Glitchy 6-switcher

[+Andrew Davie]

I wonder, though, if other machines will behave similarly, just with the application of heat from a hairdryer.

 

[Omegamatrix]

@Andrew Davie I took a look today at the routine for the 32 char and had an idea. Can you please try this out? It works on my 4 and 6 switch, but truthfully neither had a problem with the previous code.

 

32 Char Interlaced (20210130_B).zip

 

 

[+Andrew Davie]

6 hours ago, Omegamatrix said:

@Andrew Davie I took a look today at the routine for the 32 char and had an idea. Can you please try this out? It works on my 4 and 6 switch, but truthfully neither had a problem with the previous code.

 

32 Char Interlaced (20210130_B).zip 9.9 kB · 0 downloads

 

 

 

This test version (all 3) appears to be working perfectly on the glitch machine.  

I heated up the TIA and the PlusCart menu was super-glitchy. Ran the test version, and it had a perfectly-formed 32-char line.

Cooled down, heated, cooled, heated. Not a sign of any issues at all.

 

[+Andrew Davie]

Just to clarify, and as earlier described... but to bring it all together in one post:

The machine consistently exhibited glitchy behaviour after being left turned on for an hour or so. I could, if I waited, observe the glitch as it first started with just minor pixel issues, but fairly rapidly (over a minute or so) developed into a "fully glitchy" mode. At that point, I found that I could power-cycle, but it was glitchy immediately. If I left the machine turned off for a half-hour or so, and tested again then it was normal. But the glitchy behaviour returned after another 15 minutes or so.

 

After noticing this a few times, I was asked to check if application of heat (i.e, via a hairdryer) could accelerate the process.

 

 

I found that a "bare" circuit board took a lot longer to show glitchy behaviour - and when it did, it was relatively minor.  However, even minor application of heat rapidly caused increased pixel shifting (P0 only), and eventually lead to a reasonably stable "hot" state where P0 was mispositioned but at least consistently mispositioned.

 

I applied IPA in the form of a few squirts from a bottle, and the P0 misposition (glitch) slowly "devolved" back to the normal position. The transition period was usually about 5-10 seconds, depending on how much IPA I had applied.

 

I was able to repeat this multiple times -- heat and see the glitches (with the onset time and transition period of glitches seeming to be proportional to the amount of heat I applied) -- cool and watch the glitchiness return to normal.

 

On returning the board to the proper RF box and sealed up in the '2600 shell (i.e., assembled as normal), once again it took an hour or so for the glitch to appear, and behaviour at that point was -as far as I could see -exactly as described from the start of this clarification.

 

I have tested "natural" and "enforced" glitching many times with this machine now.

 

Right now I'm most of the way to having @Omegamatrix's replacement kernel installed. A full-screen menu testbed is currently running on the machine (which is bareboard at the moment).  I don't expect it to glitch - but even if it were to do so it might take several hours as I don't have access to the hairdryer until tomorrow.  Having said that, I successfully tested his one-line kernel (which did not glitch) alongside the regular menu kernel (which did glitch), so I'm fairly confident the new code will fix this issue once and for all.

 

 

One data point -- air temperature for most of these tests was approximately 28°C - 30°C.

 

Edited January 31, 2021 by Andrew Davie