Switchable Feedback Between Phaser Stages

[aleks_tedstone]

A completely unoriginal idea but I couldn't find an implementation of it on here, although I suspect there's one somewhere buried amongst all the usual Script/EVH mods of the Phase 90. The excellent analysis on https://www.electrosmash.com/mxr-phase90 got me wondering whether anyone here has demonstrated (and produced audio clips) putting the feedback resistor to each all-pass filter stage in the Phase90? See section 5.4 in the link for simulations of the notches produced by making this change. The way it changes the respective depths of the two notches produced in this 4-stage phaser seems like it would have a very distinct impact on the character of the effect.

It would be easy to implement with a 3-throw rotary switch or SP3T, which I have drawn onto Electrosmash's diagram of the phasing stages below:


It's clearly that some people find the 24K feedback resistor too much at going from stage 4 output into stage 2 input, so a further mod (on top of the switch) would be turning this into a pot/trimmer for additional tweaking delight. Already a well known variation in the EVH edition, so a logical extension of this experiment...

Presumably the good folks who added this into later iterations of the Phase 90 tried it in all positions and settled on stage 2, but I am curious! If I ever make it I will post it here.

[Rob Strand]

A long time ago I did a DSP phaser where you could set the stages, the output tap, the feedback points and also the phase of the feedback.   By phase of the feedback I mean you invert the phase of the signal that feeds back.   I went a bit further in that I could also select what tap I took the output from, so I could have an 8-stage phaser with selectable feedback but only take the output off the forth all-pass filter like a 4-stage phaser.

To make an obvious statement: there's only a few combinations which (to me) sound good/natural.  The take home message is having a lot of options isn't always good.   What the DSP phaser experiment allowed me to do was to weed out the good combinations without having a zillion useless options.   There was a simple pattern to the good sounding combinations but don't ask me what it is.  With a 4-stage phaser the number good options is quite small, even if you have the option to invert the feedback.

Something that often falls between the cracks with the MXR type feedback circuit, is the act of adding feedback to the all-pass stage stuffs-up the way the all-pass filter works.   Ideally you want to do the feedback mixing before the first stage with a mixing circuit  and that way the all-pass filters are not corrupted by the feedback connection.   In your circuit the feedback switch would select the tap of the all-pass filter and take that back to the common mixing point at the input.   Some phasers actually have an input mixer for feedback, IIRC the Boss units.   I'm not saying the MXR is bad, it's a simple mixing method but not necessarily the theoretically correct way to do it.  I discussed this issue a few years back in this thread.  See the plots at the end,

https://www.diystompboxes.com/smfforum/index.php?topic=122944.0

[aleks_tedstone]

Thanks Rob, your response and the linked discussion are very interesting! I thought there was probably a reason why I hadn't seen anyone bother with this, and I certainly hadn't caught the subtlety you mention about the phase of the feedback. Did you ever make a real life version of your favourite DSP version?

Feedback/ resonance/ regen are such a core part of the character of any given flanger that I was just interested to see what possibilities existed with phasers. One of the first pedals I bought was an Ibanez AF2 that's based on the A/DA flanger, and knowing nothing about how any modulation effects worked at the time I was only able to use my ears. The ringing metallic sound with 'enhance' cranked was always intriguing, if a little hard to apply. 10 years on I'm leaning towards the 'less is more' school of thought but I do love tweaking.

[Rob Strand]

Thanks Rob, your response and the linked discussion are very interesting! I thought there was probably a reason why I hadn't seen anyone bother with this, and I certainly hadn't caught the subtlety you mention about the phase of the feedback. Did you ever make a real life version of your favourite DSP version?

The motivation for the DSP version was to see what was useful/useable.  The DSP set-up also had number of waveforms for the LFO and ways to bend the LFO waveform.   The idea was to "answer all the questions" regarding phasers.    It's very easy to make complex systems in DSP which would be quite unwieldy in the circuit world.   It's also a real bummer making a massive circuit to find out it doesn't offer much.

After playing around with the DSP version it pretty much detracted me from building a complex phaser.  The complexity didn't offer much that that I liked.   Once you get over 6 stages it starts to sound like a flanger.   The different feedback connections sounded different but usually only one or two connections sound good.   The LFO shape was very important in shaping the feel of the unit.   The hyper-sine or hyper-triangle waveform definitely sounded more even.   However, that didn't mean an uneven sounding one was bad.

In short 4 and/or 6 stages does the job.  Throw in one feedback connection if you like, perhaps with adjustable level.    Beyond that you can play with adjustable LFO depth or play with the mix of the clean and vibrato paths.    You can see why most phasers follow this pattern.

One of the Boss Phasers has an option to switch in a lot of stages.   If you really wanted that sound it would be easier just to buy one.

Perhaps the best step-up from the phaser is not to make a complex phaser but to add Leslie characteristics.

[j_flanders]

Although not a Phase90, you could check 'The Worm' by EHX.
Re-routing the feedback it gives: phaser, vibrato, tremolo or Wah.

