BUG: AC3filter 0.70b has wrong DPL2 downmix

[frank]

@Valex
Please debug.

The new Dolby ProLogic II downmix matrix in AC3filter 0.70b has wrong parameters.
The right downmix parameters (unscaled) are

L  C       R  SL       SR       LFE
1  0.7071  0  -0.8660  -0.5000  0    L'
0  0       0  0        0        0    C'
0  0.7071  1  0.5000   0.8660   0    R'
0  0       0  0        0        0    SL'
0  0       0  0        0        0    SR'
0  0       0  0        0        0    LFE'

My preferred settings:
To maintain the original dialogue level set the Master Gain to 6 dB.
Set the DRC Power to 0 dB (= BeSweets normal compression)

[bleo]

AC3Filter currently has LFE * 1 -> L' & R'.

If you don't want LFE downmixed, then frank's matrix is what you want.

But if you do want LFE downmixed, the usual practice is LFE * 0.7071.

It's easy to switch between the two by moving the LFE gain slider between -inf (-9999) dB and -3.01 dB. So the current AC3Filter matrix is not really wrong. It just depends on how you want to downmix LFE.

Also, is there a BSI downmix for LFE? I've never seen one, so what does the LFE BSI lock in AC3Filter do?

I'm hoping for a more elegant solution to switch between downmixing LFE or not, but of course I can just customise the matrix to whatever

[frank]

You are right with the parameters, bleo.

Downmix is a great compromise. Acoustic power of 5 channels is packed into 2 channels (+4 dB)! -> There must be a compression applied.
LFE can be encoded to +10 dB above dialogue channel and has no function like a subwoofer (spreading bandwith)! It's useful in the cinema (very deep bass effects). That loudspeakers have membranes of 1-2 metre in diameter.
That's why Dolby never included LFE channel into Dolby Surround downmix. (See Dolby's Downmix Manual)

Consequently no DS filter expands LFE channel back from 2/0. So we don't need the LFE.