General question about failure analysis

Hi everyone,

I currently have a device (an alarm) that is being sold and I have 54 of them in the field. Everything is going well but my customer called me this past Monday and said one of them didn't work. When I got the unit back, I found the 16F688 was shorted. This wasn't installed on anything yet, he was doing his functional test to verify the correct operation so I am kind of ruling out noise and other things like that. All of the units are soldered professionally by Screaming Circuits and of course they are all designed the same. Should I just chalk this up to infant mortality? I don't "burn" them in but rather do a quick functional test before they ship, which takes less than 2 mintues to verify the correct operaton of the alarm. Like I said, this is the first failure I have had. Any help is greatly appreciated.

Thanks,

Chris

I've been involved with designing products that use PICs for around five years... in that time I've seen close to half-a-million products shipped. The products that have come back with a dud PIC I can count on the fingers of ONE HAND, and each one was attributable to a cause (invariably the installer). So the actual number of returns attributable to Microchip is ZERO.

We have seen duds that failed to program... but that doesn't count. However, in each case, those PICs were purchased from a 2nd-tier distrubutor or third-party reseller and we can't vouch that they hadn't played with them before we got them. We've had ZERO DOA's on PICs bought directly from Microchip.

I would suspect your client had been playing with the PCB or sticking his sticky staticky fingers where he shouldn't. This always assumes you've got a solid bullet-proof design and your programmer is half-decent and doesn't abuse the PIC to start with.

For your own piece of mind, check which pin(s) are shorted, check your software and TRIS settings that drive those pins (ie if it's an INPUT pin, ensure you're not accidentally setting it to OUTPUT) and look through the schematic connections to that pin to ensure you've not designed something silly that could cause an overcurrent, overvoltage or external short. If the pin is driving an inductive load, ensure you have protection clamping any possible back-emf's. Once you've satisfied yourself that guilt can't be laid at your doorstep, suspect the client did it... call CSI in your town and let them solve it (they usually solve any mystery however improbable within an hour - including commercial breaks)!

Melanie,

Thanks for the reply. I currently don't use a protective coating over the PCB when it's finished being soldered. Do you think this may have contributed to the failure? I am guessing it would be a good idea to spray a coating over the board before it is shipped. What do you think?

Thanks,

Chris

Most of our boards don't have any conformal coating, so that should not have any effect on reliability.

This really is a wide-open question... namely "What caused the PIC to fail"?

1. Was it a Microchip manufacturing defect? Most Unlikely but still could be...
2. Was it a poor hardware design? Only you can answer that.
3. Was it a software conflicting with the Hardware design (eg pin set to Output when it should be an input)? Again only you can answer that one.
4. Was it mishandled in shipping? Probably unlikely, these things would tend to blow an input port, rather than cause an output short.
5. Was it the customer poking test probes or screwdrivers around the board? Well was it?
6. Was it a customer miss-wire that he's denying? Most of mine fell into this category!
7. Was there some ambient condition that caused the failure? By ambient, static or EMP or high levels of localised high-power RF or excessive damp... But those again tend to blow-up input ports and not short outputs...
8. Has the customer taken out the PIC, tried to read the program and put it back in the socket initially the wrong-way round so Power Supply pins now don't line up and destroy the pins they are now connected to... Yes, I've had this happen too... if it's on a socket, someone, somewhere will unplug it sometime...

Well, there's the list... it's probably one of those... but which one?

Melanie,

Thanks for the reply. Since it's my first failure, I am not too worried about it but I am concerned. To replace it, it costs $1.25 plus my labor of 5 minutes. I'll just keep an eye on the problem to see if it happens again. As a side note, they have miswired the positive and negative leads but I have protection for that. The unit works fine after this happens. I'll chalk this one up to the customer damaging it because it worked fine when it l tested it.

Chris