Re: [LAD] [64studio-devel] [64studio-users] Correlation of alsa -p value and hw MIDI jitter - Linux-audio-dev

14 Jul 2010

On Wed, 2010-07-14 at 14:12 -0700, Devin Anderson wrote:
...
  On Wed, Jul 14, 2010 at 12:43 PM, Ralf Mardorf
 &lt;ralf.mardorf(a)alice-dsl.net&gt; wrote:
  On Wed, 2010-07-14 at 12:30 -0700, Devin Anderson
wrote:
  On Wed, Jul 14, 2010 at 10:29 AM, Ralf Mardorf
 &lt;ralf.mardorf(a)alice-dsl.net&gt; wrote:
  Hi :)
 delayed by a thunder-storm I could do another test.
 --snip-- 
 So, what you're saying is that your MIDI device and software synth
 sync up less and less as you raise the period size. 
 Yes :).
  I had presupposed
 before that your MIDI device was triggering *after* your software
 synth, but it occurs to me that it might be the other way around.  Do
 you hear the audio from your software synth first, or from your MIDI
 device? 
 I can't say it today, now I do some office work. I had the impression
 that it might vary. Sometimes the virtual drum sampler and sometimes the
 standalone drum sampler was played earlier, I need to check this ASAP.
 For older tests with my USB MIDI device it was exactly that way, that
 jitter had positive and negative delay. At least the recorded waveforms
 of external MIDI equipment (when I used USB MIDI, now I'm using PCI
 MIDI), were recorded by Qtractor, before theoretically the MIDI event
 was send ;). Note! Qtractor had no latency compensation, all recorded
 audio of external MIDI instruments should have (positive) delay, but
 negative delay. 
 If it ends up being the case that your MIDI device is being triggered
 before your software synth, then I'm guessing that the issue here is
 not MIDI jitter.  I'm guessing the issue is that the latency that's
 imposed by JACK on incoming and outgoing audio is not imposed on
 incoming and outgoing ALSA MIDI.  So, while the audio coming out of
 the software synth is delayed by a certain amount of frames imposed by
 JACK, the audio coming out of your MIDI device is only delayed by the
 latency of the ALSA drivers, the latency of the MIDI ports, the
 latency of your MIDI device.
 This would certainly explain why the problem gets worse as you raise
 the period size, and could explain why you had positive and negative
 delay in your older USB MIDI tests, as the reported MIDI jitter in
 your tests was *far* worse in your older tests than it is now.
 At the moment, I happen to be doing some work in JACK 2 that could
 potentially solve this issue by enabling MIDI to sync more closely
 with audio, so I'm very curious to know if my suspicions are correct.
 Please keep me updated. :) 
Should I build JACK dummy packages for 64 Studio and daily get JACK2
from svn co http://subversion.jackaudio.org/jack/jack2/trunk/jackmp ?
I wonder if this should be cross-posted to LAD?
On LAD and the 64 Studio list are people with much knowledge and your
reply might hit the nail on the head.
- Ralf