You are on page 1of 5

Accelerated Temperature Cycle Test and Coffin-Manson Model

for Electronic Packaging


Helen Cui, RF Micro Devices

Key Words: Accelerated Reliability Testing, Temperature Cycle, Coffin-Manson Model, Activation Energy

SUMMARY & CONCLUSIONS Product life could be significantly affected by the


maximum temperature (Tmax) the temperature change (∆T)
Temperature cycle profiles at various stress levels were dwell time and ramp rate. The larger of Tmax, ∆T, dwell
investigated for accelerated reliability testing of electronic time, and ramp rate indicate the higher stress level.
device packaging. Failure Analysis was conducted for test Temperature Cycling tests are used to characterize product
failures to determine their root cause failure mechanisms and capability and to detect unknown failure modes (e.g., die
failure modes. Weibull analysis was conducted for failures crack, via crack) during technology development, product
with the main failure mechanisms (such as solder fatigue, design verification, and product qualification.
mechanical cracks). The Coffin-Manson model has been used Often, the stress level of the test temperature profiles far
to model crack growth due to repeated temperature cycling. A exceeds the product field application stress level. This leads to
good correlation was obtained between the Coffin-Manson accelerated test. Test results at the high stress level need to be
model and the test results, and Activation Energy (EA) was extrapolated to the low stress level, i.e., field application stress
determined. The limitation and application of the Coffin- level. It is critical that the main failure mechanism remains
Manson model are further discussed. the same for the high stress level and low stress level. For
The main failure mechanism determined by failure new technology (or designs for which the field failure
analysis was Via Cracking in the package substrate. The mechanisms are unknown) it is critical to conduct failure
acceleration factor between different stress levels was analysis to determine root cause failure mechanisms and
determined by Weibull analysis of test data. The Coffin- failure modes.
Manson model was used to correlate the test data and
Temperature
determine the activation energy related to via crack failure
mechanism. The Coffin-Manson model with determined
Activation Energy could be used to estimate product reliability
under different application conditions for the same main
Test Duration
failure mechanism. Tmax

1. INTRODUCTION Ramp Rate

Today’s electronic packaging continues to shrink in size


and reaches higher packing density and higher reliability.
∆T
Accelerated test is needed to save test time and cost and
reduces cycle time to market. Various temperature cycling
profiles are employed to evaluate the effect of stress on life
and to detect the unknown failure modes.
Tmin
Dwell Ramp Dwell Ramp
The temperature cycle profile can be characterized by Time Time Time Time

• High extreme temperature (Tmax),


Figure 1 Temperature cycle profile schematic
• Low extreme temperature (Tmin),
• Temperature change ∆T, ∆T = Tmax - Tmin
• Ramp rates, In this study, a 6mm x 8mm electronic packaging module
• Dwell times at extreme temperatures. with three die was tested with variety of temperature cycling
profiles for accelerated tests. Via cracking was verified
Figure 1 shows a schematic of temperature cycle profile. through temperature cycle tests. Weibull analysis was
conducted on failures corresponding to the via crack failure

RAMS 2005 556 0-7803-8824-0/05/$20.00 ©2005 IEEE

Authorized licensed use limited to: Universite du Quebec en Outaouais. Downloaded on February 14,2021 at 16:50:03 UTC from IEEE Xplore. Restrictions apply.
mechanism. The acceleration factor between the high and low
stress level temperature cycle was determined.

2. ACCELERATED RELIABILITY TEST

2.1 Temperature Cycle Profiles

Three temperature cycle profiles were employed:

Temperature cycle -1 (TC1) -40 to 85C Crack


Temperature cycle -2 (TC2) -40 to 125C, and
Temperature cycle -3 (TC3) -40 to 150C

