Date   : Thu, 25 May 2006 08:43:48 +0000 (WET)
From   : pcoghlan@...
Subject: Re: S18 problems?

>
>Thanks to Wouter's rather smashing webpages listing what ROMs I can
>remove in order to help fault-find, I've managed to get this
>model-A-upgraded-to-a-B working ... almost. The only problem I have
>left (apart from a keyboard fault, but I'll take that to bits later)
>is that the Beeb gets confused about how many DFS ROMs are present! It
>can vary between 0 and about 4 - repeated combinations of ctrl+break,
>power-cycling and then *HELP can list the DFS 0.9 a random number of
>times.
>
>The problem /seems/ to go away if I make S18 south instead of north,
>which according to the service manual limits the access speed of the
>sideways ROMs. I say "seems" because the fault is fairly random, and
>I've not spent hours trying to find a definitive way of making it
>happen. As the supplied DFS in this machine is actually on an EPROM
>I'm wondering whether it is an old, slow one. But then, why didn't the
>original owner complain about this? Is it a fault which would have
>gone away with a sideways ROM board (because I believe this computer
>

Hi Kris,

On a bog standard model B without a sideways rom/ram board, each paged
rom appears four times in the paged rom space. For example, the rom in
socket 15 can also be accessed as if it was in non-existant sockets
11, 7 and 3. This happens because only the bottom two bits of the rom
select latch at &FE30 are decoded on the beeb motherboard.  There is
code in the OS which runs at reset and compares each rom image (the
first few pages or the first few kb - I forget which) and avoids logging
the presence of any duplicates found (in the table at &2A1-&2B0?). It
treats the highest socket number a particular image is found in as the 
real copy.

If you are seeing the same rom apparantly in more than one socket, then
the comparison routine is failing to spot that the two images are the
same. I think you are probably right about the cause being an old, slow
eprom. When accessed too fast, this could be returning invalid data
occasionally and messing up the comparison.

The fault would probably less visible with a sideways rom board in that
if &FE30 is fully decoded, there would never be any duplicate rom images.
However, if the comparison routine occasionally reads garbage from the
rom, odds are that normal usage would have too and strange things would
have happened from time to time.

Another possibility that springs to mind is that if a sideways rom board
was previously fitted, it may well have been connected to the beeb
motherboard with a flying lead or two going to one or more of the jumper
locations. If a removed jumper was not replaced or was replaced loosely
after taking away the sidewasy rom board, this might cause random problems
selecting sideways roms.

Is the OS version 1.0 or greater? I vaguely recall that the older OS 0.1
didn't do paged roms or maybe made a half-hearted attempt to do them.

Regards,
Peter.