kscharf wrote:Well some of the crystals for that frequency are no better that +/- 100ppm themselves.
Yeah, I don't think the 20 vs 50 vs 100ppm or more will make a visible difference. You can be pretty sure that most home video games back in the day were using as cheap of parts as they could get without ill effect. (Color's referenced from the burst, so any error on a per line basis will be tiny.)
We're probably also talking about pixel clocks in the low ~4MHz range, so even clock jitter in the <10ns magnitude will probably be invisible. (certainly with composite out anyway).
Interestingly enough, at room temperature my Mega644 starts up and runs from a 3.3V (and even lower) supply just fine. I suspect that 5V will give plenty of margin for the clock.
I have noticed that coming out of reset might be a little iffy on power-up (depending on power supply characteristics). I might put a little R/C or a reset generator on the AVCore board to make sure we're getting a clean reset after the power rail is up.
Knowing that 3.3V @ 28.6MHz might be OK for some applications-- it might have have been better to run the pixel port as current sinks and power the output ladder from a separate reference voltage so that video brightness is independent from the power supply. (That would allow running at lower voltages for battery operation without needing to mess with the DAC resistors to compensate the brightness. You could connect directly to the output of a LiIon cell for example and have constant brightness over battery voltage without a boost regulator to 5V)... But for UZEBOX1, simple/cheap is probably best.
-Clay