Gary Eickmeier wrote:
[...]
"KH" wrote in message
...
I don't think you want to know "how stereo works", you appear to want some
theory that ties everything together, and that when adhered to results in
*A* correct, optimally realistic reproduction. There is no such single
solution, and there never will be for the myriad reasons presented to, and
apparently ignored by you. A few:
1. The recorded signal does NOT contain all the information from the
original space, and short of an infinite number of mics and channels,
never will. This is a simple fact.
OK, I understand what you are getting at here, but my smartass answer would
be, where did it go?
It was lost. Simple as that.
The microphones heard just as much of the audible
happenings when recording in stereo as the binaural head would have.
Binaural head records only that one head's perspective. Nothing more.
With
the binaural head, the problem we found was that the head could not turn
while listening, so there was an IHL problem, but there was no "missing"
information, it was just not presented in quite the same way as we can
listen live.
It was just an information for a pair of ears at that particular, frozen
position. It simply does not contain all the information for all the other
poistions within that venue.
The stereo microphones, no matter which pattern except for AE's
nemesis the close up multimiked affair, record the direct sound, and the
early reflected and some of the reverberant,
Those data are convolved i.e. intermixed together. And they could not be
loslessly separated anymore.
just like you would if you were
there.
And if your ears were exactly at the postion of those mikes (let us ignore
different directional and freqiency characteristics for now). Note just one pair
of ears both at fixed points in space.
There is enough information going into your two ears at the time, so
why not into the microphones and therefore out of the speakers at the time
of playback?
Because information going into the ears is just that information going into that
paricular ears and nothing more.
What goes into the ears for example depends on those ears position within the
space.
With two mikes you don't record soundfield. You only record two onedimiensional
projections of the soundfield at two arbitrary points in space.
As you like visual analogies. Look at the photography -- it's a record of visual
information about a space it was taken in. But it's just record of one point
perspective of that space. You can't easily recreate whole information about
that space -- even by making that photgraph big enough and surrounding it with
varius mirrors and other tricks won't make it look like a real thing. For one
simple reason it does not contain the all information about the whole space --
it only contains 2-dimensional projection of that space, projection taken at one
particular point in space and time.
If for example there were two objects in that space and one of them occluded the
other it won't be on the photograph, while someone actually present in that
space could walk 2 steps to the side and see that object.
Now there are techniques (algorithms) which could recreate some 3D information
out of 2D picture but:
1. they are much more complex that setting few mirrors -- they require a lot of
computing power
2. they work for typical situations but they're unexact and make a lot of errors.
3. they couldnt recreate information there is no clue about in the 2D imeage,
for example that aforementioned fully ocluded object would not be recreated at all.
In case of two mikes you have two one dimensional images and unlike photograph
those two dimensions are not orthogonal. But as our hearing is much less precise
and much less data intensive than our sight I could not exclue that some smart
algorithm could convincingly recreate a soundfield similar enugh to original
one. But I'm certain that, assuming it exists at all, it's not some simplistic
addition of few delayed and attenyuated copies of the original.
What you are getting at is found in Blauert, in fact right on the cover,
where it shows a zillion small speakers around the subject's hotseat during
an experiment in which multiple microphones recorded an event of the sound
arriving at a single spot, and there is an attempt to reproduce a "you are
there" impression, kind of like binaural only with speakers and for this one
listener only. Amar Bose told me of an experiment in Eindhoven where Philips
had used about 120 channels and some signal processing in an anechoic
environment to cause the reproduction sound fields to be more of a duplicate
of the original and have no interference from the listening environs.
And those are the "easy" ways to recreate a soundfield which from some narrow
set of points resembles soundfield of a recording venue.
2. The stereo effect is an *illusion*, and always will be, again, unless
multitudinous speakers are used to replay the recording from equally
multitudinous mics.
3. There is no objective reference for realism. There is only
*preference* for various implementations of the illusion.
Yes... well... as I agreed earlier, the recording is a new work of art,
using the original as a stepping off point, but I think there are ways we
can study the whole process and get closer, if not all the way there, which
I believe is a false goal. Architects agree about the design goals of a good
hall, but audiophiles have no agreement on the design goals of a good
speaker/listening room system. This is because most think that stereo is a
"trick" that may fool your ears into hearing an illusion of the live event -
Since it is
kind of like binaural.
Binaural is not a trick -- it's just one point static perspective. It works but
only for that very point and that particular orientation. Hece one problmem with
it is that one have to not move their head.
[...]
rgds
\SK
--
"Never underestimate the power of human stupidity" -- L. Lang
--
http://www.tajga.org -- (some photos from my travels)