Abstract In order to elucidate how the degradation of

individual components affects the state of the photovoltaic

inverter as a whole, we have carried out SPICE simulations to
investigate the voltage and current ripple on the DC bus. The
bus capacitor is generally considered to be among the least
reliable components of the system, so we have simulated how the
degradation of bus capacitors affects the ripple at the DC
terminals of the PV module. Degradation-induced ripple leads to
an increased degradation rate in a positive feedback cycle. By
understanding the degradation mechanisms and their effects on
the inverter as a system, steps can be made to more effectively
replace marginal components with more reliable ones, increasing
the lifetime and efficiency of the inverter and decreasing its cost
per watt towards the US Department of Energy goals.

Index TermsBOS, Capacitor, Inverter, Photovoltaics, Power
Electronics, Reliability
I. INTRODUCTION

he Department of Energy (DOE) SunShot Initiative aims
to decrease the costs of photovoltaic (PV) systems by 75%
before the end of the decade, in order for PV systems to reach
economic competitiveness with conventional energy sources.
It is estimated that with an installed system cost of $1/Watt
($0.05-0.06/kWh) [1], PV penetration into the US power
market will be upwards of 18% by 2050.
As of 2010, the installed cost of PV systems was
$3.40/Watt [1]. While most research, both historically and
currently, focuses on the production costs of PV module
technology, the cost of the necessary grid-connected and/or
stand-alone DC-to-AC inverters has been largely ignored. As
the price of PV modules drops, the price of inverters becomes
more important. Inverters and associated power conditioning
components now constitute 8-12% of the total lifetime PV cost
[2] at $0.25/Watt [3], well above the DOE 2017 benchmark of
$0.10/Watt [1].
One of the key price drivers of inverter costs is
reliability [4, 5]. PV modules now have long lifetimes with
warranties offered up to 20 years and mean time between
failures in the field of up to 522 years for residential and 6,666
years for utility systems [6]. In contrast, the inverter has
shown a field mean time between failure of 1-16 years [7]
with typical warranties lasting only 3-5 years [5] and
occasionally (e. g. certain units by PV Powered) up to ten
years.
In short, inverter technology must become half as
costly and twice as reliable to facilitate grid-parity of solar
energy in the future [5, 8]. Inverter reliability typically tends
to be short because the inverter is both expected to serve a
large number of functions (e.g. PV power management, grid
connection/disconnection, VAR management, and/or island
monitoring [2]) while operating in relatively harsh, changing
conditions (-30-70C, 0-100% humidity, high salinity, etc).
Many sources [2, 9] consider bus capacitors to be among the
most unreliable elements of the inverter, decreasing the
inverter lifetime as much as 50% [10] due to temperature and
power cycling [11] resulting in high internal temperatures
[12]. Capacitors represent the constituent that can most easily
be altered in the short term to increase inverter lifetime and
decrease lifetime PV system cost.
The purpose of an inverter is to transform a DC
waveform voltage into an AC signal in order to inject power
into a load (e.g. the power grid) at a given frequency and with
a small phase angle (! ! 0). A simplified circuit for a single
phase unipolar Pulse-Width Modulation (PWM) is shown in
Figure 1 (the same general scheme can be extended to a three
phase system). In this schematic, a PV system, acting as a DC
voltage source with some source inductance, is shaped into an
AC signal through four IGBT switches in parallel with
freewheeling diodes. These switches are controlled at the gate
through a PWM signal, which is typically the output of an IC
that compares a carrier wave (usually a sine wave of the
desired output frequency) and a reference wave at a
significantly higher frequency (typically a triangle wave at 5-
20kHz). The output of the IGBTs is shaped into an AC signal
suitable for use or grid injection through the application of
various topologies of LC filters.


