Glitches, partially solved by swapping the CPU?

[hjf]

So I built this RAD myself and I've used it a number of times but never really got it to work. No matter what I do, I'll get screen glitches. Some @ character here and there but mostly it's random white pixels in the menu.

I tried changing timings, overclocking the Pi Zero, etc, nothing. The timings menu did help a little but the glitches never went away.

So I decided to take more drastic measures and swapped the CPU. boom - perfectly clean screen. No glitches at all. So i start playing a Nuvie and got it to finally play for several seconds until it crashed. I went into the menu again and it was glitching.

I got some freeze spray and started spraying every IC on the board and there was no change. But if I turn the C64 off for a few seconds and turn it back on, it'll run glitch-free for a while again

I checked voltages on my machine and everything seems fine. I'm powering the Pi from an external supply. I also tried a different PLA. I'm using a GAL PLA now but I put the original (faulty, reports wrong free memory) and the glitches are the same.

The SideKick64 doesn't have menu glitches and it seems to work just fine. I also tried swapping the Pis between these and nothing changed.

[frntc]

The Sidekick64-menu works completely different, you can't compare the two. There are reports that the Phi2-signal of the 6510s might be coming too slow (see https://retrospective.shop/phiagra-module-v2/). I haven't experienced that, but the entire timings starts with sampling that signal. If Nuvies run for a while then you're pretty close to working settings. Are the timings set back to default? (and you're using which firmware version?)

[hjf]

Actually the original 6510 in this C64 had the problem of PHI2 too low. I measured it with an oscilloscope and it was around 1Vpp which strangely was enough for the C64 to work, but barely.

I'll probe the level of this signal and comment back. If the level is too low maybe an improvement can be made in a further hardware revision to amplify this signal. But I'll check first.

Does the cartridge use the pixel clock? The very small white dots could be explained by this rather than by PHI2. They are too short, I think, for a CPU clock.

When the nuvies crash, I just hit the menu button. The timings are where I left them.

[frntc]

Dot clock is not used. Maybe you want to try the default timings with the 2nd CPU?

[hjf]

Third CPU actually.
The original CPU didn't work even with Sidekick64.
Second CPU made both carts work (but problematic with RAD)
Third CPU made helped a little more, i was able to run a NUVIE for a few seconds.

I did go back to the original timings. The problem is that the glitches either aren't there at all, or they appear and won't go away. Turning the computer off for a couple of minutes seems to make them go away.

I thought it may be temperature related but freeze spraying all chips on the board one by one had no effect.

[hjf]

Well, I checked with an oscilloscope. PHI2 is perfectly fine, 5Vpp amplitude:

R/W doesn't look great though:

And, fwiw, the menu music plays with a real SID, but not with a Sidkick Pico?

Anyways, I tried reinstalling everything and using default timings, and not much changed. At one point it was glitch free and I was able to play a NUVIE for a long time (almost 1 minute?). Then I tried turning the C64 off and on again, and when I entered the menu, the glitches were back.

This is very strange.

[insanity213]

What version mainboard is in your machine? I have very similar experiences with 2 different 250466 machines and multiple C128's. I was able to get the RAD stable on 250407 based machines and also on an Evo. The Evo had wildly different timing tweaks compared to the 407's. I've given up trying these on anything other than 250407 boards.