[vorbis-dev] True surround sound for Ogg -- a proposal

[David Carter]

On Sat, Jul 28, 2001 at 09:12:04AM -0700, Michael Paine wrote:
> thank you for your information!!! This rocks.  So, will ogg compress 4
> channels without using stereo-pairing?  I hope so.. probably part of the
> surround sound support they are working on.  I bet could even compress 2
> songs into 1 and get better compression than 2 seperate compressed songs :-)
> 
> I'll look into helping whatever is needed.. I'd be lovin' life if two
> sblives, with 8 channel sync'd output, could be used for second-order
> ambisonics !!

That will be harder to do -- Ambisonics relies heavily on the phases of the
various signals being correct.  With two cards running simultaneously, their
clocks would not be quite the same, and would drift as the sound played,
which would eventually get to the point that the relative phases of the output
of the two cards were far enough apart that the Ambisonic effect would be
lost.  For Ambisonics to work right, sample-accurate synch is required.  Since
the SBLive cards have no way of running off of an external clock signal, this
would probably be impossible.

(This will probably work with one card because both DACs on the card share the
same clock source, keeping them in sample-accurate synch.)

An uber-hacker might be able to desolder the oscillator on one card, and then
patch in the feed from the other card's oscillator, but that would be very
nontrivial.  If someone tried it, they would also be likely to need some
type of expensive test gear (two-channel oscilloscope, etc.) to see if the
cards really did stay in phase.  If this did work, perhaps someone could make
a bit of money by modifying other peoples' cards like that, but again it would
never get to be widespread.  Most people will probably make do with a square
speaker array.  (I did notice that the ALSA driver seems to be able to drive
the 5.1 analog output of the newer SBLives, so perhaps a pentagon array would
also be possible.  More testing would be needed by someone who has one of these
newer cards.)

If you want to experiment with 2nd-order Ambisonics, you may very well have to
buy one of the fairly expensive sound cards used for professional audio work,
like the RME Hammerfall.  This would really be overkill (and expensive), but
it can handle (IIRC) 24 channels at a time.  This card is what Richard Furse
(who wrote the CMT Ambisonics modules) uses for his experiments.

BTW, regarding your idea of getting better compression by compressing two
songs into one 4-channel file, that probably wouldn't work.  Once channel
coupling is implemented in the encoder, it will work because there are
similarities in the two channels of most stereo signals, and these can be
represented in a more compact fashion.  However, it wouldn't work the same
way with tracks from two different songs, as there would not be many
similarities to compress.

Last night, I also ran across the most refined multitrack recording program
I've ever seen on Linux (ardour.sourceforge.net) -- this would be a good
candidate to use for making multichannel test material.  It uses LADSPA
plugins, so it will do Ambisonic processing -- oggenc would just need an
input plugin written that can read its file format (32-bit PCM).

When I got your message yesterday, I didn't think much progress had been made
towards the goal of Ambisonic surround in Ogg -- I'm glad I was wrong! :)


-- 
David Carter ** dcarter at sigfs.org ** dcarter at visi.com
PGP Key 581CBE61: E07EE199C767C752 8A8B1A9F015BF2EA
Key available by finger or www.keyserver.net

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.