Hi Defex
I tried the overflow discard method and it makes quite a difference to the LFO. I makes it really easy to get sync with my step sequencer which is actaully a big improvement! I think I'll switch between the two methods depending on which mode I'm in. The more frequent your LFO updates, the less significant this change is. I'm using really coarse LFO updates becuase I like the steppy sound and I'm too lazy to implement something smoother.
On the serial thing - I actually presumed that when I sent something on serial it was just added to a buffer somewhere to be clocked out by the serial library on an interrupt basis when there was idle time. If it's just being sent on demand then that's pretty worrying for timing! I've just confirmed that you're right by looking at the docs. I'm going to have to either implement a buffer (ie. print out 10 chars every loop if waiting, once all important stuff is done) or turn off the diagnostics by default and initiate with a console keypress.
That's a good point about the memories. I have the luxury of the "memory" of my analog pots which set up all the Fatman control voltages, so that the only things lost on power-down are the midi-related settings like sequencer mode (arpeg, step, chord trill) and the contents of the sequencer buffer. I was thinking of using the onboard 512 bytes of Flash to store sequences - you could do the same with your CC interface. Then you can use a pickup-latch method for incoming CC - reject incoming CC until equal to stored value, then latch up and accept CC.