IEEE Industry Applications Society

Annual Meeting
New Orleans, Louisiana, October 5-9, 1997

Electrolytic Capacitor Life Testing and Prediction

V. A. Sankaran F.L. Rees C.S. Avant

Ford Research Laboratory Consultant Ford Research Laboratory
MD 1170/SRL, P.O. Box 2053 7317 E. Heather Way, Everett MD 1170/SRL, P.O. Box 2053
20000 Rotunda Drive, Dearborn WA 98203, USA 20000 Rotunda Drive, Dearborn
MI 48121, USA MI 48121, USA

Abstract - The aluminum electrolytic capacitor is this is not common. In this paper, first the details of the
widely used in various power electronic circuits experimental arrangement, its operation and ESR
and systems. Its functions include, bus voltage estimation methodology are given. Next experimental
stabilization, conduction of ripple current due to results are presented, followed by a comparison against
switching events, etc. In automotive applications, modeling results based on a recent model reported in
one of the big issues is the extreme and harsh literature.
temperatures they have to withstand, underhood,
which during the summer months would reduce
their life. Most of the capacitors have maximum II. CAPACITOR CONSTRUCTION AND ISSUES
o
temperature ratings of either 85 or 105 C and a few
o The aluminum electrolytic capacitor is composed
are rated at 125 C. According to literature found,
of an anode foil, a cathode foil and separator plate
one of the primary wear-out mechanisms in
wound together and impregnated with an electrolyte [1].
electrolytic capacitors is mainly due to the loss of
The electrolyte contains a solvent such as ethylene
electrolyte by vapor diffusion through the seals.
glycol and a solute such as ammonium borate. This
Other reports suggest that the main wearout
impregnated assembly is then placed in an aluminum
mechanism is deterioration of the electrolyte.
can and sealed. The seal is made up of a material
Experiments conducted to date and the results
such as, rubber, rubber backed phenolic, molded
from the same are compared against modeling
phenolic resin or polyphylenesulfide [1]. The primary
results based on a recent model reported in
wear-out mechanism for aluminum electrolytic
literature.
capacitors is vaporization of electrolyte through the end
seal [2]. As the volume of the electrolyte decreases,
the ESR of the capacitor increases and its capacitance
I. INTRODUCTION
decreases. The increase in ESR is a cumulative effect
The aluminum electrolytic capacitor is widely used since the increase in ESR leads to increased heating
in various power electronic circuits and systems. Its and this in turn leads to further evaporation of
functions include, bus voltage stabilization, conduction electrolyte.
of the ripple current due to switching events, etc.
Aluminum electrolytic capacitors typically come in
cylindrical cans. There are surface mountable (low III. EXPERIMENTAL ARRANGEMENT
voltage and small size), through hole/snap-in (low
III.1 Introduction :
voltage and medium size) and screw terminal versions
(large size). Recently some are available as flat packs When a three phase Pulse Width Modulated
and stud-mountable cylindrical cans. For motor drive (PWM) inverter is used to drive an AC motor (either
type applications the capacitors four main ratings are : permanent magnet or induction), a pseudo-square
voltage, capacitance value, ripple current and wave of current is injected onto the inverter’s DC link.
temperature. Capacitors have equivalent series internal Even if the inverter is supplied from a battery,
resistance (ESR) which leads to heating when ripple electrolytic capacitors are usually placed across the DC
current flows through the same. bus to act as a source/sink of this current, Fig. 1.
In automotive applications, one of the big issues is For applications where the inverter is operating
the extreme and harsh temperatures they have to continuously at rated load (eg. an industrial drive
withstand. Most of the capacitors in the market have controlling a pump), the capacitors on the DC bus must
maximum temperature ratings of either 85 or 105 oC. A be rated to continuously handle the injected ripple
few capacitors have a temperature rating of 125 oC but current. If aluminum electrolytics are used, the amount

