[sdiy] Top Octave Modular

[rsdio at audiobanshee.com]

On Apr 2, 2017, at 2:58 AM, Elain Klopke <functionofform at gmail.com> wrote:
> Oh, ok, here's another question... Would it be better to have square wave outputs (what comes from the NoteDiv chips) or convert the square waves to saw waves between the NoteDiv chip and the switch matrix?
> 
> -Ian

I think that Tom or maybe someone else has already answered this quite well, but here's my take:

For organ, you want sine waves;
for strings, you want sawtooth waves;
and I'm not sure for what you would want square waves, except straight up synth sounds.

Of course, for organ, those sine waves represent individual drawbars, and it's typical to mix up to 9 sine waves together to get a full pipe organ sound. Starting with sine waves and adding harmonics allows for precise control of the timbre, and maps quite directly from "real" organ pipes to organ synth controls.

For string synths, you don't really stack harmonics because the sawtooth already has a full complement of harmonics. Instead, two or three parallel top-octave synths might be mixed in parallel, with the important factor being that each individual top-octave sawtooth generator has its own independent clock source. This independent clocking allows slight detuning between the unison notes, much like multiple violins are stacked to get more volume even though they're not all perfectly tuned together. Preferably, the clocks are analog, not digital, so that their frequency can vary continuously without being limited to whole-number or rational fractions of some fixed input clock. An obvious way to cheat here is to use a single top-octave generator with a chorus effect to simulate the detuning, but if the top-octave generator with sawtooth output is cheap enough then you can get that chorus effect by duplicating the source oscillators without all the noise and expense of a BBD or other thickening agent.

For square waves, which are much easier to manipulate with CMOS digital / analog hybrid chips, you also don't need to create harmonics with additive synthesis, and you'll have something like a string synth without the same harmonic spectrum. As this thread has already discussed, filter banks can take the typical square wave harmonic spectrum and shape it slightly to get close to the desired sound. I just don't think you're going to get a Hammond sound when starting with square waves, but Hammond is not the only useful sound.

So my answer is this: it's not "better" to have square or sawtooth because it really depends upon the desired sound. Sine waveforms are probably best for pipe organ sounds, but there are many ways to get similar end results. I think it would be better to have options. A module should be basic while allowing complexity to be built on top of the basics.

Brian Willoughby