On 10/01/14 00:40, Zenaan Harkness wrote:
Now I have a "simple" audio setup, and still getting dropouts: $ ps x | egrep -vi disk\|gvfs | egrep alsa\|jack\|pulse\|mix\|vol\|aqu 3946 ? Ssl 0:00 xfce4-volumed 7849 pts/17 SLl 0:14 qjackctl 7854 ? SLsl 0:34 /usr/bin/jackd -dalsa -dhw:PCH -r44100 -p1024 -n2 -D -Phw:PCH,7 7865 pts/17 SLl 0:22 /usr/bin/alsa_in -j cloop -d cloop -q 1 7866 pts/17 SLl 0:17 /usr/bin/alsa_out -j ploop -d ploop -q 1 7867 pts/17 SLl 1:21 python /usr/bin/jack_mixer 9504 tty1 SLl 0:31 aqualung There are NO jackd XRUNs (now that I've upped -p anyway). The drop outs sound like they go for around 50-150ms, I can't be sure. They 'sound' like there's a bit of fading going on, rather than a full drop out. Perhaps there's some sort of bit-rate conversion library which is running out of data? I would like some suggestions on approaches to debugging this. Willing to learn :) TIA Zenaan
Zenaanthis war against audio drop-outs ;-) has a long-standing history. Look through a few "audiophile" sites and fora. Keywords like realtime kernel, low latency, ... sadly, lots of out-of-date stuff.
I'm watching this thread eagerly since my dedicated little MPD server has occasional hiccups as well. Note that so far I haven't noticed a correlation between absolute CPU power and drop-outs: whether MPD runs on a desktop pc, a relatively modern ARM board (Mele A2000), or even the good ol' slug (Linksys NSLU2), there are drop-out every now and than (about one or two an hour). If anybody else here knew how to debug this further, say eg. how to identify and then log these occasional events, and then try and see what else the system is doing at that moment, that'll be fantastic. These drop-out last for far less than a second, ie. manual detection is almost impossible. (Beyond hearing them, obviously :-)
PS. 'dedicated' above means that I've removed everything I could from the system except those bits that are needed for mounting an NFS share, running MPD, and piping the digital stream directly (ie. via ALSA) to the output (S/PDIF in my case, with an external DAC). No re-encoding, re-sampling or re-formating. Currently this is on the Mele box, running headless, wheezy, kernel 3.0.
-- Klaus