exxos's Atari, Amiga & retro forum

stephen_usher wrote: 14 Apr 2026 09:28 The rest of the 74Sxxx chips arrived yesterday and I had about an hour yesterday evening to do some testing....

They've made things worse rather than better as the nearly stable system using 80ns 41256 RAM is as unstable as with 4164 chips now. Also there's a new display artefact with red "ghosting" which varies across the screen changing every 8 characters.

I'm wondering if there's some sort of capacitative coupling going on somewhere on the board, a bit like the phantom short on one of the RAM sockets which went away with IPA and a slight brush. I've noticed on some of the signals they seem to be picking up about a 0.5V echo of another signal. No idea if this is just cross-talk or something more sinister.

Oh what a pain.

It's not interference from the PSU thats messing with the video is it? I ask as I've had this issue several times now with various things.

The instability is a bit annoying, what speed RAM is expected? 80ns seems more than good enough when you consider the ST will use 120ns. Are RAS/CAS stable?

The original used 200ns RAM and this is a direct reproduction. It must be a signal timing issue of some kind or maybe the Samsung 8ns RAM chips are less sensitive to whatever the issue is.

I had a bit of time to look this morning. The red fringing wasn’t there when I first turned on the machine but slowly faded in over time and also faded in nad out over about 10 seconds. Not sure how this can happen with TTL. RGB. The machine was more stable as well this morning.

I’m wondering if it’s related to the issue of the phantom short on one of the RAM sockets I had which went away by flushing with IPA and a light brush. Maybe some capacitor i’ve coupling somewhere.

stephen_usher wrote: 14 Apr 2026 11:50 The original used 200ns RAM and this is a direct reproduction. It must be a signal timing issue of some kind or maybe the Samsung 8ns RAM chips are less sensitive to whatever the issue is.

I had a bit of time to look this morning. The red fringing wasn’t there when I first turned on the machine but slowly faded in over time and also faded in nad out over about 10 seconds. Not sure how this can happen with TTL. RGB. The machine was more stable as well this morning.

I’m wondering if it’s related to the issue of the phantom short on one of the RAM sockets I had which went away by flushing with IPA and a light brush. Maybe some capacitor i’ve coupling somewhere.

Definitely worth a look at but one of those ethereal faults that can be hard to track down.

Annoying.

Well, the RAM problem is fixed...

How was that done? Well, inject the raw Z80 refresh signal into the /CAS line for the main memory chips using a 74LS08 to and the signal in with the other CAS signals.


Obviously the original design relied upon RAM chips with a long persistence time, far beyond the specification, as the generated refresh was not regular enough. I'm not sure why the original designers didn't use this simple approach to begin with, other than it reduced the chip count by one.

Where the machine in the MON debugger would previously crash after about 30 seconds now it's rock solid.

So, the "final" task is to finish off the case.

I've started the hopefully final print of the top case this evening after I got home and installed the new grey filament which arrived today. It should be ready about 10am tomorrow. Then I can print the two halves of the lower case.

Interesting, so basically boosting CAS to ensure the RAM stays refreshed?

rubber_jonnie wrote: 15 Apr 2026 09:59 Interesting, so basically boosting CAS to ensure the RAM stays refreshed?

Well, more mixing in the CPU refresh so there's a deterministic refresh of the chips rather than hoping that normal memory access will be enough.

stephen_usher wrote: 15 Apr 2026 11:00

rubber_jonnie wrote: 15 Apr 2026 09:59 Interesting, so basically boosting CAS to ensure the RAM stays refreshed?

Well, more mixing in the CPU refresh so there's a deterministic refresh of the chips rather than hoping that normal memory access will be enough.

Understood

Swapping out the 80ns 41256 chips with 150ns 4164 chips made the system unstable again. I managed to find a set of 120ns 4164 chips and the machine is working correctly again. So 120ns seems like the slowest speed it will work with, which is surprisingly fast.

I've also had a quick play with Lynx BASIC... Lots of advanced features such as procedures, while/wend, repeat until BUT only single character (case sensitive) variable names and it seems that FOR-NEXT loops break if you use GOTO within them! This makes it seriously restrictive.

Well, the top case print overnight was a bit of a disaster.

I obviously used the "wrong type" of supports which won't come off the print without destroying it.

Secondly, I'd modified a post to allow for the 74LS145 to fit, except I'd modified the one on the opposite side of the case to the chip.

Thirdly, the spool got stuck at ~4:30am so I lost 5 hours before I found out that the print was a bit naff and it made a line on the two parts where it stopped..

Other than that(?!) it looks quite good and the cut outs for the ports and the keyboard are just right.

@stephen_usher that does look pretty good actually. Shame about the supports though. I'm sure you'll sort it out.