Once I start adding things, I have to consider how it will impact what I'm trying to observe. An LED will draw considerable power, so it would be inappropriate to power it from Vcc on the board. A typical small LED would want between 4 and 10 mA. I think the 'scope with its 10 m-ohm input will do as well. When I say there's sufficient noise, I mean there's sufficient activity with the trigger voltage set to what should be a valid low level, though just barely so. All I do is set the 'scope to trigger on a positive edge of nWR but low enough that it should cause a write. I haven't yet gone so far as to try to isolate the occurrence of nWR within that Vcc range between when RESET occurs (and should, but apparently doesn't, stop all MCU activity) and when the BBRAM should inhibit all writes. Since I'm presently not so interested in finding bugs in the BBRAM's write protection, I want to do this in a way that points to unwarranted activity on the part of the MCU rather than weaknesses in the BBRAM.

Because of the sort of 'scope I use, I can see the very infrequently occurring nWR pulse when I'm looking. It has a photomultiplier on the display, so I can see events that only occur once or twice per second, even when the instrument triggers an multiple MHz. I haven't studied this very thoroughly yet, but I'll look for exactly where the difference in the BBRAM content after corruption lie and see whether that's where they occur next time. That will potentially shed some light on the problem, though I suspect the activity is pretty random. I'm trying to do something else too, however, so this series of experiments unfortunately won't get all of my attention. I have a logic analyzer that will probably enable me to set threshold levels within the range that occurs during periods of invalid Vcc, and that may shed some light, too. Setups are always a time-consuming task, though, and I want to ensure I'm looking at something that will actually yield useful information. I'm even considering a separate power source for the observation hardware. That may take more time but produce more useable results.

I might find it useful to perform a series of exercises under automation, so I get a statistical report. If I can set up a PC to run the exercise, I can use an ADC to monitor the behavior as Vcc drops. At the same time I can sample nWR, for example, and record where in the decay of Vcc the activity, if any occurs, and, after removing the large (1000 uF and 330 uF) decoupling cap's, I can compare rise times. This will also allow me to track the rise/fall time of Vcc in the entire sequence without having to make individual measurements each cycle.

RE