[sdiy] [synth-diy] numerically controlled superoscillator without hard sync

[Scott Nordlund]

Actually, I think what he means by the "combo organ" comment is that 32 samples per period offers only very rudimentary control over spectrum. You get to define only the first 16 harmonics, and the rest are just copied images of those 16. The result is similar to the buzzy "fake sawtooth" waveforms you get in divide-down machines by summing octave divided square waves. This isn't necessarily bad in itself. I actually quite like the sound of high frequency images, but I think the sweet spot is about 64 to 512 samples per period. You get useful control over the spectrum but can still hear the higher harmonic images. 32 is just too coarse. Phase-locked octaves in divide down organs are a different issue (and actually you even tend to get beating on divide-down organs that have drawbars for 5 1/3'...).

With a divide-by-n variable rate system, the problem with higher resolution waveforms is that it becomes difficult to generate high enough clock frequencies while maintaining decent frequency resolution. The PPG Wave 2 solves this if you can find its incomplete and unreadable service manual. Basically, it does octave selection by injecting clock pulses into different stages of a divider chain. So the lowest octave increments the waveform by 1 sample (128 samples per period), the next by 2 samples (64 samples per period) etc. The clock frequencies needed are the same for each octave, and the master oscillator frequency doesn't have to be particularly high (although that part isn't in the service manual, so I don't know what it actually is). The problem is that, like multisample switching, switching octaves makes an audible transient.

The Wave 2.2 and 2.3 I think used ordinary fixed sample rate phase accumulators with drop-sample playback at about Fs = 195 kHz. Palm explained that it would have to be time-multiplexed in order to add a second oscillator without requiring an unreasonable amount of hardware (though this isn't really true; the Korg DSS-1 does this in a very clever way). I still haven't found a schematic for the Wavecomputer 360 (or even high resolution photographs of the boards), but it seems to do something else...

Anyway the Gleeman Pentaphonic also used waveforms that were 32 samples per period...

----------------------------------------
> From: tom at electricdruid.net
> Date: Mon, 10 Feb 2014 10:56:03 +0000
> To: cheater00 at gmail.com
> CC: synth-diy at dropmix.xs4all.nl
> Subject: Re: [sdiy] [synth-diy] numerically controlled superoscillator without hard sync
>
> As Roman mentioned, this type of "wavetable" technique has quite a distinctive sound, probably due to everything being locked into tight harmonics. If you use more than 32 samples, you can get past "combo organ", but I still know what he means. Those combo organs used frequency division to get harmonics that were locked together, and this technique has the same fault.
>
> However, you're going to have to be careful - if you've got complex non-integer ratios, the imaged frequencies won't fall neatly into the same sample every time, and you're back to having jitter in your signal. Think about why you don't get that when everything is harmonic.
>
> T.