[sdiy] Synthex Oscillator

[Tom Wiltshire]

==================
       Electric Druid
Synth & Stompbox DIY
==================




On 12 Jul 2017, at 02:44, rsdio at audiobanshee.com wrote:

> On Jul 10, 2017, at 12:28 PM, Tim Ressel <timr at circuitabbey.com> wrote:
>> FYI there is a similar circuit in Chamberlin (2nd ed, pg 627) called a hybrid voice module. It does waveform lookups and some morphing of the addresses to get different effects. It looks interesting.
> 
> Thanks for the reference, Tim. I still haven't read every page of the Chamberlin book, but I do have the 2nd edition.
> 
> Note that the oscillator DAC in his circuit is a current output device and is followed by an op-amp current-to-voltage converter. Thus, it's clearly a digital-controlled voltage, making at least that part distinct from the Synthex. The SSM2040 serves as a variable antialias filter, whereas the VCF seems to be built from discrete op-amps and CA3080 circuits (instead of another SSM2040?).
> 
> The Synthex has no buffering of the DAC, so its resistor ladder interacts with the capacitor that Mario describes as the sawtooth waveform integrator. I noticed that the counter feeding the Synthex DAC is reset by a combination of SYNC and MR2 (Master Reset?), so it seems that the current always starts at a fixed value at the beginning of each period of the sawtooth. I want to see whether the various counters are coordinated enough that the DAC counter always reaches the same end value at the end of each sawtooth period, or if the slopes can be different.
> 
> Both voice designs use inversion of the counter (waveform?) output, but only Chamberlin inverts the address lookup to facilitate 4-quadrant reflection, reducing waveform storage (a technique used in the TX81Z, I'm sure).
> 
> Another curious difference is that Chamberlin starts with a very high clock rate, 17.145893 MHz, and handles the octave division first - before handling the note pitch selection. I would be tempted to cut the rate in half to reduce the fundamental from his range of 65.4 Hz - 33 kHz down to 32.7 Hz - 16.5 kHz, or maybe even closer to 20 Hz because I LIKE BASS. I'm not really clear on how the quad, dual, and single waveform patterns affect this range, so perhaps it reaches lower than the text describes. The Synthex starts with a 4 MHz clock and handles generating the top-octave note frequency first, then handles octave division second. I wonder if the Synthex has as fine of a pitch resolution with these lower clock frequencies. The 12-bit counter in the Chamberlin design may allow it to operate with higher clocks and finer pitch resolution. A modern Synthex redesign could make improvements here, I assume.

You assume correct - the waveform "mirroring" reduces the pitch by either one or two octaves.

In many ways, the Chamberlin design is simpler than the Synthex - or at least, more straightforward. There's none of the tricky fractional division, and the output DAC is *just* an output DAC as you've pointed out, no funny integrating cap with reset. The Chamberlin design just uses a higher clock frequency to avoid the need for some of the division tricks the Synthex uses. Chamberlin's analog side is a bit weird, using a 2040 as an anti-aliasing filter, and then using a pair of CA3080s as an SVF. That's probably not a design choice we'd make these days!

Tom