#2228 G.722 shift bug not quite fixed yet?

MO
Martin Oberhuber
Wed, Mar 18, 2020 6:19 PM

Hello PJSIP List,

I have made a simple test to validate the G.722 shift fix from
https://trac.pjsip.org/repos/ticket/2228  :

My device produces a perfect sine at 440 Hz and 0dB (float 1.0).
With a G.711a encoder, I get the exact expected output.

But with the G.722 encoder, I get a sine of -6dB (float 0.5) initially, and
only after a varying delay - I have seen 0.26 sec upto 9 seconds - it
suddenly jumps up to 0dB (float 1.0). Screenshot from Audacity is attached.
I have used ffmpeg version 2.8.15-0ubuntu0.16.04.1 to decode the stream:

ffmpeg -f g722 -i agcb_0dB_sinus_g722dc.raw -acodec pcm_s16le -ar 16000
agcb_0dB_sinus_g722dc.wav

I can provide the Wireshark pcap / RTP raw / wav if needed. To me, this
behavior is quite undesirable because it may produce a sudden jump in
volume on the receiving end. Can this be the expected behavior, or am I
looking at a defect?

Thanks a lot for your consideration!
Martin

Hello PJSIP List, I have made a simple test to validate the G.722 shift fix from https://trac.pjsip.org/repos/ticket/2228 : My device produces a perfect sine at 440 Hz and 0dB (float 1.0). With a G.711a encoder, I get the exact expected output. But with the G.722 encoder, I get a sine of -6dB (float 0.5) initially, and only after a varying delay - I have seen 0.26 sec upto 9 seconds - it suddenly jumps up to 0dB (float 1.0). Screenshot from Audacity is attached. I have used ffmpeg version 2.8.15-0ubuntu0.16.04.1 to decode the stream: ffmpeg -f g722 -i agcb_0dB_sinus_g722dc.raw -acodec pcm_s16le -ar 16000 agcb_0dB_sinus_g722dc.wav I can provide the Wireshark pcap / RTP raw / wav if needed. To me, this behavior is quite undesirable because it may produce a sudden jump in volume on the receiving end. Can this be the expected behavior, or am I looking at a defect? Thanks a lot for your consideration! Martin