Figure 1: Pulsed Width Modulation (PWM) single-phase
inverter setup. The IGBT switches, along with LC output filter,
shape the DC input signal into a usable AC signal. This induces a
!
"!
$

%

&'()*+
,-.(*/0-*+
1'0.
2(3 4
5
!
"!
1
1
"67 "67
PV Inverter Performance and Reliability:
What is the Role of the Bus Capacitor?
Jack D. Flicker, Robert Kaplar, Matthew Marinella, and Jennifer Granata
Sandia National Laboratories, Albuquerque, NM 87185, USA
Manuscript received May 21, 2012. This work was funded by the DOE
Office of Energy Efficiency and Renewable Energy. Sandia National
Laboratories is a multi-program laboratory managed and operated by Sandia
Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for
the U.S. Department of Energy's National Nuclear Security Administration
under contract DE-AC04-94AL85000.
Jack D. Flicker, corresponding author (505-284-6810; fax: 505-844-2991;
e-mail: [email protected]).
T
deleterious voltage ripple across the PV terminals. The bus
capacitor is sized in order to reduce this ripple.
The operation of the IGBTs introduces a ripple
voltage onto the terminal of the PV array. This ripple is
deleterious to the operation of the PV system, since the
nominal voltage applied to the terminals should be held at the
max power point (MPP) of the IV curve in order to extract the
most power. A voltage ripple on the PV terminals will
oscillate the power extracted from the system [13], resulting in
a lower average power output (Figure 2). A capacitor is added
onto the bus in order to smooth out the voltage ripple [14].


Figure 2: A voltage ripple introduced onto the PV terminals by
the PWM inverter scheme shifts the applied voltage off the max
power point (MPP) of the PV array. This introduces a ripple in
the power output of the array so that the average output power is
lower than the nominal MPP [13].
The amplitude (peak to peak) of the voltage ripple is
determined by the switching frequency, PV voltage, bus
capacitance, and filter inductance according to:
!
!!!
!"##$%
!
!
!"
!" ! !
!"#
! ! ! !
!"#
!


(1)
where:
V
PV
is the solar panel DC voltage,
C
bus
is the capacitance of the bus capacitor,
L is the inductance of the filter inductors,
f
PWM
is the switching frequency.


Equation (1) applies to an ideal capacitor that
prevents charge from flowing through the capacitor during
charging and then discharges the energy located in the electric
field with no resistance. In reality, no capacitor is ideal (
Figure 3) but is composed of multiple elements. In addition to
the ideal capacitance, the dielectric is not perfectly resistive
and a small leakage current flows from the anode to cathode
along a finite shunt resistance (R
sh
), bypassing the dielectric
capacitance (C). When current through the capacitor is
flowing, the pins, foils, and dielectric are not perfectly
conducting and there is an equivalent series resistance (ESR)
in series with the capacitance. Finally, the capacitor does
store some energy in the magnetic field, so there is an
equivalent series inductance (ESL) in series with the
capacitance and ESR.


Figure 3: Equivalent circuit of a generic capacitor. A capacitor is
composed of many non-ideal elements, including dielectric
capacitance (C), a non-infinite shunt resistance through the
dielectric that bypasses the capacitor, series resistance (ESR),
and series inductance (ESL).
Even in a component as seemingly simple as a
capacitor, there exist multiple elements that can fail or
degrade. Each of these elements can affect the behavior of the
inverter, both on the AC and DC sides. In order to determine
the effect degradation of non-ideal capacitor components has
on the voltage ripple introduced across the PV terminals, a
PWM unipolar H-bridge inverter (Figure 1) was simulated
using SPICE. The filter capacitors and inductors are held at
250F and 20mH, respectively. The SPICE models for the
IGBTs are derived from the work of Petrie et al. [15]. The
PWM signal, which controls the IGBT switches, is determined
by a comparator and inverting comparator circuit for the high-
and low-side IGBT switches, respectively. The input for the
PWM controls are a 9.5V, 60Hz sine carrier wave and a 10V,
10kHz triangular wave.

