• Please review our updated Terms and Rules here

IBM EGA Card Memory Expansion Boards

20200321_155517.jpg20200321_155554.jpg20200321_155903.jpg20200321_162218.jpg


thanks to ab0tj for a great design! I went off reservation with my standoffs, just used some crap I had lying around so I didn't have to drill out the expansion PCB

notice how the expansion board has a tiny amount of bowing? might want to pull the standoff holes away from the pin header by just a fuzz, like a couple thou
 
Awesome! I'm glad to see it worked for you. I think I might need to rebuild mine but with the same header you used. Yours has much better fitment than mine.
 
I built one of these a few weeks ago. Worked perfectly first try!

The mounting holes are just a hair off, but my mounting screws still make it. I didn't make any adjustment to the layout. I don't have a bend issue with all the dram's sitting flush. The mounting holes are only a minor issue, so the board is perfectly usable, and I have extras here.
 

Attachments

  • 20201003_201935b.jpg
    20201003_201935b.jpg
    185.6 KB · Views: 3
  • 20201213_115930b.jpg
    20201213_115930b.jpg
    177.7 KB · Views: 2
  • 20201213_121446b.jpg
    20201213_121446b.jpg
    79.5 KB · Views: 2
I managed to get an original NOS EGA expansion card from a friendly chap who had scored some from Computer Reset. Thought I might chime in to help with whatever info I can add.

The male pin headers are approximately 0.4" / 1cm tall, from the PCB surface to the tip of the pins. That's about 0.3" / 8mm of actual mating surface, with the height of the plastic shroud consuming the ~0.1" remainder. It's a 64-pin header in 2x32 arrangement. When you're looking at the component side of the card, with the header along the right edge of the card, pin 1 is the top right corner. Odd pins on the right side, even pins on the left.

FWIW, my expansion card came with a plastic guide installed over the pin header. It looked, at first, like a female-to-female coupler, but there are no contacts in it -- just plastic. Might just be there to protect the pins from getting bent in shipping. (Nice touch!)

The standoffs have a 1/2" board-to-board spacing. That is, if you snapped the expansion card on to the EGA card, it would be 1/2" between the PCB surfaces. The expansion card holes are roughly 4.65mm in diameter, give or take a bit due to measurement error. My calipers say the neck of the standoff (where it rests in the PCB when installed) is about 3.6mm. NOTE: The EGA card side of the standoff is smaller in diameter.

Best I can tell, the standoff holes are 3.5" C-to-C on the left side of the board, and it looks like 5.6" from the standoff center to the center of Pin 1 of the header. Pin 1 center is about 0.3" from the top board edge, and the opposite side stand-off center is about 0.2" from the top board edge.

Board size is roughly 6.125" x 3.9". That ought to be enough of the critical dimensions to help anyone designing their own, or modifying a design for the best fit.


Now a question -- mine looks to have been reworked quite a bit. Seems factory. My board is stamped "6480099XM".

In addition to the typical yellow 10u/16v tantalum caps at C17, C18, and C19, there are three additional tantalum caps on the back side of the board -- one bodged in on the rails running down around dead center of each "bank" on the left and right sides of the board, and one on the bottom left (looking at the solder side, header on left.) They are labeled "15-20" and orange in color.

There are a couple 100n ceramic caps on the bottom left, and a whole cluster of 15p black film(?) caps on the top of each bank, soldered to the chip pins.

It also appears the positive rail has been cut, where it runs from the pin header to the bottom edge of the board. A ferrite choke has been installed in-line between the two new ceramic caps.

Anyone seen this before?
 
The above sounds close to the one I got from someone who got them from Computer Reset. Mine has some hand written part numbers on it. My expansion card only seems to detect 192kb of memory.
when I run the IBM advanced diagnostics memory test, I get error 030201. I've replaced that module and moved others around in that bank but still get the error.
Does yours do the same?
 

Attachments

  • photo64027.jpg
    photo64027.jpg
    145.3 KB · Views: 1
  • photo64028.jpg
    photo64028.jpg
    151 KB · Views: 1
I posted pictures of the normal one. Here's the early? One.
 

Attachments

  • photo64029.jpg
    photo64029.jpg
    171.3 KB · Views: 1
  • photo64030.jpg
    photo64030.jpg
    158.5 KB · Views: 1
  • photo64031.jpg
    photo64031.jpg
    169 KB · Views: 1
  • photo64031.jpg
    photo64031.jpg
    169 KB · Views: 1
I got one of those last year from an ebay seller, he was selling like 70 or so of them.

