Glitchy ANTIC chip

[+Nezgar]

I've noticed in one particular ANTIC chip of mine (NTSC) seemed to have occasional subtle pixel flickering issues on some applications & games.. Was never really able to pinpoint it, until I happened to try "Encounter!" which presents the glitchiness prominently.

 

This ANTIC chip behaves exactly the same in different machines, even with socket replacement. (Which I was doing anyway in preparation for tf_hh's 512KB SRAM board) Done a lot of chip swapping between the machines, so I'm not sure exactly which machine it was originally in now (one of 1 600XL, or 2 800XLs)

 

Anyone else seen an ANTIC chip behave this way? Is it at all common? I presume there's nothing that can be done to 'fix' it other that replacing it, but I'm curious what specific graphics mode / trick is triggering it.

 

[Rybags]

Strange one - first guess was character glitching but Encounter uses bitmap mode D for the playing area graphics.

 

Tricky stuff? Not a bit of scrolling to be seen, so the CPU would be moving all the data around. Maybe glitches when successive accesses are occurring? But it might be the case that the game uses double-buffering.

AFAIK the pillars are part of the bitmap, the enemies use PM graphics.

[_The Doctor__]

use de oxit and a tooth brush and scrub right up to and on the plastic between legs and under them... I see what looks like either rust or HEAT discoloration of the package... I have seen residue cause issues in the past.... make sure the ANTIC matches the system also

[+Nezgar]

@Rybags, Indeed it does look like double buffering is used in Encounter. It's less apparent in the posted video, because it looks like youtube converted my 60fps video to 30FPS, so for quite a few second intervals, you don't see it at all - but it was pretty persistent throughout the original video. This tells me that most of the time, the pattern only shows every other frame.

 

Maybe some of ANTIC's register RAM is bad/flakey.

 

@_The Doctor__ I'll try cleaning the chip. I doubt that will achieve anything since it is so consistent. The darkness you see is probably just my finger grease from repeated removals, and pushing to insert.

[Rybags]

I should have seen this before but here's the puzzler:

 

The glitches are single scanline though it's displaying a double scanline mode (OS Gr. 7) - in this mode the bitmap lines are only read once.

There does seem some correlation - it looks like the noise is fairly consistent, ie appearing in the same horizontal positions. Vertically it's hard to tell but it might also be on the same scanline for the graphics mode.

The fact it's the same on multiple systems would suggest it's definitely the Antic at fault.

 

It'd be interesting to see if it has the same problem with demos that use stretched Mode F lines in GTIA mode to get the 80x48 resolution (in this case Antic reuses the data 3 extra times instead of 1 with Encounter).

[+Nezgar]

Yes the horizontal (and vertical) position of the glitch pixels stay constant in the same spot (on alternate frames) if I do not move within the game. The positions of the glitches only change depending on the composition of other elements when moving around the game.

 

I'm trying to find some other examples, but indeed I have seen subtle glitching on other mode F screens in other demos, barely noticible only because I'm looking for it. Ie the end of the bad apple demo where the girl paints down at the end, and the 'twirlee' demo, there seems to be extra 'fuzziness' on the edges of the animated column, which also looks like a big mode F screen manipulated with line by line DLI's.

 

Will you try to find some more that show something worth capturing - open to suggestions.

[+Nezgar]

Tried some more titles this evening to look for and try to capture some other examples.

 

Looking closely at Encounter! we can see that the glitches happen on the 2nd of the two mode 7 scan lines. I also highly suspect they horizontally line up with character boundaries. The color of the dots is the same as the enemy ship. They fade to dark red in sync with an enemy explosion.

 

This game plays music and does a rainbow pattern on the bottom text during the title screen. Music interrupt related? This one also seems to show the dots lining up horizontally with character boundaries, and the 2nd scanline of each Gr.7 line.

 

Molecule man doesn't play much for sound, but plays blips and beeps while moving the character. These two shots were hard to catch the brief glitches, also seemingly related to the audio (maybe interrupts?)

 

The twirlee demo sometimes seems to add some ghost pixels around the 'column' in this case we can see the loose pixels off to the right.

[Rybags]

Bottom line - multiple machines and same failure means it's definitely the Antic, unless by some fluke you've got a bad PSU and it's manifested the same problem by coincidence which I feel is highly unlikely.

 

Unless it's something overlooked like sufficient corrosion on a pin to produce a current drop or spurious I/O on one of the buses, then there's not really anything you can do other than replace it.

[+Nezgar]

I fully agree it's a fault in the chip, at this point was just curious about it's failure mode, and thoughts about what routines / modes trigger it.

 

Cleaning chip no diff as expected. Since the chip is 'mostly' good and doesn't crash the computer, it will still be useful as a spare for testing, etc.

[Level42]

Why is there an extra socket between that ANTIC and the motherboard socket ? Wouldn't help I suppose ?

[+Nezgar]

Good eyes. I found sometimes that helps improve contact when troubleshooting before replacing an entire socket. No difference with/without in this case, even with the original socket replaced with machine socket on a different board. (Round pins/holes)

[Level42]

Aha !!

 

I kinda find it funny to find VLSI's with (slight) issues. I have a POKEY that "sings of key"....

 

I'd love to how that happens...inside the chip...probably some production flaw....sume tiny particle of dust managing to escape the clean-room equipment ..

[_The Doctor__]

could be connection migration, the materials in the chips have impurities that lead to crystal like protrusions... I am sure someone here has the technical terminology.... like everything else. I just state things at their base....