Q About Phasers Vs. Flangers

[Paul Marossy]

So, after just building my Phase 45, I was thinking about the differences between flangers and phasers. I did read the verbage on phasers/flangers at GEO, but I still have a question about it. Depending on how I adjust that trimpot on the Phase 45, it sounds like a flanger to me - albeit, not as exaggerated sounding as a flanger, but still sounds much like a flanger. Is the difference strictly how many notches there are in the circuit? And do flangers also use feedback to exaggerate the notches?

[Rodgre]

Flangers do use feedback to create sharper effects. Often it's labeled Resonance, but it's still feeding the effected signal back into itself.

Flangers and Phasors do a similar, but different thing. Flangers work on a principle of a small delay (usually a BBD analog delay circuit) modulated and blended back in with the dry signal. Phasors do not use a delay, but (and I'm sure someone else can explain it better) a series of filters that are essentially shfiting the phase at a varying frequency. The circuits do a similar effect electronically, but with two different approaches, which gives them unique sounds compared to each other.

Roger

[Mark Hammer]

The is a great article on phasing vs flanging by Marvin Jones in a 1978 issue of Polyphony that I need to scan and post one of these days.

Essentially the deal is this.  Phase-shift (or allpass) produced notches remain fixed in number, while time-based notches change in number.  Most other things flow from that.  Both can use regeneration to enhance or exaggerate the notches and peaks, and both can use overall phase inversion of the signals to produce negative/positive phasing/flanging or simultaneous sum/difference outputs.  The fixed vs variable number thing is crucial, though.

With phase shifters, the number of notches/humps produced depends on the number of stages used (divided by 2), and the cumulative phase shift produced by those stages.  Most of us will experience phase shifters as having 12 or fewer stages (typically 2-6), typically yielding just a small handful of notches.  Because there are just a few notches and peaks clustered together, there is a more pronounced "focus" to the sound as it sweeps up and down, because it is all happening at one "place" in the audio spectrum (although that place moves around).

In contrast, the number of notches produced in the audible spectrum increases as delay time increases, so that when you get out to 10-12msec delay times, the signal is covered with notches and peaks, justifying the term "comb filter".  The key to the flanging sound is that as the swep starts from its "highest" point (i.e., shortest delay), the signal starts to get increasingly "infected" with notches, until you reach that point where you get well over a dozen and the delay starts to sweep the other way, gradually removing notches and "uninfecting" the signal.  With phasers, there is a striking similarity to sweeping a bandpass filter up and down, whereas that does not appear with flangers unless there is some serious regeneration around a particular delay time.

In theory, if you had enough phase shift stages to produce a LOT of notches, then as you swept it higher, many of those notches would move outside of hearing range, leaving just a few audible.  Then as you swept down again, the spectrum would start to include more and more notches, producing that "infected signal" phenomenon.  In fact, this is not only true in theory but in practice.  Mike Irwin made and demonstrated a 24-stage phase shifter to me (I have a sound sample somewhere), and it starts to sound very much like flanging at that point.

On the other hand, the demo uses a white noise signal, in which it is not possible to also detect the detuning that comes with varying the time by as much as 10msec.  Had he used a music signal, I think the phase/flange difference would still be perceivable.  Although phasing CAN mimic the psychoacoustic effects of changing the number and ditribution of notches, the flanging effect also comes with audible time delay, which phasing does not mimic.

So, phasing and flanging can live in the same neighbourhood and block, but they can't pass for each other flawlessly.

[Paul Marossy]

Thanks guys. That gives me a much better idea about the differences between the two. Going by what my ears tell me, it sounds like you could end up at almost the same place with a multi-stage phaser as you could with a flanger, except that the phaser would probably become a little impractical...

I would love it if you did scan that article, Mr. Hammer.  

[Mark Hammer]

I'd be happy to, but it'll have to wait for a while.  We're heading off on cross-country minivan trek for 3 weeks, and it'll take at least a week to clear out all the corporate spam from my mailbox when I get back.

