If it happens 1 time in 1000 it's far too often, if it's flash corruption. If it's just a runaway MCU, well, there's probably a reset button, or, at least, th toy can be shut off and powered back up. That, if it's just a toy, is a sparrow. Is that what you design, Erik ... toys? You've repeatedly stated it's not.

If I can make one MCU fail once while keeping all conditions within specified limits, that is, Vcc below the maximum and operating above the minimum, GND above the minimum and below the maximum, and all the signals within specified limits and timing, oscillator within specified limits, then it's an anomaly. If I can make that MCU fail repeatedly under those conditions, then that MCU is flawed. If I can make multiple identical MCU's fail in the same way, then the MCU design and fabrication are flawed, and the entire device, as a product should be rejected.

What I've gathered here is that there are numerous 805x-core MCU's, from different manufacturers and of differing designs, that have failed in similar modes, under conditions perhaps not totally understood. The result has been catastrophic, namely, that the firmware has been corrupted.

The testing to establish what's working and what's not, and under what conditions is not terribly difficult, but it is time-consuming. The manufacturers all have enough money to make this testing happen. If they'd been willing to do it, they'd have published ample reports so we all understood what the necessary precautions are. They didn't do that because they wanted to push their product to market rather than waiting the extra 6-18 weeks to do thorough testing. I'm sure they very thoroughly tested everything that they documented in their datasheets. That's why there's so little information about this particular detail.

Now, YOU say everything magically "works" once you use a 39-cent supervisor. Well, maybe that's so, but how many have you rigorously tested? Yes, I know ... there are thousands of them out there in the field that have no trouble, at least that have no REPORTED trouble. If one of these things happened to be in an A380 full of physicians on their way from the U.S. to Europe for a conference, would YOU like to be the one found responsible for having allowed a part that had a 1 in 100-billion chance of malfunctioning to be used, when the plane went down in the Atlantic because of corrupted firmware? What if the chance is only 1 in 100 thousand? ... or 1 in 100? Until someone publishes firm numbers, how will we know?

YOU say everything "works" when you use this or that supervisor. Not everyone seems to agree, at least to the extent that they use the same supervisor. How do you know yours is the right one? Why don't you use the better one someone else uses? OTOH, why doesn't the other guy use the "better" one that you use? Until there are statistical evaluations made, how will anyone know?

I don't propose to use a bunch of hardware in production. For making a data collection for statistical analysis, I have to have some tools, though.

That's the reason for the RPG.

RE