative permittivity value by to benefit from a sufficient clear time and to stay out of the
using CST, and we have concluded that a structure with a near-field region. Experimentally, the measured pulse has a rise
lower relative permittivity (e.g., εr = 2) can reduce this drop time of 280 ps, a width of 2 ns, a positive amplitude (reduced to
2600 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO. 10, OCTOBER 2010

Fig. 18. Radiation patterns in the H and E planes at 500 MHz: (a) and (b) 1 GHz, (c) and (d) 2 GHz, and (e) and (f). (a) H plane: 500 MHz. (b) E plane:
500 MHz. (c) H plane: 1 GHz. (d) E plane: 1 GHz. (e) H plane: 2 GHz. (f) E plane: 2 GHz.

Fig. 19. First repetitive high-power UWB source.

Fig. 20. Comparison of measurements by using the MICHELSON method
1 m) of 141 kV/m, and a negative amplitude, in absolute terms, and calculated radiated fields (first source).
of 82 kV/m. In the end, the measured figure of merit that is
reached for this transmission unit is 223 kV. peak amplitude of the estimated figure of merit is 270 kV. This
The general rate of the estimation is relatively close to the difference in level has not clearly been identified. Additional
pulse measured. However, we can see a significant difference measurements on the characterization of the sensor must be
in amplitude in terms of the negative peak (17%). The peak-to- carried out.
CADILHON et al.: HIGH PULSED POWER SOURCES FOR BROADBAND RADIATION 2601

Fig. 21. Second repetitive high-power UWB source.

B. Second Set: George Pulsed Power Source and Fig. 22. MICHELSON experimental setup.
Valentine Antenna
The second whole emitting system (see Fig. 21) is composed
of the autonomous dc charger, the George Marx generator, a
70-cm-long 50-Ω coaxial line on which the capacitive probe is
inserted, and the Valentine HV antenna.
The XLIM Laboratory, Limoges University, has developed a
new method for measuring intense electromagnetic fields called
the MICHELSON method [17].
The principle is based on the use of a target that scatters
the electromagnetic field whose characteristics we are trying
to find out. We then measure the scattered field in a given
direction with a conventional antenna that is adapted to the field
bandwidth. The purpose is to come back to the incident field
from the output antenna voltage. With this method, high-power
sources can be studied with simple equipment, because there Fig. 23. Comparison of measurements by using the MICHELSON method
is no risk of breakdown. The conventional receiving antenna and calculated radiated fields (second source).
recovers only a fraction of the incident field, and the target The figure-of-merit value of the radiated field, which was
is insensitive to this field. The target is small in size, and no measured using this method, is 436 kV, with a positive am-
cable is used; therefore, disturbances are also limited with this plitude of 233 kV and a negative amplitude of −203 kV. The
measurement system. Another advantage is the possibility of impulse width is 1.82 ns. The theoretical estimation gives an
also measuring harmonic signals as transient signals by simply estimated value of 460 kV. The results that are obtained by this
choosing a receiving system that is suited to the situation. method of measurement are thus very close to the results that
The measurement principle is shown in Fig. 22. The target is are estimated by calculation.
illuminated by the incident field ei (t) that will be measured.
This field is diffracted by this target in different directions
VI. C ONCLUSION
and according to different polarizations. A receiving antenna
is placed at a location so that it is protected from the direct This paper has brought together the results of work that is
radiation of the high field source. This antenna receives the carried out on the production of two UWB radiation sources
field ed (t), which is diffracted by the target. The voltage based on the sole use of a Marx generator with gaseous
vR (t), which is an accurate reflection of the receiving field, is spark gaps.
measured by a sampler that is connected to the antenna output. The first pulse source associates a Marx generator
Based on the transfer functions, which are both from the target (Leonardo) of 200 kV, 34 J, and 2 ns with a coaxial line struc-
and the receiving antenna, and after few calculations, we can ture that is equipped with two forming spark gaps. Setting the
link the field ei (t) and the measured voltage vR (t). hydrogen pressure between 20 and 55 bars allows us to adjust
Fig. 23 represents the evolution of the radiated electric field the amplitude of the unit’s output voltage to between 55 kV
measured by the antenna, which has been extrapolated to 1 m by and 100 kV. The pulses have rise and fall times of 300 ps and
using this method. The metal sphere target (with a diameter of 200 ps, respectively, and a subnanosecond midheight duration.
24 cm) is located at a distance of 5 m from the emitting antenna. The peak power that is transmitted on a 50-Ω load is 200 MW.
The receiving antenna is a UWB antenna that is developed by The maximum repetition rate at 100 kV is 40 Hz. The second
XLIM, whose bandwidth lies between 300 MHz and 3 GHz. pulsed source is solely a Marx generator (George) with lower
The result of this measurement is compared with a theoretical energy, which allows higher voltage amplitudes to be obtained
estimation. at higher repetition rates. This Marx generator integrates the
2602 IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 38, NO. 10, OCTOBER 2010

