PDA

View Full Version : 8088 1MB inline memory expansion board



eeguru
May 29th, 2011, 02:36 PM
This is a reposting of a post I made over on Mike's PCjr Forums. Thought it might have more interest here since there's more visitor traffic.

-Alan.

-------------------------------------------

Speaking of theoretical, after thinking of this thread a while, I wondered if it would be possible to design a board that literally fit into the footprint of an 8088? For systems (not just the PCjr) that were ether space constrained around the processor socket or had limited or no ISA expansion slots for RAM cards (eg. most early Tandy 1000 systems and IBM 5150s). So I tried it. I can say it's theoretically possible, though to actually build it would require about $500 just for PCBs (qty 1 to 50+ doesn't matter due to NRE) since it's 4 layer, has an abnormally high via density, 6/6 tech, and small to the point you lose a lot of efficiency on panels.

It's basically the exact footprint of a 40-pin DIP socket (2.2" x .7") plus a couple millimeters overhang on the front edge for three 2mm jumpers to allow changing of the RAM map after assembly/programming. At build time, the code can be configured for a minimum mode system (eg. PCJr, Tandy's, etc) or a maximum mode system (most PC/XT clones). And the jumpers provide 8 possible memory fill maps each with 32k resolution. Basically they feed into a lookup table along with A15-A19 with the RAM chip selects as outputs. They then get gated with the control strobes and routed to pads. The 8 memory maps could be customized on a order/build basis by changing the lookup tables before programming the PLD. If not, I can think of a few generic ones that would be useful.

It has 2x 512KB SRAMs on the top side of the board that sit in the reliefs provided by this wire wrap socket (http://www.sparkfun.com/products/8118). On the bottom side, there are a couple latches, a buffer, and a 22V10 PLD. Everything is TSOP so the wire wrap pins could be cut down a bit so it adds less than a half inch to the processor height. Though with all the chips stacked, heat might be an issue - even though everything is HCMOS.

I put together a rough BOM to try and price things out if there were enough interest in a production run. It's all SMD/SMT so it would not be user assembly friendly Though it would be an open source hardware project - so you could free-build. The total parts cost including PCBs at a 25 unit run with some NRE factored in (Kapton stencils, solder paste, TSOP programming adapter, etc), is ~$45.00. I don't have a lot of time to program PLDs, assemble, clean and package boards, so I would tack on $25 (not including shipping) to help mitigate some of that; which is ~2/3 cheaper than most web assembly quotes I found and those only cover assembly. Not to mention fronting $1200. So if there is interest... and I can't imagine there would be much.. post here and maybe we can get enough people together to give me a warm and fuzzy about selling at least 20 of the boards. Otherwise, PM me for the Eagle files and PLD code if you are interested. It was more of a mind exercise than a serious idea anyway.

PCB Layout (http://www.alanlee.org/8088ram/8088RAM.jpg)
Schematic (http://www.alanlee.org/8088ram/inline_ram.pdf)
Bill of Materials (http://www.alanlee.org/8088ram/BOM%20Inline%208088%20RAM.pdf)

-Alan

ibmapc
June 2nd, 2012, 06:49 PM
Hmm.. You've got my interest. So could we map say 128k starting at D0000 to use as UMB and the rest as LIM 4.0 EMS? Of coarse, with D000 thru EFFF used for UMB There wouldn't be a lot of space outside 640k for the page frame unless EGA or VGA is not used. But I bet it could be squeezed in somewhere. Anyhow, count me in. Let me know what you want in the way of deposite/downpayment or full payment for one maybe two boards.

eeguru
June 2nd, 2012, 07:53 PM
EMS never was a possibility on that board. Straight memory hole fill in only. And due to some layout issues, it was minimum mode only - so really only PCjr's and some Tandy's.

It never really went anywhere as the holes on the original board were too small for the wire wrap socket. I found some Mill-Max double male sockets, however they are really hard to come by. And then your post above is the only interest in more than a year. I did order three boards and fully built one up to test. It does work, though the PCjr need was addressed by the side-car HD/RAM/ROM/RTC/POST expansion board I did (which are still for sale).

