Another strange idea (phaser/flanger/wah)

[11-90-an]

Hi all,

Here's another idea...

Most of us know the frequency responses of phasers and flangers... moving notches. The idea now is to invert the frequency response, instead of notches there are thin bandpass-like frequency responses... (i'm not sure how to phrase it... )

Since you probably don't understand what I'm saying, here's a quick illustration..



My "prediction" would be that it would sound like a very thick wah... but I don't think it would sound very "vocal"..

Given that this all makes sense, the only problem now would be "how to make this"...
Multiple bandpass filters would probably be a big pain to tune and there would be lots of space needed...
Would there be a way to implement this by some sort of signal cancellation like phasers and flangers?

[skyled]

This sounds similar to the Resonator section of the Korg PS-3100/3300. It had 3 filters with very high resonance that could be swept with an LFO. There's a few eurorack type clones of it and also some standard rackmount clones. I'd like to have one at some point 

[Rob Strand]

Most of us know the frequency responses of phasers and flangers... moving notches. The idea now is to invert the frequency response, instead of notches there are thin bandpass-like frequency responses... (i'm not sure how to phrase it... )
Since you probably don't understand what I'm saying, here's a quick illustration..
....
My "prediction" would be that it would sound like a very thick wah... but I don't think it would sound very "vocal"..

It's an interesting way to view it.

When you have a flanger or delay pedal with regeneration it does that.   A phaser with *correctly implemented* feedback should also be able to do that.   

Questions that come to mind are:
- can the band-pass version of the flanger be done *without* feedback?  so you don't get the reverb from the feedback loop?   
- The standard phasers only use first-order all-passes so the question is can you do better using feedback and second order all-passes.

There's a couple of multiple peak wah's out there.  Also the EHX Bass-Balls.   I guess the wah's are more inherently band-pass than the others.

[Rob Strand]

Here's an 8-stage phaser with feedback.  I didn't play around with the sign of the feedback.





Here's the inverted feedback, we get four peaks and flatter ends,

[iainpunk]

my first reaction:
this is great!! im going to try this in a DAW to see if it works or sounds good.

ill report back later

cheers, Iain

so i forgot to post this and the post reply tab is still open so i'll add my findings,

it doesn't sound that good, its like a more breath tone version of a chorus, not pleasant like the relatively broad band of a wah (compared to this anti-phaser), but harsh and more envelope filter high Q sounds, but breath like. this is in the computer world tho, i don't know how it sounds IRL tho.

cheers again, Iain

[11-90-an]

Here's an 8-stage phaser with feedback.  I didn't play around with the sign of the feedback.



Here's the inverted feedback, we get four peaks and flatter ends,

From here, might there be a way to still have 8 stages, and have some sort of creative feedback, something along the lines of this....?

(Only a rough example, some should probably have inverted feedback)

[garcho]

resonant peaks need to be tamed and/or tuned to your instrument, the key, the room, etc notches never pierced anyone's ears

[Mark Hammer]

Keep a few things in mind:
1) Phasing effects are produced by the cumulative phase shift introduced.  To be able to predict where the notches fall, as in the assorted simulations, the phase shift occurs in the exact same spot for each stage.  In real life, resistor and capacitor tolerances vary such that, even with nominally identical component values, stage A might have its 90-degree maxima at 340hz while stage B only reaches that maxima at 410hz.  With only those two stages, there is 180 degrees of shift at 410hz, but less, in total, at 340hz.  Uni-vibes make productive use of the manner in which phase shift is distributed across the spectrum, so variation in where maxima occur is not intrinsically a BAD thing.  But non-alignment may not achieve the desired sound.

2)  My experience has been that, while nominal resistor values make each stage unity gain on a simulation, in real life there can be teensy bits of gain in this or that stage.  And since phase shift stages are cascaded, a teensy bit here times a teensy bit there, and by the time you get to the last stage, that multiplicative gain has turned into something that risks oscillation when fed back.  In principle, the more phase-shift stages, the greater the risk of oscillation.  Happily, this cab be compensated for by how much or how little feedback is applied.

[Rob Strand]

From here, might there be a way to still have 8 stages, and have some sort of creative feedback, something along the lines of this....?
(Only a rough example, some should probably have inverted feedback)

I would have to check in the simulator but I'd expect it would probably move the peaks.    The height of the peaks need careful adjustment of the feedback.

