I’d consider the “holy grail” of DIY synth module to be a key-tracking (V/Oct) VCO without any specialty chips, a low part count, and an easy build. I’ve never found this - solutions always seem to use specialty chips, like the AS3340 in the LMNC VCO, or require difficult builds, like the TH-555 (which also has a few specialty parts).
So, I was thinking about different approaches. Rather than thinking about VCOs as we normally do - what about KTFs? A high-resonance self-oscillating KTF will essentially act a sine-wave VCO.
Well, it turns out there is already a DIY KTF out there! The Yubi-Synth MiniMoog clone! It does use one specialty part - the now-discontinued CA3046. But, the CA3046 is just a matched transistor array, and DIY transistor matching is already done regularly by the community here, as described in other threads. Other than that, it has no specialty parts other than a reversed potentiometer.
Reports I’ve found online suggest this has four to five octaves of accurate v/oct tracking down to 60Hz with well-matched transistors, and there is advice for boosting the resonance of the filter.
So, I guess my idea is this - modify the Yubi design to remove the audio inputs, circuitry and controls, hard-code the beefed-up resonance all the way to max, replace the CA3046 with a manually matched array - and there you go! A v/oct tracking DIY VCO with no specialty parts! Maybe add a simple ramp/square generator to the sine output for some more variety.
Anyway, this was just an idea that floated into my head, was wondering if the wizards here had any thoughts. Has anybody explored this avenue before, are there any reasons that it wouldn’t work that I’ve not thought about?
Sine wave oscillators aren’t much good for east coast subtractive synthesis — no harmonics to subtract. In principle you could add circuitry to generate pulse/ramp/tri waves, I guess.
Four to five octaves of V/oct I’d regard as a bare minimum. Most good synth VCOs can do a good deal better. The YuSynth Minimoog filter appears not to have any temperature compensation so I’d expect its tuning not to be all that stable.
Not sure how you define “low part count” or “easy build”. But it seems to me if you took the Minimoog filter and added on circuitry to make pulse, ramp, and triangle waves, it’d end up at least as high a part count and at least as difficult to build as the YuSynth VCO — probably more parts, with all those transistors. The VCO is designed to use an LM394 transistor pair but as the note says you can use other options including hand matched BC547s, and there’s a tempco resistor but those are not (yet) that hard to obtain, and you could always just use a non tempco resistor and live with some tuning drift.
I built the MFOS VCO which is more complicated and more parts than the YuSynth, but likewise not reliant on specialty parts, and it wasn’t particularly hard to build, just more time consuming than a lower part count circuit.
Sorry, I should have said “Key Tracking Filter” in the OP. Just a VCF with a lin/exp CV control, ideally v/oct.
I guess temperature was the thing I was forgetting. I hadn’t looked at the MFOS VCO before, but it does like like there is some weirdness - the “specialty”-ish chips and the tempco stuff - that puts it out of holy-grail territory, but I suppose they’re necessary for the same reasons that that would be with the proposed VCF-based design.
Only a few octaves of tracking bottoming out at 60Hz and only sine output is fine for me, musically, because that provides the one thing I’m always after: big fat subs. (Down to 3/40Hz would be nice as that’s where my hog scoops round off, but 60 is still ok).
I’m also not super concerned about temperature - though maybe I should be - because I use a recording/sampling based workflow, so I’m not doing any live shows or long jams with the synth. I think by jettisoning the controls, the tempco stuff and the rare chips, this should still should end up on the easier side of the MFOS or Yubi VCOs, though just barely, and maybe the featureset wouldn’t be enough for most people although I think it would suit me.
No specialty chips in the MFOS VCO, and as I said, you can skip the tempco if you don’t worry much about temperature drift.
I still think by the time you get pulse/tri/ramp waves out of it, the VCF would be no easier to build than the YuSynth VCO. But I’d be happy to be proved wrong.
@EddyBergman has some interesting comments about building the Thomas Henry 555 design in a recent blog post. The positive temperature coefficient thermistor is, he suggests, replaceable by a 2K resistor at the cost of stability. He also has some useful suggestions on transistor matching and on thermally connecting them (superglue is good enough.) If you haven’t already read this article I recommend it. He even gives British, mainland European and United States links for sources of the thermistor.
The Thomas Henry VCO is, at the moment, my favourite DIY VCO. Really the only exotic component is the TEMPCO PTC and they are readily available. I live in the Netherlands but I ordered mine from the UK, making sure I kept the total cost below €22 to avoid extra fees and it worked out perfectly. I got them in the mail a week later without any extra costs. The VCO now tracks really well at least over 5 octaves in my case. It’s a bit tricky to tune though but once you experiment a bit you soon get it tuned. It has sinewave included, which the AS3340 VCO does not. I can really recommend it.
Well, it has extra circuitry to implement a triangle to sine converter (done by overdriving an OTA, same approach as the MFOS VCO mentioned earlier). The oscillator itself is a triangle core.
What do you mean by “difficult”? As for specialty parts — here’s the parts list. Aside from the tempco I think everything can be bought from Tayda. Everything’s through hole.
I’ve had a think about making a simple and stable VCO too. Obviously it’s impossible to get a perfect 1V/oct current sink using transistors, because of temperature considerations, so I focused on making the oscillator itself as simple as possible.
I originally posted this on the Synth DIY group on Facebook, but I think I neglected to post it here?
The left part of the circuit is a bog standard V/oct current sink, so modify to taste. The tricky bit is the right op-amp. I’m not sure if it’s been used before, but I got it working. The two transistors make up a Programmable Unijunction Transistor (PUT). They’re a negative resistance device that clamp short when a threshold voltage is reached and don’t release until the voltage (or is it current) is removed. This makes a sawtooth oscillator, and it’s quite simple to build.
The key word is “perfect” Not going to watch a 25-minute video right now to see which compensation approach he’s using there, but even if you put all the transistors next to each other on a single chip (like e.g. the 3340 does), you’ll have temperature stability issues if you look close enough. If they matter in practice is another thing.
That’s exactly it. The compensation method in the video is the PNP emitter follower followed by an NPN current sink trick. Far from perfect. René Schmitz et al have gone into the theory of temperature compensated exponential converters quite heavily.
I’ve wondered for a while if it’s possible to use the exponential curve from a capacitor charging through a resistor to make an exponential converter, using sample and hold techniques.
I made it as far as making sure the oscillator part oscillated outside of the simulation. It does. I haven’t gone any further into it than that. As for the sound, it’s almost entirely not unlike a sawtooth sound
The matched and thermally coupled transistors eliminate one piece of the temperature dependence, as perfectly as they’re matched and coupled, but there’s a remaining piece they don’t get rid of. That’s what people use tempcos for, or NTCs. They compensate for the remaining temperature dependence, but they don’t do it perfectly. However they do it well enough for most purposes over a useful range of temperatures.
Another approach that’s been used is to heat the converter transistor to a particular temperature and hold it there.
Nothing’s ever perfect in engineering. You just get it as close as needed however you can.
Not going to watch a 25-minute video right now to see which compensation approach he’s using there, but even if you put all the transistors next to each other on a single chip (like e.g. the 3340 does), you’ll have temperature stability issues if you look close enough. If they matter in practice is another thing.