Oddly enough I am working an EMS card tonight. Though I don't have an ETA on completion.

ibmapc
June 2nd, 2012, 09:25 PM
Oops. I miss read the date on the original post. I thought it was May 29 of this year, not last. I guess I missed it last year, must have been on vacation or something. Is the EMS card the same one mentioned HERE (http://www.vintage-computer.com/vcforum/showthread.php?30185-Universal-Multi-I-O-Controller-amp-RAM-Expansion)? I'm still interested in that project!

MikeS
June 2nd, 2012, 10:00 PM
...It never really went anywhere as the holes on the original board were too small for the wire wrap socket. I found some Mill-Max double male sockets, however they are really hard to come by. Just curious since I have and am building a few similar adapters/shim boards and the links to your PCB seem to be dead: wouldn't the wire wrap socket pins also badly deform the socket they'd plug into? And alternately, how can you replace a WW socket with a double-male 'socket'? Myself, I'm a big fan of machined pins & sockets...

eeguru
June 3rd, 2012, 12:28 AM
Yes. Usually they are pretty fat and the spring tension on the MB socket wouldn't return. The Mill-Max would have worked out nicely but nearly impossible to find. Similar to these:

151-10-640-00-009000 (http://www.mill-max.com/images/products/pdf/053N.PDF)

But with upward pins clipped level to the top of the PCB and a surface mount DIP-40 socket reflowed on the top side of the board. It's just not cost effective in both cost and labor.

The original files are rehosted below:

PCB Layout (http://www.retrotronics.org/8088ram/8088RAM.jpg)
Schematic (http://www.retrotronics.org/8088ram/inline_ram.pdf)
Bill of Materials (http://www.retrotronics.org/8088ram/BOM%20Inline%208088%20RAM.pdf)

The same idea could be done for min or max mode more efficiently now - especially using only one chip. However a card solution is probably a better bet.

MikeS
June 3rd, 2012, 11:50 AM
The Mill-Max would have worked out nicely but nearly impossible to find. ...
But with upward pins clipped level to the top of the PCB and a surface mount DIP-40 socket reflowed on the top side of the board. It's just not cost effective in both cost and labor.Ah, OK; I wondered how you'd connect the male pins to the socket. The way it's usually done, especially with IC pinout converters, is to have the pins beside the socket and connected on the pcb, but I'm sure you know that.

But if you want to use those boards: what I do when they have to be in line, instead of male-male headers I just solder loose machined IC socket pins on the bottom of the socket as I solder the socket to the pcb, effectively extending the socket pins; it helps to insert the pins into another socket first to hold them in place.

Chuck(G)
June 3rd, 2012, 12:36 PM
I use the Chinese male headers for the purpose. The socket ends are small enough to fit into a machine-pin socket:

http://www.codemsys.com/eb/machinepins.jpg

Cheap and useful for other things.

eeguru
June 3rd, 2012, 01:40 PM
Where's a good source to pick those up Chuck?

pearce_jj
June 3rd, 2012, 02:07 PM
I wondered if it would be possible to design a board that literally fit into the footprint of an 8088? For systems (not just the PCjr) that were ether space constrained around the processor socket or had limited or no ISA expansion slots for RAM cards (eg. most early Tandy 1000 systems and IBM 5150s).

I'm not following exactly what we're talking about here - a processor socket module that replaces the CPU with a board containing the same CPU and some RAM is it?

MikeS
June 3rd, 2012, 02:11 PM
Yeah, that's what we use when the pins are offset (the left adapter below); note that they're also available in a socket configuration like the one on the right. When a low profile is critical as for instance in a laptop adapter I designed, I use pins from a component carrier with the tops cut off after soldering as in the top one below.

But when the pins have to be in line with the socket I still prefer female-male pins.

9196

Chuck(G)
June 3rd, 2012, 02:29 PM
Where's a good source to pick those up Chuck?

ebay. You can get a lot of them for not very much. Search for "male round header". I usually get a combo bundle--lots of male and female (socket) headers. Very convenient for prototyping using SMT ICs.

eeguru
June 3rd, 2012, 04:29 PM
Thanks! And to borrow a thought from a happy puppy like me, I love you! Can I lick your face? :)

@ pearce_jj, yes.

eeguru
June 3rd, 2012, 05:28 PM
PS, are there any that have surface mount flat tops or butterfly L's leads? Some thing that doesn't break the top copper? It appears the board in the top center of MikeS' photo has discrete pins.