[ardour-dev] quick note re: MTC

Fons Adriaensen fons at kokkinizita.net
Wed Jan 3 03:08:43 PST 2007

On Tue, Jan 02, 2007 at 09:15:51PM -0500, Thomas Vecchione wrote:

> So what you are describing would essentially be ambisonic encoding until 
> the final stage of bounce then, and the speaker layout would be 
> determined at the final stage of bounce to file, or playback, or 
> whatever, correct?

Yes. It would depend on what your end result has to be. If it is the
AMB format signals then you record these when mixing, and the decoder
would be part of your monitoring only. If it is the 5.1 signals, then
the decoder is part of the recording chain.

> So in effect how it would work is that each channel would get pan and 
> automation done in the Ambisonic 'style'.  The location of speakers 
> would mean nothing at this point.


> However if played back that audio and  locational data is then sent
> through an ambisonic decoder which would  provide the correct number
> of channels to be mixed together post decode stage

A lot of points in one sentence... There is no separate 'location data'
in Ambisionics - the location is encoded into the audio signals them-
selves, and those can be mixed like any other audio signals. The decoder
is a purely linear operation on the signals - no 'logic' involved.

> the channel fader would modify the group of channels sent to the 
> main mix level as normal, and the master fader would modify the overall 
> level of the main mix as normal, correct?

Yes, you will not see any difference. The channel fader operates before
the panner, as it would in any system. 

> In that case does it mean each channel would have to have its own 
> decoder running that would just get fed information from the overall 
> setup of the project?  How would that affect cpu useage per track on 
> average?  (Obviously not expecting a exact number as that is impossible 
> with so many possible different setups, but possibly a comparison vs the 
> stereo panner implemented now?)

No, each channel just has the panner, and that is very simple and does
not take any significant CPU. The panned signals are mixed, and that mix
is then decoded (you can still mix in signals after the decoding of course).
There is only one decoder in the whole system. Since the decoding is a
linear operation, you can do it *after* mixing.

Of course the net result is _as if_ each channel had its own decoder and
you were mixing decoded signals. 

You could see it this way. Imagine that each channel has a very sophisticated
ITU 5.1 panner, and you mix the output of those panners. Now each panner
consists of two parts, a first one that get a mono signal in, and depending
on the panning controls distributes it into some internal multichannel format,
and a second part that doesn't depend on panner controls and does the same
processing in each channel. That second part can be split off and put after
the mixing - it's then called the 'decoder'. The part that remains in each
channel is an AMB panner.

> Also how would this affect plugin useage?  For instance at first glance 
> it would seem that in order for the plugin to be useful, particularly 
> multichannel reverbs or other similar plugins, the decoding would have 
> to have already been done correct?

Depends on the type of plugin. EQ, dynamics, etc. are done on the mono
track signal before it is panned, so no problem here. Reverb is indeed
a different matter. I've always been surprised by people setting up
reverb as an insert or plugin. It can make sense for a reverb used as
an effect on a single channel, and in that case the reverb plugin would
need to 'understand' the post-panner AMB format if it has to be multi-
channel. Another way to set up reverb is to use a post-fader aux send
to route a mix of tracks to a reverb unit, and mix in the outputs from
the reverb as if they were normal tracks. You can use 'stereo' reverbs
in this way, unless the purpose of the reverb is to create a 'virtual
room' - in the case the reverb output has to be either Ambisionic and
then it's mixed in before the decoder, or specific to the speaker
layout and then you mix it in after the decoder. Sophisticated reverb
plugins such as some from Waves in fact use Ambisonic format internally,
and can produce either that format or an ITU decoded one. If your mixing
is done in AMB format, the reverb doesn't need a decoder for itself.


Lascia la spina, cogli la rosa.

More information about the Ardour-Dev mailing list