[sdiy] Synthex Oscillator

[rsdio at audiobanshee.com]

On Jul 11, 2017, at 8:40 PM, Neil Johnson <neil.johnson71 at gmail.com> wrote:
> I've just simulated the counter/DAC part in LTSpice, and I must say it
> is rather a neat solution!  Not perfect - as Brian alluded to the DAC
> is not perfect - but the result is a very usable sawtooth.

That's great, Neil. Thanks.

Would you be willing to share the LTspice file? I'd like to make my own signal graphs by varying certain parameters of the analysis, particularly the time scale.


I was thinking about another aspect of the Synthex that I'd like to check out with the SPICE output voltage graphs.

For background, I assume that most of us are familiar with the fact that audio taper potentiometers are not always logarithmic in response, but are often piecewise linear approximations to simplify manufacturing. On a similar note, many fixed sample rate samplers use linear interpolation when altering the pitch of recorded audio. In light of these techniques, I find it interesting to consider the differences between the ideal performance and the compromises made in an affordable product.

I was thinking that, if the Synthex really is an analog sawtooth integrator with a changing current sink that tries to linearize the ramp, then it would actually be a piecewise logarithmic approximation of a linear charge rate (linearly falling voltage). If it is, that would be quite interesting (to me, at least).

The typical DCO has an active integrator, and the ramp voltage changes in a linear fashion, resetting for each cycle according to some (SYNC + MR1) signal from the digital clock that guarantees an accurate pitch. Standard DCO circuits might have some nonlinearities during the reset section of each cycle, depending upon how the capacitor charge is reset, but I believe that most have a very linear ramp in between the resets.

In contrast, the Synthex connects an array of resistors in parallel, to ground and +5V, that discharge the capacitor. I haven't seen the SPICE output yet, but I'd assume that this would be the typical logarithmic charge curve seen with any simple RC circuit (even though there are many resistors in parallel, it should be equivalent to a simple voltage divider). This logarithmic curve would change parameters each time the counter changes the parallel combination of the resistor array. It might require zooming in on a fine time scale relative to the counter clock rate, but I'd be curious to see what that actually looks like.

Note that two of the codes, 0x00 and 0xFF, are pure pull-up or pull-down resistances. The remaining 254 codes are voltage dividers whose output voltage is not perfectly predictable (or constant) because the current flowing out (or into) the capacitor will cause the divider voltage to be off by some margin. In any event, I would expect the sawtooth to be a fairly jagged line, but not quite like a voltage DAC stair step.

Brian