Download modified_version_3

A Repetitive Rate Pulse System Used in Lifetime
Evaluation of High Voltage Ceramic Capacitors
Jiaqi Yan, Le Cheng, Siyu Zhang, Yanan Wang, Yang Gou, Weidong Ding
State Key Laboratory of Electrical Insulation and Power Equipment
Xi’an Jiaotong University
Xi’an, China
[email protected]
Abstract—The lifetime evaluation of High Voltage Ceramic
Capacitors (short as HVCC), especially the ones working under
repetitive pulses, is of great importance to the overall stability of
pulsed power system. However relevant studies concerning the
properties of HVCC under the conditions of repetitive rate pulse
were rare to see. In order to realize the lifetime testing for
HVCC, a high voltage, high current repetitive pulsed power
system was developed and tested. This system consists of a
repetitive pulsed power supply and a short-circuit testing cavity,
whose repetition frequency and output voltage can be regulated
easily. With this system, the lifetime test for HVCC was
conducted, and the lifetime characteristics at different voltage
levels under 25 Hz repetitive rate pulse and two kinds of failure
modes were studied. Based on the experimental results, several
measures were taken to improve the performance of HVCC. The
lifetime was extended greatly from thousands of times to more
than 105 times. The designed system and experimental results
have great significance for the extensive application field of
pulsed power technology, which would provide support for
future researches.
Keywords—repetitive pulse system; high voltage ceramic
capacitor; lifetime; magnetic pulse compression network
II. REPETITIVE RATE PULSE GENERATOR
A. Pulse Forming Network
The repetitive rate pulse supply used is based on the high
power pulse forming technology with the magnetic pulse
compressor being the core. The supply is composed of a
primary oscillator circuit, a two-stage magnetic pulse
compression network and two HV pulse transformers. For high
voltage pulse output, the rise time is ~10μs and the amplitude
is more than 70kV with the highest repetition frequency up to
25Hz. The principle circuit is shown in Fig. 1, where C0, C1, C2
are energy storage capacitors, L is the charging inductance,
MS1 and MS2 are magnetic switches, PT1 and PT2 are pulse
transformers and C is the experiment sample. In order to
prevent flashover, PT2 and the sample are soaked in
transformer oil.
I. INTRODUCTION
In the fields of long-pulse electron beams and high power
microwaves generation, HVCC are widely used in pulse
forming net (PFN) and have large application field due to its
great performance of relatively high dielectric constant and low
inductance. The failure mechanism of HVCC under repetitive
rate pulses is rather complicated. The properties of ceramics
are the primary element which determine the operation status
and lifetime of HVCC. The structure of ceramic chips and the
thickness of epoxy coating will influence the heat and stress
distribution on HVCC. The electric field strength will affect
the aging process through partial discharge and energy loss.
The combination of various elements will finally determine the
lifetime of HVCC under repetitive rate pulses. In this paper, a
repetitive rate pulse system used in lifetime evaluation of
HVCC is built[1-3].
The elaborately designed short-circuit testing cavity was
very compact in which SF6 gas can be charged more than
1.0MPa. A Rogowski coil is integrated into the cavity. The
short circuit current is up to 20kA with the 2.5nF capacitive
load. Circuit components are elaborately selected and
parameters are carefully matched. This repetitive microsecond
pulsed power system is integrated and installed in a portable
box which can realize the scale-up production[4-6].
Fig. 1. The circuit diagram of repetitive rate pulse supply.
In Fig.1, a DC voltage with certain amplitude will be
obtained by a regulator and a full-bridge rectifying circuit to
charge C0. When IGBT is turned on, the primary energy
storage capacitor C0 will charge the secondary energy storage
capacitor C1 by L. Right now, the magnetic switch MS1 is not
saturated. The extremely large impedance makes it equivalent
to a breaking circuit. During the charging period of C1, MS1
starts to undertake certain voltage. When the product of the
voltage and time has achieved its voltage second product, MS1
is turned on. C1 will discharge on the primary side of PT1,
which would result in a pulse signal with the amplitude of
several hundred volts and the rising time of several
microseconds. While C1 discharge on the primary side of PT1,
the induced voltage at the secondary side of PT1 will charge on
C2. When the product of the voltage and time has achieved the
voltage second product, MS2 will turn on, which will lead to
discharge of C2 to the primary side of PT2. By the stepping up
effect of PT2, a pulse voltage with higher amplitude and shorter
rising time will be obtained on C, which is the secondary side
of PT2.