[anotherjim]

I'm pretty sure Mark Hammer went through this experiment here but I'm not sure it involved any particular project. He usually reads anything with "Phaser" involved so hopefully he might chime in 

[Mark Hammer]

You guessed right, Jim.  Although let me be clear: I am NOT "markphaser"!!! (a name wholly unfamiliar to those who haven't been here for 15 years or so)

But let's distinguish between DSP and analog.  With analog stages, there are two basic problems/challenges to be addressed:
1) cumulative noise - More stages=more noise-risk, and recirculating noise will get you even more noise.  So taking the signal from stage 8 of an 8-stager and feeding it back to stage 2, will take all the cumulative noise of 8 stages and recirculate it through 7 of them.  That's why one frequently sees one of the stages in anything with >4 stages configured as a sort of lowpass.  A cap will be added in the feedback loop to trim cumulative hiss.  If one is aiming for >4 stages, make an effort to have those stages as hiss-free as possible.

2) oscillation - Allpass stages are supposed to be unity gain.  But even small tolerances in the resistors used can result in small amounts of gain.  For example, if the feedback resistor is 10.05k and the inverting input resistor is 9.95k (both within the 1% spec), that gives us a gain of 1.01.  Not much to speak of on its own, but now consider what you might end up with using 5% resistors, and 8 stages (though some stages could be <unity and others >unity, just be chance).  And add to that what would happen if the cumulative gain applied to the output of those 8 stages is now fed back and recirculates through a bunch of them. 

BUT whatever I knowledge I lack of DSP (and that's a LOT), I do know that neither of these more-stages issues applies when one moves from the analog to digital domains.  Besides, DSP emulation of more stages and more feedback should only pose mental-health risk to the programmer, and not to the signal.

But while you're here, consider the orange Ross Phaser.  Interesting design that takes it's output from four swept stages, but applies the feedback through a fixed fifth stage.  So the feedback behaves AS IF the unit was a 6 stage phaser, feeding back from the 6th to the 2nd stage.  I've never heard one, but the concept is interesting.

[aleks_tedstone]

Wow thanks everyone - really interesting!

There must be a million plug-ins by now that let you do these kind of experiments with multistage phasers, but the only pedal I can think of off the top of my head is the Dwarfcraft Ghostfax, must be digital given how small the box is... It has 32 stages and you can pick any multiple of two, with adjustable feedback. No idea how that's routed though... I think all Dwarfcraft stuff is out of production now, but the demo videos are pretty thorough, and lead me to agree with Robs assertion that more than 4/6 stages isn't very useful in most guitar contexts.

A friend of mine who is very into his synths has an EHX Worm that he was interested in circuit bending (his words not mine!) and he knows I like to fiddle around with pedals so maybe I'll open that up and identify some mod possibilities. Looking at the schematics it should be easy to put a different LFO in and interface it by the LDR/LED's, so maybe some possibilities for hypertriangle/saw waveforms. I'm amazed how many options that feedback rerouting switch offers, thanks for bringing it up, looking forward to hearing how it sounds stock before opening it up.


How might you add Leslie characteristics? A notch at lower frequencies with a different LFO speed? Would I be right in saying that the spinning speaker creates vibrato too via the Doppler effect? And a bit of tremolo?

[j_flanders]

I'm amazed how many options that feedback rerouting switch offers, thanks for bringing it up, looking forward to hearing how it sounds stock before opening it up.

Questions, suggestions, ideas and (factory)schematics in this thread: https://www.diystompboxes.com/smfforum/index.php?topic=123356

Although I rarely use them, I still have the Big Box version and also the XO version.
I think the pedal is an ingenious design/idea.

However, to me, it also fits the 'jack of all trades, master of none' description. But that could just be me wanting more pronounced effects in this pedal. Everything is too subtle to my taste.

What is not so subtle is the noise/hiss. Probably the noisiest pedal I ever tried.

I think the original big box version also has a design flaw, resulting in misbiased transistors, severely reducing the headroom. It would explain why so many people complained how easily this thing distorts.
They fixed it in the XO-version. The red parts is what they changed in the XO vs the big box:

Schematic above is a partial/simplified version without all the switching and feedback routings.

[aleks_tedstone]

However, to me, it also fits the 'jack of all trades, master of none' description. But that could just be me wanting more pronounced effects in this pedal. Everything is too subtle to my taste.

What is not so subtle is the noise/hiss. Probably the noisiest pedal I ever tried.

Subtle vibrato and tremolo I like, but to me using a phaser 'tastefully' seems like wasted potential, and if my wah doesn't sound like it's bearing messages from the psychedelic void count me out. That being said, your thread on the worm highlights some interesting possibilities for simultaneously switching out C6,C8 and C10 for moving the notch around, even if it is an inherently subtle effect. Sticking some dirt in front usually does the trick but lets not get into the pedal order debate here. Cool that EHX try and pack so much into their pedals, especially this one which has an expression jack in its current version.

Did you try creating links directly between the emitter of stage 3 back into the base of stage 1? Presumably this would need a resistor to avoid going over unity gain as discussed below, although the circuit bending philosophy might align with this!