II. SIMULATION RESULTS AND DISCUSSION

Figure 4 shows the results of a SPICE simulation for
a 250F bus capacitor. The output of the inverter across the
load is an AC wave at 60Hz. The bus voltage, though
nominally 310VDC, contains an AC ripple with a frequency
of 120Hz. The inset of Figure 4 shows a close-up of the bus
voltage and demonstrates that, in addition to the 120Hz ripple,
there exists a high frequency component to the ripple. This
high frequency is equal to the PWM triangle wave frequency.

!"#$%&' )!*
+,,
!
"-
.
/-
!
0122#'
,
"3$
,
%4&
!"# !"%
&
#
'(
)*+,-./0 &/1/2,3*4

Figure 4: Results of SPICE simulation of inverter circuit. The
IGBT switching transforms the nominal DC bus voltage from the
PV array into an AC signal across the load. This induces a ripple
at 120Hz. The inset shows an expanded view of the bus voltage,
showing a ripple at the PWM frequency.
Correctly sizing the bus capacitance according to (1)
is typically used to manage this DC-side ripple voltage.
Simulations correctly predict the voltage ripple on the bus as a
function of the bus capacitance (Figure 5). The ripple has a
1/C dependence, as predicted. For a 325V bus voltage, the
ripple dramatically increases as the capacitance decreases,
with relatively little decrease for capacitances above ~370F.


Figure 5: SPICE simulations of bus ripple correctly match the
expected results from (1).
Although correctly sizing the bus capacitance is
important to control the magnitude of the bus ripple, it is not
the only thing that must be considered. The component
frequencies of the bus ripple, which are influenced by the
source inductance, are quite important. Both the size of the
bus capacitor and source inductance will affect the resultant
DC-side ripple. Those components will interact to form an LC
oscillator circuit, which will cause the inverter circuit to
resonate (ring) at a frequency determined by (2). This can be
very detrimental to the system if the frequency of ringing
corresponds either to the PWM frequency or a multiple of the
carrier wave frequency.

!
!"#$
!
!
!! !
!"#$%&
!
!"#

(2)
where:
f
ring
is the ringing frequency,
L
source
is the source inductance


Figure 6 shows the magnitude of the ripple voltage
for an inverter system with a 250F bus capacitor as a
function of the source inductance. When the source
inductance combines with the capacitance such that f
ring
is
equal to either the PWM frequency or a multiple of the carrier
wave frequency according to (2), the magnitude of the ripple
increases dramatically.


Figure 6: Ripple voltage for a PWM unipolar inverter with a
250F bus capacitor as a function of the source inductance. The
bus capacitor and source form an LC oscillating circuit that will
ring at the PWM signal and multiples of the carrier wave
frequencies.
As the inverter operates, the ripple on the DC bus
will increase over time as the bus capacitor degrades. The
exact behavior of the ripple depends on exactly which of the
capacitor components degrade.
A large enough ESL will dramatically affect the bus
ripple magnitude and frequency. However, the ESL of a
typical commercial capacitor is quite small (~nH [16]) and is
largely ignored in circuit analysis of capacitors. Since the
ESL is typically small enough to go unnoticed and does not
change much over the lifetime of the capacitor, it has not been
investigated in this work.
The shunt resistance directly controls (and is
inversely proportional to) the leakage current of the capacitor.
-300
-200
-100
0
100
200
300
400
13 20 23 30 33 40 43 30
!
"
#
$
%
&
'

)
!
*

+,-' )-.*
/,-0#%1"2 "3 45"$"6"#$%,7 826'9$'9 82:0$ %2; <0$:0$
vbus
vload
282
287
292
17 17.3 18 18.3 19
0
10
20
30
40
30
60
70
80
90
0.00L+00 3.00L-04 1.00L-03 1.30L-03 2.00L-03
V
r
|
p
p
|
e
p
-
p

(
V
)

