some questions about comb filtering

[midwayfair]

I'm wondering if it's possibly to take a couple out of phase signal splits, create oscillating band pass sweeps (at different bands), and beat them against the dry signal to create destructive phase interference at varying frequencies to get comb filtering. I assume I also need a copy sweeping in phase to get constructive filtering?

In short, I'm wondering if I can make something that sounds like flanging without delay (and without the distortion of the crash sync).

[Pojo]

Sorry I can't be of any help here....but the concept seems like it would be a phaser on steroids! Would be neat to hear...

[GibsonGM]

Possible?  Yes!  Pleasing to the ear?  Well, with work in planning, prototyping, adusting/tweaking - could be!    You can do anything you can think of to signals...the results may sound very unexpected and different than what you had in mind, but I think the idea is certainly worth a shot!

Sort of like an out of phase switch on your pickups, but with moving center frequencies....the end product will be very dependent on your bandwidth (Q), I think....squeaky and weird vs. really cool and flanger-like...

[Mark Hammer]

I think the discussion needs to begin by considering the size and spacing of the "teeth" in the comb.  In other words, the systematic relationship between the notches and peaks, how they are produced, and their numerosity.

In the case of flanging, the use of time, rather than phase, to produce the notches, results in a systematic change in not only the number of notches/peaks produced, but their spacing.  At shortest delay, there are fewer audible notches, and they are spaced farther apart.  As the delay time increases, there are nore audible notches, closer together.  I like to describe this as the signal being progressively "infected" by notches over the course of the sweep. (Note that here we are talking about slow and wide sweeps, rather than the sort of thing easily mistaken for chorus).

In the case of phase-based notches and peaks, the number of notches/peaks produced is constant over the sweep cycle, and the spacing doesn't change much.

So, like I say, one has to first ponder the nature of the desired comb, and how it might behave while being swept.

But while I've got you here, I was wondering the other day whether the concept of theta processing could be applied to phasers in some fashion.  So-called theta processing was an idea first applied by Eventide, in their Instant Flanger.  It uses multiple fixed lag phase-shift stages, added to the delay path of a BBD.  Note that we are mostly familiar with phase-shift/allpass stages where a resistor and cap go from the output of a preceding op-amp to the inverting and non-inverting pins, respectively, of another op-amp.  We vary the resistance to ground from the non-inverting pin, and vary the point where the maximum phase shift of 90 degrees gets applied, using a FET or LDR or some other means.  For "lead" phase-shift sections, that maximum of 90 degrees is applied to frequencies above a point dictated by the usual F = 1 / [2*pi*R*C] formula.  In lag stages, the resistance to ground, and cap, switch places.  It still produces up to 90 degrees phase shift per section, according to the same forula, but provides for more phase shift below the corner frequency, rather than above it.

So, what's the point?  The time delay produced by the BBD obviously provides shift in phase for the bottom end when the delay is made long enough, but, as noted above, the notches are spaced very close together, yielding a very boxey sound.  By adding additional phase shift only in the low end, when the flanger sweeps down, the notches produced are spaced a little farther apart, and in ways that are not harmonically related to the notches above that.  The result is purportedly a more "musical-sounding" flanger.  The late Jurgen Haible resurrected the concept of theta processing in his Stormtide and Son of Stormtide flangers.  If you look at their respective schematics yu can see how he implemented it, having settled on 4 consecutive stages of fixed lag allpass stages.

Okay, so why am I bringing it up?  We've had a number of projects here in recent years - many from the ever-fertile mind of Rick Holt - that included fixed lead allpass sections,  This increases total phase shift as the sweep moves upward.  I wonder if a phase shifter could be made more interesting, and possibly musical, by adding in one or more stages of fixed lag phase-shift, such that both ends of the sweep - the turnarounds, as it were - could be more interesting by having just a little more cumulative phase shift at the top and bottom of the sweep.