pulse-forming device directly into its last stage. In an air-SF6 [9] P. Delmote, C. Dubois, J. Andrieu, B. Beillard, M. Lalande, V. Bertrand,
mixture, pulses have been measured with a subnanosecond B. Jecko, L. Pecastaing, A. Gibert, J. Paillol, P. Domens, R. Guillerey,
F. Monnier, and M. Legoff, “The UWB SAR system PULSAR: New gen-
midheight duration and an amplitude that can exceed 250 kV, erator and antenna developments,” in Proc. SPIE Int. Conf., AeroSense,
with rising edges of close to 300 ps. The maximum voltage Orlando, FL, Apr. 21–25, 2003, pp. 223–234.
pulse that is reached corresponds to a power of 1.25 GW, with [10] L. Pecastaing, T. Reess, J. Paillol, A. Gibert, and P. Domens, “Very fast
rise time short-pulse high-voltage generator,” IEEE Trans. Plasma Sci.,
a resistive load of 50 Ω. Due to the power limitation of the vol. 34, pt. 1, no. 5, pp. 1822–1831, Oct. 2006.
charger, the repetitive mode of operation tests at 350 Hz can, for [11] J. C. Diot, P. Delmote, J. Andrieu, M. Lalande, V. Bertrand, B. Jecko,
the moment, only be carried out at an output voltage of 170 kV. S. Colson, R. Guillerey, and M. Brishoual, “A novel antenna for transient
applications in the frequency band of 300 MHz–3 GHz: The “Valentine”
Further work must be carried out on this generator, but its antenna,” IEEE Trans. Antennas Propag., vol. 55, no. 3, pp. 987–990,
performance is nonetheless encouraging. A mechanical im- Mar. 2007.
provement that leads to a better pressure strength (greater than [12] MWS User’s Guide, Microwave Studio documentation from CST, CST
France Eurl, 10 Avenue Réaumur, 92140 Clamart, France, 2009.
40 bars) would make it possible to generate pulses of 500 kV [13] P. R. Foster and S. M. Tun, “A wideband balun from coaxial line to TEM
on the load at full charging voltage (100 kV) without altering line,” in Proc. IEEE Conf. Antennas Propag., 1995, pp. 286–290.
its temporal characteristics. We could also consider reducing [14] E. G. Farr, G. D. Sower, and C. J. Buchenauer, “Design considerations for
ultrawideband high-voltage baluns,” FARR Research, Albuquerque, NM,
the number of stages in the generator for an output voltage that Sensor and Simulation Notes 371, 1994.
is equivalent to the voltage that we have obtained. In addition, [15] B. Cadilhon, “Study and Realization of an Autonomous Source of
with a suitable charger, a repetition rate of 435 Hz could be High-Power Electromagnetic Waves,” Ph.D. dissertation, Pau Univ., Pau,
France.
reached with no loss of efficiency. [16] S. Vauchamp, J. C. Diot, M. Lalande, J. Andrieu, B. Beillard, B. Jecko,
The radiation system is a UWB antenna called Valentine, to J. L. Lasserre, A. Paupert, G. Teyssidou, and R. Pouzalgues, “Utilization
which several mechanical modifications have been made using of a target diffraction to measure a high electromagnetic field,” in Proc.
Int. Conf. Electromagn. Adv. Appl., Sep. 2007, pp. 784–787.
the 3-D electromagnetic simulation. This antenna can withstand [17] S. Vauchamp, M. Lalande, J. Andrieu, B. Jecko, J. L. Lasserre,
voltages of more than 250 kV at a repetition rate of a few hertz. L. Pecastaing, and B. Cadilhon, “Utilization of target scattering to mea-
With regard to the antennas that are developed in other projects sure high-level electromagnetic fields: The MICHELSON method,” IEEE
Trans. Instrum. Meas., vol. 59, no. 9, pp. 2405–2413, Sep. 2010.
of the same type, we could consider carrying out integration
work on this antenna, because we always observe that antennas
play a major role in the creation of compact systems of this type.
A new method for measuring intense electromagnetic fields Baptiste Cadilhon was born in Mont de Marsan,
France, in 1982. He received the M.Sc. degree in
(i.e., MICHELSON), which was developed in the XLIM Labo- electrical and electronic engineering from the Uni-
ratory, Limoges University, has been approved. It allows strong versity of Toulouse, Toulouse, France, in 2004 and
fields (> 100 kV/m) to be measured in the UWB domain, as the Ph.D. degree in electrical engineering from the
Université de Pau et des Pays de l’Adour, Pau,
in wideband and narrowband. For the second UWB radiation France, in 2008.
source (the George generator and the Valentine antenna), we From 2005 to 2008, he was with Thales for the
measured a figure of merit of 436 kV. By a direct measurement study and realization of an autonomous source of
high-power electromagnetic waves. He is currently
with a derivative field sensor, it has been shown that the first a Researcher with the Atomic Energy Commission
radiation source (the Leonardo generator and the Valentine (CEA), Le Barp, France—a French-government-funded technological research
antenna) had a figure of merit of 223 kV. organization. His research interests include high pulsed power generation, ul-
trawideband antenna, nonlinear transmission lines, high-power electromagnetic
waves, and high-voltage transient probes.

