28 Aug
2015
28 Aug
'15
10:32 p.m.
On Fri, 28 Aug 2015, Lyle Bickley wrote:
On Fri, 28 Aug 2015 20:31:16 -0500 (CDT) Tothwolf
wrote:
So the core mats were more or less dangling? Still, I would think they
would be repairable, but it would be a major undertaking since there would
probably be 1000s of tiny stubs of wire to weld in (every single
connection, basically). All of those enamel wire terminations would then
also need to be soldered to the terminals with either a low-tin or copper
containing alloy of solder too.
I also can't help but wonder what other systems which use core memory
might suffer from this type of failure as they continue to age.
Any chance someone on the team took some photos of the core memory? I
didn't see any on CHM's IBM 1620 webpage.
http://www.computerhistory.org/projects/ibm_1620/ibm1620/
For those interested, here is a paper Google turned up which explains the
problem:
Lead Alloys for High Temperature Soldering of Magnet Wire
https://app.aws.org/wj/supplement/WJ_1975_10_s370.pdf
I think I've answered my own question today
while looking at a
datasheet for another type of solder. My guess is that the solder they
used did not contain any copper, and the tin in the solder IBM used
dissolved small amounts of the already very tiny copper wires, creating
a weak point where the wires were soldered to the terminals.
Lyle, was there any discussion of possibly laser welding replacement
stubs of wire before the breaks that could then be re-soldered to the
terminals? That would seem to be a viable repair option, although it
would require special equipment.
I was not on the Team that did the memory analysis and the ultimate
"modern" replacement memory. However, when I joined the Team, I asked
similar questions and was told that the core memory was literally
"falling apart" and was not repairable.