07803-4070-1/97/$10.00 (c) 1997 IEEE

IEEE Industry Applications Society
Annual Meeting
New Orleans, Louisiana, October 5-9, 1997

of capacitance placed on the DC bus is actually

determined by ripple current/core temperature rise
considerations, and not the amount of capacitance (µF)
required to control the ripple voltage. Conversely,
applications where the inverter/motor are at rated
power only intermittently (eg. automotive applications),
the time constant of the capacitor integrates the peak
core temperature, allowing proportionally less
capacitance to be used for a given peak motor power.
III.2 Test Circuit to Simulate Inverter Operation :
It was required to develop a test fixture for
electrolytic capacitors that would allow them to be
operated in a manner similar to that seen in an
intermittent duty cycle inverter; for a long enough time
that useful longevity/reliability information could be
obtained. The requirements were then :
Fig. 1. Typical Bus Capacitor Ripple Current for
• to inject currents of similar shape and frequency as
3-Phase PWM inverter.
occur in actual inverter operation,
• to allow the current to be intermittent,
Use of a “Buck” (i.e. step down) converter and a
• to allow the capacitor to be operated at elevated
“Boost” (i.e. step up) converter cascaded together and
temperature,
with the Boost output connected to the Buck’s input
• to operate for long periods (months) with minimal seemed like the ideal arrangement; Fig. 2.
maintenance,
• and preferably not require high power input.

Icap Therm-couples
High Frequency Current Generator

“Buck” Heated
“Boost”
Enclosure
Choke
S1
Power Capacitors
Supply Under Test
Snubber

S2

Thermo-couple

Variable Pulse Width

Controller (Current Control) Heater

Temperature Controller
and
Timer Phase Controller

Fig. 2. Experimental Test Setup

07803-4070-1/97/$10.00 (c) 1997 IEEE

IEEE Industry Applications Society
Annual Meeting
New Orleans, Louisiana, October 5-9, 1997

If the two active switches, S1 and S2, were switched at Harada et. al. [3] show how the ESR of an
the same frequency and more or less in phase, the electrolytic capacitor in both a forward and a buck-boost
combined circuit would inject approximately square DC-DC converter was deduced from the capacitor’s
wave currents into the capacitor on the high side DC voltage and current waveforms while the circuit was
bus, and the required input power would be only that operating normally. The increase in the ESR as a
required to make up the losses in the switches and function of time was used as a means of quantifying the
inductors. As per usual with Buck or Boost converters, deterioration of the capacitor.
the inductance of the input inductor was chosen to limit
The equivalent circuit of a real capacitor is a series
the current ripple, and the capacitance of the output
combination of the actual capacitance(C), the ESR and
capacitor to limit the voltage ripple. The test capacitor
the leakage inductance (LX) of the package. Hence
was placed on the high side DC bus by itself, in order to
when used in a circuit that imposes current waveforms
maximize the current in that capacitor. The input Buck
that are pseudo squarewaves (as in the PWM inverter
converter had a variable duty cycle, while the Boost
of Fig. 1.), the resulting voltage waveform will be a
converter was fixed at 50/50, so that it was a voltage
combination of :
doubler. Significant current only began to flow when
the duty cycle of the Buck converter was greater than VR = I * ESR, VL = LX * dI/dT, and ∆VC = I * ∆T/C .
50%, so the combined “gain” exceeded unity. Other
The above three elements of capacitor voltage can be
details of the circuit include :
seen in the voltage waveform of Fig. 3. :
• Control of the Buck converter’s duty cycle via a
• The “instantaneous” step in voltage (∆V = 0.5V)
current feedback loop.
when the current waveform changes to a new level
• The removal of both the capacitor at the output of (∆I = 28A), ⇒ ESR = ∆V/∆I = 0.5V/28A = 18mΩ
the “Buck” stage and the inductor at the input of the
• The gradual ramp in the capacitors voltage (dV/dt ≈
“Boost” stage, as these proved to be redundant
11mV/µs) when the current is more or less constant
(shown in dashed lines in Fig. 2).
(I ≈ 28A), ⇒ C = I/(dV/dt) = 28A/11mV/µs = 2550µF
• An inductor in series with the DC input, to ensure
• The rather “nasty” voltage glitches (∆V = 1.5V)
the capacitor being tested was completely isolated
every time the current suddenly changes (∆I ≈ 28A
from any capacitance that the DC supply may have.
in 200ns), ⇒ LX = ∆V/(dI/dt) = 1.5V/140A/µs = 11nH
• A PID controller to set the temperature of the
heated capacitor enclosure.
• A data acquisition system to monitor various
parameters, record these every 20 minutes, and
shut the system down if excessive voltages,
currents or temperatures were detected.
The automotive applications can be classified into
two major categories : continuous and intermittent, and
the required operating conditions are given in Table I.

