Hi All,

Today I found an absolute new and strange behavior reparing an Atari XEGS. I never have had such an error before neither I found something similar here. So maybe this could help other people with the same strange issue.

The XEGS I got comes with the fault description "System hangs if Missile Command is started or sometimes the game just resets" (reset = start from the beginning, same if you would press START at any time). I can reproduce this error easily.

So first, I watched the SYNC signal at the CPU - but even when the game hang, thereīs activity on SYNC. So I decide first to change the ROM - I have a lot of ROM failures specially by the XEGS. After ROM the MMU was changed. But the fault remains. Ok, next step would be the CPU. CPU failures are from my experience the 2nd most related things after RAM memories. CPU exchanged, but you can imagine... that also didnīt help.

BTW: Before I change any part, of course I watched every single pin of the CPU, GTIA, ANTIC and the FREDDIE. All fine.

I connect a keyboard, start BASIC, enter some simple calculation endless loops with PRINTīing them, this works for hours. I attached an SIO2SD, load Missile Command (file version) and... it works. No problems at all.

Ok, which components might cause the CPU to make a break... ANTIC? GTIA? All signals (RDY, HALT, IRQ) looks good. After all, I didnīt found other ideas, so I decide to remove all chips, solder sockets in and put well-known working parts into the XEGS mainboard.

Start the next test, no keyboard, Missile Command starts and... hang. Now - wished I had test this first, a lot of spent hours would have been saved - I connect the keyboard, start the XEGS with SELECT hold down and Missile Command starts. I didnīt know this before, that Missile Command could be started also with keyboard instead of Atari Basic. I donīt have an XEGS for my own, so I havenīt play around much with this.

WITH keyboard attached the Missile Command works fine. No problems.

Next, I examine the POKEY more then before. Without keyboard I see on the scope, the IRQ line is sometimes low for over 10 times of a PHI2 phase! At every time IRQ goes low for such a long time, the game hangs, resets or makes other weird things. With keyboard attached this never happens, IRQ remains constantly high while the game is running - what I expect, then Missile Command didnīt read the keyboard nor SIO etc., it uses just the joystick ports and console keys, which are attached to the GTIA.

After some more examining the POKEY I saw, that pin 16 and pin 25 of the POKEY - KRx lines - are heavy floating from -0.5 volts up to 2 volts. Now it would be clear: POKEY "thinks", a key a pressed, generate an IRQ, but the OS (which is active during the game, ATTRACT will also work) canīt read the source. In case of the bad floating signal the POKEY will generate a long IRQ, longer than normal, which cause the CPU to halt - like some of the "CPU breaker" with R/C links from the 80s.

But... the XEGS is MADE for usage without the keyboard... so why this happens?

Checking with another XEGS boards, the culprit was found very fast: It was the parsimony of Tramielīs area, removing some necessary parts to save some cents. The "good working" XEGS has line filters in the KRx (and other) lines, which will prevent these open inputs from floating around and beeing using like an antenna. At this special board, Atari replaces these line noise filters by simple 0R resistors. See this board:

XEGS_without_LF.jpg

The fix is absolute simple, after all Iīm so silly not to find this earlier, but... now it works. And maybe it helps others from spending so much time to repair an XEGS ;-)

Just solder two 1nF capacitors between ground and the both KRx lines - thatīs all. Of course you can use wired capacitors, too. Now thereīs no floating inputs and the system works fine - with or without keyboard...

Attached_1nF_Caps.jpg

Jurgen