Re: CBM-II solderless memory expansion

Den Fri, 13 Apr 2018 17:46:27 +0200 skrev Michał Pleban
<lists@michau.name>:
> Hello!
> 
> Gerrit Heitsch wrote:
> 
> > A 4464 cannot be put in place of a 4164, so I don't quite see how to
> > make that 'solderless'... Could you provide a bit more detail?
> 
> The high-profile board has 128 kB soldeded and sockets for additional
> 128 kB RAM. The board with 4464's would go into these sockets.

You might aswell go for 41256 or 44256 then. 128k soldered on to the
motherboard for bank 1 and 2, 3 sets of either 8 41256 or 3 sets of 2
44256 will fill 12 additional ram banks, i.e. bank 3-14.

I don't know if there is any reason to fill bank 0 aswell. It might be
a good idea to use 1mbit 1 bit wide (or 4mbit 4 bit wide, or an 1MB 30
pin simm) to fill all of bank 0-14.

With some additional patches to the motherboard it could be possible to
fill the parts of bank 15 that's emtpy in an unexpanded machine aswell.
That patches could be included in a 6509 replacement board.

To be able to produce something at all without my usual terrible
feature creep, just leave some inputs to a GAL/CPLD connected to some
pull-up resistors and a pin header (or a pin header which can hold
jumpers while not being used for some "bank 15" add-on).

With a circuit using an 1MB simm or 1mbit/4mbit dram chips it would
also be a repair for machines with broken on-board DRAMs.

So, there would be two PCB's. One that sits on top of the ram sockets
and holds some other kind of ram. I'd suggest making layouts for both
30 pin simms and for different types of dram. Even if there will be
no shortage of some specific type of dram, it would be nice to be able
to use what's at hand. And then there would be a PCB that sits in the
relevant PLA socket.


P.S. there is another way of doing this, and that involves an
additional set of MUXes and some card that both sits in the expansion
socket and the coprocessor socket, and mimics what the muxes does on
the motherboard. With fast enough sram instead of dram it might be
possible to let the 650x and the 8088 have access to that memory almost
simultaneously, i.e. the software sees memory from both CPU's but in
practice the accesses of course takes place sequentially. Unfortunitely
the 8088 board is clocked on it's own 15MHz oscillator. This would be
more feasable if the 8088 used the same oscillator as the rest of the
computer. By looking at the schematics and at the 8284 clock generator
it seems like the 8284 could easily be replaced with a 74xx74, the
existing timing generation on the B motherboard and maybe some simple
gates. The 8284 can do a bunch of things with the timing for an 8088,
but in a B machine it seems like it only divides the oscillator with 3
and makes sure that the reset and ready signals are synchronized to the
clock signal. If the motherboard oscillator would be used, the 8088
would run at 6MHz instead of 5MHz. Maybe some circuit doing this could
be added to an updated 8088 card replica. But this is my usual feature
creep :)


-- 
(\_/) Copy the bunny to your mails to help
(O.o) him achieve world domination.
(> <) Come join the dark side.
/_|_\ We have cookies.