Id like to add to the discussion which, a bit off topic, dealing with sampling or modelling organs/pianos.
ORGANS AND HARPSICHORDS
Organs and harpsichords are, in my view very suitable for sampling. Because they dont have dynamic variance a layering of say 8 samples per note and another 4-8 of release samples make a suitable recording. In fact a sampled organ or harpsichord is what I prefer over modelling.
I love that pianoteq includes two harpsichords and I really hope for more historical instruments like more harpsichords or perhaps a clavichord ! But for the time being, I use a sampled harpsichord whenever I want to explore the art of the fugue.
PIANO
Talking about piano is a totally different story in the light of pianoteq. The sampled pianos a couple of years ago with say 4-8 layers of samples per note just didnt sound or feel right. So they added more samples, and more, and more. Until you would end up with a ridiculousy big sample library of just one single piano. And of course, even with a huge library, the sympathetic resonance was missing.
In my mind, moddart has done something extrordinary creating pianoteq. I keep and old Danemann upright from 1920 which I tune and maintain myself and love and cherish and soforth. But playing on my roland fp5 with pianoteq is something exquisit and extraordinary, I prefer it over any sample library and also over real pianos and grands, because Ive get to adjust the touch, and feel, and sound to my exact preference.
WIND INSTRUMENTS
Want to add something further to the debate about sampling vs modelling.
Talking about any windinstrument - I believe the best way to go is samples. Given that the tone is even more complex in a windinstrument then in a piano, to create a modelling of it would be strange. How to modell the small variances of airflow in a human way ?
Summing up my post:
Harpsichords and organs - sample them!
Piano - pianoteq !!!
Any windinstrument - sample it !