Qjackctl gives some illuminating information on the problem:
02:05:18.063 XRUN callback. (54) delay of 46451.000 usecs exceeds estimated spare time of 45357.000; restart ... delay of 46451.000 usecs exceeds estimated spare time of 45357.000; restart ... 02:05:18.068 XRUN callback. (55) 02:05:36.407 XRUN callback. (56) delay of 46452.000 usecs exceeds estimated spare time of 45359.000; restart ... delay of 46452.000 usecs exceeds estimated spare time of 45359.000; restart ...
Anybody see a pattern? ;-) I wonder what this "spare time of 45359.000" would mean?
Things which seem to have helped so far:
- Setting -blocksize to 64 [equal to midi timing]: reduced glitches
significantly, but not completely. After changing blocksize, I got slightly different numbers out of Qjackctl's window:
delay of 45481.000 usecs exceeds estimated spare time of 45332.000; restart ...
I guess that means I gained .01 ms by changing the blocksize from 2048 to 64 ;-)
- Shut down eth0 and/or removed USB mouse [both eth0 and usb-uhci share
IRQ 9 with the HDSP]: appeared to reduce glitches, but not completely.
After some experiments, I am a bit happier, but not satisfied. The glitches still appear most often when several controls are being used at once, but sometimes just for a single controller when I haven't touched them in a while.
Question: can someone explain -sleepgrain to me? I know it affects MIDI, but I cannot exactly understand *how* from the docs.
Also, would a higher -audiobuf than 50 or 100 help here? I would rather keep latency lower than this, of course...
Thx again for any suggestions, D.