BTW, one of the things that phasing CAN do, which flanging can't is to change the distribution of phase shift and the resulting notches.  The UNiVibe is one example of this in action.  The disparate cap values distributes the phase shift differently such that the notches are shallower and spread farther apart.  Conversely, it is possible to produce notches that are closer together than those produced in the usual P90-type pedal/design, and all of that is independent of whether one is higher up or lower down in the sweep.  With flanging, the spacing of notches is determined by the time-delay, period.

[Paul Marossy]

Mark-

Thanks for the succint addendum, that makes perfect sense to me now.
I think I may just breadboard a phaser circuit and mess around with it just for fun...

Ok, one last question, though. There is a cyclical effect with my Phase 45, which sounds kind of like a time delay effect. What causes that? Is it the LFO that does it?

[Mark Hammer]

Couldn't tell you.  I suspect I'd need to hear it to know what part of the sound catches your attention.

[toneman]

Mark,
what about the April/May 78 article by Gary Bannister??
"To Phase or Flange"...
tone

[Mark Hammer]

That is the very article, and yes, the correct author.  I was confusing it with Marvin Jones' article on Reticon chips.  That's the one I'm hoping to scan and post when I get back from holidays.

Do you have it available or do you know of it posted anywhere already?  If so, that would save me a heap of work.

[Paul Marossy]

Yeah, if it's posted somewhere, please let us know!  

[Transmogrifox]

Phasors do not use a delay, but (and I'm sure someone else can explain it better) a

Everything mentioned after this quote helps clarify, however, it is good to remember that phase shift is time delay.  The difference is that it is not constant time delay for all frequencies unless the phase shift is linear from 0 Hz to infinity Hz.  For audio purposes, linear phase from 50 Hz to 5 kHz would be effectively the same as a flanger delay with enough phase stages...although the phaser cannot change the time delay, rather it changes the frequencies at which the time delay occurs.  

The difference with a flanger, however, is that the linear phase function slope is modulated, whereas with the phasor, the slope (constant time delay) is set by the number of filter stages, then the spectral location of this linear phase change is modulated.  So the phasor has a constant delay on a certain group of frequencies, and modulates which frequencies these are. The flanger modulates the slope of the linear phase line.

If you don't quite follow this, go back up the Hammer's posts as he did a better job of simplifying the matter.  My entire point is that phase IS delay.

[puretube]

linear phase from 50 Hz to 5 kHz would be effectively the same as a flanger delay with enough phase stages...

you can get there with n "univibes", where the 4 staggered univibe stages resemble 1 linear phaseshift stage, which all in all takes 4n phase-inverter/RC combo`s...., where n should be >2, better 4 or more;  :wink:

note: a phase-inverter (like in tubeamps) does not shift the phase,
but just inverts (swapping polarity) it with no delay involved.

[Mark Hammer]

Yes, phase IS delay, although what counts as delay on the scope is somewhat different than what counts as delay in the ear.  That is probably the source of much confusion, as the delay amounts that are worthy of mention to EE-types simply won't register in the experience of most musicians.

The one exception to this principle is that the amount of "delay" created by phase relationships varies with frequency and when you get down low in the spectrum, such delays can be heard AS delays.  For instance, were I to play a 100hz tone out of the left speaker, and send its inverted version to the right, there will be an approximate 5msec difference between when the peak of the waveform hits one ear vs the other, and inter-ear differences of that sort ARE consciously detectable by humans.  The same is not true when we move up into 1khz and beyond.

[puretube]

The is a great article on phasing vs flanging by Marvin Jones in a 1978 issue of Polyphony that I need to scan and post one of these days.
...

https://www.muzines.co.uk/images_mag/pdf/pl/pl_78_04.pdf (p.12)

[Mark Hammer]

Hah!  You beat me to it, Ton!    Scrolling down through the thread, I thought "If it was still 2004, I'd have to scan it, but I know it's posted, so I can just link to that, now."  And there, when I got to the bottom, you had already done that, with the exact link I was going to use. 

[ElectricDruid]

That article makes clear what I think of as the key difference between phasers and flangers: Flangers have the notches arranged harmonically. If you happen to play a note that hits a notch, all of the harmonics of that note will fall in notches too, and the note will completely disappear. That doesn't happen in phasers because the notches aren't spaced like that.

