28 Apr
2010
28 Apr
'10
4:54 p.m.
On Behalf Of Chris Elmquist
What I have discovered is that the 8355 ROM+IO chip is flakey. It won't
work unless I freeze it with component cooler. Once I have done that,
then it starts working and stays working for hours as long as power is
not removed. If it gets powered down, then I have to freeze it again
before the device will respond with valid fetches.
...
Any guesses as to what the failure mode might be of
the 8355? It is sort
of curious that once it gets started from frozen, it keeps working nicely.
I already did my time at ETA, cooling chips in liquid nitrogen so that's
kind of behind me now ;-)
robo351 wrote:
Hi Chris,
Freeze spray fixes usually give one additional troubleshooting information
but I've never seen a case where the component continues to work until power
is removed.
By way of anecdote, an early Nixie digital multimeter I used to use once
developed a problem: it would be working fine one moment, then, intermittently,
as one was making a measurement it would decide to go into a hard all-zeroes
display. Nothing would bring it out of that state except for powering it off
and back on. Very frustrating to use; struggled for ages (years) ans several
attempts to figure out what the trigger/fault was. Difficult to diagnose as the
various feedback loops in the discrete A/D converter made it impossible to
discern where the problem had initiated once it had been triggerred into the
fixed state. Of course, when one accessed the unit to work on it, it worked
fine and refused to fail at all.
As it turned out, a junction-FET had developed a sensitivity to changes in the
nearby electrostatic or electromagnetic field, or to stray photons: the
"+"-polarity indicator in the display was a neon lamp located within a few
millimetres of the FET, when the lamp switched on the FET would trigger into a
latch-up state. (Accessing the unit for service moved the neon lamp a couple of
centimetres away from the FET, which explained why it didn't fail then.)
So, all in the realm of speculation, but marginal or spurious semiconductor
junctions can do funny things, like exhibiting weird latch-up behaviour.