What formula does FFmpeg use when converting 5.1 AC3 to stereo wav?

[orion44]

When converting 5.1 AC3 to stereo WAV:

ffmpeg -i AC3_51.ac3 -ac 2 Stereo.wav

what is the exact formula that '-ac 2' option uses in the background?

[tebasuna51]

FL' = FL + 0.7xFC + 0.7xSL
FR' = FR + 0.7xFC + 0.7xSR

[orion44]

FL' = FL + 0.7xFC + 0.7xSL
FR' = FR + 0.7xFC + 0.7xSR

Thanks.

I'm trying to exactly replicate the "-ac 2" option with pan filter command. I've found the following via google search:

ffmpeg -i AC3_51.ac3 -af "pan=stereo|FL < 1.0*FL + 0.707*FC + 0.707*BL|FR < 1.0*FR + 0.707*FC + 0.707*BR" Stereo.wav

ffmpeg -i AC3_51.ac3 -af "pan=stereo|FL = 1.0*FL + 0.707*FC + 0.707*BL|FR = 1.0*FR + 0.707*FC + 0.707*BR" Stereo.wav

Do any of the above two lines replicate exactly the "-ac 2" option?

Which of the two lines is the correct one? (note the "<" and "=" difference)

[tebasuna51]

From ffmpeg docs (https://ffmpeg.org/ffmpeg-filters.html#toc-pan-1):

If the ‘=’ in a channel specification is replaced by ‘<’, then the gains for that specification will be renormalized so that the total is 1, thus avoiding clipping noise.

The -ac 2 is equivalent to:
-af "pan=stereo|FL = 1.0*FL + 0.707*FC + 0.707*SL|FR = 1.0*FR + 0.707*FC + 0.707*SR"

FL' = FL + 0.7xFC + 0.7xSL
FR' = FR + 0.7xFC + 0.7xSR

The < option:
-af "pan=stereo|FL < 1.0*FL + 0.707*FC + 0.707*SL|FR < 1.0*FR + 0.707*FC + 0.707*SR"

Is equivalent to:
FL' = 0.414xFL + 0.293xFC + 0.293xSL
FR' = 0.414xFR + 0.293xFC + 0.293xSR

This avoid clipping noise(sum of coeficients = 1) but you can finish with low volume.

Note the SL,SR channels instead BL,BR in your samples. By default the surround channels are SL,SR (Side not Back, default longtime ago).

The canonical downmix is:

FL' = FL + 0.7xFC + SL
FR' = FR + 0.7xFC + SR

the ffmpeg default downmix lose surround volume
Normalized (sum of coeficients = 1):

FL' = 0.37xFL + 0.26xFC + 0.37xSL
FR' = 0.37xFR + 0.26xFC + 0.37xSR

To do a correct normalization, with the max volume without clip, you need do 2 pass (like AviSynth do).

[orion44]

Do popular software players, like PotPlayer and KMPlayer, use the same downmix as the "-ac 2" option when decoding 5.1 AC3 to stereo?

[tebasuna51]

You can configure the dowmix like you want.

I have attached a 7.1 system but if you have a 2.0 you can mix, in mpc-hc:

[orion44]

FL' = FL + 0.7xFC + 0.7xSL
FR' = FR + 0.7xFC + 0.7xSR

When downmixing 5.1 AC3 audio to stereo WAV, is the following allowed according to the Dolby downmix parameters:

1. Attenuate FL, FR, FC, SL, SR by 6 dB?
2. Attenuate FL, FR, FC by 6 dB, and attenuate SL, SR by 9 dB?

Is any of the above in accordance with Dolby standards?

[tebasuna51]

...
Is any of the above in accordance with Dolby standards?

Nope. The allowed values are included in BSI metadata and can be:

Center Mix Level: 0.707 (–3.0 dB), or 0.595 (–4.5 dB) or 0.500 (–6.0 dB)
Surround Mix Level: 0.707 (–3 dB) or 0.500 (–6 dB)

Front Levels can't be attenuated unless you do a Normalize over all channels.
Then Front levels must be always from 3 to 6 dB over the Center and Surround levels.

[orion44]

Nope. The allowed values are included in BSI metadata and can be:

Center Mix Level: 0.707 (–3.0 dB), or 0.595 (–4.5 dB) or 0.500 (–6.0 dB)
Surround Mix Level: 0.707 (–3 dB) or 0.500 (–6 dB)

Front Levels can't be attenuated unless you do a Normalize over all channels.
Then Front levels must be always from 3 to 6 dB over the Center and Surround levels.

Are the Center and Surround channels allowed to have differing values, or do they have to be attenuated by the same value?

For example, the Center: -3 dB or -4.5 dB, but the Surround channels: -6 dB?

[tebasuna51]

Are the Center and Surround channels allowed to have differing values, or do they have to be attenuated by the same value?

For example, the Center: -3 dB or -4.5 dB, but the Surround channels: -6 dB?

Can be different and my recommendation is Center: -3 dB, Surround channels: -6 and after Normalize to a permanent downmix 2.0.

But, of course, is a user selection and can vary with the movie.