| ??? 12/09/08 00:00 Read: times |
#160772 - RESET is not germane to this thread Responding to: ???'s previous message |
Andy Peters said:
I have to chime in with support for Erik's position.
Richard initially said said:
I've recently encountered one 805x circuit with a 1000 uF cap on its Vcc rail, that had unidentified problems which were initially called "reset" issues. By removing the 1000 uF electrolytic cap, I made the problem "disappear" though I'm not yet sure why. Now we don't know any other details here, including possible impedances in series with the micro's VCC rail and whether the design used the really bad 8x51 data sheet RC reset. But if you have 1000 uF or more hanging on your VCC rail, you slow down the VCC rise time and it's possible that this slow ramp is baffling the micro's internal logic -- parts of it come out of reset earlier than others, other parts may not be completely reset, etc. Also RTFDS to see if it has any specific VCC rise-time recommendations. I saw no justification for the large Vcc swamping capacitor, aside from the fact the power supply was probably an order of magnitude too weak. The point I was making, however, was that THIS THREAD is not the place to be hashing out issues regarding RESET, since RESET issues are not germane to THIS THREAD . Anyways, Erik's recommendation of using a reset supervisor for one of its intended purposes -- to ensure that the processor is held in reset until AFTER the power supply has ramped up fully and is stable -- is completely reasonable and should not have been dismissed out of hand! IIRC, Erik's comment wasn't dismissed. It was simply ground-ruled out as being off-topic. This thread is not about devices requiring RESET. It's about voltage regulators. I don't think the 1000 uF itself is bad, although it's likely overkill (it's not a kilowatt audio amplifier). I do think the micro did have "reset" issues which would be completely addressed by a proper supervisor. I find it more desirable to make a problem "go away", whether it's solved or not, by leaving something out rather than by adding something. As I pointed out, the "problem" which a junior designer, no longer in my client's employ, had gotten the idea that there was a "RESET PROBLEM", which he was unable to define or otherwise characterize, from reading someone's comment on 8052.COM, and not from any substantial analysis. Further, I concluded that the 1000 uF cap was entirely superfluous, as I can imagine no circumstance under which a < 10 IC CMOS circuit that doesn't provide power to any off-board loads would need more than, say, 5 uF for Vcc swamping. Once I removed that one component, I was able to run a 1000-hour test on eight boards with nary a failure. -a I've been mystified by the frequent reference to "RESET PROBLEMS" which, so far, nobody has been able to define or characterize with anything other than a handwave. This, particularly, when the only well-defined "RESET PROBLEMS" I've observed have occurred in the presence of a supervisor and what I'd generally consider excessive Vcc capacitance, resulting in runaway MCU signaling during active RESET while Vcc was decaying. I mentioned the capacitor in the context of voltage regulator failures because Vcc decay is often very slow, sometimes over two minutes with modern CMOS circuitry, while sensible rectifier/filter circuits do have bleed resistors on their raw filtered output. When that raw output from the rectifier/filter (or an input switcher) drops out, the linear regulator (which was the topic of this thread) can become reverse-biased, which can often result in damage to the regulator, particularly if a considerable amount of energy is stored in the Vcc swamping cap. I'd be happy to discuss the purported RESET problems people can characterize, but I see no point in reading or listening to stories like, "We had a problem of some sort, so I added a supervisor, and things work just fine," since that fails to verify that the "problem" was a "RESET problem" and does nothing to support the conclusion that, whatever the "problem" was, it was remediated. As time allows, I like to fiddle with this matter, provided there's enough time to devote to an experiment that will lead to a conclusion based on observable misbehavior of the MCU or other circuit components. I'm convinced that if RESET, whether from an RC or a supervisor IC is used to drive a CMOS totem pole on Vcc, which switches Vcc ON within MUCH less than 1 microsecond, and switches it OFF just as quickly, will make the phenomenon of flash corruption and other purportedly reset-related problems disappear. This is just my opinion, however, as I've had little opportunity to prove it, and, certainly, it does nothing that serves the interests of the manufacturers of these faulty IC's, bearing the designation of "805x" to research what the problem really is, else they'd have done it by now. The IC-makers like to recommend the Intel-designed RC circuit, with the caveat that you might need something better, yet say nothing about how, exactly, you'd know that you need to do something about that reset signal. In the meantime, I'm going to stick with my belief that having a positive-going RESET is a serious design error. After all, the only signal that remains fixed at all times is GND. Associating RESET with a changing Vcc, during both power-up, brownout, and power-down, is not a clever thing to do. Please keep in mind that I'm not saying this is absolutely the case, or that anyone believing otherwise is wrong. Nobody's done-and-published anything substantial to investigate the putative "reset problem" that haunts 805x applications, least of all the 805x-chip-makers. However, if a complementary mosfet pair will make the flash-corruption problem go away, such that it doesn't occur through a 10-million cycle test, with a very slow Vcc rise and fall, I'll be happy with that. RE |
