After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 636294 - [pulsesink] Introduces artefacts into sound ("scratchy" playback)
[pulsesink] Introduces artefacts into sound ("scratchy" playback)
Status: RESOLVED INVALID
Product: GStreamer
Classification: Platform
Component: gst-plugins-good
0.10.26
Other Linux
: Normal normal
: git master
Assigned To: GStreamer Maintainers
GStreamer Maintainers
Depends on:
Blocks:
 
 
Reported: 2010-12-02 13:55 UTC by Colin Guthrie
Modified: 2010-12-02 14:35 UTC
See Also:
GNOME target: ---
GNOME version: ---



Description Colin Guthrie 2010-12-02 13:55:16 UTC
Hi,

Compare the output of the following two audio pipelines:
 gst-launch-0.10 neonhttpsrc location=http://colin.guthr.ie/bong.ogg ! decodebin ! audioconvert ! pulsesink
vs.
 gst-launch-0.10 neonhttpsrc location=http://colin.guthr.ie/bong.ogg ! decodebin ! audioconvert ! alsasink


Both pipelines ultimately go via PulseAudio (as alsa is configured to go via alsa-pulse plugin), but the former (pulsesink) pipeline produces identifiably worse audio even on small samples like the above.

It's most noticeable at the start of the playback.
Comment 1 Wim Taymans 2010-12-02 14:14:47 UTC
In the alsasink case, gstreamer does the float to int conversion. In the pulsesink case, pulsesink does float to int conversion.
Comment 2 Colin Guthrie 2010-12-02 14:35:42 UTC
So some more tests:


gst-launch-0.10 filesrc location=bong.ogg ! decodebin2 ! audioconvert ! audioresample ! audio/x-raw-int ! pulsesink

also causes problems, but:

gst-launch-0.10 filesrc location=bong.ogg ! decodebin2 ! audioconvert ! audioresample ! audio/x-raw-int,rate=44100 ! pulsesink

does not.

So as wtay says the resampler in PA appears to be the problem.

Changing the PulseAudio resampler to trivial or src-* seems to be problem free, so but on different speex ones, it appears broken.

So the problem is actually in the speex resampler in PA (or in speex itself).

Therefore, closing as invalid here.