Capac|tance (I)
8us V
r|pp|e
p-p
vs. Capac|tance for
WM n-br|dge Inverter
Slmulauon
Calculauon
0
30
100
130
200
230
300
1.L-09 1.L-08 1.L-07 1.L-06 1.L-03 1.L-04 1.L-03 1.L-02 1.L-01
V
r
|
p
p
|
e
p
-
p

(
V
)

Source Inductance (n)
V
r|pp|e
p-p
vs. Source Inductance for WM Un|po|ar
Inverter
Slmulauon
9.8kPz 8lng
240Pz 8lng
60Pz 8lng
Historically, electrolytic capacitors have suffered from
increasing leakage currents over time due to corrosion effects
[17]. However, state-of-the-art capacitors suffer almost no
change in leakage current, even over long periods of time (for
electrolytic capacitors, there is a short term decrease in
leakage current due to oxide reformation effects). Shunt
resistance effects on both the inverter input and output were
investigated; however, the waveforms of both the AC and DC
sides were found to be relatively unaffected by changes in
shunt resistance until leakage current increased by a factor of
~100. As leakage current increases by 100-times above
nominal levels, the inverter efficiency decreases exponentially
(data not shown).
One of the primary degradation modes of capacitors
is an increase in the ESR. Figure 7 shows the effect of ESR
on DC-side voltage ripple. Voltage ripple magnitude is highly
dependent on the size of the ESR before leveling off at high
ESR values (>4").


Figure 7: Inverter DC Ripple voltage vs. ESR for a bus
capacitor. The DC ripple voltage magnitude is highly dependent
on the size of the capacitor ESR. The inset shows the same graph
on a semi-log plot to give greater detail regarding small ESR
values.
Electrolytic capacitors have relatively high initial
ESR values (~20m", but up to 80m" for high voltage
capacitors) compared to metallized thin film (~1m") [14].
Additionally, the ESR of electrolytic capacitors tends to
increase over the lifetime of a capacitor due to the
consumption of H
2
O during oxide reformation [18] or
electrolyte evaporation [19]. Over time, the ESR of an
electrolytic capacitor has been found to follow the linear
inverse relationship first defined by Rhoads and Smith:

This equation gives an estimate of the best-case
capacitor lifetime due only to electrolyte evaporation.
Electrolyte evaporation is highly dependent on the operating
temperature of the capacitor and the general rule of thumb
for electrolytic capacitors is that a 10C increase in
temperature decreases lifetime by a factor of two. Sandia
National Labs has monitored the temperature of inverter
components in situ. Bus capacitors can easily reach peak
temperatures of ~60C, even for fairly modest ambient
temperatures (~20C) in the late winter.
Figure 8 shows the magnitude of the DC ripple
voltage as a function of the capacitor operating time according
to (3) for a typical electrolytic bus capacitor that has an
original ESR of 50m" and operates at an average temperature
of 70C. Soft failure, defined as an ESR increase of 200%, is
denoted by the hatched line and occurs after 71,000 hours of
operation when the ripple voltage has reached a value of 40V.


Figure 8: Ripple voltage vs. time for an electrolytic capacitor due
to ESR increase according to (3). The dashed horizontal line
indicates an increase in the ESR of 200%, a common definition
for soft failure. In this typical bus capacitor, soft failure would
be expected, in the best case, to occur after 71,000 operating
hours with a ripple voltage of approximately 40V.
The expected lifetime of capacitors using this model
is quite good (!70,000 hours), especially since failure is
through slow degradation rather than a catastrophic event, so
the capacitor continues to function, albeit in a reduced
capacity. However, this is an ideal model for capacitor lifetime
and does not take into account any effects of ripple current,
which may significantly alter the interior temperature of the
capacitor compared to the outer surface. This leads to a
positive feedback mechanism between ESR and ripple current.
In this cycle, as the ESR increases, the ripple voltage and
current also increase. This increase in ripple current causes an
increase in the capacitor core temperature, which causes
further and faster degradation of the capacitor ESR.
This model also does not take into account how
electrolyte evaporation affects capacitor integrity as the
buildup of pressure inside the capacitor container may lead to
!
#!
$!!
$#!
%!!
%#!
&!!
! % ' ( ) $! $%
V
r
|
p
p
|
e
p
-
p