Industry standards, such as JEDEC Standard JESD22-


A104-B Temperature Cycling, and other customer
specifications were used a references and guidelines for these
temperature cycle profiles. The dwell time at temperature
extremes and the ramp rates for each temperature cycle profile
were as listed in Table 1. Figure 2 Typical via crack
• TC3 (-40 to 150C) test conditions has the highest stress
level, while TC1 (-40 to 85C) test conditions has the lowest 2.3 Weibull Analysis
stress level.
• TC2 (-40 to 125C) test conditions has a very short dwell The Weibull failure distribution was plotted using the data
time, which could significantly reduce creep development at points from via crack failures. The Weibull distribution is
higher temperatures. given by [1]:
• TC1 (-40 to 85C) test conditions have minimum transition β
times, or much higher ramp rate then the TC3 and TC2 test F(t) = 1 - e -[(t-γ)/η] (1)
conditions. Where

2.2 Failure Criteria F(t) = % failed (the probability of failure) at “t” number
of cycles or hours.
Parts were electrically tested after certain number of γ = location parameter.
cycles. A device was considered a failure if it does not meet β = shape parameter or Weibull slope.
the criteria in the referenced electrical test specification. η = scale parameter or characteristic Weibull life.

• Step 1 - Electrical test: Electrical test was performed at Weibull++™ 6 [2] software from ReliaSoft was used to
each read point. If a device did not meet the electrical testing plot the Weibull charts.
requirements, it was considered an electrical test failure, and Only a few effective read points were available in this
was sent to Failure Analysis. study, since continuous monitoring was not available. All
• Step 2 - Failure Analysis: Failure Analysis was conducted parts needed to be removed from the temperature cycle
for electrical test failures to determine root cause failure chamber for the electrical test to be performed.
mechanisms and failure modes. For “Test-to-failure” tests data point collection, at least 5
o If a device has “via crack” failure mechanisms, it is read points are suggested for Weibull analysis.
considered a failure.
o If device has other failure mechanisms, it is treated as 2.4 Results
a suspended data. (Note: There was no other failure
mechanisms found in this study. The Weibull plots for the test data and analytical
o The devices that passed electrical test at the end of estimates of Weibull life are shown in Figure 3.
reliability tests were treated as a time suspended data. The Acceleration Factor is defined by the ratio of the
lower stress test mean life to the high stress mean life. Based
In this study, 120 devices were tested under TC1 (-40 to on the test data, the Acceleration Factor AF of TC3 -40 to
85C), 120 devices were tested under TC2 (-40 to 125C), and 150C vs. TC1 -40 to 85C is 20.
320 devices were tested under TC3 (-40 to 150C). The
failures and time suspended data points are listed in Figure 3. 2.5 Discussion on Temperature Cycle profiles
Failure Analysis was conducted for test failures to determine
root cause failure mechanisms and failure modes. Based on TC1 and TC3 both have dwell times of 30 minutes.
Failure Analysis, the main failure mechanism was “via Dwell time is critical for effective temperature cycling tests
cracking”. Figure 2 shows a typical via crack. [3]. The TC1 and TC3 test data were used here to estimate the

