_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

It's hard to draw conclusions, given that the crashes are to some extent random. But "10MHz has never worked before" does catch my attention! And you're saying that's made possible simply by removing an LCD that wasn't doing anything anyway?? And (to be clear) it was not attached to the CPU bus, only the port pins of a VIA?

If so, it begins to seem as if the power supply was struggling -- you need to check the voltage during actual operation. Either it's inadequate or was poorly connected, or perhaps the LCD is defective in a way that causes extra current consumption. (Does the VIA get warm, btw?)

-- Jeff

_________________
In 1988 my 65C02 got six new registers and 44 new full-speed instructions!
https://laughtonelectronics.com/Arcana/ ... mmary.html

One the reasons those LCDs are very fussy about their startup sequence, I believe, is that the need, and have, internal high-voltage generators to drive the display, and it takes some cycles for a charge pump to get that internal supply to spec. It's not impossible that the charge pump is quite needy, for its own supply, and not impossible that it periodically needs to fire up to get that supply back into spec. In which case there could be a periodic power demand. (Lots of speculation, there, of course!)

Yes, the LCD was connected only to the 6522 and the VIA didn't get warm. My power supply is an old ATX one which can supply around 23 amps at 5V. Although come to think of it I have modifed it to use one of those little volt and ammeters from ebay. The voltage does fluctuate at startup (sometimes it'll be 5.1V, others it will be 4.8V, but the current used is always the same and the crashing happens no matter the voltage). Perhaps I should just use the 5V directly and see if the meter is causing an issue.

Steve

I'm not sure how much you've said about your on-board power supply decoupling. Best practice is to have an electrolytic where power connects to the board, and a smaller bypass cap across the power terminals of each device.

But more importantly, perhaps, if you want to get to the bottom of this, is to consider your experimental design. It's not clear to me if 10 minutes is always the exact life of the machine, or an average, or a single measurement, or if it's exact or approximate. If you tabulate at least a few consecutive times to crash, that might help.

Also, the question of what the machine is doing. The simplest case is a tight loop with no interrupts. Next up might be an ordinary idle state of a monitor or interpreter. Then, a simple program such as an LED blinker. Then, something deliberately meaningful, like a RAM test which loops forever.

If, on the other hand, you are using the machine for this and that, and find it crashes at some point, it's much harder to know what the cause might be.

I still wouldn't rule out your domestic appliances (or nearby industrial users) - if you bench supply reacts to a brownout by dipping the supply rail, that's an external cause. If there's a lift (elevator) in your building, or someone doing arc welding next door, that might affect your machine.