(
V
)

LSk (!)
V
r|pp|e
p-p
vs. LSk
!
#!
$!!
$#!
%!!
%#!
&!!
!*!!$ !*!$ !*$ $ $!
! # $ % & '! '# '$ '% '&
!
#!
$!
%!
&!
'!!
'#!
'$!
'%!
!(!)*!! $(!)*!$ &(!)*!$ '(#)*!+ '(%)*!+
!"#$ &'()
*
+
,
+
(
"
+
+
-
$



&
*
)

!"#$ &.()
*
+,+
("++-$
/0 !"#$ 12$ 34 567 89:($;0$
!"#
!"#
!
!
!
! ! ! ! ! ! !
!!
!
!

(3)
where:
ESR
o
is the original ESR,
k is a constant dependent on the capacitor
design, but is usually ~1.77 [20, 21]
t is time
E
a
is an activation energy ( !4700 eV [22])
T is temperature (Kelvin)

venting and leakage of liquid electrolyte long before the
capacitor fails in a degradation mode. Finally, the model does
not consider any decreased lifetime due to voltage
acceleration. A capacitor would only be expected to have a
71,000 lifetime if it is sized (for ripple current, voltage, and
temperature) well above actual working conditions, so that it
would be relatively unstressed throughout its working life.
In a real world inverter, it is unlikely that a capacitor
will only be subject to ideal conditions well below the
capacitor sizing parameters. Often, the capacitor stress is
increased for the sake of other components. For example, as
the temperature of an inverter increases above a certain set
point, the control circuitry of the inverter will de-rate the
inverter output in order to decrease the internal temperature.
For certain inverters, this is accomplished by increasing the
voltage applied to the PV array towards the open circuit
voltage. While this will successfully reduce the temperature
by exponentially decreasing the current input to the inverter, it
increases the voltage stress of the capacitor. So, for a de-rated
inverter, the capacitor is doubly degraded, both from the
increased temperature in the inverter and the increased bus
voltage. This inverter behavior can be disastrous as the
capacitor is only rated to a voltage at or slightly above the
array open circuit voltage. Such a capacitor would not be
expected to last through 71,000 operating hours.
Unfortunately, to reduce costs, some inverter
manufacturers decide to sacrifice capacitor quality and/or size.
Incorrect sizing of the capacitor or sizing without a complete
understanding of the consequences to DC-side ripple can be
disastrous and lead to shortened capacitor lifetimes. Short
lifetimes are costly, not only due to replacement parts, but also
due to system downtime, cost of replacement energy, and any
reduced efficiency before failure identification.
III. SUMMARY AND FUTURE WORK

