[sdiy] Variable rate waveform playback in NED synclavier

[Scott Nordlund]

> I bet I could do it for less than the cost of a Fairlight CMI-30A! You could use a high frequency version of the scheme that Richie proposed, then use it as a clock for a fully static microprocessor. You'd need a NCO/filter/comparator/uP per voice, and the rest is software.
>
> It'd be at least an interesting modern version of the classic variable rate scheme.

I think that's what the CMI-30A is trying to do, using FPGAs or something or other. The nice thing is that you'd be able to do all the things that normally don't work well in discrete time, i.e. PWM, drastic waveshaping, etc. without any concern for interpolation or band limiting. Everything can be implemented in a totally naive way without aliasing. Filtering wouldn't be so easy, but the Kurzweil K250 and Akai S900 used another set of clock generators to control switched capacitor filters...

I made a basic prototype in Pure Data of a Synclavier voice. It's not painstakingly accurate, but it seems to work pretty well. I've also worked out all the math for the "dither" thing I mentioned earlier. I'll be trying that in a few minutes...