[Rob Strand]

Well I set-up a "Leapfrog" feedback structure in banks for 4 and then tried all combinations of feedback including zero (so 27 combinations).   None produced the nice peaks like the overall feedback case.   There's some traces over the top of one another.






This one is clearer, it's 18 cases with the K2=0 case removed,

[iainpunk]

hey, rob, could you please simulate this idea, i think it might be interesting to put different kinds of filters in the all pass filter feed forward? with the increased gain of the allpass filter, there could possibly be some band boost action going on in between notches.

cheers, Iain

[Rob Strand]

hey, rob, could you please simulate this idea, i think it might be interesting to put different kinds of filters in the all pass filter feed forward? with the increased gain of the allpass filter, there could possibly be some band boost action going on in between notches.

There are "known"  second-order all-pass filters which use a single opamp.   The standard all-pass filters used in phasers are first order.  When we put two of those in cascade we get, in effect, a second order all-pass.   In this case the second order filter has a Q of 0.5 which is more or less fixed due to the cascade.    With the second-order all-pass filter we don't have to stick to Q=0.5 -  that's the difference between the second order and first order stages.  [slightly pedantic but you can get Q < 0.5 with the first order sections by using different RC time constants.]

Anyway here what happens when we vary the Q.    The notches (or peaks) get closer together as the Q increases.   I've shown two examples: one without feedback, which has notches, and one with feedback where the notches change into peaks. Q's = 0.5, 1.0, 2.0



For the circuit you gave I don't think it's ends-up with all-pass.  Anyway I'll try it.

As far as feed-forward goes,  I was thinking feed-forward for the Flanger case might help gives peaks without feedback.   I'd have to get the outputs from number of taps in order to get a reasonable amount of "constructive interference" to give reasonable sized peaks.    It's really like the feedback case *except* the number of taps create a finite number of echos.  OK for DSP but not very efficient if we are using individual BBD chips.

[EDIT: cleaned-up a few things]

[Rob Strand]

hey, rob, could you please simulate this idea, i think it might be interesting to put different kinds of filters in the all pass filter feed forward? with the increased gain of the allpass filter, there could possibly be some band boost action going on in between notches.

I gave it a go but I didn't get a good response.    The all-pass is acting like wide dip.   I know a like that circuit *can* work.

Leave it with me, I'll try to work out what needs to be tweaked (probably only a value change somewhere).


OK I changed the 1k's to 8.642k's to get an all-pass response from each stage.




Interestingly the notches are moved outward.

The notches become peaks when the feedback is negative.

[iainpunk]

waauw, that's interesting indeed!
but the spread apart is kind of logical, since in bridge tee filters  the two interacting LPF and HPF have a low Q factor
is the red line with the original values i provided?

thanks for the sim. falstad doesn't provide any useful frequency analysis, and i don't have the cash to dish out for expensive simulation programs.

cheers, Iain

[11-90-an]

very interesting... thanks for the effort put in these sims, Rob...

8.642k resistance seems dreadfully specific...
Do you guys think it would still work satisfactorily with normal 8.2k's with 10% tolerance?

I wonder, too, since this new circuit provides 4 peaks with only 4 opamps(unlike the old one which has 4 peaks with 8 opamps), this circuit *can* probably yield 8 peaks with 8 opamps, right? quite cheaper....

Or am I missing something?

[iainpunk]

very interesting... thanks for the effort put in these sims, Rob...

8.642k resistance seems dreadfully specific...
Do you guys think it would still work satisfactorily with normal 8.2k's with 10% tolerance?

I wonder, too, since this new circuit provides 4 peaks with only 4 opamps(unlike the old one which has 4 peaks with 8 opamps), this circuit *can* probably yield 8 peaks with 8 opamps, right? quite cheaper....

Or am I missing something?

yes, that 23 cents you spend on one quad opamp really breaks the bank.
the main thing i see is better about this is real estate on PCB's, bit if you worry about that, i'd suggest going SMD or use SIP package opamps, since you can have a quad DIP opamp in double the footprint of 2 double SIP opamps (2 duals sandwiched, with some aluminium foil connected to ground in between to eliminate possible crosstalk)

cheers, Iain

[11-90-an]

I mean, like, if it saves some money, it's still money saved...
I prefer to use the tl074s I have here, thanks for the idea of SIPs... seems interesting... No DSP yet, though. I'm not quite ready for that...

