[sdiy] simulate detune / polyphony with a single DCO

[Mattias Rickardsson]

On 10 April 2017 at 00:23, Chris McDowell <declareupdate at gmail.com> wrote:
>
> Howdy list,

Howdy! :-)

> I've been playing with DCOs, and am realizing that detune or even polyphony (which would end up paraphony I guess) could be achieved with a single DCO. The idea is to use two timers in a microcontroller, pulsing the same pin that's used to reset the DCO's integrator. A built in DAC from the micro controls the rate of the integrator, and as detune increased, would adjust the slope. I have a PCB on the way to test all this out, looks good in LTSpice.

The sum of two constant sawtooths has a constant slope with repetitive
jumps when the two oscillators reset, that's right. The main
difference from regular oscillator signals is that the jumps occur at
different voltages, since the other (non-jumping) oscillator is added
to the mix, and can be of any voltage. The jumps always have the same
height though. (Otherwise it would end up as the "metal sync"
combination in Roland's JX DCOs and certain other synths.)

I guess it would be theoretically possible, but you would have to swap
the common oscillator reset circuit for a charge injection circuit,
that produces those jumps with the same height by injecting an
identical charge into the integrator at every oscillator period,
regardless of the present voltage. (Tom mentioned the ARP/Rhodes
Chroma... didn't that one actually have a charge pump in its VCOs,
albeit for other reasons?)

> Anyone heard of this being done before? Are there any glaring dumb oversights?

Haven't heard about it at all. I've been thinking in the same
directions, but haven't developed the thoughts much further since it
feels like there will be several problems to tackle in a real-world
scenario, such as:

- The [result of the] charge injections must be identical, even when
the resets occur at the same time or very very near in time. Possibly
impossible without having one charge injector per simulated
oscillator.

- The slope must cancel the jumps in order to avoid DC buildup. Might
be possible to do with clever error feedback, but I wouldn't bet on
it.

Also, changing individual volumes would be a bit tricky, and forget
having other waveforms than sawtooth. So... having proper DCOs for all
tones feels like an extremely simple and secure solution after all.
;-)

/mr