Path: utzoo!attcan!uunet!mcsun!unido!rwthinf!beauty!windy
From: windy@beauty.informatik.rwth-aachen.de (Andrew John Stuart Miller)
Newsgroups: comp.sys.atari.st.tech
Subject: Re: stacking 256k simms
Message-ID: <3340@rwthinf.UUCP>
Date: 20 Aug 90 15:30:47 GMT
References: <2190@gould.doc.ic.ac.uk>
Sender: news@rwthinf.UUCP
Lines: 68

tmp@doc.ic.ac.uk (Trevor Peacock) writes:

>Has anyone tried (or even considered) using blocks of four 256k simms to give
>1Mb banks? Of course, this would require demultiplexing address line A9 into
>two bits to select one of the 256k blocks within the 1 meg bank. Has this
>be done? Is the ST's chippery able to handle the loading caused by the 
>simms being connected in parallel? (this would be a total of 8 per bank for a 
>4 Mb system, two groups of eight in parallel)  
>I'm sure this is a minefield of capacitance and timing problems, but would 
>like to know whether anyone has had any success.

>Trev..

>------------------------------------------------------------------------------
>Trevor Peacock,		 | Tel : 071 589 5111 X 5052 | JANET : tmp@uk.ac.ic.doc 
>Department of Computing, | DARPA : tmp@doc.ic.ac.uk 
>Imperial College,        \ or      tmp%uk.ac.ic.doc@nsfnet-relay.ac.uk
>180 Queens Gate,London SW7 2BZ |    UUCP : tmp@icdoc.UUCP or ..!ukc!icdoc!tmp


This has been done in the German 'ST' magazine, using DIPS rather than
simms. A freind of mine over here has implemented this solution - on an
ST520+ we had to use a small electric saw to remoce the old on board
chips, meaning we that we could then desolder the individual pins, rather
than the whole chips. This significantly reduces the risc of dammaging
the motherboard.  we could then solder in modified sockets, and then
insert the piggy backed dips. Some flying wires were needed to give
adequate earth shielding, which proved to be the downfall of this
experiment. The system ran OS9 with one user on a terminal for about a
week without any problems. As soon as the holidays were over and the
owner of the ST began to log in on the console, which was in a room
locked during the break, problems with timing began - We thought there
might be standing waves in the earth lines, which being wire, as
apposed to perfect conductors are of course not at earth anyway! The
fact that the system worked single user was because only a small
address range was used, and refresh hardware could cope with the
problem, whatever it was. As soon as multi user (multi tasking)
resumed, the standing waves or whatever canceled the refresh signals
or something allong these lines...

We replaced the Earth wiring with shorter connections, and more of
them, and the problem went away... for about two weeks. The 520 has its
Store ("memory" to lusers memory is cache/associative store. Store is
anything else) directly beneath the keyboard. As time went on, the
connections worked loose and all the wonderful work literally fell to
pieces. Two sessions of soldering on the modified sockets allowed the
soldered modifications to work loose, adding to the death of this
otherwise nice modification to an ST. 

We then bought a locally made addon board, and transfered the chips to
it. This has been working perfectly ever since (4 months) despite the
computer having been moved several times.

I will post more details about the memory expansion board as soon as I
can find the comercial address of the firm in that heap of unsorted
papers called my desk.

If you want more details about hacking it yourself, send Email to me,
and I will try to get a copy of the construction details of the 'ST'
magazine circuit to you.

Happy soldering, and dont let Farraday catch you unawares!!

Andrew Miller
--------------------------------------------------------------------------------
email: windy@strange.informatik.rwth-aachen.de
snail: Ruetscherstr 165  D-5100 Aachen
voice: 0049 (0)241 894-355