When IGBT is turned off, C0’s charging process on C1 will
stop. The energy will be stored on C0. The voltage applied on
C will be nearly 0.When the IGBT is repetitively turned on and
off, the repetitive rate pulse voltage is formed. The repetition
frequency will be determined by the times of turning on the
IGBT in unit time.
B. The determination of the capacitors’ parameters
While designing the magnetic pulse compressor network,
the equivalent capacitance at each level should be nearly the
same in order to achieve higher efficiency. Therefore, the
capacitance of each storage capacitor is relevant to the
capacitance of the sample. The capacitance of the sample used
is nearly 5nF. The ratio of PT1 is 18 while the ratio of PT2 is 8.
Through the conversion of impedances, the following
equations could be analyzed.
C2 =n2C  82  5  320nF
(1)
C1 =n C2  18  0.4  129F
(2)
2
2
In order to improve the amplitude of voltage on C1 and
decrease the load on pulse transformers, the capacitance of C1
should be greatly lower than C0. Therefore, the capacitance of
C0 is chosen as 4mF.
The voltage after being processed by the regulator is nearly
220V. And the voltage after the full-bridge rectifying circuit is
nearly 310V. Thus C0 is chosen as a electrolytic capacitor with
the DC withstand voltage as 450V.
By simulating the charging process of C0 to C1, it could be
known that the maximum voltage value on C1 is nearly 500V.
Considering the energy loss, voltage margin and the actual
parameters, the capacitance of C1 is 150μF and the rated
voltage is 1200V. The capacitance of C2 is 400nF and the rated
voltage is 20kV.
C. The parameters of the primary oscillation circuit
The full-bridge rectifying model selected is MDS100, with
the reverse withstanding voltage as 600V and could undertake
a power frequency current as high as 100A. The current and
charging time should be determined by L, C1 and C0. In order
to prevent the damage to components caused by large primary
current, the charging time is 300μs. From the following
equation, L is calculated to be 63μH.
T = LC1C0 / (C1  C0 )
(3)
The IGBT used is CM300DY-24A. The collector current is
300A. The withstand voltage is 1200V. The on and off of
IGBT is controlled by the signal from a single-chip through a
drive circuit and protection circuit.
D. Pulse compressor network and pulse transformer
The magnetic pulse compressor network achieves the
magnetic compression by controlling the voltage second
product of the magnetic switches. The magnetic switches are
non-linear saturation reactors, which are composed of a
magnetic core and coils bound to the core. By taking advantage
of the non-linearity, the switching could be achieved by the
quick decrease of inductance after the saturation. When the
parameters of magnetic switches are selected appropriately, the
latter rising time will be shorter than the former one, which
could realize the goal of compression. The center issue of the
design of magnetic pulse compression network is the
determination of the voltage second product, which is
described as followed.
T
0.5UT   udt  N BS  
(4)
0
However, the voltage second product is related to the
working conditions, for instance, the voltage waveform and the
power frequency. It is hard to measure precisely. Therefore, the
parameters designed could only be as reference. The actual
parameters should be adjusted based on actual operation
conditions during the construction of the platform.
The magnetic core should be U-shaped silicon steel sheet
with high saturation flux density and satisfactory low
frequency properties. From the static magnetic hysteresis loop
of the silicon steel sheet, B  1.7T . The measured crosssection area of magnetic circuit S  1.35 103 m2 . Since the
charging time of C1 is 200μs and the maximum amplitude
value of U1 is 500V, the turns of MS1 is calculated to be 22,
according to the voltage second product equation. In actual
situation, the voltage waveform measured on C1 after
paralleling C1 to the primary oscillation circuit is shown in Fig.
2. The voltage value is 496V. The rising time is 222μs. The
turns determined from the voltage second equation is 25.
Fig. 2 The voltage waveform on C1.
After connecting the magnetic switch MS1, the voltage
decreased to 316V and rising time decreased to 54μs.
