Possible design flaw in regards to Square vs Saw wave phases

[jpmartineau]

There seems (I hope I'm wrong) to be a design issue with the square wave generator in relation to the sawtooth wave generator. Sorry if this is a bit technical, I'd really appreciate input from the engineers on this one.

I've noticed that when the Saw is all the way up and I bring up the square wave (at 50% duty cycle), eventually, when both are up, I don't hear the square wave anymore.

I haven't checked with a scope, but what I'm suspecting is the following:

When adding both waves, if at phase 0º the saw is starting from -1 and ramping up to +1 at 360º and the square starts at 1 until 180º where it will go to -1 until 360º, then, instead of the expect addition of harmonics, the result is a sawtooth wave that is phase-shifted by 180º. That phase shifted sawtooth wave sounds exactly like the sawtooth alone.

What should be done is the square wave's phase should be inverted so that from 0º to 180º, the value is -1 and 1 from 180º to 360º.

If you don't understand the two last paragraphs, take some graphing paper. Draw a sawtooth and a square. Add them. Then, do the same but with the square wave inverted. You will immediately see what I mean.

Can someone confirm if this is the case, and if so, is there a hardware fix that can be done to invert the square wave's phase?

[EDIT. I added an image for clarity]



Thanks,

JPMartineau

[jpmartineau]

Can anyone, Arturia employee or not, test this and share results with me?

1) Open up filter with no resonance, remove all modulations, modifiers, brute factors and turn everything off to just hear a clean oscillator signal.
2) Turn down volume for all oscillators except sawtooth wave.
3) Turn up volume for square wave.

Do you hear the square wave? I don't. If you're not sure what a square wave sounds like, go to an arcade and play Arkanoid, or, listen to the square wave alone.

[beefinator]

same thing on mine, that's how it is.

It's probably related to the same reason why the sawtooth silkscreen label is inverted.  Some necessary change late in development.

Ah well.  I'm not really annoyed by it.  In fact, I find it kind of interesting because if you mix just the right amount of 50% duty cycle square into the saw, you can bump the saw up an octave.  That allows for a new range of timbres, e.g. you can effectively have a -3octave suboscillator, or have a unison on 3 consecutive octaves.  Or, have pseudo-detuned AND dual-octave sawtooth if you add the ultrasaw into it.

Also, the behavior you mention about the saw becoming just a 180° phase shifted version, that's only when the square is at 50% and ultrasaw is off.  Changing either of those (including the LFO to do PWM) can create other types of interesting kinda-saw-like-but-not-really timbres.



True, it seems like it might be a mistake or unintended design, but I'm not really annoyed by it, it's just allows different possibilities than you might expect.

[Bruno@arturia]

Hi,
the both of you are right!

indeed the saw and square waveform are in such a phase difference that the square can be canceled in some positions.

It's probably related to the same reason why the sawtooth silkscreen label is inverted.  Some necessary change late in development.

Ah well.  I'm not really annoyed by it.  In fact, I find it kind of interesting because if you mix just the right amount of 50% duty cycle square into the saw, you can bump the saw up an octave.  That allows for a new range of timbres, e.g. you can effectively have a -3octave suboscillator, or have a unison on 3 consecutive octaves.  Or, have pseudo-detuned AND dual-octave sawtooth if you add the ultrasaw into it.

Also, the behavior you mention about the saw becoming just a 180° phase shifted version, that's only when the square is at 50% and ultrasaw is off.  Changing either of those (including the LFO to do PWM) can create other types of interesting kinda-saw-like-but-not-really timbres.

Do you have a microphone in my office? 
that's exactly what happened, we had to change the design, forgot about the square phase, but when we noticed we decided to keep as is, because, despite it surprising to have the square disappear, it indeed nice to get the octave up saw, and the out of phase saw that gives a different result when mixed with a triangle.

anyway that the way the machine is. and that probably why some people are saying the oscillator could sound weak.

