| ??? 10/04/07 17:22 Read: times |
#145403 - maybe you can ... Responding to: ???'s previous message |
Erik Malund said:
Richard Erlacher said:
Apparently you don't read. I've been saying that supervisors are inadequate BECAUSE they don't manage the power-off transient. The MCU, actually two of them, from different manufacturers, that I observed clobbering BBRAM did it (a) during Vcc decay, and (b) while the MAX1232 drove its RESET pin.
Clearly, at least from those cases, the MCU didn't respond as desired to its RESET pin. Of course, since Vcc was slowly decaying, but not yet to the point at which the BBRAM's protection circuitry locked out nCS and nWE, this was due to slow decay of Vcc. That slow decay is caused by too much post-regulator capacitance on Vcc. It's got to decay in a microsecond or two, and it's got to rise in a very few (<10) milliseconds. I can recreate this with some or all of the below: 1) use a supervisor that (tolerance or spec) kick in AFTER uC Vcc min is passed. 2) put a 'noise suppression' cap on the reset line. 3) have LC or RC filtering of Vcc for either or both. 4) have the trac.. ehm this is Richard the wrapped wire to the Vcc decoupling for either or both too long. Vcc pins generally aren't wired, but, rather, are soldered directly to the Vcc plane. Have a look at that wire-wrap card on my "personal page" http://www.8052.com/users/richard/Small%20WWB.JPG The dry-film mask is easily removed in small portions near the supply connections, so one can bypass each Vcc and GND pin to GND or Vcc with an 0603 10 nF cap. Vcc and GND impedance is very low, inductance is very low, and capacitance is distributed. Vcc to GND noise seldom reaches 50 mv (1%). 5) I'm sure there is more
Yes, you can 'prove' the earth is flat if you use the wrong setup. Erik I'm kind of thinking that the supervisor needs to live upstream of the voltage regulator, and signal reset whenever the voltage on the regulator input goes a mV or so below the required margin (2.2 volts with a 7805) +.2 volts. Perhaps it should also suppress the regulator output and clamp it to GND whenever it does that. When a brownout ends, or the critical input to the regulator is reached, under those conditions, an RC power-on reset might actually be sufficient. RE |
