It does seem odd to me that the DCO is controlled by a square wave rather than just varying the control voltage. It seems to require a lot more contrivance to be made to work.
I was wondering if any of you had any insight into why they would have done that. Were the microcontrollers and DACs of the era just too limited in what they could do to output analogue voltages? Too expensive? Is a digitally generated square wave really all that different to a PWM pin?
Check out this video - loads of info on the workings of various DCOs:
As far as I can tell - the square wave you’re talking about goes really fast, and there is something that counts it to generate a reset pulse on the saw core you’d see in a classic VCO. How much you need to count controls how often you reset and hence the pitch. That number is often generated by a microcontroller or CPU. The fact that it’s something digital controlling the reset of the core is what gives it stable pitch.
In a pure VCO it’s a current that controls the pitch, and deriving a stable current is a hard problem. You get instable pitches, different pitches for the same CV on each voice, temperature drift etc, which is what a DCO is trying to address. Even if you drive a VCO with a high precision DAC, the analog V-I exponential converter will still misbehave. Compound that with the fact that 16bit DACs were expensive back in the day and I think you have the primary motivation to move to a DCO, especially for polysynths.
In that video there is also a type of DCO that has an analog clock(!) that is sped up or slowed down to do things like pitch bend - which is wonderful and whacky.
The exponential conversion will have temperature drift if you do it all the analogue way using discrete transistors. You have a microcontroller though. The DAC doesn’t need to be super fast to produce a single steady voltage for each note, you could just use a PWM pin with an RC filter. And maybe switch in a voltage divider to get better resolution on the bass frequencies.
My understanding of it is that the exponential nature of the current needed to drive the oscillator core makes this difficult. If you want to do it with a DAC you’d have to do that in software, take a linear CV or midi note or whatever, and then turn it into an exponential voltage. So let’s say you’ve got 10V max - the top octave would be in the 10 - 5V range: 416mV per semitone - no problems. The next octave is in 5 - 2.5V range: 208mV per semitone - ok. then 154mvV/st, 77mV/st, 38.5 - might be getting tricky, then we’re down to 20mV and things start getting difficult, any % error in there and you will be off a significant number of cents in your tone, and you’re tuning is all out. That’s 6 octaves - maybe acceptable…
Now, I’m not saying it’s impossible, you could probably do a pretty good job with switching ranges and all that kind of thing - but now you’re adding complexity and component count where it’s probably not needed.
That said - if you get something working this way I’d love to see it!
Ok that makes sense. Yeah, I’m half tempted to try breadboarding something, just as a project. I think an instrument can have fewer than 6 octaves in range and still be quite good. Maybe switching between a treble and bass capacitor could avoid dealing with miniscule voltages…
looking at the way the amplitude compensation is done, it seems like they’re adjusting the control voltage anyway. I suppose the clock pulse was just more precise.
Yeah - I think it had three ranges though - lo, med and hi. And that will only control the loudness, not the pitch, pitch is still controlled by the pulses. Human ears are less sensitive to differences in volume than pitch too, so it wasn’t such a big deal.
There’s a whole philosophical question as to whether such a clinical and digital level of precision is actually desirable in an analogue oscillator core. I suppose we think of that a bit differently after 40 years of digital synthesizers though. Perhaps at the time it seemed quite novel.
I watched a talk by Dave Smith not too long ago where he said that DCOs had a bad rap because of how they were implemented originally. I don’t know if he meant the Juno design specifically. But he was saying that he switched back to VCOs on the Prophet 6 mostly because that’s what the customers demanded, not because there was anything particularly wrong a good DCO
