[pjsip] Bug #766: last sample of each packet is zeroed?

Bram Kuijvenhoven mailinglists.bram at kuijvenhoven.net
Sat Dec 19 15:11:43 EST 2009


Dear PJSIP developers,

When I make a call to my PJSIP-based softphone application, I hear a strange noise, which is a bit of a show-stopper for my application. I think it is due to bug #766 (http://trac.pjsip.org/repos/ticket/766). Please read on.

Before I discovered bug report #766, I did some analysis and testing myself. I concluded that the last audio sample of each audio packet is zeroed. This happens in my test program as well as when I do
  pjsua --null-audio --clock-rate 8000 --play-file test.wav --rec-file output.wav
(followed by: cc 1 2).

The problem is related to the resampling in the conference bridge. It only occors when 1) you are downsampling (i.e. to a lower sample rate)
2) the source sample rate is 48kHz or 44.1kHz

I need the 44.1kHz/48kHz source sample rate because they are the only ones supported by my sound device.

My compiler is gcc-4.3.2, platform is 32-bit x86 linux, PJSIP version 1.5.

Attached you find a Python script that demonstrates the problem. It uses the PJSUA Python module. It generates several files: compare test.wav with filtered.wav and take a look at pcms.eps. BTW if you look at the latter file, then you see a slight delay in the output with respect to the input. Why is that?

I hope the bug can be fixed! Thanks.

Regards,

Bram

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: distortion.py
URL: <http://lists.pjsip.org/pipermail/pjsip_lists.pjsip.org/attachments/20091219/7e6e50a1/attachment.py>


More information about the pjsip mailing list