The question most often asked by musicians and composers (and perhaps least often by engineers) about physical modeling sound synthesis is: Why? More precisely: Why bother to simulate the behaviour of an instrument which already exists? Surely the best that can be hoped for is an exact reproduction of the sound of an existing instrument. This is not an easy question to answer, but nonetheless, various answers do exist.
The most common answer is almost certainly: Because it can be done. This is, of course, a very good answer from the point of view of musical acoustician, whose interest may be to prove the validity of a model of a musical instrument, perhaps by comparing simulation results (i.e., synthesis) to measured output, or perhaps by psychoacoustic comparison of recorded and model-synthesized audio output. Beyond the academic justification, there are, of course, boundless opportunities for improvement in musical instrument design using such techniques. From a commercial point of view, too, it would be extremely attractive to have a working sound synthesis algorithm to replace sampling synthesis, which relies on a large database of recorded fragments. (Consider, for example, the number of samples that would be required to completely represent the output of an acoustic piano, with 88 notes, with 60 dB decay times on the order of tens of seconds, struck over a range of velocities and pedal configurations.) On the other hand, such an answer will satisfy neither a composer of modern electroacoustic music in search of new sounds, nor a composer of acoustic orchestral music, who will find the entire idea somewhat artificial and pointless.
Another answer, closer in spirit to the philosophy of this author, is that physical modeling sound synthesis is far more than just a means of aping sounds produced by acoustic instruments, and it is much more than merely a framework for playing mix and match with components of existing acoustic instruments (the bowed flute, the flutter-tongued piano, etc.); though interesting, this might well be an application which would appeal to engineers only. Acoustically-produced sound is clearly a conceptual point of departure for many composers of electroacoustic music, given the early body of work on rendering the output of abstract sound synthesis algorithms less synthetic-sounding [147,195], and, more importantly, the current preoccupation with real-time transformation of natural audio input. In this latter case, though, it might well be true (and one can never really guess these things) that a composer would jump at the chance to be freed from the confines of acoustically-produced sound if indeed an alternative, possessing all the richness and interesting unpredictability of natural sound, yet somehow different, were available. This is what physical modeling is all about, at least for this author. The rest of this book is an attempt to inch closer to this ideal.