----- Original Message -----
From: "Ethan Dicks" <ethan.dicks at gmail.com>
To: "classiccmp" <classiccmp at classiccmp.org>
Sent: Friday, September 13, 2013 5:17 PM
Subject: Repairing Mac128K board - memory problems
  Hi, All,
 A list member recently sent me a broken Mac128 logic board.  I quickly
 found a bad solder joint from a previous repair where one of the
 original 64Kbit DRAMs was replaced with a socket, but I'm still having
 issues.
 The Mac does "bong" and I do get a "Sad Mac" with a hex code.  It
 claims it's the same chip that is now socketed (G8).  If that DRAM is
 installed, I get "Mac in Jail" bars.  If that DRAM is removed, I get
 black pixels where that chip was.  The other odd symptom is, between
 the bars, a semi-regular sparse grid of on-pixels.  It's not a perfect
 pattern across the screen; it's mostly on the right side, and not
 always top to bottom in a column.  It does look regular, as if it's
 every 64 bytes or some other power-of-2, and 5% of the pixels blink as
 if they are changing in software.
 I know that Apple-branded DRAMs of this era are dodgy.  I could just
 clip them all out and replace them with fresh, *tested* chips (I have
 a DIP DRAM tester), but I am concerned that the fault is elsewhere
 since a known-good DRAM always produces a stuck-on bit (the "jail
 bars").  That smells to me like a bad gate upstream of the RAM field.
 Looking at the available schematics, unbuffered DOUT D0-D15 go several
 places - a pair of '244s to buffer the data bus, a pair of '166 shift
 registers for video data out, a pair of '161s for sound data out, and
 a 16R8 PAL named ASG that appears to be related to sound and diskette
 PWM.  The DIN side of the data bus connects to the '244 bus buffers,
 the CPU, the ROMs, the IWM, the 6522 VIA, and the Z8530 SCC.  I would
 think that if that side of the bus had a stuck bit, the CPU would be
 fetching bad data from the ROMs, so there wouldn't even be a Sad Mac.
 I can do some catastrophic repair by blindly pulling all the RAM and
 replacing it, and I can even pull the CPU and socket it so I can stick
 my Fluke 9010A on the board and do some serious memory tests and
 exercise the bus and poke at it with an oscilloscope.  What I'm asking
 for at this point is any information that can help me target my
 efforts.  It seems like this sort of fault would have been common in
 the day, but mostly when I google "Sad Mac" or "jail bars", I get
 stuff about replacing individual RAM chips or changing SIMMs in an
 SE/30.
 So anyone have any advice that could help me not remove and replace
 320 solder joints on a multi-layer board?  Yes, I can do it, but it's
 a lot of time to do it well.  I'd just like to get this old Mac
 working with an original board (I have working 512K boards, but the
 point is to have an Original Mac in time for display next January).
 Thanks for any suggestions,
 -ethan
 
Ethan,
my first move would be to get rid of the socket!
It is extremely difficult to solder a socket to a multilayer board, unless
it is part of the original flow solder process, and if some of the holes are
not through plated (or have lost the plating during previous repairs), the
socket will often only make contact with the tracks on one side of the
board.
I would remove the socket, and directly solder in the chip, taking special
care (magnifying glass at least), to ensure that the solder has flowed
correctly on both sides of the board.
This used to be a problem on a number of pieces of equipment that I used to
look after, and I often found myself repairing faults that had been
generated by the use of IC sockets for repairs.
Jim.