[pulseaudio-tickets] [Bug 53892] New: Distorted sound with time-based scheduling

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=53892

             Bug #: 53892
           Summary: Distorted sound with time-based scheduling
    Classification: Unclassified
           Product: PulseAudio
           Version: unspecified
          Platform: x86-64 (AMD64)
        OS/Version: Linux (All)
            Status: NEW
          Severity: normal
          Priority: medium
         Component: core
        AssignedTo: pulseaudio-bugs at lists.freedesktop.org
        ReportedBy: david.lynam at redbrick.dcu.ie
         QAContact: pulseaudio-bugs at lists.freedesktop.org
                CC: lennart at poettering.net


Created attachment 65902
  --> https://bugs.freedesktop.org/attachment.cgi?id=65902
PA alsa sink log.

Using PulseAudio (with a null sink and loopback) as the audio input for ffmpeg
causes severe distortion in the output after 20-40 seconds of encoding. This
occurs in any case, i.e. while recording video (x11grab) or simply dumping to a
.wav file.

The ability to record/stream should be equivalent to Xsplit/FFsplit on Windows:
the latter uses ffmpeg and is mostly smooth.

Additional info:
Linux 3.4.8-1-ARCH #1 SMP PREEMPT 
pulseaudio 2.1-1
alsa-lib 1.0.25-1
alsa-utils 1.0.25-3
ffmpeg 1:0.11.1-1 and N-43565-gaec9390
pavucontrol 1.0-1

The same problem is experienced by at least one other person:
http://www.linuxquestions.org/questions/linux-software-2/pulse-audio-module-loopback-distortion-942066/

Steps to reproduce:
Set up a null sink and loopback monitor of sound card output to the sink. Run
ffmpeg (-i alsa -f pulse) and use pavucontrol to direct the monitor of the null
sink to ffmpeg.

Open the output file or view the stream (if streaming); around 30 seconds into
the video, sound will suddenly contain loud glitches, distortions and
increases/decreases in speed and pitch.

Workaround:
Modifying /etc/pulse/default.pa to re-enable interrupt-based scheduling
(load-module module-udev-detect tsched=0) seems to alleviate the distortion
problem but also increases CPU consumption and makes sound unusable in Wine.

Variables which have been tinkered with to no avail:
(PA) real-time priority
(PA) high priority
(PA/ffmpeg) sampling rate
(pavucontrol) muting different inputs, outputs, removing microphone/USB headset
(ffmpeg) building from git
(ffmpeg) changing audio encoding, trying every thread value from 0-6 (6-core
CPU)

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA Contact for the bug.
You are the assignee for the bug.