Bigger, crazier phaser with fewer parts

[Taylor]

Here's an idea that's been knocking around my head.



The idea is that a phaser is kind of a roundabout way of getting a notch filter. If we bypass the phasing concept and go straight to filters, we can get a notch or peak filter with a single active component, instead of 2 active components and 2 variable resistances per notch.

So what we have here is a bunch of bandpass filters, tuned to different frequencies, being swept around by an LFO (not shown). The resonance is hard-set at about maximum. Then we send the outputs of all of the filters to a "switch hitter" amp, which allows to vary between inverting and non-inverting, and then this output is mixed with the clean.

I think (can't really test it yet) that by mixing the inverted signal having sweeping peaks with the non-inverted clean signal, we get notches instead of peaks. Having the inversion control continuously variable allows to control resonance from max peak to max notch with a single knob, instead a knob per stage if we used the normal way of controlling resonance of a twin T.

The result is that we have something very similar to a 12-stage phaser, except with only 6 variable resistances (these are the difficult/expensive part of any phaser, whether JFET matching or sourcing optos) and just a single chip for all the "phase" stages. And it has continuously variable resonance from negative to positive.

Another trick I'll probably do is to use another hex inverter for the LFO - this can give us 2 separate triangle LFOs, which means we can switch between a single LFO for all filters, or have separate oscillators each controlling 3 filters - that would be very psychedelic.

A couple of notes: This isn't a real phaser. I'm not sure if there could be any difference in sound, but it's worth noting. Also, the ROG Phozer is using a very similar idea, of course. What I've added is doing it with inverters for lower parts count, extended it into more stages, and made it so you don't need a multi-gang pot to control resonance.

I haven't built it yet, don't really have time right now. The values shown are just "whatever", will certainly need refining. Anybody want to give it a go? Any thoughts on the idea?

[amptramp]

The phase shift in a twin-T filter takes place mainly right around the tuned frequency with a violent shift from +90 to -90 degrees in phase near the tuned frequency but little effect more than a few percent away from the tuned frequencies.  This is in contrast to the usual phaser network that provides a smoothly varying phase shift across the spectrum.  Instead of a phasor sound, I would expect notches to be swept through the audio band that will "interrupt" certain frequencies periodically.

The outputs of the op amp stages are tied together and will fight each other - these should go to the summing junction of a final stage, possibly with separate gain adjustments.

I like the idea of the swept signal going through an amplifier that can be set for positive or negative gain with a single pot.  It will facilitate checking out all the possibilities with this circuit.

It will not sound like a phaser, but it may have an interesting sound of its own, so I am looking forward to a video.

[Taylor]

I think it will sound pretty similar to a phaser, actually. Check out the sounds of the ROG "Phozer", which is the same basic concept:

http://www.runoffgroove.com/salvo.html#phozer

Sounds much like a phaser to me.

[tcobretti]

This looks badass, and different oscillators to create a Bi-phase-ish effect is rad.

[ayayay!]

I'm not a phaser guy, but I'd think this would sound a little more flange-ish, but it's cool because I don't like the typical flange circuits. 

[Taylor]

The outputs of the op amp stages are tied together and will fight each other - these should go to the summing junction of a final stage, possibly with separate gain adjustments.

The image above is just a sketch of the idea, not a fleshed out design. So it won't work right as shown, as you say. This is just to get the idea out there, until I get some time to actually build it, flesh it out, and record it.

[Gurner]

I'd thought one of the endearing characteristics of an all pass is that you can change the phase but the amplitude remains the same....same can't be said of a band pass - I'd imagine you're gonna have all sort of level attenuation/accentuation going on, which will make for a crazy output.

[merlinb]

I'd thought one of the endearing characteristics of an all pass is that you can change the phase but the amplitude remains the same....same can't be said of a band pass

I was thinking that. Haven't you drawn a set of notch-boosters rather than notch-cut filters?

[askwho69]

Another IDEA>? wow this will be great!

[Mark Hammer]

It definitely won't sound like a flanger, because the spacing of peaks will be different.  And it probably won't sound like a phaser because it treats bass differently, and the topography of the peaks is different than what is created when producing some notches with a bit of feedback.

But it WILL sound interesting and musically valid, and that's pretty much all that matters.

[Taylor]

I'd thought one of the endearing characteristics of an all pass is that you can change the phase but the amplitude remains the same....same can't be said of a band pass

I was thinking that. Haven't you drawn a set of notch-boosters rather than notch-cut filters?