TABLE I
Application Operating Conditions

Operating Continuous Intermittent

Condition/System

PWM Frequency 8-16 kHz 8-16 kHz

o o
Ambient 70-85 C (Passenger) 70-85 C (Passenger)
Temperature 115oC (Underhood) 115oC (Underhood)

Duty Cycle for Continuous ON for 0.5 to 5 min Fig. 3. Capacitor ESR from Voltage and Current
Pseudo-squarewave > 15 min OFF for 5 to 20 min Waveforms. (ESR = ∆V/∆I = 0.5V/28A = 18 mΩ).
Current Waveform

III.3 Use of Test Circuit to Measure ESR :

07803-4070-1/97/$10.00 (c) 1997 IEEE

IEEE Industry Applications Society
Annual Meeting
New Orleans, Louisiana, October 5-9, 1997

TABLE II
Sample Details and Experimental Results

Parameter \ Test # Test1 Test2 Test3

µF, 105 C µF, 105 C
o o o (1)
63V, 6800µ 63V, 2200µ 25V, 20mF, 85 C
Application Continuous Intermittent Intermittent
Capacitor Style/Size Snap-in 35Φ x 50mm Snap-in 25Φ x 30mm “Flatpack”: 3” x 1.75” x 0.5”
Specified Ripple I Rating 10.7A @ 20kHz, 70oC 2.0A @ > 10kHz, 105oC 9.6A @ 20kHz, 85oC
Test Conditions
Ambient Temperature 70oC 115oC 115oC
DC Voltage 60V 30V 23V
Ripple Current 11A rms @ 16kHz 12A rms @ 15kHz 50A rms @ 15kHz
Duty Cycle (of Ripple I) Continuous 30sec ON & 20min OFF 30sec ON & 20min OFF
Test Duration 1500 hours 1000 hours 3000 hours
Measured Change in 27% Increase 6x and 14x 10% increase
ESR over Test Period
Note : (1) 125oC rating if rectangular can is clamped

As the ESR of an electrolytic capacitor can vary by an Two capacitors were tested at the relatively
order of magnitude from 0 to 80 oC, the use of this “insane” ambient temperature of 115oC; significantly
method to determine the increase in ESR as a function outside the specified 105oC maximum, but
of time/usage requires comparative measurements be representative of worst case underhood conditions.
made at the same temperature. The duty cycle of ON for 30 seconds with ripple current,
OFF for 20 min (but still with bus voltage across C) was
IV. EXPERIMENTAL RESULTS designed to simulate some of the application operating
cycles (see Fig. 4.). Previous measurements had
Three different capacitors were tested under different
shown that :
conditions. Details are provided below in Table II.
Results from the tests are also summarized in Table II. • 12A rms of ripple current @ 15kHz for 30 seconds
o caused the capacitor’s core temperature (via
IV.1 Test1 - 63V, 6800µF, 105 C Snap-ins for
embedded thermocouple) to increase to a peak of
Continuous Duty Cycle Application :
3.5oC above ambient,
Two capacitors were tested at the relatively
• the 20 minute cool down period did in fact allow the
modest ambient temperature of 70oC (representative of
capacitor’s core to cool all the way back to ambient.
the worst case ambient temperature for the system -
not underhood) at a ripple current equal to the
capacitor’s rating for that temperature and frequency.
The internal temperature of the capacitor was
measured to be 85-90oC. The small (27%) increase in
ESR over the 1500 hour duration of the test is
consistent with results such as those by Harada et al
[3], where the ESR had increased to failure point (2-3X)
after 6000 hours of operation.
IV.2 Test2 - 63V, 2200µF, 105oC Snap-ins for
intermittent duty cycle Application :

