I would like to do a Failure Modes and Effects Analysis (FMEA) on my designs. As a first step, I make a table with every single component in my design, and with every failure mode of that component. Then I fill out what the effects are of every failure mode, and if those effects create a safety hazard. If so, I change the design, or come up with another mitigation.
If I for instance have a LED and a transistor, I would like Kicad to generate a table like this:
Seems that this could get out of hand with a design of 100+ components.
What do you do with ICs?
BTW resistors usually don’t short. They go high, low, open, smoke or explode.
one idea you can do. Add to each symbol you are using failuremodes. For example edit the resistor symbol and add additional fields starting with FMEA_ ( could be “shorted”, “open”, …). Then use KiField. To export the list where the attribute starts with “FMEA_”
So you can do it for every device.
Do there exist commercial applications which are able to do your task?
First, system level fault trees are MUCH more useful as they are top down. FMEA are bottom up, very time consuming. After you do the first level, you need to do module, unit and then system level failure effects.
If you have the top down system knowledge, just do the FTA. Only Mil jobs still waste time and $ doing FMEAs, because of tradition.
Your simple circuit is not typical of modern digital designs where failure effects can be significantly more difficult to determine.
medical designs are small and simple for exactly that reason. No human can reliable analyze a very big design without getting bored and making mistakes. Maybe ngspice can do a crude simulation of the circuit for every failure mode?
It’s too complex to analyze every single failure mode in an IC. I guess it’s sufficient to regard every pin individually like this:
shorted to the highest power rail
shorted to GND
Open
I agree film resisters will go open, but wire wound resistors can fail shorted, as they consist of overlapping coils of wire that are only separated by a thin film of insulating paint.
I agree. That’s one of the reasons why I put a fail safe circuit between any programmable IC and the output towards a patient in medical designs. That fail save circuit is simple enough to be able to be analyzed with the FMEA method. Humans are not capable to write error free programs, and microcontrollers are too complex to prove their failure modes.
