As you recomended I did the following to fix the GSP TIME-OUT error
that we've been fighting this week:

~ 1_ Power supply reads 5.11 volts on the board.
~ 2_ CR22 & op-amp at 5B have been deleted/removed.
~ 3_ The entire set of E-PROMS have been re-burned using fresh parts off
     the shelf.

Unfortunately, the GSP TIME-OUT error still crashes the program when 
going in/out of operator screen in self-test and in the attract mode
it eventually hangs.

I
'll be eager to try any more suggetions you might think of.  Bonnie and
I are kinda puzzled with this problem and need some help.

Farrokh

This is what I know so far. I am distributing it so others can comment on it.

The Big Sympton is the GSP Timeout Error.

This is definitely caused by the GSP program crashing.

The GSP program crashes because, after powerup, one or more of the VRAMs
do not work, and nothing can be done in software to make them work. (The
GSP Simple Memory Test identifies the miscreant VRAMs.)

If power is removed and reapplied, sometimes it will work and sometimes it
won't.

I do not know of anything that will cause a VRAM to permanently not work
in a circuit until power is removed and reapplied, except for Latch-Up.
(This is when the voltage on an input is greater than the IC's VCC. This
triggers an SCR that is part of the gate structure but is not a problem as
long as the input does not get higher than the IC's operating voltage.)

Things that have been tried:

  1. I made Steel Talons Self-Test ROMs with the GSP Reset time extended 
     to 2 ms. Made no difference.

  2. I had Jim remove the +15V and -22V from a board at his test station; 
     maybe it was causing the VRAMs to latch up. Made no difference.

  3. As per Ken Williams' memo, when the Samsung VRAMs are replaced with 
     Micron VRAMs, there does not seem to be a problem.

  4. Mary has checked around; only Samsung and Hitachi still make 64Kx4 VRAMs.

  5. Glenn brought a boardset back and we fired it up on a bench with a
     variable supply. 

     When we brought the voltage up slowly it screwed up every time. 

     We kept the supply on and switched it with just wire leads. It didn't 
     screw up. 

     We returned to bringing it up slowly. It had stopped screwing up. A short
     time later the VRAM in one particular position started screwing up every 
     time. 

     If, with the board powered, I remove the VRAM and plug it back in, it 
     will usually work.

     Replacing it with a Micron VRAM fixed it.

     Swapping the Samsung VRAM with another one from a different location 
     seems to fix it, at least temporarily.

  6. We removed the 0.1 uF bypass capacitor associated with one of the VRAMs
     and measured it. It was, indeed, 0.1 uF. We added a 1 uF tantalum 
     capacitor to the area. Made no difference.

  7. I thought that maybe there was a problem with the plated-thru holes to
     +5V and Ground. I measured the resistance between +5V on every VRAM and
     a +5V Testpoint with a low resistance Ohmmeter and they seemed ok. I did
     the same with the VRAM Grounds. I also did it with the Bypass Caps. They
     all seem solidly connected.

The VRAMs are now being put in sockets and I think this is a good idea.

I would like to find out if all the MultiSync Boards are made by the same
vendor. If there are several vendors, is there a correlation between the
Problem and the board vendor?


Jed

The Multisync boards are being made by two different vendors and both have
failed for the same reason. Apparently no correlation between the problem
and the board vendor.