I do have a question: is it possible to redesign the board to use two 44256 DRAM chips? I know 64K would be wasted, but still, fewer ICs would make the board much smaller and less hot.
 
I got one of those last year from an ebay seller, he was selling like 70 or so of them.

I do have a question: is it possible to redesign the board to use two 44256 DRAM chips? I know 64K would be wasted, but still, fewer ICs would make the board much smaller and less hot.

Does yours report 256k when installed?
 
I posted pictures of the normal one. Here's the early? One.

This is it exactly. Same hand-written part number on the back: 700137-000 rev. 0 and a serial number. (For the sake of Google) There's a sticker on the front that reads "International Technology Corp." -- which is not exactly a unique and identifiable name. Could be a reseller, IBM partner, or is this an early clone part? It certainly looks like a genuine IBM part.

About the detected memory, I haven't actually tried this card yet. I just had my 5170 (which has the EGA card installed in it) out on the bench last weekend, but I didn't get around to installing the expansion yet. I have a few things I want to do on that machine, and this is one of them, but I don't power-up anything with 30 year old tantalums until I've replaced them all.
 
This is it exactly. Same hand-written part number on the back: 700137-000 rev. 0 and a serial number. (For the sake of Google) There's a sticker on the front that reads "International Technology Corp." -- which is not exactly a unique and identifiable name. Could be a reseller, IBM partner, or is this an early clone part? It certainly looks like a genuine IBM part.

About the detected memory, I haven't actually tried this card yet. I just had my 5170 (which has the EGA card installed in it) out on the bench last weekend, but I didn't get around to installing the expansion yet. I have a few things I want to do on that machine, and this is one of them, but I don't power-up anything with 30 year old tantalums until I've replaced them all.

You guys are killing me :)
I NEED to know if your cards behave like mine.
A funny think I noticed when trouble shooting the 030201 issue is when I looked at the ram module in that bank, one of the legs was not in the socket. Not sure if whomever did that did it on accident....or......
 
This is it exactly. Same hand-written part number on the back: 700137-000 rev. 0 and a serial number. (For the sake of Google) There's a sticker on the front that reads "International Technology Corp." -- which is not exactly a unique and identifiable name. Could be a reseller, IBM partner, or is this an early clone part? It certainly looks like a genuine IBM part.

Here's mine
 

Attachments

  • photo64057.jpg
    photo64057.jpg
    138.6 KB · Views: 2
Is there suddenly a demand for IBM EGA + Expansion? There really isn't any advantage to other clone cards and monitors are in short supply. I wouldn't resort to dodgy computer reset versions. It makes me wonder if they were looking to repurpose unused stock for a customer request since it's being indicated there are a handful of these found there like this? They certainly don't resemble what's pictured in IBM documentation with those modifications on there.

The expansion board has been cloned and works just fine. I have almost enough for assembling a second one here from my order, collecting dust, if someone was looking for one of these.
 
Is there suddenly a demand for IBM EGA + Expansion? There really isn't any advantage to other clone cards and monitors are in short supply. I wouldn't resort to dodgy computer reset versions. It makes me wonder if they were looking to repurpose unused stock for a customer request since it's being indicated there are a handful of these found there like this? They certainly don't resemble what's pictured in IBM documentation with those modifications on there.

Probably not. This is more of an attempt to discover some history. The card I have looks like an IBM part, but with some clues that it might not be, and I'm curious about its origin story. I came across this thread while Googling for scraps of info, and figured while I was here, I might as well drop some observations and physical specs for anyone either looking for the same, or hoping to build their own. There were some questions about hole diameters and fitment that I thought I could add to, although if my card is not a genuine IBM part, it might be best taken with a grain of salt. At the very least, if someone's finding this 10 years down the road and all the hobby replicas have vanished, maybe it'll help someone cook up their own.


You guys are killing me :)
I NEED to know if your cards behave like mine.

Haha ... sorry man. I got some memory chips in the mail to feed a couple of my other machines, so I've got the Tandy 1000 on the bench today. But I'll try and make a point of getting this card installed this week, and I'll report back whether Check It and the AT Diags disk gives the thumbs-up. I don't know if it matters for the sake of testing, but I only have a CGA monitor available at the moment. I do have a 5154 as well, but I have not taken it apart yet since receiving it in the mail. After having received a Trinitron with a crack in the PCB under the flyback transformer, I'm careful to inspect them before applying power. And may as well go ahead and re-cap while I'm in there, too. For better but sometimes worse, I'm exhaustively thorough, which means I often have a dizzying backlog of work to do before I get to play with new toys.
 
