A poltergeist in my machine?
jnc at mercury.lcs.mit.edu
Tue Sep 8 05:53:09 CDT 2015
> From: Jerry Weiss
> Check if the -5v charge pump is stable and supplying the correct voltage.
> I recall a problem ... in which the on board pump went slowly bad.
> Depending on the data content and tolerance margins for the memory
> chips, we saw very erratic problems.
That doesn't sound like my problem; this board was _consistently_ failing
hard (would not write _any_ data, no-how, no-where), and then *poof* it was
pretty consistently totally working (passing memory diags, etc).
And the change from 'consistently not working at all' to 'working completely'
happened when I 'touched' (see below) that pin with a probe. So either i) it
was pure coincidence (possible, I suppose), or ii) there was some wierd
> From: Jon Elson
> 1. There was a conductive hair on the board, and you knocked it off
> with the probe.
I'd brushed the board pretty comprehensively with a toothbrush to get all the
dust off. (Yeah, I know, static - but it's pretty humid here right at the
> 2. There was a bad solder joint and the pressure of the probe broke
> through the oxide. ... This is my highest probability guess. ...
> 3. The chip has a bad internal wire bond, and the pressure on the lead
> made it work.
Well, I was using a DIP clip, so the pressure on the pin was pretty
consistent before and after. Yeah, attaching the 'scope probe to the second
pin would have wiggled the clip a tiny bit, but I'm still dubious.
My current best guess, after sleeping on it, relates to the fact that the
'magic' pin was an output from a delay line. Delay lines, in that era, were
apparently potted confections of inductors and capacitors. So maybe the extra
current drain with the probe on somehow affected one (or more) of the
capacitors in the delay line? A total WAG, but it's very mysterious!
More information about the cctech