[ardour-dev] quick note re: MTC

[Fons Adriaensen]

On Tue, Jan 02, 2007 at 10:02:21AM -0500, Dave Phillips wrote:

> Ardour2 is just so cool. As always, deep thanks to all devels. :)

Seconded, the only thing that keeps me from switching to A2 are
the multi-channel bugs that were eleminated from A.99 after the
the branch to A2.

To refresh memories: when the multi-channel panner is bypassed
(to be replaced by a plugin), the mixer strip's gain structure
goes crazy: fader gain is applied twice and there may be other
anomalies as well.

I have for a long time considered replacing the (rather useless)
multi-channel panner by a second-order ambisonic one. The result
can then be used directly for Ambisonic rendering, or decoded to
ITU 5.0 (the .1 channel is handled separately) by a plugin in the
master strip. This is actually the _best_ way to pan for ITU 5.0
- the combination of a 2nd order Ambisonic panning and a dual
frequency band AMB -> ITU decoder is equivalent to the optimal
frequency-dependent panner which would be difficult to realize
in any other way (and much too complex to be included in each
mixer strip). Such panning gives a much better surround image
than just dividing the signal between the two nearest speakers.
Sure, the current multichannel panner also uses all speakers,
but in a way that doesn't make any sense in psycho-acoustic
terms. 

I'm willing to do the DSP part of this if I can team up with 
someone to do the GTK and automation related part. The ideal
widget would be a 360 degrees rotary one (without end stops),
eventually combined with a small vertical slider for height.

There may be a problem with automation as the azimuth domain
is cyclic and there will be a discontinuity when crossing
+/- 180 degrees. The panner code can handle this correctly
as long as the automation does *not* try to interpolate such
a jump. The alternative is not to store the azimuth but sine
and cosine, wich requires two automation tracks and is less
easy to edit manually.


-- 
FA

Lascia la spina, cogli la rosa.