Compute module

[Tadeusz Pycio]

I was interested in the idea of adding another processor architecture to the RC2014 kit, in the form of a card that would work with the existing one. I once came across this drawing that captures this idea.

I have tried the layout of the required components on the RC2014 PCB and it looks like this could fit in there. The example uses an HD63C09, a slim 64kB RAM, ROM, 74HCT646 and glue logic on an EPM7032/64 (not sure yet how many cells it will use). The processor can be any other 8-bit, it's more about the concept here. An interesting topic to think about.

Happy New Year!

[Alan Cox]

On Fri, 30 Dec 2022 at 10:26, Tadeusz Pycio <ta...@wp.pl> wrote:

I was interested in the idea of adding another processor architecture to the RC2014 kit, in the form of a card that would work with the existing one. I once came across this drawing that captures this idea.

With the 6809E/6309E you can do something even more interesting. The CPU uses the bus only on one half of the cycle. The other half can be used for other things. The 6845 and the 6881 SAM used it for DRAM refresh and video fetch but it's also possible to run dual 6309E/6809E processors or dual 6502 in this fashion.

For coprocessors I built a Z80 coprocessor card with its own 512K RAM and some ROM that talks over a control link to the main host (and the main host can interrupt/reset/etc it). That was based on how some S100 setups worked. I got more interesting results out of using a 1K shared memory with built in interrupt (IDT 7130). With a bit of glue you can make it appear in I/O space on two RC2014 buses at once, allowing you to have at least three CPUs all talking to one-another. It's a small card that sits flat between two rc2014 backplanes. I have built a similar card where the other side of the 7130 is an Arduino but it lacks something in authenticity. 68HC11 would be perfect for that or an 8051 perhaps but 68HC11s are now stupid prices.

The two board joining card was a lot easier - it fits the rc2014 modular structure and it's much easier than trying to cram all the parts onto an RC2014 sized card (the Z80 one ended up a bit oversized)

There were some very neat S100 era setups where the secondary card was an entire system (serial, printer, RAM, etc) and you booted CP/N or CP/M off them with MP/M (or one of the better clones) running on the S100 host to provide storage. A few such cards were dual or even quad client systems on a card.

Alan

[Mark T]

74hct652 might be slightly easier to use as both processors then have separate controls and don’t need to negotiate which one controls the direction.

[Tadeusz Pycio]

Yes, the 74HCT652 seems like a better choice, but I couldn't get it in a slim DIP24, and the normal size doesn't allow me the PCB size. The extra two bits of arbitration will disallow simultaneous access attempts, so I'm left with the 74HCT646. I'm in the process of working out the glue logic, only when the work is complete will I have more knowledge as to whether it makes sense.

[Mark T]

I didn’t realise the 74hct652 was in wide DIP24, the ones I have are SOIC-DW, so again in a wider package than most but not so noticable in smt.

[Tadeusz Pycio]

I know there are 74HCT652's available in slim DIP24's, but I haven't been able to get one, and I don't want to use SMD components as it loses the retro spirit. I will probably soon be forced to use SMD components, they are more readily available and cheaper, but while I still can I am using the through-hole ones. :)

[Bill Shen]

 I think you’ve mentioned 32- or 64- macrocell CPLD for glue logic?  32-macrocell CPLD can easily emulate the 74652 function.  Probably can also include a serial port if you use a 64-macrocell CPLD.
Bill

[Tadeusz Pycio]

Hi Bill,

Here I am more concerned about the number of pins available than the use of the macrocell. A chip with more pins I won't be able to use, so what I can I want to transfer to the 74HCT646.

[Tadeusz Pycio]

Unfortunately, my fears were confirmed, I ran out of free EPM7032 pins. I will still try to follow Bill's advice, transfer the 74HCT646 functions to the CPLD and in the place obtained after this IC I will add GAL 20V8, which should satisfy the number of pins needed. Another option is to use another sub-processor that has less glue logic requirements than the HD63C09, I think a second Z80 would be an easier challenge here.