07803-4070-1/97/$10.00 (c) 1997 IEEE

IEEE Industry Applications Society
Annual Meeting
New Orleans, Louisiana, October 5-9, 1997

• easier heatsinking of the can (i.e., to a flat

I
30sec heatsink), and hence higher ripple current
capability,
20min
• Operation to 125oC, as long as the can is clamped
t
flat (to prevent bulging and damage to the seals).
TCAP Three “Flatpack” capacitors were suitably clamped and
∆T
tested at 115oC. Note that it is important to design the
115oC
clamp carefully because it experiences significant
t expansion force due to the build up of the internal
pressure within the capacitor and the subsequent
Fig. 4. Typical Profile of Internal Temperature Rise bulging/expansion of the can. Previous measurements
and simulations had shown that 50A of ripple at 15kHz,
As the test progressed, the peak of the ∆T (core- 115oC would result in a peak excursion in the core
ambient) was monitored as an indicator of increasing temperature of about 10oC, thought to be just
ESR, Fig. 5. After 1000 hours, it had increased by 2.7 - acceptable. After 3000 hours of operation, the minimal
3.5X, indicative of excessive ESR, so the test was change in ESR is extremely impressive.
aborted. Before and after measurements of ESR
according to the method described in Section III.3
revealed that indeed the ESR had increased
substantially, Fig. 6., to 6 and 14X its original value. As
the goal was no more than a 2X increase in ESR after
2000 hours, the test was aborted, and the capacitor’s
were declared to have failed.

14

12

10
∆ T(core-amb) C

8
Cap1
Cap2
6

0
0 200 400 600 800 1000 1200
Elapsed Time (hours)

Fig. 5. Change in Core Temperature with Time.

IV.3 Test3- 25V, 20uF, 85oC “Flatpacks” for

Intermittent
Duty Cycle Application :
Within the last couple of years, a capacitor vendor
has introduced a line of “Flatpack” capacitors; where a
rectangular can with a semi-hermetic end seal is used.
Two advantages accrue from this type of construction :

07803-4070-1/97/$10.00 (c) 1997 IEEE

IEEE Industry Applications Society
Annual Meeting
New Orleans, Louisiana, October 5-9, 1997

At Beginning of Test the end seal. The model incorporates relationships for
ESR change with electrolyte loss, ESR change with
temperature, and heat transfer value with geometric
dimensions. Also included in the model is the effect of
can diameter on the expected life. Capacitor life is
proportional to the capacitor volume divided by seal
circumference [2]. As the can diameter increases, the
circumference increases linearly while the capacitor
volume increases with the square of the diameter.

Input Initial ESR

Calculate ESRHot from (2)

ESR = ∆V/∆I = 0.6V/17.5A = 34mΩ

After 1000 hours Calculate DTr using (3)

Calculate TC as per (4)

Calculate P using (5)

Numerically Estimate Vt as given in (7)

Calculate No
ESR using Eq. (1)
Is ESR > Limit
ESR = ∆V/∆I = 3.5V/16.8A = 208mΩ (i.e. 6.1x)

Yes
Fig. 6. Test2- Change in ESR of 63V, 2200µF Capacitor
After 1000 hours at 115oC Done

Fig. 7. Flow Chart for the Model

V. MODEL DETAILS
In [2], a life prediction model for aluminum In this section the model presented in [2] is reproduced
electrolytic capacitors is presented, where the model is for convenience; the methodology takes the following
based on the physics of wear-out mechanism. In [2], approach. First the relationship between the change in
the results from this model are compared against the ESR with the reduction in the volume of the electrolyte
traditional model where the life doubles for every 10oC is established and is given below in Eq. 1.

