Matthias Pfaller wrote:
>         Here's the additional memory info I promised . . .
>             I'm using 1Mx9 Samsung KMM591000AN-8 SIMMs.  Although the data
>         sheets shed no light on the subject, I determined through
>         experimentation that they will not tolerate /WE activity during a
>         refresh cycle; however, the pc532 DRAM controller allows /WE to change
>         as soon as the CPU tries to begin a bus cycle, even if the DRAM is in
>         the middle of a refresh cycle.  The symptom I observed is that
>         following one of these "illegal" /WE refreshes, all DRAM read cycles
>         generate different data than what was put in to the DRAM on a write
>         cycle -- until a "normal" refresh (with no /WE activity) kicks the
>         SIMM back to normal operation.
>             The "cure" is to gate /WE with /RFCYC.  This keeps /WE disabled
>         during refresh cycles while permitting normal operation on normal bus
>         cycles.
>             To do this I added a new IC, call it U99, in the spare slot next
>         to U20 on the pc532 board.  A 74AS10 is perfect for U99.  U99 will
>         perform the generation of /WE that is currently performed by part of
>         U30.  I tried really hard not to make any cuts to the beautiful PCB,
>         and by bending leads over on one of the ICs I was successful.  Here's
>         what I did:
> 
>             1. Disable U30-11,12,13 by bending these 3 leads over and
>         reinserting U30 into its socket so that these leads are effectively
>         not connected.  I know there are Good Reasons not to leave a gate
>         floating, but it hasn't caused a problem [yet].
>             2. Jumper U99-7 to a convenient ground. U99-14 should already have
>         Vcc present (thanks to George).  If you haven't already done so, give
>         U99 a Vcc-to-ground bypass capacitor.
>             3. Jumper from U19-14 to U99-1  (/DDIN)
>             4. Jumper from U19-23 to U99-2  (/DDINL)
>             5. Jumper from U19-5  to U99-13 (/RFCYC)
>             6. Jumper from U18-9  to U99-12 (/WE)
>             7. Tie unused U99 inputs (pins 3,4,5,9, 10, and 11) to +V (e.g.,
>         U26-1 or U31-1, or any other convenient point).
> 
>             That's it!  I've had no problems at all running this at full speed
>         (50MHz for U25).
>             I can't tell for sure if this is the problem that Stuart and
>         Julian are having, but it sure sounds similar.  Good luck!
>         -----------
>         Mark J. Munkacsy, AJ1B    m2m@sgfb.ssd.ray.com

I decided to install 4MB SIMMS in place of the 1MB. These are 9-chip
SIMMS
but the machine did not even boot.

From the above explanation it looks safe to do it in any case.

I applied the patch and tested the machine with the original 1MB and
it seems to be just fine (BTW. I had to a 74F10 rather than 74S10 which
I think should have similar performance).

However, the 4MB SIMMS still would not boot (meaning no ROM message).

I now suspect that the memory I use may not be of the right king (FPM).
Anyone knows where I can check? In case anyone actually knows, I have
eight SIMMS (I tried 4 at a time in sockets 6/8/10/12, as well as all
eight):

4xSAMSUNG KMM594000B-6
	9xChips: Samsung KM41C4000BJ-6
2xCuBig ???
	9xChips: OKI M514100A-70SJ
2x???
	9xChips: Panasonic MN414100ASJ-07

All have '94V-0' on their PCB.

--
Eyal Lebedinsky (eyal@eyal.emu.id.au) <http://samba.anu.edu.au/eyal/>