[merlinb]

I'm wondering if it's possibly to take a couple out of phase signal splits, create oscillating band pass sweeps (at different bands), and beat them against the dry signal to create destructive phase interference at varying frequencies to get comb filtering. I assume I also need a copy sweeping in phase to get constructive filtering?

In short, I'm wondering if I can make something that sounds like flanging without delay (and without the distortion of the crash sync).

You could cut out the middle man and go straight for a bunch of notch filters, each tuned to a different frequency and swept by an LFO...

[R.G.]

You could cut out the middle man and go straight for a bunch of notch filters, each tuned to a different frequency and swept by an LFO...

Even better if they're swept by un- or loosely-correlated notch filters.

However, these are not comb filters.

As Mark said, phase shift creates a number of notches in specific places with respect to each other. Time delay generates many more notches, each at the place where the time delay creates (2*N-1) times 180 degrees of shift at that frequency. If you move the notches around with respect to one another, it's not a comb. Well, OK, it can be a comb with bent and broken teeth in a time varying fashion. 

Not that it wouldn't sound interesting or good. I personally like the sound of independent randomly swept notches.

[Mark Hammer]

The notches don't have to be random to behave in a perceptually aperiodic manner.

Let's say we have a notch that sweeps back and forth from around 350hz to around 800hz, and a second notch that sweeps back and forth from around 200hz to around 2khz.  That is, sometimes notch 2 sweeps lower than notch 1, and sometimes higher than notch 1.  They might both be driven by the same LFO, but at a certain point one notch will seem to "accelerate" more than the other. Remember that perceived speed is a function of how much territory is covered within the same time period.  That's why I say "perceptually aperiodic".

Of course, all of that is in the realm of the abstract.  It may well turn out that it doesn't sound to our ears the way I think it will.

Finally, consider that a bandpass subtracted from the original signal behaves very much like a notch created in the input signal.

[samhay]

I had a play with a two-and three- notch filter system a while ago where the notches swept in opposite directions, and due to poor component matching, from fairly different start/stop points. It sounded interesting, but not, to my ears, worth the trouble. However, if you want a flanger-but-not-quite-a-flanger-like sound, it might be worth a try. It can be easily implemented into a Phase 90-style circuit by splitting and inverting the LFO output.

[midwayfair]

a systematic change in not only the number of notches/peaks produced, but their spacing.  At shortest delay, there are fewer audible notches, and they are spaced farther apart.  As the delay time increases, there are nore audible notches, closer together.

Bugger. I completely forgot about that. And it's right there in RG's article I read before brainstorming this.

Hrm. I might still try to make something with this idea. What I really want is a time machine for the notes to go back in time a few milliseconds to create actual through-zero flanging live.*





*Yes, this IS the worst use of a time machine ever. But at least it won't create time paradoxes and cause time and space to rip apart. It might melt faces, though!

[Mark Hammer]

I had a play with a two-and three- notch filter system a while ago where the notches swept in opposite directions, and due to poor component matching, from fairly different start/stop points. It sounded interesting, but not, to my ears, worth the trouble. However, if you want a flanger-but-not-quite-a-flanger-like sound, it might be worth a try. It can be easily implemented into a Phase 90-style circuit by splitting and inverting the LFO output.

One of the potential advantages of "movable notch" system is that individual notches or peaks can be swept independent of each other.  With phasing, the end-product is dependent on the sum total of all cumulative phase shift.  The functional outcome may be notches, but the notches rely on all those phase shift stages moving together.  Similarly, with flanging, you either get all those notches in their mathematical relationship...or you don't.  There is none of this notches moving one way for the upper mids and highs, and notches moving another way for the low end.

If one has individual sweepable notches (or bandpass filters subtracted from the original at a mixing node), they are free to be moved around independently and individually, without losing the effect.  I might also point out that there is nothing which dictates that all notches/peaks HAVE to have the same Q or selectivity as each other.  The milky wash of a UNi-Vibe is a product of shallow-broad "dips" being swept, rather than the more precise notches of a phase shifter.  So what if one had wider passbands with modest gain, and narrower high-gain passbands being swept independently and subtracted from the input signal?