[sdiy] CMI III oscillator.

[Scott Nordlund]

----------------------------------------
> Date: Mon, 8 Aug 2011 15:28:34 +0100
> From: dalenkarl at yahoo.se
> To: synth-diy at dropmix.xs4all.nl
> Subject: [sdiy] CMI III oscillator.
>
>
> CMI Series III:

Wow, thanks for that. As I had suspected, the Wersi MK1/EX20 are very similar (well, more like the CMI I/II).  It has the same dual/interleaved 6809 processor thing, which works because the 6809 needs RAM access for less than half of its clock cycle, so two can share the same RAM chips and exchange data without any penalty. Also each voice is an autonomous, variable playback rate module. The Wersi stuff is like the CMI I/II where each voice has its own wave RAM, but in this case it's small enough to fit in the internal registers of a mask programmed Zilog Z8610, and I think it's synthesized by the main CPU and stored in this memory when a note is triggered (which permits dynamic voice allocation). Another difference is that the Wersi dynamically switches between a few different waveforms at various playback frequencies, to keep the clock rate within reasonable bounds, and there's some linear interpolation (so I assume it's actually storing data in some DPCM format). But these are just single cycle waveforms with 16 or 32 harmonics, or simple formant synthesis that stretches or truncates a waveform.
I'd always wondered how much the CMI III extended the architecture of the I/II versus starting from scratch. I wasn't sure how easily the design could accommodate a shared waveform memory. I think Dave Rossum was much more clever since he designed the Emulator I that way from the start.