VCO - handling 1v/oct with a microcontroller?

[jonbutler88]

Hi all,

While I'm pretty sure this won't work, I can't find anything that will tell me it won't, so please put me out of my misery!

Basically I've been reading around synth DIY for a few months now with the intention of building a few eurorack modules to go with my non-DIY collection. I'm not keen on kits as I want to get the hang of things like PCB layout (relevant skills for my job / future jobs), so I thought I'd try making a simple VCO from scratch. Of course I now realise that's a somewhat ambitious goal if I want it to track 1v/oct well, and I disappeared into decision paralysis around matching transistors vs finding a dual matched transistor pair IC vs using a V2164, and rather than making fun sounds I've spent too long worrying about accuracy and the joy has gone out of it somewhat...

I still want to make an analog VCO, but I don't really want to have to worry about the thermal properties of transistors and such like. What I really want is to handle the linear to exponential conversion with a simple microcontroller (e.g. something like an attiny), and then use that in place of the typical exponential conversion part of a VCO schematic (e.g. to feed into an OTA as part of the VCO core).

I'm assuming this won't work, because if it did everyone would be doing it I imagine. But I'm curious why.

First post, be gentle

[PHOBoS]

It's possible but there is a limit to the precision you can get depending on the used dac.

Grumble and JovianPyx can probably tell you more about it.

[JovianPyx]

What PHOBoS said.

There certainly are microprocessors that could do it - but do it well? I have doubts. There are even microprocessors that have both DAC and ADC, but usually the (cheap ones) have functional but not precise ADC and DAC.

To be honest, I started out using analog circuits to do electronic music, but I too didn't want to mess around with the expo stuff with matched transistors and tempcos. I built two Fatman synths. Those use "linear" instead of "expo" CVs and do not need special attention paid to temperature. However, expo has a clear advantage with modulation especially in a patchable synth.

If I were doing analog today, I'd want proper electronic treatment of expo signals with tempcos and matched transistors. That is the way to get VCOs that have the widest in-tune tracking range.

[Grumble]

My first few generators use a Direct Digital Synthesizer (AD9833), so it is at crystal stability and can be tuned very precise using a cheap chinese knockoff arduino nano.
I used a small piece of software to calculate the frequency register values from a given frequency, it is slower than using a lookup table but it enables me to modulate the frequency with a control voltage.
The frequency is set by a serial signal (note) derived from the midi input of the synth.
At this moment I turned to analog oscillators based on the AS3340 Memorymoog schematics from Electricdruid.

Sure it is possible to make a lin >> exp converter, but you need very accurate DAC and ADC's with a large input/output range.
I think it's easyer to build one from schematics that float all over the internet (including EM)

[AlanP]

I'm a simple man. I divide VCOs into two broad categories -- tasty, and precise.

Precise oscillators are largely digital, like the Mutable Instruments Braids oscillator, which uses a STM32 for everything. They can be analog, like Jurgen Haible's Living VCOs, but the cost tends to be higher to get accurate analog than with accurate digital.

Tasty oscillators are ones with funk and personality, like a friend who treats time as something he heard of once and turns up when he feels like, like a cat, but you like him anyway because of his personality.

[ixtern]

I suspect that simple controllers are not fast and accurate enough to make fast exponential conversion (think about frequency modulation to calculate, for example) and if you get controller with enough computing power, usually you can do other modules in software too.
So there is a boundary where using microcontrollers with analog circuits doesn't make sense.

Simple chips like Atmega 328 (or Arduino modules) may be used together with DACs(12 to 16 bits) to make linear control voltages (usually in MIDI to CV modules) to control analog exponential converters and that' all.

Making analog exponential converter is not a complicated thing - two transistors (monolithic pair, discrete or something between - two dies in one case) drived by two OpAmps and tempco somewhere.

There are still available old good CA3046, or double transistors (two dies in one case) like BC847DS or even paired 2N3904.

And there are some good tutorials about building VCOs.
The best for me was xonik's one:
https://xonik.github.io/theory/vco/expo_converter_1.html
https://xonik.github.io/theory/vco/expo_converter_2.html

[jonbutler88]

Makes sense, I hadn't run the numbers on how much precision a 14-bit ADC might get me (for example), but I just assumed it would be "enough". Comforting to know I'm not the only one that doesn't want to struggle with this though, and it sounds like perhaps a dual matched transistor pair and a tempco will be sufficient.

I guess I'll just push ahead and build something anyway, if it doesn't track at least it will have personality

[ixtern]

[Electric Druid]

It't not impossible to do Lin<->Exp conversion on a small cheap microprocessor. I've done it on a 16F PIC. The trick is that it's the same for every octave, but shifted one bit. So you can use an accurate look up table that covers an octave, and then do the right number of shifts. It's quick and it doesn't need any maths.
Which is handy, because we don't even have a multiply instruction.

Also worth pointing out that if you've got a microprocessor in your VCO, you can feed the VCO output back to the processor and add an autotune routine. So maybe you don't need that great a DAC after all - just compensate for it. It does however need enough *resolution* that you can't hear any little steps. It's only *linearity* that it doesn't need.

The worst problem (imho) is the ADC accuracy reading the incoming CV. If you're decoding a MIDI stream, you're ok because you've got fixed notes, and then you add a known pitch bend - all relatively simple. But an external CV? It could have *any* crazy modulation applied to it. It could even be the output from another oscillator as someone tries to do FM on your oscillator. Aliasing as the incoming frequency goes above your ADC sampling rate is going to cause all kinds of chaos, so you have to limit stuff like that.

It's certainly not a simple problem, but the linear-exponential conversion is the simplest bit, ironically.

[ixtern]

Ah, lookup table, how could I forgot this simple solution which I've used with Arduino frequently. Although I've went further and even for modulation steps I've used lookup tables - one table for one modulation step. Crazy but worked. Not for expo conversion but for other solutions.
Thanks for the tips.

[VA1]

Just make the transistor exp-converter.
Like iextern says its easy, you can buy the tempco at mouser and use the schematic from the tb303.
Make some change to only need 1 1K tempco.
Else you will be very unaccurate, go listen for fun to alpha juno with modulation and high notes.