|
|
Equipe Analyse/Synthèse
|
Physical Models of Trumpet-like Instruments, Detailed Behavior and
Model Improvements
Xavier Rodet, Christophe Vergez
Proceedings of the International Computer Music Conference
(ICMC'96), Clear Water Bay, Hong-Kong, August 1996.
Abstract
The behavior of a model of instrument has to be foreseeable in order
to be used by musicians. We are working in this way. We first
develop a basic model of a real instrument to capture its main
principle. Then its behavior is
studied and some properties are settled about its different ways of working:
periodic, quasiperiodic, or chaotic trajectories. Finally we try to
improve our model without loosing any information about its behavior.
We are studying four models of instruments with
sustained sounds : clarinet like, flute
like, violin like, and trumpet like instruments. These models have the
same basic structure:a non linear time delay system. But each classe
of instruments has its own specifications. For instance, the flute
model includes a second delay loop. In the same way, the linear
feedback loop of the violin model is composed of two interleaved loops with delays, and the force
exerted by the bow on the string is an instantaneous nonlinear
function of the bow's velocity. Our model of trumpet like instruments is rather more
complicated. In fact, since the mass of the vibrating lip can't be
neglected, the hypothesis of an instantaneous nonlinearity can't
stand. The nonlinear function has then two inputs.
The behavior of the four previous systems shall be studied, in order
to control the models efficiently. A further step is to consider
more elaborate models. Their behavior is also precised to avoid
loosing some control abilities. Furthermore the effects
of each improvement will have to be analyticaly weighed up to consider how closely the model's behavior
matches that of the associated acoustic instrument. For instance, we could include in our basic trumpet, a more realistic
reflexion function, taking into account the effects of both the
mouthpiece and the bell. This could be done either by identifying
parameters of a measured reflexion function, or by using truncated
cones with fractional delays. Another way of improving the model could
be to describe more precisely the wave travelling into the bore. Then, viscothermal losses
and nonlinear propagation in the waveguide should be considered.
Our work should lead to the best model, in terms of a compromise between realism and computational
complexity.