Bradley Sepos

Some samples and discussion on #1349.

Care to point me at the relevant code areas? If worst comes to worst, I can probably unhook one of my surround processors and record the decoded output.

The way I read that, there still should be no difference in level between the L+Ls and R+Rs driven. Only the direction of the phase shift is changed.

Unless I'm smoking something, this is the culprit: https://github.com/libav/libav/blob/master/libavresample/audio_mix_matrix.c#L234-L238 The spec calls for a 90 degree phase shift, negative polarity for the right channel. The code does not do this.

Indeed, impossible to do correctly without a sample buffer. Technically, this is an inversion not a 180° phase shift (which involves a time delay). A very crude approximation at best.

Indeed. If you really want to rot your brain... http://forum.doom9.org/showthread.php?t=112122

Agreed. It may be helpful to note that there are two DPL2 modes, Movie and Music. When Music is selected (usually on a DPL2 compatible receiver), Rs is inverted to...

90 degree phase shift is 1/4 cycle time delay. The reason this approximation sort of works is there isn’t enough difference over 1/2 cycle to throw off the decoder. :smile_cat:...

+90 and -90 for 180 degree phase shift between the two surround channels. Doing it a second time would yield an in-phase signal delayed by 1 cycle. Theoretically, the decoder...

Ah, I see what you mean. For music, I certainly doubt anyone would bake in a phase shift in DD5.1. That’s just silly for actual 5.1 playback, especially for music....