Design of VCO for synth with continuously variable waveform

[pitagoras]

Hi. I need to design a VCO for an analog synth.
It will work with the well (moog) stablished 1V/Octave control.
Normally these oscillators output triangle, sawtooth, square (with different width ratios).
The appreciated ones have among other things, a knob like in the image attached, allowing a continuous variation of waveform.
Any idea is appreciated.

[JPortici]

doing tri->saw->sqr should be pretty simple by altering the bias in a circuit like the minimoog triangle shaper. however how they did achieve the variable width i don't know.

[pitagoras]

how they did achieve the variable width i don't know.

We don't even know if they're cheating, in the sense of switching some things based on pot. This is from Sub 37 which is essentially a sub patthy with 3 octaves. So there might be some digital switch doing stuff there. But as an exercise I think it's interesting to think it purely analog.

[JPortici]

eh, go buy one, open it up and take a look. i actually have some kind of idea on how to do it in analog but it's complicated enough to be difficult to draw it as a block diagram

[BrianHG]

Would you be interested in a cheating MCU solution?

What frequency range & waveform would you like?  Many, many, many years back I used to use a precision VCO from burr-brown which went through more than 2 octaves, which gave you I believe a triangle and square wave, but, there would still be some temperature drift.  It was an expensive device at the time, over 10$ each. I believe it was this one: http://www.ti.com/lit/ds/sbvs018/sbvs018.pdf
I used it as a precision AD converter, 0hz to 20Khz with an input voltage of 0v-10v.  The high frequency span is configurable.
You can also look at this one which is still available today, but, not the same kind of precision as the burr-brown device ->  http://www.ti.com/product/LM331

The MCU solution today would be the cheapest and most repeatably stable and reliable, around 2 dollars, all on 1 chip, only need 5v supply, with your choice of any waveform output, but, you will need to do some programming.

[magetoo]

Is it really controlled by only one pot, or does the knob do something out of the ordinary?  My first thought is that maybe we're looking at three pots controlled by one knob, and each act in their own third of the range.  Or it's just a rotary encoder that controls some DACs.

If there's three separate signals, there could be just three separate circuits where each does only one thing; saw/triangle shaping, saw/square shaping, and pulse width.

[pitagoras]

Is it really controlled by only one pot, or does the knob do something out of the ordinary?  My first thought is that maybe we're looking at three pots controlled by one knob, and each act in their own third of the range.  Or it's just a rotary encoder that controls some DACs.

We don't know. I think it is "mostly" analog, probably some analog switch is playing a role (Moog does use 4066s, for example the "open" Werkstatt).
I think it is an interesting exercise to think it analog.

i actually have some kind of idea on how to do it in analog but it's complicated enough to be difficult to draw it as a block diagram

could you describe the idea in text?
I was thinking that starting with squarewave followed by integrator, changing the DC offset of the squarewave yo get sawtooth. And increasing even more you begin to "squarize". But then pushing even more we need to change the pulse width... I don't see how it can be done in a continuous way...

The MCU solution today would be the cheapest and most repeatably stable and reliable, around 2 dollars, all on 1 chip, only need 5v supply, with your choice of any waveform output, but, you will need to do some programming.

Thank you for your links Brian, yeah if it were me I would software synth everything, as I'm a full time software dev.
In fact another project I'm planning is a fully digital FPGA based theremin (based on some really serious works someone is doing at thereminworld).
This synth, however, is kind of a request from a keyboard player I know. The thing is that musicians appreciate "analog". I was thinking that some of the imperfections of analog "might" be desirable for them, including slightly imperfect tuning, imperfect waveforms, etc.



 

[magetoo]

We don't know. I think it is "mostly" analog, probably some analog switch is playing a role (Moog does use 4066s, for example the "open" Werkstatt).
I think it is an interesting exercise to think it analog.

It's certainly entirely or almost entirely analog, I was just saying that if you can divide the problem into three parts it would be easier to solve.  (This can be done with analog electronics too.)

