RAM fault rabbit hole

[gbejniet]

OK, so, to start with - I'm not going to ask or expect anyone to remote-diagnose my machine, instead I will ask a general question someone might be knowledgeable about:

If Minerva reports a memory fault due to a broken track, and the track is fixed, might there now be no video (TV says "No Signal") after the memory test, due to a different problem? Or is this evidence the track fix is bad?

If you want to read further, here's the story: My AH 128K QL started failing with white screen after pattern (first intermittently, then permanently) maybe after I accidentally left it on overnight. I suspected a RAM fault so put in Minerva. Sure enough, the fault was in high memory (IC14) - I seem to have thrown away the actual test result, but in any case if I'd misread and changed the wrong IC I wouldn't expect things to get any worse.

Minerva then boots (or used to) with limited memory.

As is customary I then made a pig's ear of IC14 removal, breaking tracks to pins 3 (/WE) and 16 (VSS) in both directions. I put a socket in, got a replacement HM4864P-2, and started repairing the tracks (pin 3 in both directions IC13-14-15, pin 16 only from the left IC13-14 as IC15/16/8 are also fed from the other side I believe).

With 3 fixed on the left but not the right, I get:

70827600
F082F680
00020000

i.e., IC8, as you would expect - as it is missing /WE. However once I fix 3 on the right I get... nothing, no video signal. Suspecting I have shorted something but can't find a short from adjacent pins or anything that may be touching that shouldn't.

So yes, the issue is whether I must have shorted or opened something and if so why there is any video at all before this last step? Does Minerva shut down video programmatically after the memory test (and another fault stops it going further) or is this just evidence my pin 3 fix is a mess? (Worth noting that the result is the same if there is no IC14 in the socket, so the higher bank test should be failing in that case. Instead, no signal.)

If the latter then I have a simpler problem (me). If the former then I have a more complex problem (also me).

[Chr$]

Failed ZX8301 (IC22) is (In my experience) the most common cause of no video/black screen. They are fragile chips and often die if you mess around with the RGB cable with the QL on, so it may not be RAM related. Could that have happened? Are you using the RGB connector? Perhaps remove and re-seat the ZX8301. Better still would be to try another known working one.

[Martin_Head]

When you replaced IC14, Did you manage to rip out any of the through plating in the holes.

You could have a pin on the IC socket that is soldered on the bottom side of the board, but a track on the top of the board is not connected because the through plating is missing.

[Pr0f]

The video is produced from the lower bank (in address range) of RAM chips and will try to happen whatever. However - the ZX8301 is also doing address decoding for ROM, RAM and part of the I/O - so if the chip isn't doing anything - it could be duff as Chr$ says.

Usually, if you have a missing track - you'll get some sort of Video output - but will also see a memory fault with Minerva - so no video might indicate some sort of short. Also - tracks are above and below the board - so if a track is damaged on the underside - it's easier to see. Most of the pins on the RAM chips are in fact shared across all the chips (Address lines, /WE) - so you can test this connectivity by selecting the same pin on successive chips and checking for continuity with a multimeter. The data in and out are tied together on each chip and only share a connection with the equivalent pins on the chip in the other bank that's on the same data line - so using the multimeter you can test for these and the connection to the LS245 transciever chip.

It can be a bugger to find the damaged track or bad connection - but persistence will pay off.

[tofro]

If the 8301 were good, it would produce a picture regardless if memory is good or bad, or at least a signal the monitor can sync to. I'm pretty sure it's toast.

[gbejniet]

Many thanks everyone.

When I unsolder the repair on /WE between IC14 and IC15, the Minerva RAM fault screen returns:

70827600
F082F680
00020000

I just did this again for my own confirmation. So if this signal can still come out, and comes from the ZX8301, then that chip is at least working, am I right? Or does the RAM fault video signal come from somewhere else? (I'd like to say I'm being deliberately naive here, in order to work through an issue, but I can't really be sure I am)

If I have really fried the ZX8301 I will be pretty cross with myself as I was quite assiduous about not interfering with monitor cables whenever there was any possibility either the TV or QL was on. I'd heeded all the warnings, I thought.

And yes, I probably did pull out through plating on pins 3 and 16, at least this would account for the breaks either side of IC14.

[Derek_Stewart]

Hi,

If you use Minerva, download the Minerva utility disk, file from QL Homepage or link:

http://www.dilwyn.me.uk/qlrom/min198utils.zip

Where there is a program called 'ramfail_bas'

this take the data you have and identify the faulty internal ram chip.

This program doe not work on memory expansions.

[tofro]

Many thanks everyone.

When I unsolder the repair on /WE between IC14 and IC15, the Minerva RAM fault screen returns:

70827600
F082F680
00020000

I just did this again for my own confirmation. So if this signal can still come out, and comes from the ZX8301, then that chip is at least working, am I right? Or does the RAM fault video signal come from somewhere else? (I'd like to say I'm being deliberately naive here, in order to work through an issue, but I can't really be sure I am)

If I have really fried the ZX8301 I will be pretty cross with myself as I was quite assiduous about not interfering with monitor cables whenever there was any possibility either the TV or QL was on. I'd heeded all the warnings, I thought.

And yes, I probably did pull out through plating on pins 3 and 16, at least this would account for the breaks either side of IC14.

If you're getting any sort of video output, the 8301 is likely OK. There's no alternative way to get something to the screen.

If you get video output when you remove your "fix" (and none otherwise), it's likely something with your fix is wrong.

[gbejniet]

Well after over two years and much butchering of traces, I have finally replaced enough RAM to make Minerva happy and get back to 128K. Thanks all who advised.

I now seem to have run up against an issue which I'm sure wasn't occurring even recently when Minerva would boot with 64K and I could type ok: all keypresses are registered, but the codes are wrong. Here's an example result of me booting to F2 and then pressing 'a' repeatedly:

aalaa
aa
8aaaaaaarvalaaaaaaa9

The new lines are also the result of pressing 'a' rather than intentionally hitting Enter. (probably worth noting that pressing F2 doesn't work reliably which might arguably be down to the same issue)

All keys seem to be affected. This I guess is likely something at the 8049 end, but I wonder what? if all keys are registered then I'd say it's running ok, but the result not being read correctly.

Can't find anyone else having this happen (?) Once again not asking for others to put themselves out, but if anyone knows a little about how this all hangs together that might be relevant I'm keen to learn.

[jobdone]

Please expand on F2 doesn't work reliably? The last time I saw a 64K booting minerva, the 74LS245 had failed. The reason I asked about F2 is because I have also seen a state where pressing F2 half the time gives monitor mode or doesn't boot. The ZX8302 was at fault.