[sdiy] Prophet VS a phase accumulator design?

[Rainer Buchty]

On Tue, 1 Feb 2011, Paul Maddox wrote:

> The ESQ-1 DOC chip I thought did something different, I have a feeling 
> that "drops" samples as you reach higher octaves. Rainer Buchty is 
> *the* man for ensoniq and he's on this list somewhere.

The DOC uses a 24 bit accumulator and a 16 bit incrementum. Regarding 
wave playback, there are 2 parameters setting wave size (256b to 32k in 
discrete 2^n steps) and address resolution (quick-and-dirty octave 
jumping). Depending on those settings, a maximum of 8 accumulator bits 
are used as fraction, in worst case only 2. However, the SQ80 uses a 
maximum of 4kB in wave size, not 32kB.

In order to avoid too much sample skipping for higher octaves, 
multisampling is used with 8-key-wide multisample zones (i.e. 16 entries 
per sample). In order to further avoid aliasing, the higher multisample 
waves are more or less uniformly mapped towards an octave wave, some 
more "aggressive" sine, and finally sine.

Regarding the ESQ1/SQ80 OS, the wave octave and fine-tuning settings are 
computed together in order to determine the multisample zone, whereas 
modulation is not. Hence, you can play the usual game of addressing 
"wrong" zones by extreme pitch modulation.

Btw, especially in the long waves there are audible zone splits, 
especially noticeable in the BRASS waveform(s).

The sample rate scales with the number of active oscillators and on the 
ESQ1/SQ80 is 1MHz/(24+2)=38.462 kHz -- on Mirage's SoundProcess it drops 
down to 1/(32+2)=29.412kHz due to 32 active oscillators.

So for a PPG-like rate, you'd need to restrict the DOC to 3 active 
oscillators (plus 2 refresh cycles), hence 1MHz/(3+2)=200kHz.

Rainer