Back from the Grave: Atari 2600 (Skewed, garbled screen; buzzing)

The console that effectively created, dominated and then almost destroyed the videogame industry, all on its own: The Atari 2600.  Originally launched in 1977 as the VCS (Video Computer System) before being renamed upon the 1982 release of the Atari 5200, it was originally released to play upon the name and reputation of the Fairchild VES (Video Entertainment System) released a year earlier; itself renamed to the Channel F upon the release of the VCS.

The 2600 itself was, even by the day's standards, an unimpressive machine.  While it featured a 1MHz 6507 (similar to the 6502 found in the Apple II and VIC-20, just with fewer addressing pins), it had just 128 bytes (yes, bytes) of RAM, only allowed for 4 kilobyte cartridges (which was later expanded to 32kB by way of bank switching) and had almost zero graphical capability.  But developers managed to use this meager offering to exceptional effect, resulting in games that weren't Pong and securing its place at the top of the proverbial food chain (right up until ET, anyway).
This time, the unit in question is a 1982 "Vader" unit - one of the first to feature the 2600 name - which our father purchased in the early '90s in non-working condition with ideas to fix it himself.  We never had one of the more "mainstream" gaming consoles due to cost, and the slow failure of our C128D (by this point one of the FDD heads had broken) gave him the idea to seek out a cheap alternative.  As it would turn out, it would have been extremely difficult for him to fix at that time due to parts availability.

Essentially the system would sometimes produce a picture and sometimes produce a sound.  Any time a picture was present it was in obvious horizontal (though not solid) bands which skewed further to the right as the band progressed.  Any sound was just a constant buzz.  These results, however, were fairly consistent: the display and sound, when present, would differ from cartridge to cartridge, but each cartridge would always produce the same results.

Within the 2600 there are three main ICs: an MOS 6532 "RIOT" (RAM, I/O and Timer), a custom chip designated TIA (Television Interface Adaptor) and an MOS 6507.  These are also socketed from factory, which saved a good deal of time and risk.  Pretty much everything else within the (rather spacious) shell was off-the-shelf circuitry, making it (appear) extremely easy to test.

Since an initial probing of the CPU suggested it was working perfectly (the data and address lines were all active), we decided it was one of the remaining two chips that was causing the issue.  Removing the TIA resulted in no output, which wasn't an entirely useful piece of information but still relevant.  Removing the RIOT, however, resulted in exactly zero change to the output.  Seen as the RIOT controls talking to the cartridge and contains all of the RAM, this seemed a likely culprit for failure.

As good fortune would have it, the 6532 is still being made to this day by Rockwell Semiconductor for embedded applications.  While it took some time for them to be delivered, they were cheap and brand new.  As poor fortune would have it, changing this chip did exactly nothing.

We figured from this that the next most likely culprit was the TIA chip.  Unfortunately, as mentioned earlier, this is a custom chip made for the purpose by Atari - the only way to obtain one is from another 2600, which are becoming increasingly rare, or from NOS via a now-defunct service center.  That they are also different between PAL and NTSC units presented another problem.

We did manage to find one for a reasonable price in Germany, which presumably is a PAL unit (it was unspecified) and ordered it.  Much to our chagrin, however, this would also prove to be the wrong course of action.  Exactly zero change in output was observed.  On the plus side, unless both had suffered identical failures (possible but unlikely), this meant the original TIA was still in good order.

In hindsight, we learned that a failing/faulty TIA should produce perfectly straight, two-tone vertical bars across the screen.  That the picture was skewing horizontally should have lead us to suspect a timing issue straight up, and all timing is controlled by the CPU itself.

By process of elimination, this left only the CPU as the culprit, despite it giving all the indications of proper function.  As with the 6532, the 6507 is still being manufactured and was able to be obtained easily and cheaply.  Switching in a replacement brought the entire machine to life, even with the original RIOT and TIA in place; clearly the original CPU was doing something, but not what it was supposed to be doing.

And that ends this BftG.  Admittedly this isn't a very technical example - all we really did was replace one chip after another until it worked, which means we now have enough parts to actually build another (working) 2600.  But sometimes, especially early on, without the right tools or simply without the right knowledge, this is all you really can do.  And if it works...

Remember that electricity can be dangerous to both property and life.  If you don't respect it you can wind up seriously hurting or even killing yourself or others, and chances are it's going to be painful while it's happening.  Unless you're sure of what you're doing, avoid tampering with any electrical circuit - especially when it's live or mains powered - and always ensure you have a nearby assistant who can contact emergency services.