Feedback in classic modulation pedals?

[guitarnerdswe]

Hi, new guy here!

I have long wondered why certain classic modulation pedals sound the way they do. Particularly, feedback in phasers and flangers. In a block logo MXR Phase 90, is the feedback positive or negative? In a univibe, is there any feedback at all in that circuit?

Another thing I would like to understand, what determines the starting frequency of previously mentioned pedals? Or rather, does anybody know where those 2 pedals starts their sweeps?

Phase 90 schematic:
http://www.matsumin.net/diy/bunkai/MXR_Phase90/MXR_Phase90_1993.BMP

Univibe schematic:
http://4.bp.blogspot.com/-5J3L9R1_Chk/T_WXTQl5NqI/AAAAAAAABo0/K70MXGGHHVE/s1600/fv_sch_vint.gif

[PRR]

Welcome.

> Phase 90, is the feedback positive or negative?

There must be a "Theory of Phasers" web page somewhere?

[Mark Hammer]

Vibes never have feedback, simply because it doesn't do anything useful.  The notches created in phasers are accentuated by feedback, as are the peaks adjacent to the notches.  The way that phase shift is distributed in vibes creates very wide shallow dips, rather than focussed narrow notches,  So feedback has nothing very obvious to accentuate.  You'll note that the slowest speed on a vibe is faster than the slowest speed on a phaser; another result of the subtlety of these shallow wide dips.

There is no universal starting frequency.  In fact, just about 90 minutes ago I was setting up the variable starting point for the sweep in a P90 I was finishing.  The starting point is largely determined by the joint function of the cap and resistance to Vref (or ground in some designs).  With a cap of .05uf and resistance of 22k, the notches "begin" around 144hz.  As the resistance of the FET in parallel with the 22k resistor goes low that frequency starts to go higher.  The trimmer on a P90 adjusts the biasing of the FETs to set that parallel resistance to whatever standard MXR settled on for "their" sound.  I replace the 1M fixed resistor from the trimmer to FETs with a 510k fixed resistor and 500k pot to move the starting point around.

[Rob Strand]

I have long wondered why certain classic modulation pedals sound the way they do. Particularly, feedback in phasers and flangers. In a block logo MXR Phase 90, is the feedback positive or negative? In a univibe, is there any feedback at all in that circuit?

The frequency response of the final mixed signal has notches and peaks.  Without feedback the peaks have unity gain.   When you add feedback generally you want to choose the sign of the feedback to emphasize the peaks.  You have to be careful about what negative and positive feedback means.  The all-pass filter changes the phase from DC to high frequencies.   For a 4-stage phaser there is only a single central peak. You want positive feedback at the peak frequency and this turns out to be when you have negative feedback at DC; so superficially you might call that a negative feedback connection.   If you were to flip the phase on a 4-stage phase the central peak isn't affected so much and it's the bass and treble frequencies which get emphasized.  These are not sweep so it doesn't sound as good.   When you have a flanger you have more peaks across the spectrum so you can often choose either sign for the feedback.  However, positive feedback at low frequencies will cause bass lift, which is undesirable, but you can remove the effect with a high-pass filter.

The main reason the MXR Phase 90 sounds different to the Univibe is the Univibe has quite a wide sweep.   While the Univibe looks like a 4-stage phase the wide spit in the bands makes is work more like a 2-stage phaser.   To some degree that's why you need a wider sweep (it's also why feedback isn't effective on this unit
as there is no usable central peak).  The two all-pass filters tuned to the lower frequency cause more of a tremollo effect at low frequencies.   The tremollo effect and the wide sweep make Univibe sound more pulsating than the Phase 90.

The LFO on the Phase 90 use a pseudo-triangle wave whereas the Univibe uses a sine-wave (more or less).  It's seems like a contradiction that the Univibe is more choppy yet it uses a sinewave LFO.

The Phase 90 has a bit of a quirk in that the signal has a small amount of distortion.

One thing worth adding is not all Phase 90's sound the same the differences in the JFETs affects the sweep ranges.  Also not all Univibes sound the same, that's usually due to inconsistent adjustments and variations in the optos.

The other one to throw in is the EH small-stone phaser.  That one has a little wider sweep than the Phase 90, a different LFO, and uses feedback.   It sounds more characteristic of a phaser with feedback.