ESR ESR0 = (V / V0 )
below the rated temperature. The traditional model is 2
(1)
said to give substantially more liberal predictions. The
flow chart for the model given in [2] is presented in Fig. Where, ESR - Equiv. series resistance @ 25 oC in Ω
7. The model is based on the primary wear-out ESR0 - Initial ESR in Ω
mechanism of electrolyte vaporization and loss through V- Volume of electrolyte

07803-4070-1/97/$10.00 (c) 1997 IEEE

IEEE Industry Applications Society
Annual Meeting
New Orleans, Louisiana, October 5-9, 1997

V0 - Initial volume of electrolyte At each time step the new values for ESR, ESRHot, ∆Tr,
TC and P are calculated to establish the new volume of
ESR is inversely proportional to temperature and the
electrolyte Vt for the next time step. These calculations
effect of temperature on ESR is calculated using the
and steps are carried out in the sequence shown in Fig.
formula given below.
7, until the ESR of the part exceeds a certain set limit
ESR Hot / ESR = D + Y • e ( C )
- T /F and at this time the device is considered to have failed.
(2)

Where, D,Y and F are coefficients which are VI. MODELING RESULTS AND DISCUSSION
determined by least squares fit. Using the model presented in [2] and summarized
Next the temperature rise ∆Tr which is a function in the previous section, modeling was carried out and
of the electrical resistance, ripple current flowing the variation of ESR with time was predicted for various
through the capacitor and the thermal resistance is ambient temperatures. Fig. 8 contains the results from
calculated using Eq. 3 given below. the model for the 63V, 35mm, capacitor used in Test1.
ESRHot · I 2 The effect of can size (dia 35mm) was included in the
∆T r = (3) calculations. Also shown in Fig. 8, are the results from
H ·S the two Test1 samples.
where, I - Ripple current in Amps
H - Heat transfer per unit surface area(W/m2
o 0.1
C)
S - Surface Area (m2)

Using this the operating/core temperature can be

calculated as follows. ESR70C
ESR (ohms)

ESR85C
TC = T a + ∆T r (4) ESR100C
o Test1 S1
Where, TC - Core Temperature ( C)
Test1 S2
Ta - Ambient Temperature in Kelvin

The vapor pressure of the electrolyte (in this case

ethylene glycol) is a function of absolute temperature
and chemical properties of the liquid and can be 0.01
expressed as given below in Eq. 5. 0 500 1000 1500 2000 2500 3000 3500

Time (hours)
P =e
{( -A /T ) +B}
C
(5)
where, A = 7060 and B = 21.63 for Ethylene Glycol. Fig. 8. Predicted ESR Change Over Time for Various
Ambient Temperatures With 11A rms Ripple Current
The proposed model in [2] presumes that the rate of
electrolyte loss is directly proportional to the vapor From Fig. 8., it can be said that there is some
pressure of the electrolyte and this is expressed as agreement between the 70oC test (Test1) results and
follows. the model results. However, the model seem to
overpredict to some extent. Please note that the
dV
dt = k • P (6) sample size is small and not enough data is available to
where, t - Time in hrs. draw final conclusions.
k - Leak Rate Constant (units/mmHg/hrs.) Next the 63V, 25mm capacitor used in Test2 was
used to characterize the quality of the seal modeled. The capacitors in Test2 were subjected to an
P - Vapor Pressure of Electrolyte (mmHg) ambient of 115oC and were operated in an intermittent
Eq. 6 is integrated numerically using (7), where ∆t is the mode (30sec ON and 20min OFF). The duty cycle is
integration step (hrs.). so small and hence for modeling, the core temperature
and ambient temperature were kept the same. Fig. 9,
V t = V t -1 - (k • P • ∆t ) (7) contains the modeling and experimental results. It can
be said that the model again overpredicts the life. Note

