Modeling an acoustic piano...

[joshb]

Happy New Year everyone.

I thought I'd start the new year with an interesting experiment...modeling an acoustic piano. Can anyone point me towards a good starting point?

Karplus-Strong?

Any clues how Pianoteq does it?

[Music Engineer]

Karplus-Strong?

yes - but you will at least need an additional allpass in the feedback loop of the delayline to get the inharmonicity of stiff strings:

http://www.dafx.ca/proceedings/papers/p_071.pdf

and you probably want to model sympathetic resonance, i.e. a mutual cross-excitation of all the individual strings. i guess that implies to actually run the delaylines for all strings at all times making it essentially a giant feedback delay network, but i'm really just guessing

[Ananke]

https://ccrma.stanford.edu/~jos/pasp/Pi ... hesis.html

Could be an interesting project. Maybe stick it on github after

[EvilDragon]

Pianoteq uses much more than K-S... Lots of equation solving in realtime, possibly a crossbreed between waveguides and additive synthesis as well...

[syntonica]

You've got a number of things to model: the strings, including the initial strike, sympathetic resonance and reverb, the damper, etc. Then, if you want, I suppose you could use samples for the sounds of the hammers, the pedals, etc. There's quite a bit going on under the hood. The latest Pianoteq seems to go overboard at this--I find it overly muddy and noisy. Sampled pianos seem to be more clear and open sounding, at this point, although limited in adjustability.

[EvilDragon]

Latest Pianoteq is absolutely not going overboard... you can reduce the noise volumes if you find it noisy (I don't).

[syntonica]

I've just listened to the new demos, I haven't futzed with the new plugin yet, but I still didn't care that much for the previous version. Somewhere between Pianoteq and True Piano lies modeled piano Nirvana for me.