[bluebunny]

There must be a "Theory of Phasers" web page somewhere?

Not sure if "theory of..." is the same as "technology of..." (Geofex link).

[antonis]

or "the analysis of": https://www.electrosmash.com/mxr-phase90

[Mark Hammer]

The Small Stone's "Color" switch changes several things at once.  First it changes the amount of feedback (from the last stage to the input, as opposed to feeding from a later to an earlier stage).  Second, it changes the speed range (slower and faster).  Finally, it changes the LFO waveform to suit the speed range.  Faster uses a triangle waveform, while slower uses a "hypertriangular" waveform that slows down as it approaches the "low' end of the sweep, and speeds up as it approaches the top of the sweep.  The Phase 90 form of phaser uses the same waveform for all speeds.  In some respects, it is a cruder design, although EHX didn't go dramatically farther in its attempt to produce a one-knob/one-switch phaser.

Some phasers have two outputs: dry plus wet and dry minus wet.  Feedback does not function the way for each of those combinations.  Indeed, cranking up the feedback in one gets a less rather than more intense sound.

Not only do Uni-vibes never come with feedback controls, but few chorus pedals come with feedback controls as well (unless the pedal is designed to cover both flanging and chorus delay ranges).  The function of using feedback is to shine a spotlight on a focal point of wherever the filtering action of the circuit is at the moment.  Neither chorus nor vibe have any sort of focal point, in comparison to phasers and flangers that "start" up here, and go down to there.  Vibes and chorus certainly have perceptible motion, but no discernible "here" and "there".

[guitarnerdswe]

Thanks for all the replies guys! I'm new to this, but I'm trying to learn. I do have some more questions on the Phase 90:

1. Can the feedback in a Phase 90 be both negative and positive at the same time? I thought it was either or?

2. Also, is there a way to calculate how much of the signal is sent back to stage 2 through that 22k resistor?

3. I might have been unclear about what I meant by starting frequency. Is there a way to know the frequencies of the notches (or the midpoint between them) at the bottom of the sweep, or at the top of the sweep?

4. Does that speed knob change only the speed, or does it scale the depth on the sweep in any way?

Cheers guys!

[Mark Hammer]

 Each phase shift stage in that circuit acts as a kind of mixer that combines several signals.  If the feedback signal starts to exceed a gain of 1x (unity), then the circuit will start to oscillate.  So, the feedback resistor should NEVER go much below 12k in that circuit, and maybe not even that low.  Keep in mind that, even with 1% resistors, there is always the risk that one or more stages creates the teensiest amount of gain.  So the stock 22k value (24k in some issues) keeps the amount of feedback well below the level that might result in oscillation.  I've dropped it down to 15k or so, but note that the feedback also accentuates any hiss accumulated over the 4 stages.  A 20k pot in series with an 18k fixed resistor would get you a decent range of feedback levels; more and less intense than stock.

[Rob Strand]

2. Also, is there a way to calculate how much of the signal is sent back to stage 2 through that 22k resistor?

The simplified answer is it's the ratio of the 10k resistor (opamp output to opamp -in) and the 22k (10k/22k).

3. I might have been unclear about what I meant by starting frequency. Is there a way to know the frequencies of the notches (or the midpoint between them) at the bottom of the sweep, or at the top of the sweep?

It varies on each unit. That's one of the things that makes one unit sound different to another.  Also it depends on the number of stages.

4. Does that speed knob change only the speed, or does it scale the depth on the sweep in any way?

You will often see an RC filter on the output of the LFO this smooths things out.  In some cases it reduces the depth with high speeds a tad.

[idy]

The down and dirty way to measure the top and bottom of the sweep of an actual individual circuit would be:
feed your phaser with white noise and record the output. On a looper for example. Open that file with a program like "transcribe!" There you can highlight any bit of the wave file, say the first 1/2 second of the sweep, and use the "show spectrum" function. You will see nice peaks in your white noise signal above the corresponding keys on a piano keyboard. Translate using a table...Then highlight a section from the top of the sweep...

I have experimented with an old stereo graphic eq that has a spectrum display... not very fine gradations but 10 bands gives an interesting visual of the "voicing" of distortion pedals.

