Odd screen corruption on power on. Issue remains after CLS.

[prefim]

Hi Folks, Hoping someone has seen this and can offer some guidance.


I have an issue 4 (the tiny PCB) in my 464 that I just received. I'm seeing the odd screen corruption on power on. If I perform a CLS it clears the screen and adds some odd dots at the bottom which move if you press return enough times to push them off the top but never go away as they wrap around back onto the bottom. If I change screen modes the dots change with it. It loads games fine although the dots are present even in game. Note the screen is B&W as I'm using the luminance out pin on the RGB din as a composite out so no panic there. Images attached.

[Bryce]

Hi prefim,     
That's probably a bad connection between the ASIC and the RAM. Or if you are really unlucky, a bad ASIC.

Bryce.

[prefim]

I've reflowed all the regular chip pins but haven't attempted the SMD chip yet.

[Bryce]

Try lightly pushing the ASIC pins individually and see if any have movement. You don't need to reflow them unless it's needed.

Bryce.

[prefim]

Yeah, tried that. No movement that I could detect. Gave them all a good clean in case there was anything bridging them like corrosion. Didn't seem to make a difference. Nothing is getting too hot too quickly either used to alcohol to see evaporation rates etc.

[Bryce]

Page 21 of the "Amendment Service Manual" gives the pinout and connections for the pre-ASIC. You could try probing the associated pins to establish if any are stuck.

Unfortunately, despite being the amendment, it's full of mistakes and ommisions too, just like the original schematics. 

Bryce.

[prefim]

I'll check it out. Someone on facebook has suggested pulling the 8x ram chips and replacing with to of the larger type in the empty sockets.

[Bryce]

A pointless exercise. If the RAM had an issue the CPC wouldn't boot at all because the CPU would be getting corrupted data too. In your case the CPU is getting good data, just the CRTC (inside the ASIC) is getting bad data (or corrupting the data itself internally).

Bryce.

[gerald]

Likely a bad ASIC or a bad timing when fetching the ram for video refresh.
The RAM is with no doubt OK otherwise you would have far more problem.
Connection RAM-ASIC  and ASIC-Z80 are fine too for the same reason.

Only a logic analyser on the RAM would tell us if the issue is within the ASIC or a  timing glitch during the video fetch.

[prefim]

Well i used my hot air station and some flux to reflow the asic and then went over the rest of the board in case its a bad pull up resistor or something. No dice. Its no better or worse so I guess its most likely a bad asic. Not something I plan to swap or would even assume you can get nowadays so I will wait for my next 464 to arrive and see what I can make from it.

[Fessor]

Im wondering why the Copyrightmessage is cleanly printed and there is no corruption where the spaces are and why there is a change at Cursorposition 9,20 from three pixels high to four pixels high. If it is a readout-problem, shouldnt this error occur under all circumstances?
From the Pixelpositions under the Cursor of the first Screenshot, its like that bit 0x00010000 is set.The Pixelmask seems to appear at &e8xx, &f0xx and &f8xx; what does peek(&e800) return after a reset?

[pelrun]

It can't just be in the video fetch. CLS erases the bad pixels temporarily, and the cursor is visibly inverting them - which requires a read-modify-write operation in the CPU.


Something is corrupting the actual data in the RAM. I'd be interested to see what happens if the CRTC is pointed at &4000 instead.

[prefim]

Well I've fired it up with my scart cable now so we see colour. I did a print peek(&e800) and got a 1. did a CLS and repeated and got 96.


If you can tell me how do point the CRTC at &4000 I'll try it. Note after the CLS, there's still some dots at the bottom. these move up the screen if you keep hitting return to push them but they wrap around....

[Fessor]

Ah, corruption displayed as PEN 2... therefore bit 1
After CLS you are peeking into the Pixels of the "R" of the Readypromt at Location 1,1

To point the CRTC to &4000:
out &bc00,12:out &bd00,16
Quick n Dirty, because the Interpreter still thinks the Screen is at &c000. With out &bd00,48 you can switch back to &c000.

[prefim]

Just tried this and got a completely blue screen, no corruption. pressing return about 20 times eventually causes a screen scroll so the screen comes back as was. also outing back bd00 to 48 returns the screen as was without the returns being pressed.

[prefim]

Turns out it was the ram..


I binned off the 8 standard and replaced with two larger and bingo. Clean screen and all is good with the world.

[Bryce]

Most likely a trace to the RAM, rather than the ICs themselves, but as long as it now works, that's what matters.

Bryce.

[gerald]

Most likely a trace to the RAM, rather than the ICs themselves, but as long as it now works, that's what matters.

Bryce.

I am more thinking about a timing issue like the one I fixed on remax pre-asic.
Replacing the 8 device by 2 devices change le capacitive load on the signals.

[Bryce]

I am more thinking about a timing issue like the one I fixed on remax pre-asic.
Replacing the 8 device by 2 devices change le capacitive load on the signals.

Quite possible, but why would it happen later and not from day one? What degraded enough that the timing became an issue?

Bryce.