R EFERENCES
[1] J. D. Taylor, Introduction to Ultrawideband Radar Systems,
Laurent Pécastaing was born in Bayonne, France,
W. C. Nunally, R. N. Edwards, and D. V. Giri, Eds. Boca Raton,
in 1974. He received the M.Sc. and Ph.D. degrees in
FL: CRC Press, 1995, ch. 6, p. 287.
electrical engineering from the Université de Pau et
[2] R. Hoad, N. J. Carter, D. Herke, and S. P. Watkins, “Trends in EMC sus-
des Pays de l’Adour, Pau, France, in 1998 and 2001,
ceptibility of IT equipment,” IEEE Trans. Electromagn. Compat., vol. 46,
respectively.
no. 3, pp. 390–395, Aug. 2004.
He is currently a Lecturer with the Laboratoire de
[3] V. I. Koshelev, Y. I. Buyanov, Y. A. Andreev, V. V. Plisko, and
Génie Electrique, Université de Pau et des Pays de
K. N. Sukhushin, “Ultrawideband radiators of high-power pulses,” in
l’Adour. His research interests are focused on high-
Proc. IEEE Pulsed Power Conf., 2001, vol. 2, pp. 1661–1667.
power microwave (HPM) sources, compact pulsed
[4] S. Bazzoli, “Caractérisation et simulation de la susceptibilité des circuits
power devices, including pulse-forming lines or
intégrés face aux risques d’inductions engendrées par des Micro-ondes de
Marx generators, and ultrafast transient probes.
Fortes Puissances,” Thèse de Doctorat, l’Université de Lille I, Villeneuve
d’Ascq, France, 2005.
[5] J. Benford, J. A. Swegle, and E. Schamiloglu, High-Power Microwaves.
New York: Taylor & Francis, 2007, p. 531.
[6] B. Cadilhon, L. Pecastaing, T. Reess, and A. Gibert, “Low stray induc- Thierry Reess was born in Pau, France, in 1968. He
tance structure to improve the rise time of a Marx generator,” IET Elect. received the M.Sc. degree in plasma physics from the
Power Appl., vol. 2, no. 4, pp. 248–255, Jul. 2008. University of Toulouse, Toulouse, France, in 1992
[7] B. Cadilhon, L. Pecastaing, A. Gibert, T. Reess, J. Paillol, and P. Domens, and the Ph.D. degree in electrical engineering from
“Design realization and experimental study of a 200-kV 1-ns rise-time the Université de Pau et des Pays de l’Adour, Pau, in
Marx generator,” in Proc. 15th ISH, Ljubljana, Slovenia, Aug. 27–31, 1996.
2007. He is currently a Lecturer with the Laboratoire de
[8] L. Pecastaing, T. Reess, J. Paillol, A. Gibert, and P. Domens, “Electrical Génie Electrique, Université de Pau et des Pays de
breakdown studies of pressurized N2 in nonuniform fields,” Eur. J. Phys., l’Adour. His research interests include pulsed power
vol. 16, no. 1, pp. 59–69, Oct. 2001. devices and electrical discharges in gases and liquids.
CADILHON et al.: HIGH PULSED POWER SOURCES FOR BROADBAND RADIATION 2603

