Guillaume, thanks for your answer. I enjoyed your internal jokes
*On a sidenote: Not really a piano slowly going out of tune, but VERY subtle random unison variations would certainly contribute to realism. This has been said before, and it is only logical to think about this (see below).*
Concerning fff, I am not sure if your suggestions can do the trick. I have tried these things of course, but there is a point when the sound gets too artificial when you increase the hammer hardness yet further.
Listening suggestion: Keith Jarret, Köln Concert. Part I towards the end, Part II from 7:30 (right in the middle of it). There is a "ringing", feedback-like quality to the sound that seems more than the sum of its parts (and probably not easy to express in a math equation). I haven't yet succeeded to reproduce that. By increasing the hammer hardness beyond reason, one thing that happens is that the sound gets somehow thinner, less physical. Not so in a real piano.
How about trying a kind of (subtle of course) phase distortion or waveshaper on a per-note-basis (for the really hard strokes)? That is one thing that greatly helps with drum sounds. Just a bit of it, and the whole thing really bangs! One can try this kind of thing with Logic's sculpture, and while this beast is very hard to control, it sometimes goes in that direction.
Maybe it would also be interesting to take into consideration more noises from the piano -- not just the obvious hammer noise, but also the pounding on the keybed or the amplification of the hammer noise by means of cabinet resonance. Maybe you have already done that, but also maybe it was not quite enough.
There is a funny paradoxon: In instrument design, one focus is, logically, to get as much *sound* (from the strings/air column) as possible, reducing unwanted noises (hammer, action etc. in the case of the piano), and in digital pianos, we look for those noises as factors too bring more life into an otherwise dead clean emulation.
One prominent example is the sympathetic resonance. I think this was not really intended from the start but more of a by-product of incorporating several strings into one casing. Now, after 300 years, we are loving it and could not do without it. Except maybe Glenn Gould who supposedly hated most everything about playing, including the audience Did he use the pedal(s)? I bet very seldom.
Here's a pic that looks as if he was at least tempted to pedal (legs crossed of course, or it wouldn't be him):
http://www.perfectpitchpeople.com/gould.gif
But this shows quite an adolescent Gould; maybe he got over this nonsense later
The real problem with the controller is maybe not so much that there's but 127 (although a finer resolution would certainly be desirable) but that in most controllers the 127 ceiling is reached to early. Maybe there are some that are virtually impossible to push to 127, but on those that I know, "the "hard" setting does not set the stake higher but simply activates another (exponential) curve. The headroom stays the same.
Generally, I would say "headroom" is the basic field to explore in future PTQ development (and this applies to every modeled instrument, as well as to MIDI controllers). The soft notes are already done brilliantly (and they are maybe the factor I love most about it). I am looking forward to future developments and upgrades. For which I shall be happy to pay