Some notes on AVR-DA timers

[T3sl4co1l]

FYI:

- TCB in "one shot" mode, isn't really triggered.  It's more that CNT is reset on event, and stops counting when it equals CCMP.  That's it, that's all the logic, there's no flip-flop like the text seems to imply, it's only ever counting because it's unequal, or stopped because it's equal.  So, like, STATUS[RUN] bit is a shortcut for CNT == CCMP, not actually indicating any internal state.

So you can trigger a cycle by writing CNT = 0, no events needed.

Or if you need to change CCMP to get different delays in real time, whenever CCMP is set below CNT, CNT keeps on counting and overflows and wraps around.  You have to write both equal to prime it for another go (but notice, because the comparison goes unequal then equal, it triggers a new output event / WO pulse / interrupt).

I haven't checked out the other modes in detail yet but I think they will be consistent with similarly simple logic.  I suspect there is no hidden on/off register, it's either stopped because compare, or because it's gated by something else like an input/event.  Why the text doesn't just go out and say this, I don't know.

- TCD is written confusingly.  The thing that stuck with me is the use of words "command" and "synchronization".  Is it talking about some special, additional logic or function?  Is every transaction routed through some sequential gate, and you can only write one register at a time until each operation goes through?  Is it really just the usual clock-domain-crossing registering?  Is it like at the bus entrance to the peripheral, instead of to each register?  (Actually come to think of it, that's probably what they're doing.)  I think they mean "synchronization" as, simply writing to any register on the clock domain, and "command" is only specific bits (CTRLE and CTRLA[ENABLE]).  Maybe that doesn't really matter much (unless you're using it at a much lower clock rate, any register write will just about instantly go through), but one of those things that, if you don't do it just right, what one thing did you screw up, how do you know?

Take note of the errata as well.  Like the AVR64DA28-64, enabling TCD ouputs, enables all four outputs, not just the one and other pair they're documented as.  Not really very useful anyway (they only mirror the first pair?) but it's two pins you can't use for other things.

Tim