Dear everybody
I am trying to understand. Here is the case:
Some electronic boards work at high temperature conditions as 150-
The failure time estimation using FIDES at limit stress conditions, lF, is expected to be smaller than lREX (lF < lREX) as some special failure modes due to high temperature may be involved, failure modes not priori counted by FIDES (so we should add a certain value lMD to lF in order to include the MD failure modes not counted).
My question is: if the reasoning is correct, why have I found the FIDES failure time estimation lF greater, much greater than lREX (lF >> lREX)? More, lF is outside the confidence interval at 5% risk (95% confidence).
Should I accuse the small number of failures? Should I consider that the failures are "early time failures"?
Can I validate the maximum failure rate from REX ? Have other ideeas?
Thank you.
Sorin
First, this high temperature are outside the normal range of use of FIDES models (this range will be specified in the next version of the Guide).
But also, what kind of technologies are you using at so high temperatures ? The usual components don′t work at this temperatures. Are you using special components, designed or hardened to sustain high temperatures ? In that case, it is normal that the FIDES prediction doesn′t apply.
It is true that FIDES gives a strong influence to temperature. Much stronger for example than the MIL-HDBK-217. But the state of the art (for example JEP 122C) and the very good accuracy of numberous predictions made with FIDES (at more usual temperatures) indicate that this tendancy is right. Nevertheless your case may be very interessant to analyze in details because it can give new idea on this question.
Also remember you should compare your field return (REX) to a prediction at the mean conditions, not the limit conditions. At high temperature the variation may be very big.
Lambda