31 Oct
2010
31 Oct
'10
11:51 a.m.
On Sun, 31 Oct 2010, O. Sharp wrote:
Date: Sun, 31 Oct 2010 14:22:35 -0400 (EDT)
From: O. Sharp <ohh at panix.com>
Reply-To: "General Discussion: On-Topic and Off-Topic Posts"
<cctalk at classiccmp.org>
To: "General Discussion: On-Topic and Off-Topic Posts" <cctalk at
classiccmp.org>
Subject: Re: Repairing core memories....
On Sun, 31 Oct 2010, Jos Dreesen wrote:
If you look at core planes, many have repairs visable in even in "new" planes,
probably ones that failed margin tests. I looks like the way cores in the
middle of the mat are replaced is by snipping a bit of the drive and sense
wires a couple of cores away from the faulty core, replacing the core and
threading in a bit of new wire for the drive and sense lines, with enough
slack to solder the splices above the plane. You can see the little splices in
many core planes (usually covered with a tiny drop of glytol or some such)
Peter Wallace
So I have these 2 PDP-8/L core stacks I am trying
to recover:
One would be perfect, if bit 3 @ adr 0 would be alive...
Although this is 99.99% OK, it is of course not good enough.
Just to mke sure I'm hearing you right: Are you saying one individual core is
100% dead, and everything else is 100% okay? If so - if it's one single,
individual core which is broken - I'd suspect this is pretty much
irrepairable. The options would be to A) try to form/mold/create a new
ferrite core around the wires, which would be insanely difficult assuming it
could be done at all, or B) to unwire that plane of the core stack, replace
the broken ferrite core with a new one, and then rethread/rewire it. That's
extremely nontrivial work, I have to believe. But they were originally
threaded by hand, albeit by people with microscopes and _very_ good eye-hand
coordination and sewing skills, so it's within the realm of possibility.
If I've read you wrong, though, and the whole thing is working except address
0 bit 3 is working _some_ of the time, that would be really good news. :)
You could try scoping out the sense amps and inhibit drivers for bit 3 and
see if they were just borderline to specifications, something which might be
just tweaky enough to affect one bit.
The other stack seems to be a total loss :
2 sense wires are open circuit, more than 20 select diodes shorted. Of
course the further quality of the cores is unknown.
Oddly enough, this one might be easier to deal with. Have you tracked down
where the sense wires go open-circuit? If they're failing at a solder-joint
near an outer edge, you could potentially repair them without having to open
up the whole damn stack. The diode replacement would likely mean partially
disassembling the stack assembly to gain access to both sides of the diode
PCB, but at least you're only having to dig in one level.
O'course making those repairs might simply reveal more problems at the core
level. But you've gotta start somewhere, I s'pose. :)
Is there any realistic way of getting one fully
functional stack out of
these ?
Removing a single core from the really bad stack to the almost OK stack
would seems almost feasible to me, since address 0 is bound to be on a edge
of a core mat.
I suspect I'm not the only one on the list who:
-thinks opening up a core-stack and repairing it is
theoretically possible;
-also thinks it would be a hell of a daunting project;
-is somewhat amazed at the dexterity and patience of the people
who originally hand-wired them at manufacture; and
-thinks pulling off a repair of a core-plane by rewiring it by
hand would give significant bragging rights. :)
-O.-