That said, sweeping a *lot* of notches produced anyhow you like is going to sound "flangey", and sweeping a few notches is going to sound "phasey". The rest is largely details, hohoho

[Rob Strand]

That article makes clear what I think of as the key difference between phasers and flangers: Flangers have the notches arranged harmonically. If you happen to play a note that hits a notch, all of the harmonics of that note will fall in notches too, and the note will completely disappear. That doesn't happen in phasers because the notches aren't spaced like that.

If you listen to the delay/phased signal they also sound different, and there's no notches since we haven't mixed the signals yet. (There's many extraneous reasons that can cause this as well eg. sweep shape.)

[Mark Hammer]

That article makes clear what I think of as the key difference between phasers and flangers: Flangers have the notches arranged harmonically. If you happen to play a note that hits a notch, all of the harmonics of that note will fall in notches too, and the note will completely disappear. That doesn't happen in phasers because the notches aren't spaced like that.

That said, sweeping a *lot* of notches produced anyhow you like is going to sound "flangey", and sweeping a few notches is going to sound "phasey". The rest is largely details, hohoho

It does.

[Rob Strand]

If we use the starting point that a 10-stage phaser using the common first order allpass filters starts to sound like a flanger we can see what matches up in terms of notches and delays.

An all-pass filter is an analog circuit.  However, it has an effective delay.   The effective delay (in time) is the group delay.  The group delay is defined as Tg = - d(phase)/dw.   LTspice can plot group delay.

https://en.wikipedia.org/wiki/Group_delay_and_phase_delay

I took four circuits to compare:
- a straight delay  0.68ms to 3.4ms, representing the flanger
- a conventional phaser constructed from 10 first order allpass stages.   
   From experience, this is the minimum phaser which starts to sound like a flanger.
- a cascade of 5 *identical* second order all pass filters.
   The Q was chosen to be 0.6 which gives a fairly flat group delay over a wide bandwidth.
   The second order all-pass circuit are discussed in this thread.
   (It doesn't matter which one we used as we are only interested in the response.)
   https://www.diystompboxes.com/smfforum/index.php?topic=129676.0
- a full-blown 10th order Bessel allpass.  This is made up of the same form circuit
  as the previous one.  The difference is the f0's and the Q's are not equal.
  They are tuned to given the flattest group delay and widest bandwidth.
  Basically these are the best approximation of a delay by an analog circuit.

What I've done is tune the delay and lower notch frequency to roughly match between the
Flanger (delay) and each of the Phasers (all-pass).

Schematic:


Group Delay:
- we can see at the low end of the sweep the delay is flat in the lower part of the guitar spectrum.
- at the high end-of the sweep we can see the delay extends to nearly to whole guitar spectrum.
- in short the all-pass base-line first order pass circuit just approximates a delay over the sweep.
  That kind of lines up as the 10 stage circuit does start to sound like flanger.
- We can also see the 5x identical stage Q=0.6 circuit provide a match over bandwidth.
- The Bessel has further improvements.  Nice and flat over a substantially wider bandwidth.



Notches:
- this very much follows the pattern of the group delay.
   The frequencies where the group delay is flat are exactly where the notches agree best.
- the 10 stage simple circuit just keeps up with the flanger at the lower notch points (both lo and high sweep)
   and spreads over a reasonable part of the guitar spectra.





As far as the high frequencies go.  I suspect our ears are less able to decode the mess in the high frequencies from the flanger.   From the guitar spectra we might expect we need to get some degree of matching upto 3kHz to 4kHz.   The circuits are crudely holding on at the top of sweep.  So it does support to some degree why the 10th order filters sound like flangers.

If use higher order filters then we would expect an even closer match between the flanger and phaser.

For common low-order phasers these clearly sound like phasers.   The above matching patterns definitely fall apart for low filter orders  (small numbers of stages.)   This follows from the group delay not having a flat region within the range of guitar frequencies.

[Steben]

What about the statement that a 10+ stage phaser might be used in a rotary woofer part sim, but a delay is slightly more simple anyway?