Antoine Silvestre de Ferron was born in Tarbes, Joël Andrieu was born in Figeac, France, in 1964.
France, in 1977. He received the M.Sc. degree in He received the Ph.D. degree and the Research
electrical and electronic engineering from the Uni- Directorship Habilitation from the University of
versity of Toulouse, Toulouse, France, in 2002 and Limoges, Limoges, France, in 2003 and 2008,
the Ph.D. degree in electrical engineering from the respectively.
Université de Pau et des Pays de l’Adour, Pau, Since 1991, he has been a Lecturer with the Uni-
France, in 2006. versity of Limoges. His research work is performed
From 2006 to 2008, he was a Researcher with in the Department of Waves and Associated Sys-
the Atomic Energy Commission (CEA), Le Barp, tems (OSA), XLIM Research Institute, University
France—a French-government-funded technological of Limoges–National Center for Scientific Research
research organization. He is currently an Engineer (CNRS), Brive, France. His research interests in-
with the Laboratoire de Génie Electrique, Université de Pau et des Pays de clude ultrawideband metrology for various applications, radar, electromagnetic
l’Adour. His research interests include high pulsed power generation with compatibility (EMC), and high-power microwaves (HPMs).
military and civil applications and combined high-voltage transient probes.

Michèle Lalande (M’92) was born in Noth, France,

Pascal Pignolet was born in Rennes, France, in in 1962. She received the Ph.D. degree in electronics
1951. He received the Ph.D. degree in gas and from the University of Limoges, Limoges, France,
plasma physics from the University of Orléans, in 1986.
Orléans, France, in 1984. His dissertation was de- She is currently a Professor with the University
voted to the spectroscopic study of nonresonant of Limoges. She is a Member of the Department of
laser–atom interaction. Waves and Associated Systems (OSA), XLIM Re-
He is currently a Professor with the Université de search Institute, University of Limoges–National
Pau et des Pays de l’Adour, Pau, France, where he is Center for Scientific Research (CNRS), Brive,
also the Head Manager of the Laboratoire de Génie France. Her research interest includes antennas and
Electrique. His main research interest is focused on transient measurement applications.
pulsed power domain, materials for electrical engi-
neering, and laser applications.

Stephane Vauchamp was born in Limoges, France,

in 1979. He received the M.Eng. and Ph.D. de-
grees in electrical engineering from the University
of Limoges, Limoges, France, in 2003 and 2008,
respectively.
He is currently a Contractual Engineer with
the Department of Waves and Associated Systems
(OSA), XLIM Research Institute, University of
Limoges–National Center for Scientific Research
(CNRS), Brive, France. His research interests in-
clude the metrology and measurement of pulsed
fields for high-power and detection applications.