RAMS 2005 557 0-7803-8824-0/05/$20.00 ©2005 IEEE

Authorized licensed use limited to: Universite du Quebec en Outaouais. Downloaded on February 14,2021 at 16:50:03 UTC from IEEE Xplore. Restrictions apply.
acceleration factor. ∆TL = ∆T of low stress level temperature cycle
TC2 has dwell time of only 1 minute. A 1 minute dwell fL = cycles per day at the low stress level temperature
time will result in much less damage than a 30 minute dwell cycle
time due to less creep development, especially at high fH = cycles per day at the high stress level temperature
temperatures.
cycle
The TC1 ramp rate (42 C/min, acting as thermal shock) is
TKL = max absolute temperature at the low stress level
much higher than the TC3 and TC2 ramp rate (10 - 14 C/min).
A higher ramp rate has more damage than a lower ramp rate. temperature cycle, in ºK
TKH = max absolute temperature at the high stress level
3. COFFIN-MANSON MODEL temperature cycle, in ºK
For this study, a=-1/3 and b=1.9 are used.
3.1 Coffin-Manson Model
3.3 Determine Activation Energy
The Coffin-Manson model has been used successfully to
model crack growth in solder due to repeated temperature Activation Energy EA is the most critical parameter.
cycling. Coffin-Manson model has typically considered 3 Activation Energy EA is specifically related to certain failure
factors, maximum temperature Tmax, temperature change ∆T, mechanisms and failure modes. Activation Energy EA was
cycling frequency f. It has been successfully used for determined by correlating thermal cycling test data and the
mechanical failure, material fatigue or material deformation. Coffin-Manson Model. In this case, the following steps were
This model takes the form [4] used:
Acceleration Factor AF using test data
N = A · f -a · ∆T -b · G(Tmax) (2) AF_Test TC1 vs. TC3 = 4365 / 217 = 20
Where Acceleration Factor AF using Coffin-Manson model
AF_Model TC1 vs. TC3 = (∆TH/ ∆TL)b · (fL/fH)-a ·
N = the number of cycles to fail exp((EA/K) ·(1/TKL – 1/TKH))
f = the cycling frequency
∆T = the temperature range during a cycle Set AF_Test TC1 vs. TC3 = AF_Model TC1 vs. TC3, then
A = coefficient Activation Energy was determined to be: EA=0.42.
a = cycling frequency exponent, typical value is around -
1/3 3.4 Discussion on Coffin-Mansion Model Limitations
b = temperature range exponent, typical value is around 2.
G(Tmax) = exp((EA/K)*(1/Tmax) is an Arrhenius term The Coffin-Manson model gives a semi-empirical
evaluated at the maximum temperature Tmax reached estimation only. It only gives a trend.
in each cycle. The Coffin-Manson model only consider the effects of
maximum temperature Tmax, temperature change ∆T, cycling
K = Boltzman’s constant 8.623 x 10-5 eV/K
EA = Activation Energy frequency f.
The Coffin-Manson model considers the effects of dwell
time and ramp time using frequency of usage f. It cannot tell
Activation Energy EA is the most critical parameter. It can be
the difference between dwell time and ramp time. Dwell time
determined by test. and ramp rate are critical factors in temperature cycle test.
Dwell time is important for creep development, especially at
3.2 Calculate Acceleration Factor high temperature.
The Coffin-Manson model cannot tell the difference
Calculate Acceleration Factor AF using Coffin-Manson Model between Temperature Cycle and Thermal Shock. Ramp rate is
Term 1 Term 2 Term 3 the key difference between Temperature Cycle and Thermal
b -a
AF = NL/NH = (∆TH/ ∆TL) · (fL/fH) · exp((EA/K) Shock. Thermal Shock has much higher ramp rate than
·(1/TKL – 1/TKH)) (3) Temperature Cycle. A higher ramp rate has more damage
where than lower ramp rate [5].
Term 1 = effects of temperature change
Term 2 = effects of frequency of usage 3.5 Discussion on Coffin-Mansion Model Application
Term 3 = effects of the maximum temperature
(determined by Arrhenius model) The Coffin-Manson model with determined Activation
AF = acceleration Factor Energy (EA) can be used to estimate Acceleration Factor of
NL = the number of cycles to fail at the low stress level temperature cycles at different stress levels; to determine
temperature cycle whether one can meet a new customer requirement (such as a
NH = the number of cycles to fail at the high stress level new temperature cycle profile and cycles requirement) and/or
temperature cycle to correlate temperature cycle tests versus field usage for the
∆TH = ∆T of high stress level temperature cycle same failure mechanism. Figure 4 shows a schematic of

RAMS 2005 558 0-7803-8824-0/05/$20.00 ©2005 IEEE

Authorized licensed use limited to: Universite du Quebec en Outaouais. Downloaded on February 14,2021 at 16:50:03 UTC from IEEE Xplore. Restrictions apply.
reliability analysis flow chart including baseline study of 5. Agarwal, R., Tuchscherer, L., Cui, H., “Thermal Cycling
accelerated reliability test and data analysis, and application of and Thermal Shock for FCOB (FCOB - Flip Chip On
the Coffin-Manson model for the same main failure Board) Testing”, ASME Interpack Conference, 1999
mechanism.
ACKNOWLEDGEMENTS BIOGRAPHIES

