epoch1970 wrote: > I have tried many things To me this is an extremely complex topic, I've > been turning dials at random so let's not consider I have the final say. > I wouldn't say it is drifting, it's more like I can't get both to align > perfectly on the same tick. I won't try synching 2 sw players, my other > Alix is performing its regular job as a router! > > Yesterday I compiled a linux kernel with real-time extensions. This > should be the end of any drifting, according to this OS. After applying > this: http://subversion.ffado.org/wiki/IrqPriorities I still can hear > some echo. > The machine is a bit unstable, I have squeezelite locking up if it is > set to a higher priority than SBS (this happens to me on non-RT kernels > too), and a high priority SBS means using the web GUI interrupts music > > So I'm just starting to find out if this setup can perform better than > my non-RT one (which works well, as said above.) > > My kernel is compiled with rt-preempt, 1000Hz tick (tickless disabled), > no power management. Uname: Linux prades 2.6.33.7.2-rt30rt-1khz-nopm #1 > PREEMPT RT Fri Jan 11 23:54:25 CET 2013 i586 > The RT kernel comes from pengutronix.de, they have debian repos. > The only shared IRQ is between usb1.1 and usb2.0, and there is no usb1.1 > device connected. > The audio card is a cheap C-media dongle, this one has usb2 when my > regular V-Dac mk 1 doesn't. > I invoke squeezelite as this: chrt -f 65 /usr/local/bin/squeezelite-pa > -n Prades -b 16384:16384 -o front:CARD=default,DEV=0 -a 10:4 -z > 172.17.255.87 (I think you can disregard the -b option, and I'd like to > try -a 5:4 next and see if the system can follow.) > I am synching an SB3 and the alix, in the same room. The sources are > about 5 m apart. > > Right now I need to recover from the last lockup ;) and continue testing > on highly syncopated music, but I am almost certain this setup cannot > work the same as the hw players. I have another SB3 hooked to my V-Dac > where squeezelite is located, I can compare easily. > It's getting very, very close, mind you. The value of Triode's > contribution is exceptional. > But I'm ready to admit we were not giving good people good money for > nothing, and that the linux pc-based $30 Transporter is a dream.
There's a fundamental difference between usb audio and hardware players with tuned device drivers - that is what current delay does alsa report for the device. In the case of the hardware players I believe its well tuned and accurate as Logitech could focus on tuning their own device drivers. In the case of usb-audio it depends on the version of alsa (more recent alsa kernel drivers have more code in this area, but I've not tested) + any buffering in the device which linux will know nothing about. The recent usb-audio code appears to think it knows when the latency is less than 2m and will log otherwise, but I would want to try multiple versions and also there will be an uncertainty due to the device which must do some level of buffering. ------------------------------------------------------------------------ Triode's Profile: http://forums.slimdevices.com/member.php?userid=17 View this thread: http://forums.slimdevices.com/showthread.php?t=97721
_______________________________________________ discuss mailing list discuss@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/discuss