07803-4070-1/97/$10.00 (c) 1997 IEEE

IEEE Industry Applications Society
Annual Meeting
New Orleans, Louisiana, October 5-9, 1997

that in Test2 the capacitors were operated outside their electrolyte leakage based model presented [2] cannot
temperature rating. Moreover, the capacitors in Test2 be used to predict the life of these capacitors.
were operated in an intermittent duty cycle mode as
described in Section IV.2. In the model presented in
[2], the author has used the vapor pressure data of VII. CONCLUSIONS
Ethylene Glycol for the temperature range from 20 to
Electrolytic capacitors used in a three phase pulse
100oC. This data was used to establish a relationship
width modulated inverter have to carry a pseudo-square
between vapor pressure and temperature. However,
wave of current. A test fixture for the electrolytic
the ambient temperature for Test2 was 115oC. This
capacitor that would inject square wave currents into
could be one of the reason for the difference between
them and operate them in continuous and intermittent
the experimental and modeling results at temperatures
mode was designed and built. In this paper, results
above 100oC. Also it is important to note that under
from high temperature testing of electrolytic capacitors
such high temperatures, apart from electrolyte
are presented. Using a physics based model reported
vaporization and leakage there could be electrolyte
in literature, based on the physical wear-out
degradation. This could also explain why the model
mechanism, the change in ESR with time for various
overpredicts the life.
operating temperatures was modeled. The experimental
and modeling results were compared and analyzed. It
1 was found that the model over estimated the life for the
two cases when electrolyte leakage was the main cause
of failure. The electrolyte leakage (dryout) and
electrolyte deterioration (wearout) characteristics have
ESR70C
to be further understood and models need to be
changed or adjusted to properly account for these
ESR (ohms)

ESR85C
0.1 ESR100C phenomena. Further models should account for
ESR115C continuous and intermittent operation of the
Test2 S1 component. These models are essential for designers
to design and guarantee the reliability/life of the
inverters and converters.

0.01
0 500 1000 1500 2000 2500 3000 3500 ACKNOWLEDGMENTS
Time (hours) The authors would like to thank Dr. John M. Miller
and Dr. Xingyi Xu for their valuable suggestions.

Fig. 9. Predicted ESR Change Over Time for Various

Ambient Temperatures With No Ripple Current REFERENCES
[1] United Chemi-Con, Inc., Understanding Aluminum Electrolytic
The capacitors used in Test3 are “Flatpack” capacitors Capacitors, Second Edition, Richmond, IL, 1995.
and according to the manufacturer these are almost [2] M.L. Gasperi, “Life prediction model for aluminum electrolytic
hermetically sealed capacitors (zero weight loss) and capacitors,” 1996 IAS Conf. Rec., Vol.3, pp. 1347-1351.
hence there is no evaporation of electrolyte, i.e., dry out [3] K. Harada et al., “Use of ESR for deterioration diagnosis of
is not a contributing factor. electrolytic capacitor,” IEEE Trans. on Power Electronics, Vol. 8,
No. 4, pp. 355-361, Oct. 1993.
It is mentioned in [4] that “Wearout” is the best
term to use in describing capacitor end of life and this is [4] L. Macomber, “The truth about thermal design of computer
grade capacitors,” presented at APEC ‘97.
due to chemical changes which occur over the life of
the capacitor. The chemicals in the capacitor material
breakdown and the effective electrolyte resistivity
increases, thereby leading to an increase in ESR and
core temperature. This would explain the results from
Test3, where the capacitors have shown very little
change (<10%) after 3000 hours of intermittent
operation at an ambient of 115oC. Therefore the

07803-4070-1/97/$10.00 (c) 1997 IEEE