Subject: Re: Le mat?riau Sonore, hidden place and physical models
From: Philippe-Aubert Gauthier (Philippe-Aubert.Gauthier@USherbrooke.ca)
Date: Sat Sep 17 2005 - 11:24:46 EDT
Their is some other approaches in synthesis that could be applied to physical
models (this might even be done). As suggested below, even if it is possible to
change all the anisotropic paramaters like stifness, mass, geometry etc., it
could be a real mess to get something that you like ... or just fit what you
have in mind. Playing for hours with damping might not be so exciting ...
To simplify the programmation of synth (and potentially physical-model based
synth) by the musician/composers (and to get a nice headache to the designers),
some peoples have tried to run optimization algorithms (we can imagine genetic
algorithms, gradient measurements, or whatever you want) that try to reduce the
difference (time signal, frequency spectrum, wavelets, I don't exaclty remind
what it was ...) betweem a target sound and the synth sound. An example would
be to put a set of constraints like : I want only AM oscillators, simple
filters, complex envelop generator, a wave shaper and connect all these
potential modules as you which to create a sound like this: "vroooohtictictoc"
(where this nice excerpt would be a record of your voice trying to explain in
sound the type of sound that you like to have) ... and then, under the set of
constraints, after some hours or days of computation you get the best fit that
can be achieved with this sound. (okm why then just not use you voice?)
You can even extend this to hardware evolution. Like the evolvable motherboard
presented by Bird and friends in Leonardo 36 no 2, 2003. They have tried to
create an oscillator on this motherboard and they get, after some genetic
algorithm applied to the evolvable motherboard a new type of oscillators which
was new to the traditionnal engineering knowledge of oscillators.
Imagine, if you apply this to the new promising hardware avenue of FPGA (field
programmable gate array) that can make so many complex circuits.
That's only some examples ...
Selon Richard Wentk <email@example.com>:
> At 23:55 16/09/2005, you wrote:
> >Well, if they made it correctly, you would be able to adjust
> >parameters which would allow you to MAKE it a good piano, or not. If
> >it only had one sound, what's the point? Well... there is one, but, it
> >would be vastly more useful if there were parameters.
> I think you'd have to have so many parameters - right down to the
> metallurgical characteristics of the alloys used in the frame and strings,
> and the physical properties of the case and soundboard, including the
> layers of lacquer - that you could spend forever tweaking everything trying
> to get the perfect sound.
> And then you have the problem of how to diffuse it. A piano is not a point
> source, and a stereo simulation will lose a lot of relevant information.
> Speakers are always the weak link in the chain, and only become tolerably
> realistic if you have a five or six figure budget. Which means instead of
> trundling around a big flat harp on a truck you have to trundle around a
> huge speaker set instead. Oh well. :-)
> Meanwhile at the other extreme today's sampled pianos seem good enough for
> recording anything that isn't a commercial solo classical performance.
% Philippe-Aubert Gauthier, ing. jr , M.Sc.
% Étudiant au doctorat en reproduction de champs acoustiques
% GAUS (Groupe d'Acoustique et de vibrations de l'Université de
% [ Sherbrooke)
% CIRMMT (Centre for Interdisciplinary research in Music, Media
% [ and Technology)
This archive was generated by hypermail 2b27 : Sat Dec 22 2007 - 01:46:11 EST