Probably not. This is more of an attempt to discover some history. The card I have looks like an IBM part, but with some clues that it might not be, and I'm curious about its origin story. I came across this thread while Googling for scraps of info, and figured while I was here, I might as well drop some observations and physical specs for anyone either looking for the same, or hoping to build their own. There were some questions about hole diameters and fitment that I thought I could add to, although if my card is not a genuine IBM part, it might be best taken with a grain of salt. At the very least, if someone's finding this 10 years down the road and all the hobby replicas have vanished, maybe it'll help someone cook up their own.

Sure history is good to see. I'm sure they are originally IBM based on the markings, but they look modified. I thought the original daughter card came with a plastic shield for the solder side of the card, to protect it from making contact with the card in the next slot during install. Components soldered on the back side would mean it could not have that. ab0tj's clone is a good working design, and gerbers are available, and the holes can be adjusted if needed. If you had to order the ram anyway, and assembling not an issue, all you need is a pcb, which I have extra, and a few other inexpensive parts. You'd probably only spend about 10$ for the parts. Might be a better way to go.
 
I do have a question: is it possible to redesign the board to use two 44256 DRAM chips? I know 64K would be wasted, but still, fewer ICs would make the board much smaller and less hot.

Hard no. EGA uses a "Planar" memory organization in which the 16 color graphics modes each draw 1 bit of color information from four parallel banks of memory. (These bits are then combined into a single pixel through the palette register.) This memory is fetched a byte at a time from each bank and fed out pixel-by-pixel through four separate shift registers, so in essence from the video hardware's standpoint the RAM interface is 32 bits wide, not eight.

If you look at an old 256K VGA card with 8 RAM chips on it you'll find it's four sets of 4464s (64k by 32 bits), not one set of 41256s (256k by 8 bits). Same deal. Clone EGA cards also often (I'm actually going to go so far as say "usually") use this arrangement because, yeah, it saves a ton of space, with the only downside being that you need to sell the card with 256k already on it. (In theory you could probably make an EGA card that had sockets that could take either 4416s or 4464s, as long as you're happy with the only two memory options being 64k and 256k, nothing in-between; this is frankly what IBM probably should have done.)

A memory card with eight 4464s (still wasting a quarter of it because of how the memory is arranged) instead of the 24 4416s *might* be an option? You'd have to add some tricky logic to turn the separate CAS/RAS signals for the three sets of 16K in each of the four banks into memory addresses within the single 64K set you're doing instead. And there may be some additional complication relating to refresh I'm not thinking of, so I'm not going to go any further than saying it's a theoretical possibility.

(You know, if you really wanted to be fancy you might be able to use a 72 pin SIMM for an EGA expansion card, but it'd have to be a really tiny one, and you'd still of course need the aforementioned logic to fudge CAS lines into address bits.)
 
Last edited:
When I said "tiny" in reference to a SIMM I was thinking in terms of capacity. The thing that makes me worry that adapting larger chips would be even harder than just turning CAS signals into address bits is you might also need to somehow regenerate/mangle the RAS signals to get proper refresh. And that problem will get harder the more rows you add? Someone more skilled in DRAM black magic would be better to answer that question.
 
A memory card with eight 4464s (still wasting a quarter of it because of how the memory is arranged) instead of the 24 4416s *might* be an option? You'd have to add some tricky logic to turn the separate CAS/RAS signals for the three sets of 16K in each of the four banks into memory addresses within the single 64K set you're doing instead. And there may be some additional complication relating to refresh I'm not thinking of, so I'm not going to go any further than saying it's a theoretical possibility.

So it has to be two SRAM chips 32-bit x 256kwords? Something like this one: https://www.digchip.com/datasheets/parts/datasheet/740/HM62W16255HLJP-12.php
 
So it has to be two SRAM chips 32-bit x 256kwords? Something like this one: https://www.digchip.com/datasheets/p...255HLJP-12.php

For this in addition to deriving additional address bits from decoding the CAS lines you're also going to have to demultiplex the address lines entirely. Which might not be too terrible. There are some things about the planar memory address generation I don't know without puzzling over the EGA manual(*), though, so I can't say how complicated it'll be. Does anyone know off the top of their head if the same memory address applies to all four planes *always*, or if there are cases where one or more planes have different effective addresses? I assume its the former because later VGA cards often used 16 bit wide DRAMs (with byte select lines to control which 8 bit half is read/written if not both), but IBM's EGA supporting 64k/128k/192k configs might make it a special case?

(* And of course IBM's EGA used some gate arrays, not discrete circuitry, for most of the really juicy parts, including DRAM address generation, so some of it still might not be entirely clear.)
 
Back
Top