PRISM Forum - Message Replies

Posted by: Jim Ramsden (jimmy.ramsden@lmco.com )
Date posted: Mon Jun 12 15:12:12 US/Eastern 2000
Subject: Hermetic vs Nonhermetic ICs
Message:
In trying out the demo I stumbled upon something that puzzles me. I took a linear IC and changed it from hermetic to nonhermetic. Its failure rate increased as expected. But I also noted that the default thermal rise increased from 13C to 25C. So I played a what-if, and left the part as non-hermetic but forced the thermal rise to be 13C. When I did, the failure rate of the IC decreased from 0.04261 to 0.02897 in going from hermetic to non-hermetic. I was wondering if I am misunderstanding something about the RACRate model? And this leads to a more general question, are the details behind the RACRate models (what the lambda and pi factors are and what impacts them) documented in detail somewhere ? The on-line help refers to Appendices A-G, but I can't seem to find them anywhere with the demo. Is there a way I can get a copy ? Thanks...jrr

Subject: Hermetic vs Nonhermetic ICs
Reply Posted by: David Dylis (ddylis@alionscience.com )
Organization: Reliability Analysis Center (RAC)
Date Posted: Tue Jun 13 14:06:04 US/Eastern 2000
Message:

As you mentioned, performing a PRISM prediction on a linear IC in hermetic and non-
hermetic package styles with the same temperature rise can exhibit failure rates as you
 describe.  I performed a prediction of a Linear IC in a Ground Stationary, Outdoors
 environment using a telecommunications operational profile and accepting all PRISM
 defaults.  The following highlights analysis results:
 
 IC, Linear (hermetic):  0.004088 failures/10^6 hours
 IC, Linear (Non-hermetic):  0.007286 failures/10^6 hours
 
 I then, similar to your analysis, modified the temperature rise for the hermetic and non-
 hermetic to be the same (13C).  This resulted in the following:
 
 IC, Linear (hermetic):  0.004088 failures/10^6 hours
 IC, Linear (Non-hermetic):  0.004014 failures/10^6 hours
 
 The result, though different than yours confirms that the failure rate of the non-hermetic
 does decrease substantially with a decrease in temperature rise and approximates (or may
 be slightly lower than its hermetic counterpart).  Plastic encapsulated versions of the
 same microcircuit will typically have a higher temperature rise due to therma resistance
 differences than a similar hermetic version;  therefore, it would be unlikely that the same
 device operated under the same conditions would experience this issue.  In addition, it
 should be noted that processes used in the selection and quality of components are not
 factored into the failure rate.  These factors are addressed later in a PRISM analysis when
 components are integrated into a system/assembly.
 
 RACRate component models are shown at the top of each component failure rate screen
 in the PRISM tool and demo.  The PI factors used in the calculation of each failure rate
 prediction contributor to the overall PRISM predicted failure rate are also provided.
 Models used in the calculation of PRISM assembly and system failure rates are also
 provided on the failure rate screen for a given component or assembly.