Thanks to Chris Calabro, Danielle Wiebking, Craig Helen Cui


Callahan for initiating and coordinating the tests, to Preston RF Micro Devices
Scott for the Failure Analysis, and to Terri Gilbert, Wade 7628 Thorndike Road
Cole, Steven Smith for providing customer related Greensboro, NC 27409
information.
e-mail: hcui@rfmd.com
REFERENCES
Helen Cui is a senior reliability engineer in Reliability and
1. Kececioglu, D.B., Reliability Engineering Handbook, Failure Analysis Lab at RF Micro Devices. She received her
Prentice Hall, Inc., Englewood Cliffs, New Jersey, Vol. 1, B.S. (1988) and M.S. (1991) in Mechanical Engineering from
1991 Xian Jiaotong University, and M.S. (1995) in Reliability
2. Wiebull++TM 6, ReliaSoft Corporation, Tuson, AZ, 2002. Engineering from University of Arizona. She has more than
3. Zhai, J.C., “Board Level Solder Reliability vs. Ramp Rate 10 years working experiences in Reliability Engineering. Her
& Dwell Time During Terperature Cycling”, IEEE 41st experienced areas include Reliability Testing, Accelerated
Annual International Reliability Physics Symposium,
Life Testing, Failure Analysis, Reliability Data Analysis,
2003
Product Qualification, and Electronic Packaging.
4. Engineering Statistic Handbook, www.itl.nist.gov

Temperature Cycle Tmin Tmax Dwell Time Ramp Time Ramp Rate Cycle Frequency of
Profile (C) (C) (min) (min) (C/min) Duration usage
(min) (cycles per day)
Temperature Cycle TC1 -40 85 30 3 42 66 22
Temperature Cycle TC2 -40 125 1 17 10 36 40
Temperature Cycle TC3 -40 150 30 15 13 90 16

Table 1 Temperature cycle profiles

Data Set Failure Time Suspended Slope Characteristic Location Mean Life
Suspended at Cycles Parameter Weibull Life Parameter γ (Cycles)
β η
Temperature Cycle TC1 9 111 500 1.20 4644.7 17.7 4365
Temperature Cycle TC2 11 109 750 1.82 2915.8 0 2592
Temperature Cycle TC3 z 78 242 114 1.86 224.9 0 217

Table 2 Data for Weibull plot of temperature cycle

RAMS 2005 559 0-7803-8824-0/05/$20.00 ©2005 IEEE

Authorized licensed use limited to: Universite du Quebec en Outaouais. Downloaded on February 14,2021 at 16:50:03 UTC from IEEE Xplore. Restrictions apply.
Weibull plot of temperature cycle test data
99.9
99.0
90.0
70.0
50.0

10.0
% Failed

5.0

1.0

0.1

0.01
1.00 10.00 100.00 1000.00 10000.00

Cycles

Figure 3 Weibull plot of temperature cycle test data

Reliability Test and Data Analysis – Baseline Study


Reliability Test (Test to Failure)

Failure Analysis to determine the main failure


mechanism

Test Data Weibull Analysis

Determine AF using Test Data

Determine AF using Coffin-Manson Model

Determine Activation Energy EA for the Main


Failure Mechanism

Application of Coffin-Manson Model for the same Main Failure Mechanism

Estimate AF of Determine whether we can Correlate temp


temperature cycles at meet a new customer cycle tests vs. field
different stress levels requirement (a new TC profile usage for the same
and cycles required) failure mechanism

Figure 4 Reliability analysis flow chart

RAMS 2005 560 0-7803-8824-0/05/$20.00 ©2005 IEEE

Authorized licensed use limited to: Universite du Quebec en Outaouais. Downloaded on February 14,2021 at 16:50:03 UTC from IEEE Xplore. Restrictions apply.

You might also like