Distorted DTS-HD to Stereo Downmix

Finally had a chance to test the latest Infuse Beta 7.5.7 (4452). Had to put aside a few hours to let me disconnect my Apple TV from the back of my TV, setup the testing station again, and then record all the clips.

Add into the mix a weird bug with Audacity on Linux was recording everything slightly too slowly (some kind of resampling going on somewhere??) – the SPDIF captures didn’t line up with the FFMPEG WAV file. I initially thought the new Infuse Beta was doing something funky, but even the old version had the same issue. I gave up and redid everything on my Macbook.

In the screenshots below you can see a comparison of the following captures of my test Interstellar clip:

  • FFMPEG output (-ac 2)
  • Infuse 7.5.6 (current public release)
  • Infuse Beta 7.5.7:
    • Downmix: Original, LFE in Stereo Downmix: off
    • Downmix: Original, LFE in Stereo Downmix: on
    • Downmix: Normalized, LFE in Stereo Downmix: off
    • Downmix: Normalized, LFE in Stereo Downmix: on

Notably none of the options result in audio output that is identical to 7.5.6.

The “Downmix: Original” versions now have an overall volume that is slightly lower which seems to prevent clipping.

To my eyes the “Downmix: Normalized, LFE in Stereo Downmix: off” version looks most comparable to the FFMPEG output. It’s worth mentioning however that there are many approaches to downmixing and they all have tradeoffs (see this stackoverflow thread for discussions around this), and it partly comes down to taste. I’m in no-way claiming the FFMPEGs method is “the best”. I’m just using it as a reference point.

From my testing it seems like the DTS-HD to Stereo Downmixing is fixed in 7.5.7. Clipping no longer occurs.

It’s great to see that Firecore added multiple options here too, so that the output can be adjusted to taste.

5 Likes