Random QL glitch: no video signal and refuses to boot

[bwinkel67]

So a few times over the last week the QL has had a weird behavior and I haven't tracked it down. It's pretty standard system (US model with DB9 connectors and NTSC video) with a small mod and an add-on. I'll describe them and then the glitch:

The mod:

• Added two-way switch running a wire off of pin 9 of MC1733 to pin 3 of RGB connector to give me both monochrome and color composite at a flick of a switch.
  • There's a whole thread of why -- US models don't give color-composite on pin 1 of RGB for some reason.
  • I recently double checked all connection and the solder joints are good, no shorts and the switch is new.
  • Also, I never flip the switch until power is off to both the computer and the monitor -- very careful about that

Add-ons:

• Hermes chip
  • had that since the 90's

The glitch:

So here's what happens, after a short period of time the picture goes away (no signal). If I reboot (either side switch or remove power) I get nothing, no picture and the computer won't boot since F1 or F2 don't spin up the microdrive. I do get a power led (yellow) but that's it. The Hermes gives keyboard audio feedback but after about 8 or 10 key-presses it also stops accepting them. If I leave it off for a while things come back to normal. At present the video switch is set to normal (i.e. the mod is turned off and I'm only getting monochrome-composite from pin 3 / 2).

It happened last night after the machine only being turned on for 10 minutes total and it took about 30 minutes to come back after I powered it down (completely unplugged both QL and monitor). It just happened today, again after being on a short time. This time I immediately opened up the case (maybe 5 minutes between next power up) and the video picture came back immediately. I then proceeded to poke around, pressing on chips and wiggling connectors (including the mod wires and switch) and nothing, the picture is solidly in F1 monochrome mode and has not left me (been on now for a few hours).

Any idea what to look at? I checked the power on the brick and it gives 15 volts AC and about 10 - 11 volts on the DC line, which is a bit high but it's not under load. Should check it now that I have it open. I assumed somehow I damaged the MC1733 chip with my mod but it is working presently and also, I'm bypassing the mod to be back to factory normal. I also thought maybe the ZX8301 but again, picture signal is working perfectly now.

So that's why I'm calling it a glitch since it isn't as repeatable. But what causes no video signal and no boot up on in a QL? What should I be looking at?

I do have replacements for the ZX8301 and the MC1733 but if my mod is causing it then I'd rather not blow a second and I'd like to keep the mod if possible (plus my ZX8301 is wired with diodes like a Christmas tree so that will take a bit of work to swap out). All I'm doing is grabbing color composite off of pin 9 of the MC1733 and ground pin 2 off of the RGB connector so that shouldn't damage the chip right (assuming I didn't do something bonehead like short connections -- again, double checked that I didn't)? And I added a 47 ohm resistor from pin 9 to help the color image.

[bwinkel67]

Still going...have rebooted it but not turned it off and the case is still open. Just weird that it happens immediately and now not. Is suppose it could be a solder joint or re-seating is required in sockets...they have never been pulled and I don't have a IC puller so that won't be easy. Could I use a toothbrush and clean the tops of them with isopropyl in the hopes some of it would flow down the pins and clean any oxidation?

I will next turn it off for 10 minutes and back on and see what that does.

[FrancoisLanciault]

I am experiencing exactly the same behavior.

Same thing, no video and the hermes click stops after about 10 keystrokes. That QL historically had many problems, mainly freezing just after the F1/F2 prompt. But even when frozen, each key press would always produced a sound with Hermes fitted.

The problem appeared two weeks ago. Because there is no image, I though the issue was with the cabling or the VGA converter. But when I noticed the new keystroke behavior, I realized the problem might be something else. I had no time to investigate further.

My setup is different: Goldcard + QL-SD + Hermes

Let me know if you find the cause. I can also do test on my setup (like probing signal with a scope)

Francois

[bwinkel67]

I hate to say this but I feel better now that two of us are in the same boat though obviously I don't want your computer to be broken. Hopefully we will figure it out. At least I know it wasn't just a bonehead mistake on my mod. Am I correct to assume you've been careful not to pull your RGB cable out of the computer or monitor while the QL & Monitor were on since the ZX8301 (and likely the MC1733), both of which process video, aren't buffered properly and can throw a electrical charge back on the wire.

[Pr0f]

I can answer one of your mysteries at least.

I've been looking at the Hermes code - and the keyclick will work whilst there are spaces left in the buffer - after so may character presses, if nothing is reading the keystrokes out of the Hermes chip - the buffer will fill and no longer accept keys - hence no more click

That tends to confirm that the QL is not operating (as in no processor activity - could be in some tight loop, stalled in some hardware way). If you can operate it with the case undone - you may want to check if the smaller of the 2 large chips near the expansion connector on the right hand side is getting hot - this is chip that produces video - ZX8301, and the only thing I can think might be tied to your video modifications...

[bwinkel67]

Thank you for the explanation on the Hermes behavior. So I have an infrared thermometer and here are the temp readings on each chip with the case propped open (lid is on but propped at 70 degrees). The computer has been on for hours and currently the video signal is fine:

• 68008 - 106 degrees center of chip
  • There is a capacitor on top that's like 110 degrees
• ZX8301 - just under 97 degrees
• MC1733 - mid 97 degrees
• Hermes - mid 94 degrees
• ZX8302 - mid 91 degrees
• 2 x JSU ROMS - around 95 degrees
• RAM chips closed to the edge 80 degrees
• RAM chips one row in 89 degrees
• HAL 16L8CN is about 107 degrees, hottest among the other smaller chips which are in the 90's
• Heat sink at its hottest is 113 degrees but varies between that an 105 degrees.

Note that the room they are in is a comfortable 74 degrees. All temps are in Fahrenheit.

So I think those values seem all in the normal range. I don't know about cold chips since the RAM ones at 80 degrees are equal to just measuring the plastic in the front which is also 80 degrees, so they are not basically generating any heat. The memory test on boot-up though seems fine.

I'm guessing the CPU would run the hottest. I doubt closing the lid would cause a chip to drastically go up in temp.

[bwinkel67]

Since currently my mod is off (switched to the monochrome-composite mode which is factory setting) I will next test it in color-composite mode to see if the chips go up in temperature. The computer today glitched in monochrome-composite mode after having been off for the night so I don't think that's it but it's worth a try. Unless there is some internal non-volatile memory that sticks around between boots and is used by the system, I don't see how color-composite signal (i.e. the mod) could somehow stress the MC1733/ZX8301 to put itself in a mod where it cannot boot until after some time.

Of course, that also begs the question, is it safe to grab the MC1733 pin 9 signal to drive composite paired with ground? We assume that's what the normal pin 1 PAL-composite on most QL's (UK version at least) does but is there anything else that helps protect/buffer it. I put on a 47 ohm resistor to help the color. I could try something higher I suppose to drive the chip less (if I read correctly why SCART video cables on the QL add 330/390 resistors).

[Martin_Head]

If you think it's a temperature issue. Get a tin of freezer, and when the QL locks up. Open the lid, give a few things a squirt of freezer, and see if it will then reboot.

If it does, then it's a case of trial and error to isolate the area of the problem.

[FrancoisLanciault]

I can answer one of your mysteries at least.

I've been looking at the Hermes code - and the keyclick will work whilst there are spaces left in the buffer - after so may character presses, if nothing is reading the keystrokes out of the Hermes chip - the buffer will fill and no longer accept keys - hence no more click

That tends to confirm that the QL is not operating (as in no processor activity - could be in some tight loop, stalled in some hardware way). If you can operate it with the case undone - you may want to check if the smaller of the 2 large chips near the expansion connector on the right hand side is getting hot - this is chip that produces video - ZX8301, and the only thing I can think might be tied to your video modifications...

Good point. How long is the hermes buffer. In my case, the key clicks stop after excatly 7 strokes, always.

Francois

[bwinkel67]

I don't think it's a temperature issue. I've had it running for hours in a warm room. Waiting for it to repeat so I can use the temp gun to figure out if any chips are hot. On mine I'm assuming one of two things, a solder joint that's bad and when it heats up it disconnects/crosses (maybe even my jumper wire) or chips just need to be resocketed/cleaned. I did try and press the board and/or wires to address the first issue. Then I cleaned various chips with isopropyl to see if that helps. I haven't pulled any as of yet since I'm trying to be minimally invasive but if it returns (esp after I put the lid back on) that'll be next.

Also, I did a continuity test with my multi-meter though it's an old analog one. It's weird that almost every pin is somehow connected just with resistance whereas if it's a direct connection (i.e. a wire directly crosses) it measures at 0 ohms. Even monochrome-composite (what was originally going to pin 3 of the RGB) and color-composite (pin 9 on the MC1733) have some connectivity value (250 ohms) along the board (i.e. they aren't fully isolated). Just hard to see with a multi-meter if a trace is jumping since on mine 250 ohms is close to 0 on the analog scale.