I'd like to be in the situation where I can confidently order one or more of GARTHWILSON's 4MB SRAM modules. They are lovingly made and the capacitors are affixed and tested individually before the surface mount memory chips are added. However, I don't want to hassle GARTHWILSON and then not use it - or worse - destroy it through my own ineptitude. I presume that I'm not the only person in this situation and this may account for the slow sales of such modules. To overcome this impasse and as a circuit CAD beginner exercise, I devised a stub implementation of the Wilson Mines Co RAM Module which is intended to work with the so-cheap-they're-not-worth-a-socket 50¢ RAM chips. This is not a direct competitor to GARTHWILSON's work. Quite the opposite because the stub implementation is about 3x larger, 1/8 volume, slower and lower quality.

For some applications, one of the cheapest RAM chips will be sufficient. However, it is nice to know that considerable headroom is available, even if it is never used. If system requirements change considerably, it is very re-assuring to know that up to seven additional address lines are available without changing the memory interface.

My first design used 62256 32KB SRAM possibly in an erroneous 0.3 inch width DIP. It is possible to bodge 0.6 inch width DIP into such sockets but it ain't pretty. Nevertheless, it is possible to fit four up to four chips in contiguous memory. With some fudging of the address lines, it may be possible to obtain four chip stacks of four chips in contiguous memory and obtain 512KB SRAM. This dove-tails nicely with the official GARTHWILSON memory module which is available in increments of 512KB up to 4MB.

I suspect that two or more variants may be required for 0.6 inch 32KB RAM, 0.3 inch 64KB RAM and 0.3 inch 128KB RAM. Unfortunately, four 0.6 inch DIP may be larger than 100mm*100mm and this may increase the cost of manufacture. This is important because the stub implementation is very much a cheap 'n' cheerful option with a USD1 board, USD0.50 74HC139 and one or more USD0.50 RAM. It is also DIP only and possible to assemble with a cheap, 15 Watt soldering iron.

Unfortunately, I am unable to test any modules because I don't have a working system which uses a conformant interface. Indeed, it was this dependency loop which partially inspired an alternative implementation. It would be great if someone with a working module could test a stub implementation. You might have to patch software because the stub implementation has partial decoding and therefore anything expecting 512KB blocks of RAM may not function correctly. Alternatively, if you're boot-strapping towards a Wilson Mines Co 4MB RAM Module purchase, a stub implementation may aid prototyping even if it is found to be non-conformant. I hope you appreciate the deliberate angled design which is in the style of StarTrek iso-linear memory. I believe this style is popular with Planck 6502 boards and RC2014 boards.

If my work is conformant with the official interface, I will publish KiCAD files. However, if it is not conformant, I don't want to dilute an established standard.

Finally, if I'm sending one board, I might as well send others. Choose what you want from the full list but try to keep it under 2kg. No charge. Pay it forward.