My previous blog described risk mitigation for tin whiskers
using conformal coating. Now I will
continue on the subject of tin whiskers by looking at some software that
attempts to quantify the reliability risk of tin whisker failures. I have found two software packages that
provide some assessment of potential risk due to tin whisker growth: the CALCE
tin whisker calculator (1) and the Pinsky tin whisker assessment calculator (2)
from Raytheon Analysis Laboratory. The
CALCE model actually gives a quantitative reliability assessment while the RAL
spreadsheet only gives a 1 to 10 relative risk assessment.
As an exercise I applied the CALCE tin whisker calculator to
an electronic module. This assembly uses
several larger QFPs but is not considered a leading edge high density
module. The CALCE calculation was
performed after entering the appropriate input parameters. These parameters include how many pins of each
device, the lead spacing are various lead dimensions. Using the calculation software the predicted
reliability after 5 years was only predicted to be 12%! That is, 88% of the modules of this design
are predicted to have failed due to tin whiskers. The actual assembly analyzed was already 5
years old and no tin whiskers were observed on the module. This seems to suggest that the prediction
calculator is overly pessimistic in giving a life estimation. The one positive aspect of being overly
conservation is that if the predicted reliability is good then you are standing
of firm ground. The downside is that I cannot really use the predicted lifetime
to make risk decisions.
The second assessment algorithm is the Raytheon Analysis
Laboratory algorithm by David Pinsky.
This calculator uses an Excel spreadsheet and is not as detailed as to
the input variables. For example it does
not asked for exact spacings between pins.
Interestingly it does require inputs for variables such as tin plating
thickness, tin annealing history (if any) and forced or convective
airflow. This calculator is more suited
to single component analysis rather than system analysis. Therefore only one component was analyzed: a
204 pin QFP. In the CALCE calculator the
probability of any two pins shorting was 1.05% which means that the probability
of any pair of pins within the 204 pins failing is high. The Pinsky calculator does not give a
quantitative assessment but rather a risk number between 1 and 10 where 10 is a
very high risk. For the 204 QFP, the
risk number is 8.41 which is considered a high risk. This could be considered in accord with the
pessimistic CALCE prediction.
The specific module that I analyzed has been in the field
for over five years and no known tin whisker failures have been observed. Furthermore examination of modules that were
several years old have been examined with no evidence of tin whisker
growth. Therefore both the CALCE and the
RAL calculators should be viewed very conservative. While there could be some value as a relative
risk comparison between modules, these models cannot be used to accurately
predict field reliability and cannot be input into reliability prediction
programs. Considering this, it appears
that more research needs to be done in order to develop these models
further. The CALCE calculator is more
precise but the tin whisker growth rates must be modified and it needs to
include some of the input variables from the Pinsky method. As of the date of
this blog posting I am not aware of a planned major revision to the CALCE
calculator. Tin whiskers have been
studied since the 1940s (4) but it does appear that more papers need to be
written and more degrees awarded. Then we can move on to study zinc whiskers!
References
No comments:
Post a Comment