A tale of a chip and a socket

Noel Chiappa jnc at mercury.lcs.mit.edu
Fri Apr 8 16:19:14 CDT 2016


So I just had the incredibly amusing experience of managing to repair an
-11/04 CPU by un-soldering a chip, putting in a socket, and putting _the same
chip_ back in that socket!

Before you go 'WTF?!?!', let me explain what happened.

The CPU wouldn't run, and in poking around, I stumbled on the cause: all the
registers would not 'take' 1's in the 0360 bits. Hmm, 4 contiguous bits -
sounds like it might be a bad register file chip. But before I pulled it, I
wanted to make sure it wasn't some other part of the data path - Mux, ALU,
etc.

So I put a DIP clip on the chip, whipped up a 3-instruction 'scope loop that
would exercise it, and... while I was looking at it, the problem went away!
WTF? So I pull the clip - and the problem comes back. Repeat. Clearly there's
a bad connection in the chip, and the pressure of the clip is 'fixing' it.

So I pull the chip, put in a socket (I always use sockets on repairs, I'm
paranoid I'll overheat the parts - I don't mind living with an potential
eventual bad contact from corrosion), and figure what the heck, let me see if
fiddling with it fixed the bad connection - and sure enough, it now seems to
work!

And if it eventually fails, no problem - it's in a socket, I know where to go
if the machine stops working, those P3101A's are rare and expensive, etc! :-)

	Noel


More information about the cctalk mailing list