While I am not an absolute fan of watchdogs and am convinced that their usage must be considered and that NOT using one IS an option, I would also prefer external watchdog over the mostly flawed internal ones, where security is more important than cost. After all, I've expressed already my credo in this regard in more than sufficient (and maybe more than bearable) length in the "reset" pdf (which I still did not finish).

And, Erik, a BADLY WRITTEN runaway code may kick an external dog as easily as a BADLY WRITTEN runaway code may kick the internal dog. Besides, if you strive for perfections, have you heard about window triggered watchdogs?



What is the duration of the reset pulse, then? And, have you verified that the clock is running? (on ALE, perhaps, or, with sufficient care (high impedance probe) directly on the crystal pins)



While I don't claim there is no internal flaw in the logic in the said mcu, there are certainly also mechanisms how to kick a modern CMOS to - temporary or permanent - death. If, as a consequence of some strong pulse, some of the internal CMOS gates latches up (do you know this effect, Farshad?), that gate cannot perform its logic function anymore, and stays dead - either until power cycle, or permanently.

Just a remark - funnily, in CMOS you might fry _some_ of the transistors inside (in any way), and the logic (or most of it) may function normally, although for example the power consumption will be elevated, and such "half-dead" chip is certainly more prone to various obscure errors than a fresh one...

JW


PS. I am eagerly waiting Kai Klaas's comment, IIRC he mentioned subjecting Atmel '51s to strong EMI...