15 Jul
2010
15 Jul
'10
8:15 a.m.
Transport issue for Qtractor - has impact to the jitter issue
So the advice to use amidiplay is something I'll follow soon.
Hi all :), hi Robin :), hi Devin :)
Robin, for 64 Studio 3.3 alpha the group has got read and write access
to /dev/hpet too. Btw '[...] | sudo tee [...]' for 3.3 alpha isn't good,
regarding to the enabled root account. Anyway, for this recording test I
kept the value 64 for hpet/max-user-freq, but I'll test higher values
soon.
Devin and some others want to know if the drum module is played before
FluidSynth DSSI is played.
JACK2 doesn't start with -p4096, so I started JACK2 with -Rch -dalsa
-dhw:0 -r44100 -p2048 -n2 to increase the unwanted effect, to get a
clear result.
Without recording it's clear audible that the external drum module
always is played before FluidSynth DSSI is played, plausible regarding
to the audio latency (stupid idea to use such high latency ;), so
there's the need to do an audio test with lower latency, to see if
jitter might change what instrument is played first.
Instead of a rhythm I did record 4 to the floor at 120BPM to 'see' the
jitter.
I did record FluidSynth DSSI by the sound card too. Left channel the
drum module, right channel FluidSynth DSSI and regarding to Qtractor's
graphic of the wavefrorms, there is jitter for both!
Ok, next recording with -Rch -dalsa -dhw:0 -r96000 -p512 -n2.
Without recording it's already audible that the drum module is played
first all the time ...
and it's visible too. Again there's jitter for both. The audio recording
of the drum module always is before the MIDI event. The recording of
FluidSynth DSSI sometimes is before and sometimes after the MIDI event.
There's no offset for the audio track.
I kept -Rch -dalsa -dhw:0 -r96000 -p512 -n2 and recorded FluidSynth DSSI
alone, internal Linux, without using the sound card.
The audio recordings are before the MIDI events and there's jitter. I
never noticed jitter internal Linux before.
I need to repeat the test ASAP, but by using 64 Studio 3.0 beta and
perhaps an older version of Qtractor.
Playing FluidSynth DSSI by MIDI and the recording made internal Linux in
unison, there isn't audible jitter. But after starting playing sometimes
MIDI and the audio recording are perfectly synced and sometimes there's
delay, real delay between the recording and MIDI, not only an early
reflection like effect (but without audible jitter, the jitter only is
visible by the waveforms).
$ qtractor -v
Qt: 4.5.2
Qtractor: 0.4.6
More maybe tomorrow.
Cheers!
Ralf
15 Jul
15 Jul
8:33 a.m.
New subject: Transport issue for Qtractor - has impact to the jitter issue
On Thu, 15 Jul 2010 10:15:53 +0200, Ralf Mardorf wrote:
Transport issue for Qtractor - has impact to the
jitter issue
what transport issue? oh i see... :)
qtractor has in fact a known trouble with jack transport sync when slaving
to period/buffer sizes greater than 2048 frames. it skids. maybe that's
exactly what you're experiencing when running jackd -p4096.
i'll recommend for you to turn jack transport mode to "none" or
"master"
only, never "slave" nor "full" on those situations (in
view/options.../audio/jack Transport).
another recommendation would be to grab the latest from svn trunk where
the said trouble has been already mitigated :)
cheers
--
rncbc aka Rui Nuno Capela
rncbc(a)rncbc.org
8:57 a.m.
Hi Rui :)
no, a misunderstanding, at -Rch -dalsa -dhw:0 -r96000 -p512 -n2 there's
an issue internal Qtractor, no JACK transport.
There's a MIDI track with FluidSynth DSSI and there's an audio track
with a recording of FluidSynth DSSI, done by Qtractor audio out directly
connected to audio in.
The recording is several bars. If I start somewhere within the bars
Qtractor might play the MIDI track and audio track in sync. If I stop
and start playing again, somewhere else, the same tracks are played with
delay. The delay can be around 1/16 note (this is around 125 ms @ 120
BPM, not measured, I'm able to here what a 1/16 note is ;).
It's a kick played at each beat. Sometimes when playing the MIDI and
audio kick are played in unison and after stopping and starting again,
there can be delay of around 1/16 note or sometimes less, but 1/16 note.
The MIDI and audio track aren't synced all the time after pushing play.
Cheers!
Ralf
6 p.m.
New subject: Transport issue for Qtractor - has impact to the jitter issue
On Thu, Jul 15, 2010 at 1:15 AM, Ralf Mardorf
<ralf.mardorf(a)alice-dsl.net> wrote:
Devin and some others want to know if the drum module
is played before
FluidSynth DSSI is played.
JACK2 doesn't start with -p4096, so I started JACK2 with -Rch -dalsa
-dhw:0 -r44100 -p2048 -n2 to increase the unwanted effect, to get a
clear result.
Without recording it's clear audible that the external drum module
always is played before FluidSynth DSSI is played, plausible regarding
to the audio latency (stupid idea to use such high latency ;), so
there's the need to do an audio test with lower latency, to see if
jitter might change what instrument is played first.
As we discussed yesterday, I don't think this is due to jitter. I
think this is due to the fact that the (necessary) latency imposed by
JACK on audio is not imposed on ALSA MIDI.
The MIDI path to your soft-synth probably looks something like this:
MIDI controller -> MIDI in port -> ALSA -> Fluidsynth
... after which Fluidsynth generates audio that travels a path that's
something like this:
Fluidsynth -> Jack -> audio out ports
Jack will impose a certain amount of latency on the audio before it's
output to the audio out ports. In the case above, I think it would be
around 90 milliseconds or so. There might be some other, minor
latency (i.e. internal buffering, etc.), but IDK, as I haven't seen
the code.
The MIDI path to your external drum module probably looks something like this:
MIDI controller -> MIDI in port -> ALSA -> MIDI out port ->
external drum module
... after which the drum module generates audio that travels a path
that's something like this:
drum module -> drum module out ports
The drum module imposes some sort of latency, but it's clearly not as
high as the latency imposed by JACK with the settings above. Note
that nothing is routed through JACK on the way to the drum module.
Solving this problem is complex. It's unlikely that your drum module
allows you to specify its latency. AFAIK, you can't directly specify
the latency of ALSA MIDI ports, but I have very little experience with
ALSA and could be totally wrong.
If Fluidsynth supports JACK MIDI, then you might solve this problem by
routing your MIDI through JACK using the ALSA 'seq' or 'raw'
interfaces. MIDI that runs through JACK is subject to the same
latency as audio up until it reaches the MIDI drivers, after which
it's up to the driver to sync audio and MIDI if the MIDI port isn't
subject to the same latency as audio. I know that the FFADO driver in
JACK 2 syncs MIDI closely with audio. After looking over the ALSA
drivers, I'm pretty sure they try to sync MIDI with audio too by
adding a delay of `nframes` to the time that the MIDI message goes out
(I could be wrong, as the code for the ALSA MIDI drivers is a bit
complicated).
--
Devin Anderson
devin (at) charityfinders (dot) com
CharityFinders - http://www.charityfinders.com/
synthclone - http://synthclone.googlecode.com/
5282
days inactive
5282
days old
linux-audio-dev@lists.linuxaudio.org
3 comments
3 participants
participants (3)
-
Devin Anderson -
Ralf Mardorf -
Rui Nuno Capela