[sync]

The RSF Kobol has variable waveforms. The schematics are available here:
http://www.univertron.com/SAS/manuals/RSF/Kobol%20Expander%20Schematics/ (7.jpg)
http://elektrotanya.com/rsf_kobol_sm.pdf/download.html (You have to wait about 2 min until the "Get Manual" download link appears)

They used some op amps and OTAs to get continuously variable waveforms controlled by voltage/pot.

[Fungus]

It's certainly entirely or almost entirely analog

It was definitely 100% analog in the 1970s when Moogs were invented.

Wikipedia has quite a bit of info...

https://en.wikipedia.org/wiki/Voltage-controlled_oscillator#Control_of_frequency_in_VCOs

This page is interesting as well.

http://www.birthofasynth.com/Thomas_Henry/Pages/VCO-1.html

[JPortici]

could you describe the idea in text?

Certainly. You can have different ways to morph from sawtooth to triangle, the difference in how the waveform gets "folded".

One idea is you have the three waveforms: sawtooth (core), triangle and variable pulse width. Two crossfaders (or a three input crossfade, three or four OTAs) where the three inputs are the three waveforms.
Vcontrol controls the crossfade position, where
Vc1 = 0 -> output = triangle
Vc1 = V1 -> output = sawtooth
Vc1 = V2 -> output = VPW
Vc1 > V2 doesn't make any difference / is clamped at V2.
the VPW shaper has its control voltage made so
for Vc1 < V2 the DC is 50%, for Vc1 > V2 up to V3 the DC increases

sure at some point there is a crossfade or selection between sawtooth and square. All the saw-to-tri shapers i know of can go from saw to tri to square (50% dc, fixed) by varying a bias voltage/current (the simplest being the minimoog triangle shaper) knowing if it's all crossfading or actual waveshaping should be obvious by looking at the waveform while sweeping the shape knob

[pitagoras]

It was definitely 100% analog in the 1970s when Moogs were invented.

Thank you for your links.
In my case I would like to design a VCO whose waveform is continuously variable from triangle to pulse. See knob in first post.

[BrianHG]

The MCU solution today would be the cheapest and most repeatably stable and reliable, around 2 dollars, all on 1 chip, only need 5v supply, with your choice of any waveform output, but, you will need to do some programming.

Thank you for your links Brian, yeah if it were me I would software synth everything, as I'm a full time software dev.
In fact another project I'm planning is a fully digital FPGA based theremin (based on some really serious works someone is doing at thereminworld).
This synth, however, is kind of a request from a keyboard player I know. The thing is that musicians appreciate "analog". I was thinking that some of the imperfections of analog "might" be desirable for them, including slightly imperfect tuning, imperfect waveforms, etc.

I understand.   But, if you sample the POT input on a cheap DSPic, of even 3 inputs for 3 pots if needed for that matter, drive a mono 20 or more bit dac, say a around 500Khz or 1Mhz sample rate on the pots and dac, with simple R/C filter on the inputs and outputs, coded for continuous real-time sampling of the ADC at each overkill clock on the DAC will mimic a true analog circuit when synthesizing analog audio waveforms from 10hz to 20Khz.  Now when I say, 500Khz/1Mhz with a low pass filter on the DAC filtered, I mean filtered down to 100Khz flat audio bandwidth, not 20Khz, so you get those beautiful sounding ear piercing saw ramp with that razor sharp fall transition of a pure analog synthesizers which sound so incredible.  Now, since no one in their right mind would think of synthesizing an audio signal, or use 32 bit floats in a MCU running a sine/triangle/saw with a 500Khz/1Mhz DAC sample rate, with at least 16bit ADC updating every DAC cycle, you would be in a domain with a digital synthesized audio oscillator using a 32bit DSPic at 32 mips compared to others with a 8/16bit MCU running at 8 mips, 10bit ADC on the pot & an internal 10bit DAC, or PWM running at the base 44Khz if even so, output filtered to a low 20Khz or less, your design will match an all analog oscillator design.

