[FFmpeg-trac] #4175(swresample:new): Dolby Pro Logic II / Dolby matrix downmixing level balance

[FFmpeg]

#4175: Dolby Pro Logic II / Dolby matrix downmixing level balance
-------------------------------------+-------------------------------------
             Reporter:  ranutso      |                     Type:  defect
               Status:  new          |                 Priority:  normal
            Component:  swresample   |                  Version:  2.4.4
             Keywords:  downmix      |               Blocked By:
  surround stereo dplii              |  Reproduced by developer:  0
             Blocking:               |
Analyzed by developer:  0            |
-------------------------------------+-------------------------------------
 Summary of the bug:

 Hello guys. I am trying to downmix a 5.1 audio to 2.0 (stereo) using the
 audio resampler filter "matrix_encoding=dplii". The resulting stereo audio
 presents some audio level bias towards the right speaker when there is
 more than one active channel on the surround stream. That is, it seems
 that the audio volume on the right speaker is higher than it should be.
 This is, of course, using a test signal where all surround channels
 produce the same sound at the same volume.

 How to reproduce:
 {{{
 % ffmpeg -i ChID-BLITS-EBU-Narration441-16b.wav -ac 2 -filter:a
 "aresample=matrix_encoding=dplii" dplii.wav
 ffmpeg version 2.4.4 Copyright (c) 2000-2014 the FFmpeg developers
   built on Dec  4 2014 14:13:11 with Apple LLVM version 6.0
 (clang-600.0.56) (based on LLVM 3.5svn)
 }}}

 You may freely obtain the surround test file I used on the command above
 from Fraunhoffer's website.
 Direct link to the surround wave file: https://www2.iis.fraunhofer.de/AAC
 /ChID-BLITS-EBU-Narration441-16b.wav
 Fraunhoffer's page where that file's URL is located:
 https://www2.iis.fraunhofer.de/AAC/multichannel.html

 If you analyze both the surround input file and the resulting stereo wave
 file on Audacity, for example, you will notice that by 19 seconds into the
 audio stream all channels are active with a test signal. The same point in
 time on the resulting DPL-II downmixed shows a "higher volume" on the
 right speaker. This effect is even more noticeable by 22 seconds when they
 change the test signal. It is interesting to note that in the initial
 seconds of the file, when a male's voice is announcing each channel
 individually, I could not notice this right speaker bias. The channel
 downmixing in that segment sounds correct. This effect is very noticeable
 on sound tracks with very active surround channels, they all seem to be
 stronger on the right speaker.

 Thank you.

--
Ticket URL: <https://trac.ffmpeg.org/ticket/4175>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker