> I am not an expert on these things, but I believe that all the numbers
> that mChip thrwos about in this data are the raw materials from which
> one can compute an MTBF, if one knows the appropriate incantation.

MTBF computation requires some sort of failure mode prediction or
analysis. If you accept a hardware that is perfectly working then you can
predict a software glitch by using the S/N in the circuit and the
respective logic levels and their margins. Then you can come up with a
number of seconds required for a 50% probability that a logic state will
be misread and cause a glitch, for each pin. The total will refer to all
the pins. So having a very low number of IOs and redundant IOs helps. As
Alan Pearce said, it is related to the number of joints...

With a computer, that can try to re-read inputs and use redundant outputs
this computation becomes very complex. However normal logic circuits stop
here and computers can be more reliable that normal logic because of it
(computers can try harder). Chip internal logic is considered very
reliable (unless it is made or designed by certain manufacturers <g>).

I think that no manufacturer who has the disclaimer about his products not
being usable for 'life support' and 'critical' applications will ever give
a MTBF figure without the lawyers smothering it instantly.

Peter

--
http://www.piclist.com hint: To leave the PICList
mailto:piclist-unsubscribe-request@mitvma.mit.edu