Allpass filter for distortion?

[PBE6]

Does anyone know of a distortion pedal that uses an allpass filter to affect the phase of the signal? I know that phase is not supposed to be particularly audible unless mixed with the original signal to create frequency notches as in phaser effects, but I was messing around with a basic circuit on Circuitlab (https://en.m.wikipedia.org/wiki/All-pass_filter#Active_analog_implementation) and it does some funky looking things to a square wave. I figured that if it sounded cool it would show up in at least a few distortion designs, but I can't seem to find any.

[antonis]

I've tried sometime ago to design a Vibrato pedal with some all-pass filters and JFETs with LFO (to make a variable R) but it resulted to a VERY UGLY distortion so I gave it up..

[Mark Hammer]

When I whipped together the "Woody" acoustic simulator some years back ( http://www.aronnelson.com/DIYFiles/up/M._Hammer_s_Woody.pdf ), I threw on a single allpass stage between the clipping stage, and the mixing stage.  The reasoning was rather specious.  Mostly it was because I had recalled reading many years before then that the Aphex Aural Exciter had used phase-shift to somehow align the treble-brightening with the rest of the signal.  The original information may have been misleading, or my recollection of it mistaken, or both.  But, I had an op-amp to spare and stuck it in, and it sounded alright, so I left it.

That's about as close to what you're asking as I've ever seen.

But that doesn't make it necessarily a pointless question.  The alignment of harmonic content and the fundamental can be a very big part of what gives a signal its character, and/or presence.  So I think it is worth experimenting with using allpass stages to tinker with that.

[R.G.]

As noted, the phase of harmonics in a signal is not particularly detectable. The human ear mechanism seems to be based on detecting the magnitude of components, not so much their phase.

However - the shape of a waveform does have a lot to do with the relative phases of harmonics and the fundamental. The same mix of harmonics and fundamental, with just the phase of harmonics altered sounds the same, but has a very visibly different shape on an oscilloscope. So far, no surprise.

But clippers that do distortion are waveform-sensitive, not frequency-domain sensitive.  They start distorting at a voltage level. That is - they are purely time-domain devices, not frequency domain. The classical formula for making a different-sounding distortion is to filter before clipping so that the biggest components left before the clippers are clipped most, then filter after clipping so that you affect the frequency domain results of the hash created in clipping.

Allpasses before clipping will mess with the time-domain shape of the waveform before clipping, and so will change the clipping products. Allpasses after clipping might not be audible at all.

Maybe that's a good way to get rid of some of the humpiness of the tube screamer circuit. 

[PBE6]

Great! Thanks for the input all. Interesting suggestion RG, I'll try putting an allpass before a simple distortion and see if it changes. I'll see if a "reverse" allpass sounds any different too.

[Mark Hammer]

Point taken.  What if we were to marry the multi-band clipping approach with phase alignment?  That is, split the signal up into several separate bands, and clip each band, but introduce different amounts of phase shift for the different bands such that the harmonic components they add sum to a different sort of waveform?

The accompanying question is, much like the difference between this diode and that, would the difference only show up on the scope, or would it show up at our ear drum as well?

[ashcat_lt]

RGs words are very much along the lines I was thinking. 

A not-so-secret trick in radio for getting that bigger-than-life broadcaster voice is phase rotation.  Human voices, especially male voices, can be very assymetrical waveforms.  Phase rotation can redistribute that energy to make it much more symetrical, which tends to buy some headroom because you're not hitting one rail before the other.  It is an extremely subtle difference, but can make it so that you need less compression overall, but also helps the compressor to work more consistently when it does.  I use all pass on vocals and other assymetrical waveforms in mixing pretty often.  An all-pass before the clipping section will change what gets clipped and when.

OTOH - a parallel all-pass creates that whole comb filter thing which is an interesting way to implement a "character" knob.  It will, of course, do different things depending whether it's before or after the clipping.

[puretube]

Audio signal peak energy equalization by L.R. Kahn

(Hammond had a similar thingy for "compression"...)

[puretube]

Well, maybe not exactly for "compression",
but for indeedy: "violent phase-distortion"

kinda compressive, though...

[puretube]

But in the end, anything that "slows down" the slope of a waveform
kinda "mellows" it,
while something that steepens the wave,
makes it sound "harsher" or at least "rougher"
(at least in "our" terms of "distortion"...)

[PRR]

> anything that "slows down" the slope of a waveform kinda "mellows" it

If that were very true, radio stations would not use phase-twisters. They need their sound to Stand Out as you scroll across the band. Today they universally do phase-twist. So not that simple?

[ashcat_lt]

You can change the "steepness" of a wave just by turning it up or down.    It's the sudden changes of slope that cause (are?) harmonics.  It would be easy enough to test this in software, if I was near a decent computer with which to do it.  I tend to think that an all-pass after a clipping stage won't have much audible change except maybe if the signal gets distorted again after, and an all-pass before a clipping stage will have a subtle-at-best effect.

It may, however, be a reasonable way to get a little more headroom out of a circuit that's supposed to run clean.

[ElectricDruid]

I like the idea, and I don't think it's at all pointless to give it a try. As you've found, an allpass will bend a waveform all out of shape by shifting the relationship of the harmonics. That on its own won't change the *sound* audibly (only the wave shape), since it doesn't alter the harmonic strengths, but once that signal has been run through a distortion/overdrive/clipping stage, it certainly should. Like RG said, the clipping is purely a time-domain signal affair, so changing the shape of the waveform going in should change the contents coming out.
Another possibility would be to use an all pass in parallel with the straight signal to get a notch filter to make the "character filter" mentioned above (by changing the frequency response in this case) and then do the distortion after that.

Give it a try and do let us know what you find out.

Cheers,
Tom