Can someone confirm if this is the case, and if so, is there a hardware fix that can be done to invert the square wave's phase?

A hardware fix would require to swap two SMD 0805 resistors, remove two others, and replace them with  through hole, soldered on smd pads.
So doable if you are used, and tooled for SMD.

Regards
Bruno


edit: corrected typo

[stevism]

it's threads like these that make these forums awesome to read!

[beefinator]

Do you have a microphone in my office? 
that's exactly what happened, we had to change the design

naw, I'm not some deductive genius, I just remembered happening to see it mentioned before on the forums I think...   

A hardware fix would require to swap two SMD 0805 resistors, remove two others, and replace them with  through hole, soldered on smd pads.
So doable is you are used, tooled for SMD.

Interesting.
So if it is something that simple, then what exactly was the reason for the change?
Seems like an interesting process, designing an analog synth.

[RichMakeGame]

this is very interesting I've just been playing around with patches making use of this saw octave switch trick and I like the results. I think I noticed previously *something* was going on but I would not have had a clue what exactly that was if it weren't for this thread. nice one

[JnC]

In other words:

It's not a 'bug'. It's an 'undocumented feature'.



It lends credence to what the designers have said all along about mixing waves and not maxing out everything.

[jpmartineau]

A hardware fix would require to swap two SMD 0805 resistors, remove two others, and replace them with  through hole, soldered on smd pads.
So doable if you are used, and tooled for SMD.

Is there any chance I could get the specifics about the modification? I have access to a lab with everything I need to do this mod...

Thanks!

- JP

[Bruno@arturia]

Hi here are the required informations.
you'll find original schematics, and modified version.
plus a picture showing where U31 is located on the PCB.

Please note that

1/ this operation will void your warranty. And Arturia can no be made responsible of any damaged caused by this mod or it's realization.
2/ theses informations are provided as is, with no guaranty it will be working.
3/ this mod can destroy your machine if not properly done
4/ I and Arturia will not provide support for this mod: this information are released to the DIY community, any electronic technician should be able to understand it, do it, and explain it to others.

The enclosed documents show the modification do to done on the schematics side, it's up to you to realize it the way you want on the PCB.
I enclosed a picture of the original PCB to show where is U31 located, as there is no silk printing showing U31. Resistors designators are silkprinted

Enjoy

[jpmartineau]

Thanks Bruno!

Very kind of you to share this!

— JP

[beefinator]

wow!

I'm not planning on doing this, but I think it's really cool that you're putting it out there.
 

[JnC]

Hello Bruno and Yves,

I'm trying to understand the circuit shown in the SquareWavePhase Original.png image. It's been many years since I studied analog circuits. I actually cracked open a book tonight that had been shut for about 25 years.

I think the circuit type is a subtractor, meant to subtract the voltages at pins 5 and 6.

I was initially confused by why one of the inputs was labeled "saw". But then I found this article which explains how to make a square wave from a sawtooth wave and a copy of the sawtooth that has been phase shifted by 180 degrees:

http://crca.ucsd.edu/~msp/techniques/v0.11/book-html/node186.html

But for that method to work, the sawtooth wave would need to be out of phase, further upstream of the voltage divider that is on pin 6. (Voltage dividers don't shift phase, so I'm thinking that had to happen before this circuit.)

Is that how the circuit works: by making a square wave from a sawtooth and its phase shifted copy?

[Bruno@arturia]

It's a comparator, not a subtractor(open loop, no feedback).
It basically compares the saw voltage with a threshold, that is composed of the sum of PW knob, env and lfo voltages.
If saw in higher than threshold, output square is up, and down otherwise.

[Isak]

Hi Bruno.

thank you for posting the way to "fix" the phase of the saw and square, this mod is working great!
now i'm really happy with my minibrute
tomorrow i'll try to upload a vid that shoes how it work.
one thing...the saw became a bit lower in Vol then the square, i really dont care about it, but maybe some people will.

cheers,
Isak E.