Comments on How To Fix The Sound Issues Between Skype 2.0 And Pulseaudio On Fedora 9

How To Fix The Sound Issues Between Skype 2.0 And Pulseaudio On Fedora 9 The main problem with Pulseaudio and Skype on Fedora 9 (maybe also on other distributions) is stuttering, crackling sound. This document describes how to fix the sound issues between Skype 2.0 and Pulseaudio on Fedora 9.

6 Comment(s)

Add comment

Comments

By: Andrig

I found your post when trying to get audio issues worked out with Skype. I am also running Fedora 9, and I have PulseAudio selected in the GStreamer Properties for sound in and out. I can only get audio working with the default device selection. If I select Pulse at all it doesn't work. If I select the default device, then things work until I play a Flash video (Flash 10), then it stops working.

I tried your changes (I already had the groups for my user), so I just changed the daemon configuration changes. I cannot get any sound at all with pulse. Any ideas?

By: Kjetil Birkeland Moe

I found this to help - though the daemon.conf already had those lines, I didn't need to add them. I simply altered the realtime-scheduling to 'yes', and that fixed it. This was on Fedora 10.

By: Pas_

Ok it was niece, everything but last line. Oh, And everything is default settings except "realtime-scheduling = yes" and "resample-method = src-sinc-best-quality".

Set resample-method with caution because it is able to consuming lot time of yours cpu. In default it is set to src-linear in ubuntu 9.04 and it met standard users requirements. In my case *best-quality take 100% CPU, in src-linear ~20% CPU. If you using PCU only to talks you can tray trivial, or maybe speex-float-N where 0<N<9, the 3 is default.

Im recommend to take look at two settings
default-fragments= in documentation say it is set to 4 in ubuntu 9.04 i have 8 and it is nice.
default-fragment-size-msec= they say default 25 i have 10. In my opinion it should by 25ms because specification of VoIP.
that settings mean how long bufor i have. 8*25ms = 200ms its long, 100ms in sys with lack of ram is sufficient.

If have more trouble take look at :
; default-sample-format = s16le
; default-sample-rate = 44100
; default-sample-channels = 2
i leave out that.

all my settings is like that:

high-priority = yes
nice-level = -11
realtime-scheduling = yes
realtime-priority = 5
resample-method = src-linear
default-fragments= 8
default-fragment-size-msec= 25

now it take about 23% of CPU in test speech on AMD Athlon 64 3000+ an ubuntu 9.04 64bit.

basic info: http://linux.die.net/man/5/pulse-daemon.conf

By: Robin

On my machine running Fedora 14, these advices did not solve the problem. However, it could be solved by replacing the line

load-module module-udev-detect

in file /etc/pulse/default.pa with

load-module module-udev-detect tsched=0

and restarting PulseAudio afterwards. By this, the timer-based scheduling of the PA Server is disabled.

Maybe this tweak will be useful for someone.

By: matthias

thanks a lot robin!

on my fedora16/64 bit/macbookpro5.5 this helped me improving my sounds.