[sdiy] dsPIC Stuff

[Scott Nordlund]

> I bet this would sound really nice with multiple K-S models for each
> note to give a slightly detuned sound as each note decays away.  It
> sounds great as it is, but would be a bit less static with voices
> doubled up or chorus applied to the voice mix afterwards.

I've noticed that you can get nice beating effects by coupling two string models together with a rotation matrix. It should behave similar to two coupled strings (not quite the same as two detuned strings). Something I've been messing with is coupling a network of string models this way in a sort of ring topology to simulate sympathetic resonance. Even with small amounts of coupling, the result is much more rich sounding than simple models. Each resonant mode gets sort of split into a number of "sidebands" that beat against each other. A simpler way to do it is to just add some sort of global feedback, but this can be unstable or have weird frequency response. A more mathematically correct way is a Householder feedback matrix, but the degree of coupling here is inversely proportional to the number of strings. I guess if there are too many strings it could be split into subsets, but that's not really something you can adjust on the fly. I can look up a Yamaha patent that might help with that, if anyone wants to go down that road...

> PS. Adding some peaks and dips to the overall frequency response might
> also sound nice to mimmic body resonances in a real stringed instrument.
> Does anyone know how close you can get to an acoustic guitar sound
> using a bunch of Karplus-strong algorithms and a few IIR filters to
> model the key body resonances?

Well, the Moog string filter could be translated pretty easily to a bank of IIR filters, but I think body resonance has some time domain characteristics that can't be simulated this way.   Also simple feedback delay networks seem to be useful, if not terribly accurate. I've had good luck with just randomizing delay times and feedback matrices. I think the fashionable way to do it these days is to commute the body resonance to the "stimulus" signal. You can use some sort of additive synthesis or filtered sample playback, depending on how accurate you want to be regarding velocity sensitivity, etc. Or the output can be convolved with a pre-recorded impulse response. It's also very easy to model model pluck position using a feed forward comb filter. Another feed forward comb filter and second order IIR can model a guitar pickup.

What would sound great is coupled waveguides, FDN resonantors and some sort of waveguide-like reverberator, basically the same thing at 3 different scales (string, body and room). That might be a little much for a dsPIC implementation though.

(different Scott)