Monday 5 March 2012

Atari 800XL Memory Repair

This is an old post, preserved for reference.
The products and services mentioned within are no longer available.

I was tidying up and came across a bag of 30 pin SIMMs. Definite relics of the past, some were 256K and you needed 4 of these to get 1MB memory, enough to comfortably run Windows 3.11. I can't really see anyone coming to me with a computer for which these would be an upgrade. The rest were 1MB sticks, and with 4 of those, your 4MB was just enough to scrape your way into Windows 95. I'm a hoarder at heart, so I couldn't possibly throw them out, but I was wondering how I could justify keeping them, what possible use could they be?
Then I remembered I'd found a use for one back in the 1990s when I didn't have access to replacement DRAM chips. My dear old Atari 800XL, the first computer I had with sound and colour (or color as having to repeatedly type COLOR and SETCOLOR didn't help my spelling back in the day). After a long, hard life, it had become a bit flaky. A unique feature of those Atari's was a built in self test, you could test the sound, the keyboard the ROM and the RAM. The RAM test should show up 48 green squares (1 for each KB of memory), but I was getting about a dozen RED ones scattered around.
This is where the 30 pin SIMM came in. I basically stuck it on top of the existing chips, and wired it in place. This was a while ago, so I don't have the wiring diagram to hand, but from what I remember the address and data rows / columns were all driven through resistors, so all I had to do was cut the end of the resistor where it went to drive the old chips and wire it to the nearest appropriate pin on the SIMM - although ideally D0 goes to D0, D1-D1, D2-D2 etc. with RAM it doesn't really matter if D0 goes to D3, D1-D6 and D2-D2 etc. as it will only be reading back what it wrote, so it will be in the correct order. The 1MB SIMM I had used was overkill in terms of capacity, I just tied the spare address lines to ground - at some point in the future, I never got around to looking at memory banking to make use of the extra capacity. The timing was also a bit more than high enough, the originals looked to be 150nS, the replacement were 70nS, so more than fast enough.
Not the neatest job in the world, it was sort of proof of principle, I was interested to see if it could work. No problems at all, memory test came up with 48 green dots and ran for a couple of hours with no problems reported. And most importantly, Spellbound worked!
This is now very much a spare as the keyboard and button panel are both incomplete, but it does actually run.