The Sinclair QL Forum

Hello all! So, like the title indicates, I've got a faulty QL which I am trying to debug.
First things first, it's an issue 5 board, with some modifications (which I didn't make). Namely, all the usual TLC with resistors and stuff recommended by Sinclair, but also one 5V trace cut roughly between the microdrives and under the 7805 (but not where indicated by the Service Manual), plus a wire from the 5V out pin of the 7805 to another 5V trace, just under the left keyboard ribbon socket.
Second things second, I've checked all the removable ICs on another working QL and also tried the corresponding ICs from the working QL onto the 'faulty' one. All ICs from the faulty one work perfectly on the working one, and the ICs from the working one make no difference to the faulty. So we can be relatively sure it's not a problem of the ICs. Obviously, I'm talking about the 68008, the 8301 (well, 8310 in both cases), the 8302, the 1377, the 8049 and the ROMs (JM for the working one, AS for the faulty).
Moving on to the symptoms, the QL is stopping at the 'white screen' stage, after the typical random red and green pixels. It does react as it should to a reset, only to reach the white screen again.
That in itself indicates that everything is *almost* working, only the RAM write test fails -- at least, according to the service manual.
I've checked, with the rather basic multimetre that I have here, the voltages and well, all ICs appear to be getting the correct 5V that they should. I've also checked and double-checked the data lines from the RAM chips to the 8301, the 74LS245, the CPU and the ROM chips, all present and accounted for, also the lower address lines, again all present; no shorts between adjacent pins either.

Now the interesting part. After staring at the boards for hours with my nose practically 1 inch away from the chip tops and the multimetre probes in my contortionist hands (you know what I'm talking about, right?) I accidentally connected my Gold Card to the faulty QL (instead of the working one) and turned it on. Lo and behold... it worked. Kind of.
The random green-red test shows... twice... as it should with a Gold Card, then the "Gold Card v2.49" thing... and a second or two later, the "F1...monitor, F2...TV, Sinclair Research etc." prompt. Only, rather garbled, as you can see in the following image: So, it looks as if there's a "feedback loop", roughly 4 times. [please forgive the horizontal stretch, the TV I used was set to "widescreen"] The funny thing is, you can see bytes being overwritten slowly, so the image settles after 10 seconds or so.

My interpretation: bearing in mind that the RAM is implemented on 1-bit RAM chips, the fact that the general layout of the display is correct (i.e. you can still see the "Gold Card" or the "F1 monitor, F2 TV" pixels being in the right place), that means that the RAM output is, at least, all present. Also, the fact that the display is such, means that the ROM access from the Gold Card, i.e. 68000, is fine, otherwise, it would be reading 'garbage' from the ROM. Now, to me, that looks like... a timing error, somewhere. Given that the (whatever) CPU has access to the RAM array through the LS245 buffer, I was thinking that maybe that IC is malfunctioning, or 'bouncing' its output and producing artifacts on writes. Alternatively, the oscillator feeding the 8301 could be off? Of course, it could always be the RAM chips themselves, but somehow, I don't find it very probable.

In any case, I would welcome any insight you guys might have. I don't want to start unsoldering chips without being (reasonably) convinced.

Just some thoughts...

With a Gold Card, the 68008 is not used. The CPU on the Gold Card runs the system. The CPU copies the data from the ROMs to the Gold Card RAM and then all normal memory reads and writes by the CPU use the RAM on the Gold Card. The only exceptions are writes to screen memory (which are copied to the original RAM) and I/O accesses. All screen data is still stored and then read (by the QL hardware) from the QL RAM. Reads by the Gold Card CPU are from the Gold Card RAM I think.

In your post, are you saying that the screen display is messed up but improves with time?

Do you get the same results after a reset once it has been on for a while?

Mark

1024MAK wrote:Just some thoughts...

With a Gold Card, the 68008 is not used. The CPU on the Gold Card runs the system.

Yes, I know, but since the 68k8 is verified working, that's not an issue.