[guitarnerdswe]

2. Also, is there a way to calculate how much of the signal is sent back to stage 2 through that 22k resistor?

The simplified answer is it's the ratio of the 10k resistor (opamp output to opamp -in) and the 22k (10k/22k).

3. I might have been unclear about what I meant by starting frequency. Is there a way to know the frequencies of the notches (or the midpoint between them) at the bottom of the sweep, or at the top of the sweep?

It varies on each unit. That's one of the things that makes one unit sound different to another.  Also it depends on the number of stages.

4. Does that speed knob change only the speed, or does it scale the depth on the sweep in any way?

You will often see an RC filter on the output of the LFO this smooths things out.  In some cases it reduces the depth with high speeds a tad.

Aha, so roughly 41.66 % if it's a 24k resistor? I typed 22k by mistake earlier. You said that was the simplified answer. How complicated is the complicated one?

The down and dirty way to measure the top and bottom of the sweep of an actual individual circuit would be:
feed your phaser with white noise and record the output. On a looper for example. Open that file with a program like "transcribe!" There you can highlight any bit of the wave file, say the first 1/2 second of the sweep, and use the "show spectrum" function. You will see nice peaks in your white noise signal above the corresponding keys on a piano keyboard. Translate using a table...Then highlight a section from the top of the sweep...

I have experimented with an old stereo graphic eq that has a spectrum display... not very fine gradations but 10 bands gives an interesting visual of the "voicing" of distortion pedals.

If I had a Phase 90 here, this would be so much easier  I'm actually trying to recreate the Phase 90 in my Axe-Fx III. If somebody who has a Phase 90 could post a noise clip on a slow setting (and perhaps a fast), that would render all calculations unnecessary  I'm basing most of my "knowledge" on this article: https://www.electrosmash.com/mxr-phase90

I've tried matching the bottom of the sweep to the mentioned 58.5/340.8 hz notches (using noise and a RTA), and it sounds a bit low compared to recordings and clips, but I could be wrong. The important part is probably the top of the sweep though, since that is more audible on guitar. I tried using LTspice to simulate it, but I couldn't get it to work (user error).

[Rob Strand]

You said that was the simplified answer. How complicated is the complicated one?

The long answer is the presence of the 24k doesn't just add in the feedback, it totally screws up how the all-pass filter works.  If you want to model that you would have to write down the transfer function of that part of the circuit.  There will be two transfer functions one for the audio path input and one for the feedback input.

If I had a Phase 90 here, this would be so much easier  I'm actually trying to recreate the Phase 90 in my Axe-Fx III. If somebody who has a Phase 90 could post a noise clip on a slow setting (and perhaps a fast), that would render all calculations unnecessary  I'm basing most of my "knowledge" on this article: https://www.electrosmash.com/mxr-phase90

If you said that in the first place I would give you this link,

http://www.dafx14.fau.de/papers/dafx14_felix_eichas_physical_modeling_of_the_.pdf

It does the vintage Phase 90 without the feedback.  It also shows the method idy gave using the spectragram.     For the *vintage* phase 90 the lower notch spans from about 150Hz to 850Hz.   It does vary from unit to unit.   For a 4-stage all-pass with all-pass frequency 'f0'  the lower notch occurs at 0.414 f0 and the upper notch at 2.414 f0.  So from the lower notch frequency you can work out f0 and then the higher notch frequency.

[guitarnerdswe]

You said that was the simplified answer. How complicated is the complicated one?

The long answer is the presence of the 24k doesn't just add in the feedback, it totally screws up how the all-pass filter works.  If you want to model that you would have to write down the transfer function of that part of the circuit.  There will be two transfer functions one for the audio path input and one for the feedback input.

If I had a Phase 90 here, this would be so much easier  I'm actually trying to recreate the Phase 90 in my Axe-Fx III. If somebody who has a Phase 90 could post a noise clip on a slow setting (and perhaps a fast), that would render all calculations unnecessary  I'm basing most of my "knowledge" on this article: https://www.electrosmash.com/mxr-phase90

If you said that in the first place I would give you this link,

http://www.dafx14.fau.de/papers/dafx14_felix_eichas_physical_modeling_of_the_.pdf

It does the vintage Phase 90 without the feedback.  It also shows the method idy gave using the spectragram.     For the *vintage* phase 90 the lower notch spans from about 150Hz to 850Hz.   It does vary from unit to unit.   For a 4-stage all-pass with all-pass frequency 'f0'  the lower notch occurs at 0.414 f0 and the upper notch at 2.414 f0.  So from the lower notch frequency you can work out f0 and then the higher notch frequency.

