"Something" is not "tracked down." What is the "something"? Which signals were out of specified limits? How are they related to 805x RESET function?


that's not how it's spelled! Look it up!


or, possibly, lack thereof. The principal complaint about the RESET issue when I took an interest in it, years ago, was that there were related oscillator startup problems. I don't know how this tied in to the RESET discussion, but if Vcc rises slowly, and RESET relies on Vcc, what do YOU think happens to RESET? Likewise, if Vcc rises and falls slowly, how would you mitigate the risks? A supervisor doesn't respond quickly to slowly decaying Vcc, no matter how well-regulated it may be. Linear regulators stop regulating at some point. You have to stir that into the mix at some point. Further, switchers don't regulate all that well in the first place, and people are tempted to filter their output with excessive capacitance. That goes into the mix too. Am I barking up the wrong tree? Well, maybe, but I've done some research. I suspect it's more than you've done.


With an electronic power switch activated by a push-on/push-off switch (probably with others too) you can control the behavior of Vcc quite precisely and at no more cost than one of the popular supervisors. The difference is that the electronic switch absolutely prevents the usually excessive capacitance on the circuit board, due to inadequate power supply, from slowing the decay of Vcc. If you stop your MCU oscillator, or pull down Vcc within fewer than ten machine cycles, you should have little "RESET problems." If you raise Vcc within a millisecond of adequate available input to your linear regulator, you should have little startup or "RESET trouble." If Vcc has slow rise and fall times, all bets are off. Don't take my word for it, though. Find one of those boards that gave you all that trouble and try driving it with a moderately powerful external linear supply of, say, 20 amperes. Remove all decoupling caps larger than 0.1 microfarads from Vcc-to-Gnd. There's probably only one. If you now have lots of power-to-gnd noise, the large caps' absence isn't the cause. I'll bet you can't find any RESET problems with that setup.

You once described a problem with a board that the customer complained required power be cycled twice before it would operate properly. That's typical of exactly what I'm addressing. What happens is that an ultra-weak supply, often a switcher, can't drive the on-board Vcc rail up fast enough to give the on-chip oscillator a "kick-start." It also holds the Vcc up when powered down, and if it's powered up again within a short time, it is high enough to allow the oscillator to start because there's already plenty of charge in the big cap's, but less than enough to run the oscillator. I can't guarantee that this is what was going on in your case, but I'll be you didn't check for that.

Unfortunately, that risky voltage interval lies between the specified limits of the MCU and the trip point of the typical supervisor. I've done a little experimentation and found that the circuitry doesn't mind if you take down Vcc to below 0.25 volts within a microsecond of the first bounce on the power-down switch or drop below specified Vcc.

I've also observed run-on of the MCU during active RESET when the supervisor drove RESET after the MCU Vcc was out-of-spec.

I'm not sure what the problem is, but it's certainly not just RESET. I think it's all more dependent on power supply behavior. I've had code-space corruption occur with a power supply that decays slowly, but not with one that decays quickly (< 10 microseconds). This, of course, has to be done with a switch or with something a bit more sophisticated.

I once fiddled with all this stuff when it was convenient to do so, but, since MCU makers don't care, I don't see why I should expend time, resources, and effort on it. After all, you guys don't mind just sticking a finger to the wind and using whatever add-ons the industry salesmen recommend. I don't have RESET problems because, in new designs for my clients, I don't use MCU's that reference RESET to Vcc. If I have an old design that uses an 805x MCU, I control Vcc with an electronic switch. RC reset works just fine under those conditions.

RE