1024MAK wrote:The CPU copies the data from the ROMs to the Gold Card RAM and then all normal memory reads and writes by the CPU use the RAM on the Gold Card. The only exceptions are writes to screen memory (which are copied to the original RAM) and I/O accesses. All screen data is still stored and then read (by the QL hardware) from the QL RAM. Reads by the Gold Card CPU are from the Gold Card RAM I think.

Actually, this is where it gets interesting. Indeed, the ROMs get copied onto the GC RAM, patched for new addresses, and the system is reset. So, that means that the ROMs and the data and address buses are ok.
However, the QL RAM data and address buses are *not* directly connected to the CPU address and data buses (regardless of whether the CPU is the 68k8 of the QL or the 68k of the GC). Instead, the data bus gues through the 74LS245 (8-bit bus transceiver) and the address bus through two 74LS257 (quad 1-of-2 data selectors).
So, my interpretation is that up to the 74LS trio that isolates the CPU buses from the RAM buses, the system is fine. The problem *should* be... oh, wait a minute... just had an epiphany!

The end picture shows that data is being shifted/duplicated by one or two addresses : note the duplicate "F1" and "F2" in the image I posted! So, that *could* be a problem with the lowest bit or bits of the RAM address bus, i.e. A0 and A1... which, conveniently, have to get selected one-at-a-time from one of the two '257 selectors! If that chip's specific output is malfunctioning, it could be screwing up the lower bits of the address bus, thereby making the RAM chips writing data coming from the CPU randomly one or two addresses back and/or forth! Which would be a perfect match to the observed artifacts!

1024MAK wrote:In your post, are you saying that the screen display is messed up but improves with time?

No, it sort of worsens with time, as you can see in the screenshot I posted.

1024MAK wrote:Do you get the same results after a reset once it has been on for a while?

Yes, the results are pretty much the same, no matter how long you wait.

All in all, after the epiphany I just had, I think I'll go ahead and check or replace the '257 that's responsible for the lower address nibble...
Thoughts?

Sounds good to me.
But first check for dry joints and broken tracks around the multiplexer chip

Mark

Skagon, how did you get on after your epiphany?

Funny you should ask; all this time, while waiting for some chips and sockets to arrive, I wrote a test programme and checked my assumptions. You see, my previous conclusions were drawn thinking in "PC" mode, not in "QL" mode. So I hadn't taken into account how the pixels are stored in the QL's video memory. After I switched thinking mode... it became clear what I had to do.
Funny thing is, with the Gold card the QL is working nicely, except the display issue, obviously.
So I POKEd the video RAM and drew some patterns, while taking more photos from the faulty and the working QLs, to see if I could see any visible patterns.
So, it's clear now that it's the address line A2 that's not working -- or maybe floating so it's randomly registering as high or low.
That can give us some conclusions:
1. Since the A2 line from the CPU is multiplexed with A0 on the 257, and there's no problems with A0 bytes, the output of the 257 is ok.
2. Since the ROM is being read correctly, there's no problem with the A2 line outside the resistor (R74), i.e. CPU and ROM and expansion slot are ok.
3. Since I changed the 8301 and the results are the same, it's not a problem of the 8301.

Therefore, my conclusion is that there's either a problem in the input of the 257, or one of the RAMs has blown its A2 line and is contaminating it with random noise.
Obviously, the first thing I'll check is the 257. I've already got enough sockets to re-seat all the RAM chips and both the 257s and a bunch of new 257s.
The plan for tomorrow is to desolder the 257 and see if a new one fixes the problem. If not, I'll have to desolder all the RAM chips, solder the sockets and order replacements (if I can find some on eBay).
For obvious reasons, I hope it's the 257.

If anyone's even following the thread, news flash.
So... like I said, I got my el-cheapo soldering iron and braid, and replaced the old TI 74LS257 with a nice... socket. Then I got one of the NOS Signetics 257s that I got here and here's the result: So, after the foreplay was over... moment of truth. Well, the photo is a bit *after* the moment of truth, but you get the point... Diagnosis was correct, now I've got another fully functional QL to play with!
I think I'll just add some caps on the Vcc lines, just to keep the ripple down, like the US QLs have, before closing up.

Thanks to Mark for actually bothering to offer some help. Cheers, mate!

Having just read this post, I have to say great diagnosis and also the screen display looks pretty clean now.