Joseph and Christian (the latter is a colleague of mine at the IKA http://www.rub.de/ika
), started a discussion on simulating pipe organ signals in conjunction with appropriate room acoustics to achieve flexible physical organ models. Here are some thoughts about that issue:
There are several possible ways to generate pipe organ sounds artificially without the use of samples. A widespread approach is Digital Waveguide Synthesis (the inventor of DWG, J.O. Smith of the CCRMA offers an E-Book on his institute's web site http://ccrma.stanford.edu/~jos/
which contains all necessary information). This technique is a time-domain simulation method which enables astoundingly realistic generation of instruments with linear resonators and nonlinear oscillators (such as strings, pianos, horns, flutes, drums and -of course- organ pipes). Some years ago, we used DWGs to analyse the characteristics of free- and beating-reed organ pipes. I know that there are several flute models implying DWGs that sound really nice. However, flue organ pipe modelling may be even more efficient by simply adding up the partials in frequency domain, because the harmonic structure of organ pipe sounds is rather simple. The same holds for the temporal structure, after the attack transient time is over, the sound is practically static (except for "Schwebung", of course). However, I think this shows the most critical aspects, when a special organ pipe is to be modelled: 1) the attack phase has to be reproduced very carefully, as it essentially determines the perceived timbre, 2) the formant balance which determines the spectral features of the sound has to be adapted to that of the given pipe. The first aspect can be simulated rather using a time domain approach, while the latter aspect is more easily controlled in frequency domain. Perhaps a hybrid model works best?!?
It may be very difficult to create a model that is able to synthesize the sound of a given pipe *exactly*. Of course, when size and length are provided, you will come close to the desired sound. However, the "special" cues that make pipes of, let's say, Silbermann and Cavaille-Coll or even of two different Silbermann instruments distinguishable, are determined by some "other specifications". Here, things become complicated. How does the metal or wooden body of the pipe influence the sound? How is the exact airflow geometry when the sound starts? And what does it influence at all? How is the chest constructed? Again: perhaps a hybrid model that combines DWGs and filtering works best.
By the way: in my opinion, noise reduction should be applied only very carefully to organ pipe sound samples, because the flow noise is (may be) an essential part of the sound (especially for narrow stops). How do you think about that?
And finally: Congratulations, jeuxdorgues.com is a great idea and a formidable project! The sound examples that are provided are very nice!