17 Jul
2013
17 Jul
'13
10:09 p.m.
On Wed, Jul 17, 2013 at 05:21:12PM +0200, IOhannes m zmoelnig wrote:
when running zita-a2j resp. -j2a in verbose mode, i
understand that
the 1st number (the "loop error") should be as close to zero as possible.
in practice they are varying wildly betweeen +16 and -10.
how bad is this?
It's not that bad, it just indicates that there is some timing
jitter. This can happen for a variety of reasons, see below.
more troubling however, is that whenever i start an
alsa application
(e.g. alsa_delay), one or both zita-*2* applications complain with:
Detected excessive timing errors, waiting 15 seconds.
This may happen with current Jack1 after freewheeling.
You could try playing with the value in jclient.cc, line 437
to avoid this. Note that the delay error value printed by -v
is averaged over 1/4 second, while the one tested is not.
(btw, i'm using robin's
jackd2-1.9.10+nobus+20130114gite4730c6c-NMU1)
after 15 or so seconds the waiting application recovers, only to
hick-up again after some seconds (see end of email).
i'm running the programs as follows:
jackd -dalsa -dhw:0,0 -r44100 -p1024 -n2
zita-j2a -v -d hw:Loopback,0,1 -c2 -r44100 -Q16 -p2048 -n2
zita-a2j -v -d hw:Loopback,0,0 -c2 -r44100 -Q16 -p2048 -n2
alsa_delay hw:Loopback,1,0 hw:Loopback,1,1 44100 2048 2
This is really a combination of all possible worst-case scenarios.
* Does your HW have a high resolution timer that Jack can use
for its microseconds timer ? If not there will be trouble.
* Same for the aloop device. It emulates a HW device at some
configured sample rate, and therefore depends on an accurate
timer in order to do this without loads of jitter.
* I have *no* idea of the correct implementation of the DLL used
in Jack2, but zita-ajbridge does critically depend on it.
* For anything USB, 44.1 kHz is really asking for trouble. With
48 kHz you have at least a a simpler relation to the USB interrupt
rate of 1 kHz, and if that doesn't help using -n 3 probably will.
I don't know 'Andrea Electronics'...
* Keep in mind that zita-ajbridge was not designed to be used
this way. Its primary use is to add an additional (decent)
soundcard to Jack, and I have my doubts about aloop.
finally: i'm wondering a bit on how to choose
proper values for
period-size and fragement size.
that is: should zita-*a* period/nfrags match those of jack and/or
those of alsa_delay?
For Jack vs. zita-*2* there are no restrictions. In normal use
(that is adding a real sound card to Jack), you would make the
period size of zita-*2* shorter than the one used by Jack
(unless that is already quite small) to avoid additional latency.
I don't know how having different period sizes on both ends of
an aloop would affect it.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)