Re: [Octal-dev] 0,1,infinity redux (yet again :-)

[Neil Nelson]

It seems to me the n-channel problem can be seen as where do we
want the sound to come from in a given environment/context. For stereo
(two speakers), the choices are left, right, or somewhere in between.  In
two dimensions, a horizontal plane, a circle can be drawn with the listener
at the center with the result that stereo represents a pie slice of possible
sound source locations. The problem can then be described as not which
channel but where on the circle the sound is selected to be: a sound and
circle position correspondance.  This can then be extended to three
dimensional sound in much the same way a position is described on the Earth
using two, perpendicular circumferences. In this way, how many speakers/
channels one has becomes irrelevant; they just need to be specified as being
at particular positions so that the sound position can be approximated by
amplitude and phase adjustments for the available speaker positions.

Given that we only have two ears, study of sound source location suggests
that location determination is also affected by the delay of sound reaching
one ear before the other and the spectrum differences caused by the shape of
the ear and other damping and modifying effects of the head/body. And then
there is the wave pattern interference in the listening field resulting from
multiple sound sources (using more than one speaker to represent the location
of an intended single sound source).

But of course, many of these issues are secondary in importance or outside
useful capability/utility.  However this does suggest that we could
encode our music for sphere locations and then have the run-time speaker/
channel implementation provide that specific translation.  I.e., you obtain a
composition encoding and then select the physical equipment locations or
type (as in earphones) to reflect the sound locations intended.

P.S.: My encouragement to Dave O'Toole on his planned development objectives.

Neil Nelson address@hidden