OT: Having a FIT getting PIC FIT.


This is mildly related to PICAXE but off at a slight tangent... that's my excuse :)

Does anyone have a link to details of Microchip FIT / MTBF data for PICs?
I already have Microchip's web-published reliability documents but they don't define exactly what was tested and how - merely a vague paragraph.
Microchip technical can't (won't?) supply details of the nitty-gritty. (Yes, I asked). I'm going to do a Stan 'Sigh'.

Has anyone seen any documents that can fill in the blanks?

Note: I'm just after info about the PIC (and obviously PICAXE) hardware reliability, not EEPROM/Flash erase/write endurance or the potentially monstrous task of code testing. i.e. theoretically the easy bit ;)


Thanks Technical.
I have that one and it gives some impressive figures... I'd just like to know what tests they did but Microchip not very helpful.
I'm not sure it's thorough enough to control a nuclear reactor (just kidding).
I'll just set one going and be back to you (statistically) in a few millennia..;)


New Member
I am an R&A engineer (Reliability & Availability) so I'd be happy to offer some insight if I can.

Microchip does what most component manufacturers do to estimate the failure rate of their components. They run HTOL tests (High Temp Operating Life). In these tests, hundreds to thousands of components are operated at high temperature, and sometimes higher than normal voltage, to "accelerate" the failure mechanisms. Then they use some equation to calculate the acceleration rate and then use that to estimate the failure rate at some nominal operational temperature. In this case they use the Arrhenius equation with an Activation Energy of 0.7.

Statistically with a large number of failure free device hours, one can estimate what the maximum failure rate would be, to some confidence level, at an unaccelerated temperature. Most manufacturers provide an unaccelerated failure rate estimate at 55C (ambient, not case or junction) and 60% confidence. Using the same Activation Energy one can recompute that down to a typical operational temperature for your application.

For an entire PCB, this is done with all components on the BOM, then added up. A single IC like this typically has a failure rate of around 1 to 10 FIT. A typical motherboard might be 1000 FIT.

In order to reduce the amount of test time, or the number of components in test, manufacturers use what I consider to be excessive and unrealistic acceleration factors. Microchip is no exception. IMHO anything over around 50 is getting into unrealistic territory. Microchip uses around 250. But as I say this is pretty common. If I am analyzing a board with a Microchip component I would use one of the more conservative numbers.

I do applaud Microchip for providing their data on their web site. Many do, like Maxim or TI and others. But some are very secretive about it.


I'm sure things have progressed quite a bit since I was in the industry (~15 years ago) but in my day ANY semiconductor (used for critical purposes) was deamed to have a life expectancy of ~30 years. This figure was derived from the time it would take for the dopants and impurities to migrate through the substrate.


Thanks Revolvr, I have seen their test summaries.

But, there lies the slight rub; define "operated"....
Microchip use the word 'exercised'- probably a test procedure which may, or may not, include running code.
If so, what did it do?
I didn't get the vaguest clue from Microchip technical.
And it's not an easy DIY job.

Beaniebots; There are some monster FIT times around - they use a some glue now to hold the dopants in longer.
"Lies, damned lies, and statistics" ;)


So... srmet's next satellite is an interstellar probe...

It spends 50,000 years at about 3K before reaching the warmth and light of the destination star.

What's the chance of the batteries being recharged and the PicAxe waking up - "Hello denizens of Alpha Centauri! We bring you greetings from Planet Earth!"

What's going to stop this working? I suspect the battery. How could we make it work?