Activation Energy Concept
Activation Energy Concept
There is probably no more crucial total of 692 devices were tested at 125°C, and the total
issue than Reliability at the system device hours were 867,520. Failure Rate and MTBF are
level to both the component vendor given by:
and the systems house, and surprisingly there is very limited No. of Failures
FR = ---------------------------------------------------------- (EQ. 2)
understanding of the mechanisms that yield to failures. This Total Device Hours
paper sets forth simply what the electronic Design Engineer
needs to know with regard to calculating a given 1
MTBF = ----------------------------------- (EQ. 3)
Failure Rate
component's Mean Time Between Failure (MTBF), Failure
Rate (FR), Failures In Time (FITS), and what all this jargon Since we had two failures in 867.5K hours, the FR is 2.3 per
means. million hours, and the MTBF is 433.8K hours (at 125°C).
Suppose we wanted to know what the FR and MTBF would
Reliability Overview be at 25°C. Using Equation 1 and assuming that the
Contrary to popular opinion, all integrated circuits begin Activation Energy is 1eV, we calculate an Acceleration
“dying” the moment they are born, and in general, raising the Factor of:
average junction temperature will result in increasing the -------- = e – ----------- – -----------
R2 EA EA
R1 KT2 KT1
failure rate. The Arrhenius relationship which is common in
many physical and chemical processes has been found to fit 1.6E-19 1 - – ----------------
= e – ------------------------ 〈 ---------------- 1 -〉
the failure rates in IC's as well. Equation 1 expresses the 1.38E-23 398°K 298°K
relative failure rate for temperatures, T1 and T2, and the = 17, 698
ratio, R2/R1, is often referred to as the Acceleration Factor. Then we multiply the 125° MTBF of 433.8 hours by 17,698 to
obtain 7.7 billion hours and corresponding FR of 0.13 per
-------- = e – -----------
R2 EA
EA – ----------- (EQ. 1)
billion hours.
R1 KT2 KT1
Where: Another often heard term is FITS which stands for Failures in
Time and is defined as the number of failures per billion
R = Failure Rate hours. For the example above, FITS is equal to 0.13.
EA = Activation Energy
(typically 0.5eV to 1eV) The Real World
It turns out that the foregoing analysis isn't quite right, and
T = Absolute Temperature (°K)
the reason is that our calculation was based on a relatively
K = Boltzmann's Constant (1.38 × 10-23) small sample of devices. To prove the point without thinking
about it too much, suppose that we had observed zero
One very significant issue is the assumed Activation Energy.
failures in the earlier example. That would lead us to the
Illustrated in Figure 1 is the Failure Rate for 0.5eV and 1eV
false conclusion that the MTBF was infinite and the FR was
as a function of temperature. Significantly, at junction
zero. What do we do now?
temperatures above 100°C, the failure rate at 1eV is 1,000
times that of 0.5eV. Fortunately, we can turn to Poissan statistics to bail us out.
And we all thought that the statistics course in school would
Reproduced in Figure 2 are the reported Activation Energies
never be of any benefit. Equation 4 predicts the probability
for various kinds of components. At Elantec, we use 0.8eV to
(of failure), P(X), of finding X failures in a sample whose
1.0eV, which is best suited to the kinds of processes that we
average failure rate is A.
employ.
–A X
In order to calculate MTBF we will also need to obtain
e A-
P ( X ) = ----------------- (EQ. 4)
X!
reliability data from the IC vendor. Virtually all manufacturers
routinely run life tests on devices which span their product P(X) = Probability (of failure)
line and package repertoire. Life test usually means placing X = Failures observed
devices in a burn-in oven, under power, at temperatures
A = Average number of failures
which are typically set at 125°C for 1,000 hours or more.
