Saw shifter evolution, Linear Detuning and TZVCO

[Jaba]

Some time ago, as many of you, I grew an interest in the "saw shifters" family of circuits, such as the ETI Waveform Multiplier, Scott Bernardi's Obesifier, the Yusynth Saw Animator, Ken Stone's CGS38 Saw Pitch Shifter and many others.
I built and modified a couple of them, two more are still on my "to do" list.
They all share the same core circuit. The simplicity and versatility of that core saw shifter circuit is magic to me.
It actually is sort of a specialized saw phase shifter, but it can work well as a saw pitch shifter too.

I soon realized that if the modulation wave is a sawtooth too (or an inverted sawtooth) then the circuit IS a frequency shifter, and if the modulation saw is slow, the result can be made identical to what Jurgen Haible used to call "linear detuning", and actually is the same effect; the input (audio) and modulation waves and the internally generated PWM rectangle (comparator's output) must be carefully trimmed to show exactly the same peak-to-peak amplitude, and the reset time of the LFO must be short, at least comparable with the reset time of the audio sawtooth, but it is not that difficult, it works.

This is probably not new to some of you, I know I did not invent it, but I made some experiments with a CGS38, and then with a modified Yusynth Saw Animator and I can confirm that it works and sounds fine.

If you want to try it, my suggestion is NOT to use a sawtooth core LFO, since they usually are too slow in the reset phase; I used the classic triangle LFO circuit (actually the two that were already included on Yusynth's PCB) and a tri-to-saw converter copied from Thomas Henry, it worked well.
My modified version of Yusynth's Saw Animator now has UpSaw/DownSaw/Tri/Square/Sine LFOs and built-in adders so that it can even output the "linearly detuned up" saw, the "linearly detuned down" saw and the original saw already mixed without any additional module; just it and a VCO, and it sounds great !

After this, my next thought was: if it acts as a frequency shifter and Fout = F1 - F2, then Fout can go negative; -> instant Thru-Zero FM for free !?!
(Well it is not actually "for free", some work still has to be done in designing the control part)
I tried FMing a VCO while sawshifting it at audio rate (another unmodulated audio saw oscillator), and it seems to work ! the typical time-reversal images on my scope were right as I hoped ! the output frequency actually goes negative smoothly and easily.

Has anyone else tried this ? I may be proved wrong, but I see it as a very promising and probably cheap way to achieve TZFM, audio PM and maybe more.

please comment. I need some friends to play with, or I soon get lost

[Tim Servo]

Hey Jaba,

I'm also working with some Saw phase shift circuits, but it had never occurred to me that they could also produce a frequency shift too. Do you have any audio examples you could share with us? One thing that I'm seeing (at least in the circuit simulations I'm running) is that there is a substantial 'reset glitch' where the original input saw and the phase shifted output overlap. Is this something you hear in your circuits?

Cheers,

Tim (Mr. Glitch) Servo

[Jaba]

Hi Tim,

no, sorry, no audio demos available yet; I'll try to make some, as soon as I find an easy way to record my modular directly.

And yes, the "saw shifters" can be used as saw pitch shifters too; Ken Stone calls his CGS38 version a "Saw Pitch Shifter" and clearly explains how to do it on his CGS38 page. I (and many others) can confirm that it works fine.
If you replace the LFO with an audio range oscillator or a VCO, you simply extend the range of the pitch shift. No side effects, AFAIK.
The same can be done with any other similar comparator-based SawShifter. I tried with a modified YUSynth Saw Animator, with good results. I removed (jumpered) the resistors at the modulation input (R11 and R26) from YUSynth's version of the circuit, because this makes it easy to achieve a higher modulation index.

I can think of two causes for your glitch:

1) Amplitude mismatch. The glitch can then be cancelled by matching the amplitude of the two saw waves.
First you have to properly calibrate the circuit as a phase shifter; this means to carefully match the amplitude of the rectangular wave (from the comparator) to the amplitude of the input saw wave, cancelling the "step" in the new shifted saw.(*)
Then, if you want to use it as a pitch shifter too, you need to match the "modulating" saw amplitude too. This removes the glitch.
Of course, if you want to use another source as one of the saw waves, you'll have to recalibrate everything from start, unless it has exactly the same amplitude.

On a dual trace oscilloscope, if you look at the output (=phase shifted) saw while triggering from the input saw, if the modulation saw is slow, say 1 Hz, you can see the waveform slowly shifting and then suddenly jumping back, or stopping for a while. You are observing the phase shift. We want it to shift the whole 360 degrees, and we want the "jump back" to exactly match a saw period, so that when the wave "jumps back" it has been shifted one period and jumps one period, and you can't see or hear the difference, then it looks and sounds as one continuously shifting saw. By matching the amplitude of the modulating saw we do that. You can even use this as a method for calibration: you look at the scope and move your trimmer until you have a smoothly shifting saw.

Since frequency is nothing more than the speed of variation of the phase (from theory), your shifting saw has a slightly different frequency (difference=1Hz in the example) and if you increase the frequency of the modulating saw there is no limit, you can have any pitch shift, even below zero !

2) If your glitch comes instead from a too slow reset in the modulation saw, then I don't think it can be cured by calibration alone. I would use a different LFO, or an audio VCO pushed into the low frequency range.
A classic saw/ramp LFO is often made by a simple triangle core LFO with a diode shunting the Rate pot. As I mentioned before, this approach does not work here, since the discharge of the big timing capacitor through the diode is far too slow to be masked and lost among the "vertical" transients of our audio sawtooth waves. The same triangle core LFO followed by a static triangle-to-sawtooth converter worked fine instead.

If you tell me which circuit you are using and what are your input signals, I can probably be more specific about how to cancel your glitch.

(*) If I am not mistaken, it turns out that even without this first calibration the output signal is not that bad: it is definitely not a shifted saw, but the "error" (i.e. the difference between the desired phaseshifted saw and the actual output) is a PWM rectangular wave with the same frequency of the input, i.e. the same note, so that if you are just trying to make a saw "fatter" it does not hurt, does not sound "wrong", just a bit different, but still in tune, and fat. Scott Bernardi's Obesifier circuit actually has no calibration trimmer, I suspect that this may be the reason; I would like to have comments on this from someone who built the Obesifier circuit, since I still have not done it.

cheers
Paolo(Jaba)