This work has investigated capacitor degradation
effects on inverter operation, namely with regard to the
voltage ripple onto the nominal DC supply voltage of the PV
array. This ripple is deleterious, since it decreases average PV
power output. SPICE simulations have been used to simulate
the input and output of a PWM unipolar PV inverter utilizing
IGBT switches. These simulations have shown that voltage
ripple on the DC-side is composed of two distinct frequency
components. There is a ripple at twice the carrier wave
frequency and a ripple at the PWM frequency.
The magnitude of the DC ripple voltage has been
shown to be affected both by the capacitor size and source
inductance. The combination of these components forms an
LC oscillator that, if the resonant frequency occurs at one of
the ripple voltage component frequencies, can drastically
increase the magnitude of the ripple voltage. Leakage current
does not impact the inverter input or output for changes up to
two orders of magnitude from nominal values. However,
ripple voltage was found to be highly dependent on capacitor
ESR. For electrolytic capacitors, this ripple voltage as a
function of ESR can be transformed to a function of time
through a linear inverse formula. Using this formula, it is
estimated that an average inverter would be expected to
operate for 71,000 hours before failure, at which point the
ripple voltage would be 40V. However, this best-case formula
does not take into account voltage stress or ripple current, so
lifetimes of real world capacitors are expected to be much
lower.
Sandia National Labs is currently conducting
experiments on bus capacitor reliability. These efforts focus
on metallized thin film capacitors for three reasons. First,
there is a trend towards replacing electrolytic capacitors with
metallized thin film capacitors, even for smaller residential
systems. This push is driven both by the widely held view that
thin film capacitors are inherently safer than electrolytics and
that the increased cost of the thin film capacitors is
compensated by their tolerance of higher ripple currents.
Second, most reliability data on metallized thin film
capacitors are for pulsed power applications. While these
experiments show excellent lifetimes (up to 150,000 hours
[23]), some experiments have shown much shorter (<10,000
hours) lifetimes under constant voltage conditions [24-28].
Finally, thin film capacitors are said to be inherently
safe compared to electrolytics due to self-clearing properties
that prevent short-circuiting. Therefore, thin film capacitors
should not fail catastrophically. However, since the number of
clearing events increases exponentially after soft failure, the
pressure inside the capacitor increases and may eventually
lead to flashover and catastrophic failure. Future experiments
will examine the failure behavior of thin film capacitors if
they are used beyond their soft failure points to determine if
they are significantly safer compared to electrolytic capacitors.
ACKNOWLEDGMENT

The authors thank Mike Fife of PV Powered and
Fenton Reese of Fenton Rees and Associates for useful
discussions.
REFERENCES