[PRR]

....i don't have the cash to dish out for expensive simulation programs.

Uh, LTspice is dead-free (if there is a higher tier it has never pimped it to me). I hate it, but many people use it, so massive peer support is out there.

There are others. Is TINA-TI still available? Old farts should hunt for Pspice 1999. Many pay-wares are available as "trial" and SOME of them are dead-free up to maybe 99 components, which is more than about any pedal. (Especially if you don't do stupidstuff like simming the power supply and its bypasses.)

MicroCap is abandonware, and looked very good, and mature. Digi-Key has web-based Scheme-it.

Or is it a religious conflict? (House of Mac?)

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

There ARE cell-phone sim apps but the way Pspice fills my desktop screen I don't see how any useful work can be done in 3 inches.

[Rob Strand]

There's one key point about all these circuits.   The phaser circuit require an all-pass filter, that's how they work.   It doesn't matter what *circuit* you use to get the all-pass function it will still work.

In the standard phaser circuit the all-pass blocks are first order circuits.    We can mentally bundle the these first order all-pass circuits in pairs to "make" a  second order all-pass circuit.   The key aspect about the result of that is the Q of the second order all-pass circuit is 0.5.

The addition of feedback to get peaks etc. is a completely separate idea that we add on to an existing circuit.   The idea should work regardless of the all-pass circuit we choose.

So far we have:

Reply #3                  the standard circuit, first-order all-passes and effectively Q=0.5

Reply #11                Second order all-pass circuit.
                                  This circuit lets you choose Q's greater than 0.5.
                                   Conclusion: higher Q's brings the notches (or peaks) closer together.

Reply #12                Second order all-pass circuit.
                                  Here we saw the the notches and peaks get further apart, which implies the Q is less than 0.5.
                                  After analyzing the circuit  the Q was something like Q=0.18.
                                  As it turns out this circuit cannot

We could actually get Q=0.18 a lot easier using the standard circuit in reply #3 with unequal cap values.
One advantage of using the second order circuit is less opamps.  However, the control of the resistors is not as convenient as the common circuit.    The resistors are not ground and in the case of the Reply #12 circuit the resistors aren't of equal value.
Another disadvantage of the second order circuits is odd part values.  The standard circuit in reply #3 requires no effort with part values.

[Edit I think the maximum Q is sqrt(2) / 4 = 0.35 not 0.47]
Just to prove the point I took the circuit of reply #12 and redesigned it to have the maximum Q.  This turns out to be  Q = sqrt(2) /3 = 0.47, which is lower than the standard circuit on Reply #3.    As you can see the notches are now closer to the circuit in reply #3. The main point here is it's the Q of the  circuit not the circuit itself which provides the response.

but the spread apart is kind of logical, since in bridge tee filters  the two interacting LPF and HPF have a low Q factor
is the red line with the original values i provided?

It actually didn't make sense to me unless the Q was less than 0.5, which turned out to be the case.

8.642k resistance seems dreadfully specific...
Do you guys think it would still work satisfactorily with normal 8.2k's with 10% tolerance?

I wonder, too, since this new circuit provides 4 peaks with only 4 opamps(unlike the old one which has 4 peaks with 8 opamps), this circuit *can* probably yield 8 peaks with 8 opamps, right? quite cheaper....

As mentioned above, the second-order circuits have good and bad aspects.   If you were to use  standard values you need to check how much effect it has on the response.   If you were using a circuit with feedback to get peaks, 10% error in the gain is enough to completely stuff-up the peaks.   They become very touchy circuits, not good for production and it's likey the feedback needs a trimpot - just like Flanger's have!

It is possible to change the circuit in reply #1 to have equal resistors and different cap values.   That helps the problem of needing different resistor for the parts that need to be swept.

[iainpunk]

....i don't have the cash to dish out for expensive simulation programs.

Uh, LTspice is dead-free (if there is a higher tier it has never pimped it to me). I hate it, but many people use it, so massive peer support is out there.

i did not know that, i just assumed it be expensive like the rest

i saw a thread about LiveSpice that runs in DAW's, is that a recomendable program/plugin?
i do a lot of electronic music production for a friend who's been working on a game for the last 7 years (i doubt he'll ever finish it tho)
i'd love to do some weird ass sounds and effects i have in real life, but not in my DAW.

cheers, Iain