According to calculation, the number of turns at the primary
side of PT1 is 2. The designed ratio is 18. In ideal situation, the
turns of PT1 should be 2/36. In actual conditions, the maximum
output voltage value was obtained when the turns were 4/96.
The maximum voltage value on C2 is 8.4kV. The rising time is
18μs, as shown in Fig. 3.
inductance of the arc is determined by the length of gap. The
inductance of the copper bar is determined by its size and
distance. The length of copper bar is 8.2cm. The width is
4.04cm. The thickness is 0.01cm. The diameter of the copper
sphere is 2cm. The gap distance and the distance between the
copper bars are determined by the experiment voltage.
When the gap distance is 2 cm, the inductance of the circuit
could be estimated as followed. The inductance of the metal
strip is shown in (5).
L =4l (ln
Fig. 3 The voltage waveform on C2.
According to the waveform shown in Fig. 3, the turns of
MS2 is calculated to be 33. By adopting the same testing
method, the maximum efficiency of power supply could be
obtained when the turns of PT2 is 9/75. When C=2.5nF, the
voltage waveform on C is shown in Fig. 4. The voltage value is
77.3kV. The rising time is 15μs. For better insulation, PT2 is
positioned in transformer oil.
d
d
bc
 1.5   0.2235
)  41.7 nH
bc
l
l
(5)
The inductance of a 2 cm long arc is shown in (6).
L  14s  28nH
(6)
By adding the inductances together and taking the
estimation error into account, the whole inductance is about
several hundred nH. The resistance in the circuit is mainly
caused by the arc, which is relatively small. Therefore, the
discharge period is determined by the inductances and
capacitance, as shown in (7).
T =2 LC
(7)
By testing the discharge period, the value of inductance
could be obtained, as shown in Fig. 5. When the capacitance of
HVCC is 2.5nF and the gap distance is 1.87cm, the discharge
period is nearly 100ns. The inductance of the discharge circuit
is 101nH.
Fig. 4 The voltage waveform on C.
III. SHORT-CIRCUIT TESTING CAVITY
By using the discharge circuit as shown in Fig. 1, the
discharge current is rather small, the discharge period is at the
scale of microseconds and no sparks are observed due to the
large inductance at the secondary side of the pulse transformer.
No obvious heating effect is observed at the discharge circuit
through the infrared thermal imager, which is beneficial to the
precise measurement of temperature rise of HVCC after the
experiment.
However, concerning certain applications of repetitive rate
pulse power technology, for instance, high power microwave, a
compact discharge circuit, a low inductance of discharge
circuit, a short discharge period and the ability to withstand
extremely high discharge current are required. In order to
simulate these actual working conditions, another type of
discharge circuit is designed.
The structure of B type discharge circuit is rather compact.
The inductance is rather small. However, the repetitive
breakdown of the air gap between copper spheres under
repetitive rate pulse voltage will cause arc, which may further
result in strong light and large amount of heat. This would
interfere the temperature observation of the infrared thermal
imager. In addition, the heat of arc would be transmitted to oil
and HVCC through air and copper bars, which may influence
the temperature distribution.
The research on the thickness of coating and the electric
field strength mainly concentrate on the lifetime and change of
capacitance. The temperature rise doesn’t need to be measured
precisely since it is only used to judge the operation status of
HVCC. Therefore, type B is adopted. In the experiment, the
gap distance should be adjusted according to the environment
conditions, which may guarantee a breakdown when the
voltage is elevated to the designed value.
In this type of discharge circuit, the gap between bronze
spheres is paralleled to HVCC sample by copper bar. When the
voltage on HVCC is higher than breakdown voltage of the air
gap, the gap is broken down, which assists in HVCC’s shortcircuit discharge through the gap.
The inductance of the discharge circuit is composed of the
inductance of the HVCC, the copper bar and the arc. The
inductance of the HVCC sample is nearly 15nH. The
Fig. 5 Type B: another type of discharge circuit of HVCC.
