18 May
2009
18 May
'09
5:51 p.m.
Tony Duell wrote:
There's nothing inherently bad about putting a crowbar circuit on a 12V
supply, but in a microcomputer, the 5V rail is the 'important one' in
that if it goes high it wipes out just about every chip in the machine
whereas the 12V rail is used by the RAM and not a lot else. So it makes a
lot of sense to sense the 5V rail for any protection circuits (like the
crowbar).
Having the crowbar short out the 12V rail (rather than the 5V one) is
quite common in that the former can supply a lower current so it's easier
to short out, and shorting any output will shut down a SMPSU.
This PSU seems to have a crowbar circuit (SCR
Q5, zener CR20, etc), but
amazingly it seems to be connected to the 12V output. This is, IMHO,
ridiculous.
I have a few ideas as to why it could be considered a bad thing, but
I'd be interested to hear more details! When a PSU output sits at about 1V, I often
suspect that a
crowbar is triggereing, which will shut the supply down, thus enabling
the crowbar to turn off, the cycle then repeats. So what you see is an
average of the 'too high' and 'shorted' output voltages.
OK - I'll check the outputs with the oscilloscope when it's in that
mode, and see if I can see a waveform that looks like that. If the output capacitors have high ESR (a likely
problem due to old age),
there will be spikes on the output, sometimes around twice the normal
I've grabbed new replacements for all the capacitors, apart from the big
high-voltage input-side ones. One of the output capacitors has shrunk
its heat-shrink wrapping quite considerably, so I'm guessing that one
has been running quite hot. Might as well replace them all though - the
replacements only cost me NZ$10 (something like 3 pounds, maybe US$5, at
a guess) so it's hardly breaking the bank.