I am producing some 5.1 mixes for short films but am also want to supply clients with a Dolby Digital 2.0 surround file as a back-up for archival purposes. A few years ago, someone once told me how to encode four separate channels (L, C, R, S) from a 5.1 mix into Dolby Surround:
Lt - Left 100%, Right 0%, Center -3db, Surround -3db, +90 shift
Rt - Left 0%, Right 100%, Center -3db, Surround -3db, -90 shift
From my 5.1 mix stems, I took my L and R surround stems and converted them into a new dual mono stereo file, before applying the phase shifting as described above. I mixed this file with the Left, Center and Right channels as described above, creating a new stereo file and encoding it to an AC3 file with Dolby Surround mode enabled in Surcode For Dolby Digital.
I burned a test DVD and the surround channels indeed played the audio I had encoded specifically for surround using the above method. However, there was also some bleed into the surrounds from some of the audio in the front channels. Why is this?
I assume its a phasing issue. I am using Audition - do I need to use the Graphic Panner to correct the phase in the left and right channels before encoding to Dolby Surround to stop this from happening? If so, what settings do I need to apply?
Thank you to anyone who understands and can help with this. I have never had this happen before.
+ Reply to Thread
Results 1 to 8 of 8
No, if you do the math, there will ALWAYS be bleed-through. ALL matrixed multichannel methods have some amount of bleed-through, it is just the nature of matrixing (different methods/algorithms have different proportions of bleed-through in different channels). Otherwise, there would be no need for discreet 5.1 channels. But you can't get something for nothing.
Live with it. The rule (for Dolby Surround) is: any particular channel only has 3dB of separation from an adjacent channel, but does have infinite separation from the opposite channel. That is the reason ProLogic (and later ProLogicII) got invented - to partially overcome the loss of channel separation (look it up if you don't believe me). PL does this by constantly varying the volume of the non-primary channel(s) based on the instantaneous statistics of the signal in the primary channel(s), thus artificially augmenting/magnifying the separation (but at some cost of "pumping"). This works semi-successfully when PL is the final decode link in the chain, but becomes a problem in this modern digital world where people (possibly such as yourself) attempt to use it in the middle of the playback chain, particularly when compression is involved (because compression bases their statistics, and thus their compressibility, on some of that separation).
I don't remember ever having bleedthrough before (although it's been a few years since I mixed for PL). The odd thing is, its only a certain stereo percussion loop in the film's musical soundtrack that it happens with, which looks like this in the phase window:
The other sounds and instruments stay in the front but this particular loop suddenly comes out of the surrounds when it plays.
I'm not going to comment on the difference between what you perceived before vs. now, or why. I just know that If you go:
Discreet 4.0 (or 5.1) ->Matrixed 2.0 -> 4.0 again (whether in realtime playback only, or saved as files/streams), you will ALWAYS have bleed-through (aka crosstalk, aka bad channel separation). It's physically & mathematically impossible to avoid.
Maybe you just notice it more this time because of the transient nature of percussion...
Is that 3db of separation why you need to apply a -3db attenuation to the center and surround channels?
No, that is so the sound will end up with equivalent power/loudness, regardless of where you streered it.
To get back to L, C, R, S from Lt & Rt (which is all you would have to work with at that point), no matter how you combine them, you would have crosstalk elements in the resulting equation.
Fair enough - but I am still interested in the science behind the technology, even if it is old. Cheers for the enlightening answers.
Last edited by Xoanon; 25th Mar 2014 at 23:38.