Thanks! By my attempt to calculate, I got that the higher notch move between 873 and 4948 hz. Did I get that right? I actually looked at that very document the other day, but didn't realise that it contained the info you gave about frequencies and ratios. Did I simply miss it?

Looking at my real time analyser, I can see that adding feedback changes the frequencies of the notches, like you said. I have a simple feedback parameter in the Axe-Fx III that allows me to add positive or negative feedback (positive sounds more like what the resistor does on the pedal). So basically, if I get the notches right with no feedback, then the final step I just to add the correct amount of feedback. That's why I wondered if there was a more exact or "better" way to calculate what percentage of signal gets sent back to the second stage? 

[Rob Strand]

Thanks! By my attempt to calculate, I got that the higher notch move between 873 and 4948 hz. Did I get that right? I actually looked at that very document the other day, but didn't realise that it contained the info you gave about frequencies and ratios. Did I simply miss it?

That looks right to me.   The article probably doesn't contain that info.   About 15 to 20 years ago I did a lot of work on phasers.   A lot of the finer aspects about phasers don't appear in any articles!

Keep in mind the 2.414 and 0.414 values are for the *vintage* Phase 90 where there is no 24k resistor.  Adding the 24k resistor will change those numbers a bit.

Looking at my real time analyser, I can see that adding feedback changes the frequencies of the notches, like you said. I have a simple feedback parameter in the Axe-Fx III that allows me to add positive or negative feedback (positive sounds more like what the resistor does on the pedal).

If you want to emulate the actual pedal there's a bit more going on.

If you want an *accurate* phaser emulation, there's a quite a few subtle points which make a difference.  It's actually hard to explain why.

I have limited understanding of the Axe-Fx III but I suspect it has a built-in phaser block with feedback.   If it is like that you don't actually make your own all-pass filters.     Is that correct?

So the main problem is the circuit *doesn't* act like a perfect "block diagram" phaser.     The Axe-Fx III does act like a perfect block diagram phaser.   You can't just tweak the feedback to move the notches of the Axe Fx III to make it like the circuit.   If you move the notches you will change something else, like the level at the peaks or the levels at high and low frequencies. 

The reason the *circuit* doesn't act like a perfect "block diagram" phase is because adding the 24k resistor upsets one of the all-pass filters.   With that resistor in place the circuit behaves slightly differently.   The hard thing to understand is separating feedback from the addition of the resistor because the circuit links the two aspects.     Adding feedback moves the notches but adding the resistor *also* affects the notches (and other aspects).   The way to see the effect of the 24k with *no* feedback is to connect the 24k resistor to ground.  In spice you can see this case is not the same as removing the 24k altogether.   While there is no feedback in either case, the notches are moved.  If you look close at the 24k case there is some high frequency boost and the depth of the upper notch is reduced.

So the short story is for an accurate Phase 90 emulation with feedback you need to consider a lot more stuff.  On the other hand you can simplify the goals and set the amount of feedback so the level of the peak is the same as the circuit and you just live with the notches being moved.  You could scale the frequency range so the lower notches match-up as close as possible.

Another finer point is at what point the feedback mixes back into the circuit.   For the Phase-90 the feedback mixes back to the second opamp.  From a block diagram point of view this is roughly equivalent to mixing a negative feedback *after* the second opamp (and before the third opamp).   I don't know how the feedfback is implemented on the Axe-Fx III.   If the feedback connects back to the input of the first opamp, ie around the whole bank of all-pass filters, then the feedback will behave quite differently to mixing it midway along the all-pass bank like the Phase 90.

So there's a lot of places where you can get wrong results!


So here's a pic of the different phaser configurations for a 4-stage phaser:
VINTAGE = no feedback resistor at all
MODERN_FBZERO = feedback resistor grounded
MODERN_WFB = feedback resistor (24k) in normal feedback connection
IDEAL_WFB      = ideal block diagram where feedback connects to a summing node; feedback level -10/24
(The ideal feedback case with Zero feedback will give the same plot as the VINTAGE.)