Right. If you read the first post, you see that the plan is to mix the inverted filtered signal with the non-inverted clean signal to produce notches. And by using this switch hitter amp, we can still have peaks, which you get with negative feedback in a standard phaser.

[Top Top]

Taylor you have actually done the thinking through of an idea I had myself but never took the time to sit down and figure out. I had a similar idea because of getting phaser-like sounds at times from building a wah/inverted wah out of two bandpass inverter stages. I had thought, "what if you built it on all six stages tuned differently?" which you have done there.

An additional 4069 can provide two independent LFOs or one LFO with both positive and inverted outputs (with two inverter stages left over), or even two LFOs with positive and inverted outputs (if fed to LEDs powered in an inverted fashion to one another).

[Taylor]

An additional 4069 can provide two independent LFOs or one LFO with both positive and inverted outputs (with two inverter stages left over), or even two LFOs with positive and inverted outputs (if fed to LEDs powered in an inverted fashion to one another).

Right, in my first post above I mentioned that I wanted to do a dual LFO using a second hex inverter - should make for some pretty unusual sounds with only 2 hex inverters and a quad or dual opamp. I'm excited to build this, just don't know when I can find the time. I really hate the fiddliness of breadboarding, so I'll probably perf it up and see how it sounds.

[PRR]

> similar to a 12-stage phaser, except with only 6 variable resistances

Yes, maybe.... but the frequency shift is at best square-root of LDR resistance, whereas in the classic all-pass F varies directly as R.

So you may need a wider range of R, which may force you to futher high/low extremes. Matching mid-range is hard enough, holding a match to extremes is harder.

And when changing one R in a 3-R 3-C network, you must have some other parameter shift. Gain or Q or both. (It's also unlikely that one leg can vary 2000:1 1nFd-2uFd while the other leg stays the same value; but you'll work that out on the bench. Also isolating the twin-Tee inputs from each other may be needed.)

[Taylor]

Well, for sure this is a funky design which won't work or sound exactly like a phaser and will definitely have some quirks. I just thought it might be a fun idea to play with.

[PRR]

Frequency sensitivity is square-root at best, much worse when R_var is way-away from the other leg's resistance.

1 ohm = 28KHz
10 ohms = 5.3KHz
100 = 2.4KHz
1K = 977Hz
10K = 342Hz
100K = 112Hz
1Meg = 23Hz

Shape-shifting is bizzare beyond simple tweaks:



Base gain is fairly stable but peak-gain and Q are all over the place.

While you may not need 1000:1 change of F, there's no good way to get a 1000000:1 range of resistance in a simple LDR or FET. (LDR will go that far, VERY slowly and with very uncertain final value. FET will get to an unknown resistance pretty quick.)

[Strategy]

Good work Taylor. There are various synth modules and things which operate on the principle of multiple bandpass filters swept independently with the capability to set different cutoff frequencies for each and different control voltage inputs. The result is typically used to create vocal-ish 'formant' type sounds. I think some of them are sort of phasey sounding as well. I think Mark Hammer is right it probably will be uncategorizable but highly musical.

Way to go!

Strategy

[Taylor]

Here's the LFO. It allows all stages to be fed by one triangle LFO, or to split them into 2 groups of three, each with their own independent LFO. This LFO can also have a square wave tapped from the output of the leftmost inverter. Not sure if square wave phasing would be all that interesting, though, so I haven't shown that in the schematic. Although, with the slow response of optocouplers, I suppose it wouldn't be all that square anyway. It'll be worth messing with anyway.

I'm not positive how well the inverter will do driving all 6 LEDs in mono mode.

[Taylor]

Frequency sensitivity is square-root at best, much worse when R_var is way-away from the other leg's resistance.

1 ohm = 28KHz
10 ohms = 5.3KHz
100 = 2.4KHz
1K = 977Hz
10K = 342Hz
100K = 112Hz
1Meg = 23Hz

Shape-shifting is bizzare beyond simple tweaks:
Base gain is fairly stable but peak-gain and Q are all over the place.

While you may not need 1000:1 change of F, there's no good way to get a 1000000:1 range of resistance in a simple LDR or FET. (LDR will go that far, VERY slowly and with very uncertain final value. FET will get to an unknown resistance pretty quick.)

I didn't completely understand the significance of this at first, but it's a problem. I pondered it and am wondering about doing switched resistors using CMOS switches. According to RG's ASMOP page, these resistors vary between 100 ohm and infinity, assuming a 100 ohm series resistor. Unless there's a gotcha here (...) that seems like a fruitful way to go with it.