I read somewhere here that Vorbis Multichannel Coupling for more than 2 stereo channels are, um, lacking.
Sooo... just in case (big if here...) I want to dabble and try my hands at trying to 'improve' on this situation, where should I start? E.g. :
- What resources/documentations/specs should I read
- What code to tweak
- Etc
I may or may not do it, as I'm not sure about my free time allocation in the future, but who knows...
Full Version: Multichannel Coupling?
Make yourself at home:
vorbis-dev mails '07 January
vorbis-dev mails '07 February
vorbis-dev mails '07 March
Check specificially the threads with "multichannel" and "5.1" in the subject.
This reflects my view of the multichannel situation and possible solutions for Vorbis I
I guess the first things you need to do is get familar with the Vorbis I spec and the libVorbis source code.
I noticed that the 2nd paragraph of the 5.1 bounty wiki entry ist just baloney since an ambisonic channel mapping hasn't been defined in the Vorbis I spec nor a mechanism to tell the decoder how to upmix it to 5.1. I guess some Ambisonics zealot added this without knowing much about Vorbis I.
Until now I only addressed the "coupling" (in terms of redundancy removal). You should also think about how and when "point stereo" can be used when coupling L with LS, R with RS, L with R or LS with RS. This requires some listening tests. To further improve things you should include inter channel masking. I'm not sure whether inter channel masking is already done anywhere in the source or not. But it makes sense in a 5.1 environment for closely spaced speakers (like L, C, R).
So, to recap:
- get familar with the source and the spec on coupling/psy model related matters
- do some experiments about sound localization in 5.1 and how pairwise intensity coding can (if) be applied without messing too much with the spatial image
- introduce pairwise coupling for 5.1, possibly adaptively like I described in the mail
- introduce inter channel masking
- introduce new floor / residue modes for the .1 channel
Good luck!
SG
vorbis-dev mails '07 January
vorbis-dev mails '07 February
vorbis-dev mails '07 March
Check specificially the threads with "multichannel" and "5.1" in the subject.
This reflects my view of the multichannel situation and possible solutions for Vorbis I
I guess the first things you need to do is get familar with the Vorbis I spec and the libVorbis source code.
I noticed that the 2nd paragraph of the 5.1 bounty wiki entry ist just baloney since an ambisonic channel mapping hasn't been defined in the Vorbis I spec nor a mechanism to tell the decoder how to upmix it to 5.1. I guess some Ambisonics zealot added this without knowing much about Vorbis I.
Until now I only addressed the "coupling" (in terms of redundancy removal). You should also think about how and when "point stereo" can be used when coupling L with LS, R with RS, L with R or LS with RS. This requires some listening tests. To further improve things you should include inter channel masking. I'm not sure whether inter channel masking is already done anywhere in the source or not. But it makes sense in a 5.1 environment for closely spaced speakers (like L, C, R).
So, to recap:
- get familar with the source and the spec on coupling/psy model related matters
- do some experiments about sound localization in 5.1 and how pairwise intensity coding can (if) be applied without messing too much with the spatial image
- introduce pairwise coupling for 5.1, possibly adaptively like I described in the mail
- introduce inter channel masking
- introduce new floor / residue modes for the .1 channel
Good luck!
SG
QUOTE(SebastianG @ Mar 13 2007, 16:13) 
I guess some Ambisonics zealot added this without knowing much about Vorbis I
http://wiki.xiph.org/index.php?title=Bount...&oldid=1352
I remember Monty saying Ambisonics was planned for Vorbis II. Just search the forums for "Ambisonics" written by xiphmont
Sorry, I don't see any relevance to what I said. Are you just being informative or did you try to point out a mistake/misunderstanding of mine?
My point was that the ambisonics suggestion in the 5.1 for Vorbis I bounty wiki entry -- quoting "This should be implemented with ambisonics as the internally coupled channels, and then use ambisonic g-format filters for en/dec coding from and to 5.1. This will allow correct playback on many different speaker layouts and an easy path to headphone support." -- just doesn't make any sense because the Vorbis I spec doesn't support this.
Cheers!
SG
My point was that the ambisonics suggestion in the 5.1 for Vorbis I bounty wiki entry -- quoting "This should be implemented with ambisonics as the internally coupled channels, and then use ambisonic g-format filters for en/dec coding from and to 5.1. This will allow correct playback on many different speaker layouts and an easy path to headphone support." -- just doesn't make any sense because the Vorbis I spec doesn't support this.
Cheers!
SG
QUOTE
the point was that the ambisonics suggestion in the 5.1 for Vorbis I bounty wiki entry -- quoting "This should be implemented with ambisonics as the internally coupled channels, and then use ambisonic g-format filters for en/dec coding from and to 5.1. This will allow correct playback on many different speaker layouts and an easy path to headphone support." -- just doesn't make any sense because the Vorbis I spec doesn't support this.
How many recording engineers actually use ambisonics though unless they have a soundfield microphone?. I understand it's easy to mix from Ambisonic-B to 5.1 surround sound, but not vice versa.
QUOTE(pepoluan @ Mar 13 2007, 10:33)
- What resources/documentations/specs should I read
You can also have a look here:
http://wiki.xiph.org/index.php/Ambisonic
QUOTE(SebastianG @ Mar 13 2007, 18:33)
Sorry, I don't see any relevance to what I said. Are you just being informative or did you try to point out a mistake/misunderstanding of mine?
Informative
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.