Shown in Figure 3 is data taken from a slice in time for a Suppose we ran a large number of such life tests (say 1,000)
variety of devices manufactured by Elantec. As indicated, a which actually had an average failure rate of 3.12. Figure 4
summarizes the statistics of Equation 4. The table predicts
that 4.4% of the time (or in 44 life tests) we would observe no
1 CAUTION: These devices are sensitive to electrostatic discharge; follow proper IC Handling Procedures.
1-888-INTERSIL or 321-724-7143 | Intersil (and design) is a registered trademark of Intersil Americas Inc.
Copyright © Intersil Americas Inc. 2004. All Rights Reserved. Elantec is a registered trademark of Elantec Semiconductor, Inc.
All other trademarks mentioned are the property of their respective owners.
Application Note 1104
failures, 13.7% of the time we would observe one failure, θJA obtained from the datasheet is 95°C per watt which
21.5% of the time we would observe 2 failures, etc. We could results in a junction temperature of 116°C. The datasheet
say that 39.6% of the time we would observe either no states that the maximum allowable junction temperature is
failures, 1 failure, or two failures. Probably a better way of 150°C, so the application is okay from that point of view.
looking at this data is that in 1,000 tests we would anticipate
The life test circuit for the EL2044 indicates that the test is
observing more than 2 failures 60.4% of the time. Therefore,
done with 15V supplies and essentially no load at 125°C
if we ran only one test and observed 2 failures, we would
under “ambient” conditions; hence, the life test data reported
have to say, “with a confidence level of 60.4%, that the
by Elantec would be under these conditions. The average
actual failure rate is 3.12.”
power dissipation using Equation 5 would be 228mW, and
We should use 3.12 average failures in all of our calculations the corresponding junction temperature per Equation 8
instead of our 2 observed failures, and we should always would be 147°C.
add, “to a 60% confidence level” to all the numbers we
We can now examine the predicted impact on MTBF and FR
quote. So, our experimental data from Figure 3 boils down to
resulting from operating the junctions at 116°C. Using the
an MTBF of 867,520 hours divided by 3.12 or 278,051 hours
Arrhenius Relationship we derive an Acceleration Factor of:
at 125°C to a 60% confidence level.
AF = R2
--------
Fortunately, we don't have to go through all this convoluted R1
reasoning each time we want to make calculations because = e – ------------------------ ----------------- – -----------------
1.6E-19 1 1
the statisticians have calculated fudge factors for us which 1.38E-23 389°K 420°K
are summarized in Figure 5. = 9.0
Note that this solves our “zero observed failure problem” by The corrected (for finite sample size) MTBF with a 60%
assigning 0.916 average failures to the case of zero Confidence Factor that we calculated earlier was 278K
observed failures to a 60% confidence level. hours. To obtain the “worst case” MTBF, simply multiply by
9.0 to obtain 2.5 million hours with a corresponding Failure
On the other hand, 60% confidence level doesn't sound very
Rate of 0.4 per million hours.
confident. If we wanted to be more conservative, we could
use a fudge factor from a 90% confidence level. Now our 2.5 million hours seems like a long time, but presumably
3.12 average failures become 5.3, and that makes our there could be many devices in the system. If, for example,
failure rate and MTBF look a lot worse. Most semiconductor there were 100 amplifiers, we would expect an MTBF of
manufacturers have historically used 60% confidence levels. about 34 months.
2
Application Note 1104
FIGURE 1.
Activation Energy
REPORTED EA (EV)
FAMOS Transistors
Charge Loss 0.8
Source: Burn-In, F. Jensen, N. Petersen, Wiley and Sons, New York, 1982
3
Application Note 1104
0 1 0.044157 0.044157
1 1 0.137770 0.181927
2 2 0.214921 0.396849
3 6 0.223518 0.620367
4 24 0.184344 0.794712
Intersil Corporation reserves the right to make changes in circuit design, software and/or specifications at any time without notice. Accordingly, the reader is cautioned to
verify that the Application Note or Technical Brief is current before proceeding.
For information regarding Intersil Corporation and its products, see www.intersil.com