SO BE WARNED: If you don't buy an original synth and model your generator's output exactly the same as the reference unit, the musicians you are selling to will hate your device and never buy it, they will hear the difference.  You need to mimic response delay from every source signal as well as all the distortions and frequency sweep behaviors of the original.  This is a true undertaking to be done right, just making an AD fed DAC oscillator wont cut it unless you are trying to market your own type of sound synth module with it's own characteristics.

[BrianHG]

I just realized that you have no sine wave in the volume control knob in your image.  It doesn't look like a frequency control, just like a triangle waveform with varying duty cycle, then it gets amplified to a square-wave, once again with varying duty cycle.  It looks like what one of my 30 year old all analog function generators did.
Here is the https://www.maximintegrated.com/en/products/digital/clock-generation-distribution/silicon-crystal-oscillators/MAX038.html
You got your 3 waveforms, adjustable duty cycle and frequency range.
Perfect for audio range and you get the waveforms your looking for.  As for mixing them on a pot, you will have to figure that one out on your own...

My above warning stands, whatever circuit you use, if it doesn't match the behavior of the original device your trying to replicate, musicians wont buy it, even if your circuit is analog...

[JPortici]

My above warning stands, whatever circuit you use, if it doesn't match the behavior of the original device your trying to replicate, musicians wont buy it, even if your circuit is analog...

[rant]
i think you don't know synthetists at all. they are always buying the replica and throwing it away because it's not the same even if it is (but with a different faceplate), but still, they buy it. they have this crazy idea that with the same instrument you can become good at it.
The eurorack world is a madhouse, full of noobs that start learning electronic because of synth diy (great AWESOME! ) and after one week they read a tutorial on eagle and propose you their new module, which is a copy of a copy of a copy of a copy of a copy of a 70's design with this transistor because it's cheap, this expensive opamp because they say it sounds good, all in an atrocious layout (very bad )
and storms of noobs buying those diy projects.[/rant over]

Said that with respect to the big guys which are always coming out with new, crazy, musical, fantastic designs, which are always tempting but i got out of that world and went back playing regulars boring old keyboards

[BrianHG]

Ok, maybe I was misunderstood.  Yes, analog synths create sounds which cannot be replicated any other way and they do sound amazing.  My comment was on the basis that pitagoras wanted to replicate precisely the existing control knob in his photo which already exists.  I did say in the post above "unless you are trying to market your own type of sound synth module with it's own characteristics."  What I meant is if pitagoras has basic programming MPU C lang skills, and he want to create his own line of synth modules with his own unique characteristics, I would recommend choosing the MPU route along the lines I outlined to generate that analog feel, and now, he can now offer multiple sounding products with one circuit board, with just different software code.

[JPortici]

i should have written RANT with a bigger font no, i understood what you meant. just wanted to point out that 95% or more of the knob fiddlers want to buy replicas because they think they'll write better songs (the very small percentage of actual musicians don't care and want to experiment with new stuff. they also get many of the new things for free because publicity so...)

I already went down the route of making a digital oscillator that also had waveform modulation that achieve the exact thing he wants to do.. and it wasn't exactly an easy task (seemed like at the beginning but if you don't want aliasing to create too many problems and at the same time acheive modulation rates at audio frequencies, crazy high sampling rates like 1 MHz are probably what you have to resort to) altough it proved to be an excersise that pushed my skills and lack of. Expecially putting four of those in a single dspic, had to cut corners somewhere but who cares, result was good for me.
doing this in analog? a nice excercise too, to test your creativity

[BrianHG]

doing this in analog? a nice excercise too, to test your creativity

Doing this with vacuum tubes, now there would be a testimony to devotion...

[Lee Leduc]

Here's some music synth items based on a PIC micro.
http://electricdruid.net/

VCO
http://electricdruid.net/wp-content/uploads/2015/07/VCDO-Datasheet.pdf