[1] "$1/Watt Photovoltaic Systems," US Department of
Energy, White Paper 2010.
[2] Y. Xue, K. C. Divya, G. Griepentrog, M. Liviu, S.
Suresh, and M. Manjrekar, "Towards Next
Generation Photovoltaic Inverters," 2011 IEEE
Energy Conversion Congress and Exposition
(ECCE), 2011.
[3] L. Bony, S. Doig, C. Hart, and E. Maurer, Achieving
low-cost solar PV: Industry Workshop
Recommendations for Near-term Balance of System
Cost Reductions: Rocky Mountain Institute (RMI),
2010.
[4] B. Jablonska, H. Kaan, M. Leeuwen, and G. Boer,
"PV-prive Project at ECN: Five Years of Experience
With Small-scale AC Module PV Systems,"
Presented at the 20th European Photovoltaic Solar
Energy Coference and Exhibition, 2005.
[5] A. Ristow, M. Begovic, A. Pregelj, and A. Rohatgi,
"Development of a Methodology for Improving
Photovoltaic Inverter Reliability," IEEE Transactions
on Industrial Electronics, vol. 55, pp. 2581-2592,
2008.
[6] A. Maish, "Defining Requirements for Improved
Photovoltaic System Reliability," Progress in
Photovoltaics, pp. 165-173, 1999.
[7] A. B. Maish, C. Atcitty, S. Hester, D. Greenberg, D.
Osborn, D. Collier, and M. Brine, "Photovoltaic
System Reliability," Conference Record of the
Twenty Sixth IEEE Photovoltaic Specialists
Conference, 1997, 1997.
[8] R. Morgolis, "A Review of PV Inverter Technology
Cost and Performance Projections," National
Renewable Energy Laboratory, 2006.
[9] W. Bower and D. Ton, "Summary Report on the
DOE High-tech Inverter Workshop," DOE Office of
Energy Efficiency and Renewable Energy, 2005.
[10] C. Rodriguez and G. A. J. Amaratunga, "Long-
Lifetime Power Inverter for Photovoltaic AC
Modules," IEEE Transactions on Industrial
Electronics, vol. 55, pp. 2593-2601, 2008.
[11] T. Von Kampen and E. Sawyer, "Reliability
Considerations of Inverter/DC Link Capacitor Using
PP Film and 105C Engine Coolant," Proceedings of
the International Microelectronics and Pakaging
Society (IMAPS), 2008.
[12] S. G. J. Parler, "Thermal Modeling of Aluminum
Electrolytic Capacitors," Conference Record of the
1999 IEEE Industry Applications Conference, 1999.
[13] S. B. Kjaer, J. K. Pedersen, and F. Blaabjerg, "A
Review of Single-Phase Grid-Connected Inverters for
Photovoltaic Modules," IEEE Transactions on
Industry Applications, vol. 41, pp. 1292-1306, 2005.
[14] M. Salcone and J. Bond, "Selecting film bus link
capacitors for high performance inverter
applications," IEEE International Electric Machines
and Drives Conference (IEMDC), 2009.
[15] A. F. Petrie and C. Hymowitz, "A SPICE model for
IGBTs," Applied Power Electronics Conference and
Exposition, 1995. APEC &apos;95. Conference
Proceedings 1995., Tenth Annual, 1995.
[16] A. Albertsen, "Electrolytic Capacitor Lifetime
Estimation," Jianghai Europe, GmbH, 2010.
[17] J. Both, "Leakage Current Properties of Modern
Electrolytic Capacitors," BC Components, Aug 30
2001.
[18] S. Parler, "Application Guide, Aluminum Electrolytic
Capacitors," Cornell-Dublier.
[19] M. L. Gasperi, "Life Prediction Model for Aluminum
Electrolytic Capacitors," Conference Record of the
1996 IEEE Industry Applications Conference (IAS
1996), 1996.
[20] R. C. Rosenberg and D. Karnopp, Introduction to
Physical Systems Dynamics: McGraw-Hill College,
1983.
[21] G. Biswas, X. Koutsoukos, and K. Goebel, "Physics
of Failure Models for Capacitor Degradation in DC-
DC Converters," The Maintenance and Reliability
Conference (MARCON), 2010.
[22] G. Rhoads and A. Smith, "Expected Life of
Capacitors With Non-Solid Electrolyte," IEEE 34th
Electronic Component Conference, 1984.
[23] E. N. Power. (2008). Capacitors Age and Capacitors
Have an End of Life.
[24] H. Fuhrmann, M. Carlen, D. Chartouni, T. Christen,
C. Ohler, and T. Votteler, "Novel measurement
methods for in-depth analysis of AC metallized film
capacitors," Electrical Insulation, 2004. Conference
Record of the 2004 IEEE International Symposium
on, 2004.
[25] W. Sarjeant, F. MacDougall, D. Larson, and I.
Kohlberg, "Energy storage capacitors: Aging, and
diagnostic approaches for life validation," Ieee
Transactions on Magnetics, 1997.
[26] M. W. Carlen, P. Bruesch, and R. Gallay, "Electrical
Endurance Characterization of Polypropylene
Winding Capacitors for Traction Applications: a New
Experimental Method," Dielectric Materials,
Measurements and Applications, Seventh
International Conference on (Conf. Publ. No. 430,
1996.
[27] D. SHAW, S. CICHANOWSKI, and A. YIALIZIS,
"A Changing Capacitor Technology - Failure
Mechanisms and Design Innovations," Ieee
Transactions on Electrical Insulation, vol. 16, pp.
399-413, 1981.
[28] J. B. Ennis, F. W. MacDougall, X. H. Yang, R. A.
Cooper, K. Seal, C. Naruo, B. Spinks, P. Kroessler,
and J. Bates, "Recent Advances in High Voltage,
High Energy Capacitor Technology," IEEE
International Pulsed Power Plasma Science
Conference, 2007 (PPPS 2007), 2007.