posted 20 March 2003 03:16            

struckplates.kym clampedcircplate.wav freecircplate.wav

This is accomplished by spinning off the first 15 modes of vibration (m,n)=(0,1)..(0,3) to (4,1)..(4,3). The frequencies of these modes are fixed by the boundary conditions on the plates. But the relative amplitudes of the various modes depends on where you strike the plates.

Using Mathematica, I computed the eigenfunctions of these plates (the shape of the vibrations as functions of radius), and stored them as 15 contiguous sample sets of 256 samples each, all packed into a 4096 element WAV file. There is a WAV file for each of the boundary conditions. As the mod wheel (!CC01) varies from 0 to 1, the strike position is moved from the center of the plates to the edge.

A multiplying waveshaper fed by a constant object is used to select the appropriate amplitude from this wavetable, given the mode indices (m,n) of the vibration. A script spawns off a bunch of envelope controlled oscillators for each of the first 15 modes.

You will notice some difference between the two plate configurations -- different frequency sets. But most notably, when you strike a clamped plate at the edge, you get no sound -- the edge is clamped and so it can't vibrate or respond to any strikes.

This is a DSP hungry Sound, and so I dropped the MIDI polyphony to 1 for presentation here. That allows these sounds to run on a 4 DSP Capy. If you set the polyphony to 3 you will need more than 16 DSP's.

[I don't understand the scaling here... if one instrument runs on 4 DSP's, then why can't 5 instruments run on 20? All I know is that the max I can run on my 20 DSP Capy is polyphony 3.]

I'll have to think about how to do this for the metal filters... Right now these sounds are really 15-oscillator monsters, each with their own ADSR envelope and amplitude lookup table (multiplying waveshaper). There must be a way to do this amplitude scaling for filterbanks...