manual/include/vbap-panner.html

107 lines
3.9 KiB
HTML

<p class="warning">
Ardour's VBAP panner is currently in development, and its semantics may
change in the near future, possibly affecting mixes using it. It is advised not
to rely on it for important production work while the dust settles.<br/>
The Panner only works in fixed static mode, it does not support
automation playback.
</p>
<p>
<dfn><abbr title="Vector-base Amplitude Panning">VBAP</abbr></dfn>
is a versatile and straightforward method to pan a source around over an
arbitrary number of speakers on a horizontal polygon or a 3D surface,
even if the speaker layout is highly irregular.
</p>
<h2>Basic concepts</h2>
<p>
VBAP was developed by Ville Pulkki at Aalto University, Helsinki, in 1997.
It works by distributing the signal to the speakers nearest to the desired
direction with appropriate weightings, aiming to create a maximally sharp
phantom source by using as few speakers as possible:
</p>
<ul>
<li>one speaker, if the desired direction coincides with a speaker
location,</li>
<li>two speakers, if the desired direction is on the line between two
speakers,</li>
<li>and three speakers in the general 3D case.</li>
</ul>
<p>
Thus, if the panner is moved onto a speaker, only this speaker will get any
signal. This is handy when precise 1:1 routing is needed.
</p>
<p>
The drawback of VBAP is that a moving source will constantly change its
apparent sharpness, as it transitions between the three states mentioned above.
</p>
<p>
An <dfn>horizontal</dfn> VBAP panner has one parameter, the <dfn>azimuth
angle</dfn>. A <dfn>full-sphere</dfn> panner offers an additional
<dfn>elevation angle</dfn> control.
</p>
<p class="note">
More elaborate implementations of VBAP also include a
<dfn>spread</dfn> parameter, which will distribute the signal over a
greater number of speakers in order to maintain constant (but no longer
maximal) sharpness, regardless of position. Ardour's VBAP panner does not
currently include this feature.
</p>
<h2>Speaker layout</h2>
<figure class="right">
<img class="mini" src="/images/VBAP-panner-5.png" alt="The VBAP panner with 5 outputs">
<figcaption>
The VBAP panner with 5 outputs
</figcaption>
</figure>
<p>
Each VBAP panner is specific to its <dfn>speaker layout</dfn>&mdash;the
panner has
to "know" about the precise location of all the speakers. A complete VBAP
implementation must therefore include the possibility to define this
layout.
</p>
<p>
Ardour currently uses a simplified approach: if a track or bus has more
than two output channels (which implies stereo), it assumes that there are
N speakers distributed in a regular N-gon. That means that for
irregular layouts such as 5.1 or 7.1, the direction dialed in will
differ a bit from the actual auditory result, but any desired spatialisation
can still be achieved.
</p>
<h3 class="clear">Experimental 3D VBAP</h3>
<figure class="right">
<img class="mini" src="/images/VBAP-panner-10.png" alt="The VBAP panner with 10 outputs, in experimental 3D mode">
<figcaption>
The VBAP panner with 10 outputs, in experimental 3D mode
</figcaption>
</figure>
<p>
For tracks with 10 outputs, Ardour will currently assume a 3-dimensional
speaker layout corresponding to Auro-3D 10.1, which is a horizontal 5.1
system, four elevated speakers above L, R, Ls, and Rs, and an additional
"voice-of-god" speaker at the zenith.
</p>
<h2 class="clear">N:M panning</h2>
<figure class="right">
<img class="mini" src="/images/VBAP-panner-4in5.png" alt="The VBAP panner in 4 in, 5 out mode">
<figcaption>
The VBAP panner in 4 in, 5 out mode
</figcaption>
</figure>
<p>
For tracks and busses with more than one input, Ardour will (for now) assume that
the inputs are distributed symmetrically along the latitude around
the panner direction. The width parameter controls the opening angle of
the distribution sector.
</p>