IV. OPERATION
The signals that need to be monitored during the lifetime
experiment are the temperature of the sample, the voltage
waveform applied on the sample, the capacitance of the sample
and its loss factor. Through analyzing these parameters, the
sample could be determined its failure. The voltage waveform
and temperature signal are monitored online. The pulse voltage
produced by the repetitive rate supply is monitored offline
through the voltage divider and the high voltage sensor. The
heating phenomenon is measured by the infrared thermal
imager. The capacitance and loss factor are measured offline
by a LCR. The resistance is measured by a megameter.
The criteria for failure is the basis of a lifetime experiment.
The failure of HVCC is accompanied by the defects on the
appearance and the change of electrical parameters. Since the
whole sample is placed in transformer oil, when there are
obvious defects appearing, for instance, the cracking of epoxy
coating and the falling off of copper electrode, the HVCC
could be determined as failed. The change of electrical
parameters will influences the voltage waveform and the
temperature of the sample. In specific, the change of
capacitance would result in a change of voltage waveform
applied on the sample. The obvious increase of loss factor and
decrease of resistance will bring about the abnormal heating of
the sample. When there are abnormal phenomenon appearing
and a sharp increase of temperature, the sample could be
determined as failed. The measurement of the electrical
parameters should be suspended. When the capacitance
changes over 20%, or the loss factor is higher than 1% or the
resistance is lower than 1000MΩ, the sample should be
determined as failed.
TABLE I.
The Group of
Sample
A
B
C
D
LIFETIME RESULT OF SAMPLES
The preliminary lifetime experiment of HVCC has been
carried out and the results are shown in the tableⅠ. Samples of
different groups have different improved aspects. According to
the table, different improved measures influenced the lifetime
significantly. In the future, more detailed and comprehensive
researches will be carried out.
Several measures, like coating semiconductor at the edge of
silver layers, adding coupling agent to the ceramic-epoxy
interface and improving the structure of brass terminals, were
taken to improve the performance of HVCC. The effectiveness
was observed and he lifetime was extended greatly from
thousands of times to more than 105 times.
V. SUMMERY
With this platform, the accelerated lifetime test for HVCC
was conducted. The investigation of the failure models would
guide the manufacturing of HVCC, and the improvement
would enhance their performance, which is very important to
the reliability of pulsed power systems. Analyzing the factors
that influence the lifetime of HVCC and studying the failing
process could provide theoretical guidance for improving the
properties of HVCC and its reliability, which may furthermore,
be quite valuable for the improvement of high power repetitive
rate pulsed power technology.
REFERENCES
[1]
[2]
[3]
Lifetime (Number of Shots)
A1
A2
A3
A4
51000
104250
45000
58500
B1
B2
B3
B4
B5
31250 27000 22500
9000
12375
C1
C2
C3
C4
C5
50
100
3500
250
9000
D1
D2
D3
D4
300
375
50
3500
A5
28250
B6
250
C6
9000
D5
3125
[4]
[5]
[6]
B. Caballero, S. Luis and P. W. Smith, “Discrete pulse forming lines for
a compact z-pinch pulsed power generator,” Pulsed Power Conference,
2007 16th IEEE International, Albuquerque, NM, vol. 1, 2007, pp. 546549.
Young AL and Hilmas GE, “Mechanical vs. electrical failure
mechanisms in high voltage, high energy density multilayer ceramic
capacitors,” J. Mat. Sci., vol. 42, pp.5613-5619, 2007.
J. Su, X. Zhang and G. Liu, “A long-pulse generator based on Tesla
transformer and pulse-forming network,” IEEE Trans. Plasma Sci., vol.
37, pp. 1954-1958, 2009.
T. Shao, D. Zhang, and Y. Yu, “A compact repetitive unipolar
nanosecond-pulse generator for dielectric barrier discharge application,”
IEEE Trans. Plasma Sci., vol. 38, pp.1651-1655, 2010.
S. D. Korovin, I. K. Kurkan and S. V. Loginov, “Decimeter-band
frequency-tunable sources of high-power microwave pulses,” Laser Part.
Beams, vol. 21, pp.175-185, 2003.
K. Oyamada, S. Konosu, and T. Ohno, “Electric field concentration in
the vicinity of the interface between anode and degraded BaTiO3-based
ceramics in multilayer ceramic capacitor,” Appl. Phys. Lett., vol. 98, art.
no. 072905, 2011.