Hans Pelleboer wrote:
> 
> Hello,
> 
> In recent weeks I have read a number of questions regarding
> stereo microphone techniques and the decomposing of auditory images
> in their constituting components. Usually the problem is worded somewhat
> as follows: How does one write a csound script to emulate a M-S microphone
> image? Or: how to derive center information out of a stereo recording?
> The attempts to answer these questions center around the simple arithmetic
> of phase relationships, as used in the summing-matrix for a M-S pair.
> What is glossed over here entirely is the problem: What is an auditory
> image, and how do we perceive one as such?
> In the process of the build-up to the illusion which we call `stereo,'
> auditory sources are imprinted --arbitrarily!, because set up through the
> subjective hands of a recording engineer-- 

[snip]
	Aha, If you will, let me recall the one of first major stereo recordings to
really punctuate the use of stereo... The Beatles with Rubber Soul for
example.  Listen to instruments isolated to channels and I see what seems to
be the point here.  The recording *doesn't* sound like a recording of a room
from a distance, but rather assigns tracks to different speakers.  So perhaps
todays' stereo sound tries to mimic a room, but still the point seems to be,
how can you get the middle of the sound when it is really an acoustic
invention of a sound engineer?  How does one reverse the artistic techniques
of the mix-down?  Maybe it could be done algorithmically using similar
techniques as a physical model, the physical model being parameters that
emulate the techniques of a mixing engineer:)

> manner, that play-back over loudspeakers gives us some suggestion of
> direction, size and distance. Even if we simplify drastically by leaving
> out all time relationships, as caused by reflections and such in real-world
> acoustical environments, then still the sound pressures picked up
> by a single pair of microphones are the resultants of a vectorial addition
> of a -long- row of terms.
> Just manipulating the resultants does not reach far enough to alter the
> underlying relations of their components -- for the same reason that you
> can't make grapes out of wine!
> 
> The idea, therefore, that there ever will be a simple operator --something
> like a reciprocal `pan' knob-- that will deduce a source from the complex
> streams that it generates, is probably futile.
> 
> A way more fruitful approach, I think, is to start from the other side of
> the loop and build auditory images through SYNTHESIS, not attempting to
> derive
> the essential parameters through analysis. This will not lead to a
> clean-cut acoustical model, but it will definitely generate a bundle of
> useful parameters for image synthesis, which will be close to the controls
> of
> a typical mixing console and peripherals.
> One can think of massaging the direct/reverb ratio, diffusion with
> allpass filters or through convolution with noise. Early reflections and
> such are easily made by adding different delays.
> For this type of synthesis, csound provides us with an ample set of tools.
> Multi channel output allows the use of discrete center channels and the
> like.
> 
> Hope these remarks will help to elucidate the issue somewhat.
> 
> Hans Pelleboer
> 

-- 
Michael Coble, Time Inc. New Media, Pathfinder
Gallery: http://www.panix.com/~coble,  representing various artists
Home:    http://pathfinder.com/pathfinder/staff/mcoble/
Work:    http://www.pathfinder.com/