Print

[stepleton]

Some of the Lisa cloners I mentioned in my last post have turned to the Sun Remarketing 2MB memory boards for RAM. They've encountered troubles with a few of the bare SunRem-original PCBs available for sale online, a problem that apparently boils down to a missing connection between one chip and ground. This is an easy flaw to bodge-repair with a bit of wire. Here are some of the threads about it that I've seen on Mastodon:

https://mastodon.social/@iconicbits/111257483656871396
https://mastodon.social/@paulrickards/111224219350765681 (scroll up and down)

Evidently the 74LS logic family (not F, HCT, etc.?) is also required for some of the support chips on the board, per the second thread.

I think there may be reverse-engineered CAD files out there for this board that don't have the missing trace problem, but I had a hard time finding them.

[RolandJuno]

Can confirm. I built up my bare board but it didn't work, at first giving a bus error (with the addition of an Apple 512K RAM card) which had be chasing my tail. I bought so many different kinds of 256k 30-pin RAM SIMMs (9 chip and 3 chip variety) to troubleshoot the issue (at great eBay cost).

So far, the "NMBS" 9-chip variety seems to work (and may have been originally provided as part of a kit years back) although there was a forum post that mentioned 3-chip variety had success. Unfortunately the link to the eBay auction showing the chips doesn't work. https://68kmla.org/bb/index.php?threads/apple-lisa-sun-remarketing-2mb-expansion-what-ram-do-i-need.2598/

Adding the bodge wire on U4 74LS157 pin 15 (enable, which was floating) to pin 8 (GND) and the card suddenly began to work.

I also mistakenly used 74S373 for U9 and U11 instead of 74LS373 (note S vs LS) as noted on the silkscreen. It was easy to miss as U5, U6, and U7 use 74S373. With the wrong chips in U9 and U11, I got random parity errors and system instability. Replacing them with the correct parts fixed the issue. So use the exact parts on the silkscreen!

There's a white silkscreen marking in the bottom left corner (that looks like JLCPCB marking) that reads 42779GY151-190514 which I believe means the boards were produced on Apr 14, 2019. If your board has this, it needs this bodge to make it work.

[sigma7]

many different kinds of 256k 30-pin RAM SIMMs (9 chip and 3 chip variety) to troubleshoot the issue

The problem (possibly not the only one, but a major DRAM compatibility issue) is that the Lisa design implements 256 cycles per 4ms of refresh. DRAM chips larger than 256K x1 typically need more (eg. 512 cycles per 8ms), and so are not refreshed adequately. They may appear to work at first but will fail in a few minutes.

A 256K x9 SIMM with 3 chips may be made with one 256K x1 chip (for the parity bit) and two 256K x4 chips. The 256k x4 chips probably (maybe always, IDK) require more than 256 cycles of refresh. I speculate that an over-performing part may still work if it hosts the video page.

A modification was in the works to convert the SR design to "hidden" refresh to support most DRAM chips, but since 30 pin SIMMs of any size (and sockets) are no longer inexpensive, we really need a new memory board design that uses current RAM chips.

TLDR: Check the DRAM chip's datasheet: if it specifies more than 256 cycles of refresh, it probably won't work in the SR board.

[patrick]

You should be able to use 1M x9 SIMMs, which were more common than the 256k types. Address line A9 (pin 18) must be tied low so that only 1/4 of the capacity is used. The upper half of the memory is not refreshed, but since it is not used, this is not critical.