[Vorbis-dev] 5.1 surround channel coupling

[Richard Lee]

Please excuse me for being dense but 

>> Does this mean there is an existing B-format to Vorbis encoder?

>Yes, I posted an initial version of the encoder here several days ago.

What you posted was an Ambisonic panner which allowed any direction (speaker) to be coded into B-format.

Does this mean there is a logical and tested path for a *.AMB file (the Ambisonic studio B-format file) to be encoded into a Vorbis stream ?

>> And a Vorbis to B-format decoder?

>Current Vorbis Ambisonic streams do not need to be decoded into
B-Format.

What I was asking is if there is a logical and tested path for a Vorbis Ambisonic stream to recover the original B-format signals.

B-format is the definition of the soundfield.  ie what it should all sound like.  It makes no reference to speakers.  

Once you have B-format, you derive however many speaker signals you need depending on how many speakers you have.  This is a separate process.  The Speaker Decoder.

If you think about it, this is how you need to do surround sound properly if you are interested in more than sound coming out of zillion speakers.

A big advantage of Ambisonics is that you only need 3 channels to define a complete horizontal stage.  Hence you can encode 5.1 7.1 ... zillion.1 into 3 channels.

With 4 channels, you can define sound coming from any direction on a sphere including up & down.

Of course you take full advantage of Ambisonics only by encoding directly into B-format.  Your Ambi panner allows this.

Starting with an original 5.1 file means you are stuck with the faults of 5.1 eg side images are very poor.

>Vorbis started without any stereo channel coupling, and it worked and
was much better than other lossy audio codecs. I believe (well,
actually I have the opportunity to know it) the same applies to
Ambisonics.

If we don't need the efficiency gains from coupling, that's good.  But I believe from the nature of the Ambisonic signals, we will gain a lot; maybe 2x or more with "lossless coupling".

What I'm trying to get a handle on is whether "lossy coupling" would be acceptable at HF (me to answer) and whether there would be much gain in efficiency (Vorbis guru to answer).  This is why I'm asking the questions :
________________

Is it more usual to adopt eg "8 phase" at a lower frequency band before going to "4 phase"?

Do the terms "8 phase", "4 phase", "point" still make sense when we are coupling 4 channels?

Is there much advantage doing "4 phase" coupling above 18kHz only compared to "lossless" coupling of the 4 channels right up to 20kHz?
________________

>So maybe you could think about how to use Vorbis mechanisms to do UHJ; perhaps it will take much less work than a regular multichannel coupling. UHJ is proven to work and it may be valuable to do listening tests with it.

UHJ is a stereo signal, a matrix system like Dolby Surround but more sophisticated.  But when B-format is encoded into UHJ, a lot of information is lost.

You can get quite good results if you have a clever Speaker Decoder but no one would use a UHJ chain if a modern compressed format like Vorbis could take advantage of lossless in B-format signal to transmit accurate B-format.

I speak as someone who was involved in UHJ trials more than 20 yrs ago.

IMHO, the only reason for UHJ today is to get a good stereo signal from B-format.  But there are much simpler ways of getting a stereo signal with advantages & disadvantages compared to UHJ

However, if you do use UHJ for whatever reason (even just for stereo) it MUST be "losslessly coupled" over the whole frequency range.
_______________

The Speaker Decoder is accurately described in 2 documents

"Ambisonic Surround Decoder"

and 

"SHELF FILTERS for Ambisonic Decoders"

both available from 

www.ambisonicbootlegs.net\Members\ricardo

There are a lot of faulty equations for Ambisonic (Speaker) decoders on the www which will give sub-optimal results.  I've just corrected one on 

http://en.wikipedia.org/wiki/Ambisonic_decoding

You can find the listening tests proving the above decoders at

http://www.ai.sri.com/ajh/ambisonics


-- 
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.441 / Virus Database: 268.18.3/700 - Release Date: 24/02/07 20:14