Pedal Chain Global Feedback (thought experiment)

[Matthew Sanford]

So, you know I think on crazy, probably useless, things, like have a phase switch and being able to put pick ups in any order, things like that.

Recently I was thinking on global negative feedback, the linearizing effects, how it can remove things, etc., and wondered what it would be like to set up a pedal that you could plug the guitar in and out to amp, with voltage followers for each end and a switch to put the first one's negative feedback from itself, or the output buffer, sort of like this



I am wondering if you plugged your effects chain in the middle of that, would global feedback make much change to the sound coming out? Would it take the resonant peaks of the flanger and remove them from the signal at the start? Or would it really not do much of anything that could be workable?

I was thinking on the SVF in druid's FilterFX and how it changes depending on where the feedback goes to, so felt maybe something similar could happen this way, but I suppose it could just delete the whole signal too...?

Sore-E, just more crazy talk from yours truly...

[Rob Strand]

In principle the idea works but there's a whole lot of caveats.

You would probably want the phase of the forward path to the non-inverting to work as planned in many cases.  It's not a requirement but if you overlooked it things would behave quite different.

A big one would be loop stability.   As drawn the feedback loop would be unstable in a lot of cases.  So a modification would be to add a feedback resistor from the output of the first opamp, then add a resistor from the overall feedback point to the opamp - input.   That will weaken the feedback, achieve what you want to achieve, and avoid a lot of cases of instability (not all).   For a simple case perhaps look at minimal feedback where the two resistors are equal value.

With frequency dependent circuits the overall feedback will generally weaken frequency response.   If you have a treble boost it will reduce the amount of treble boost.   If you gave treble cut it will reduce the amount of treble cut.   The two cases are not affect equally though.   The amount of boost will be reduced more than the amount of cut.  For peaking EQ I'd expect the cut to become more narrow and the boost to become wider; the amount of boost/cut similar to the treble boost/cut.

If you put a non-linear circuit like the TS-9 in there then the filtering effect will be reduced.   It reduces in a manner which pushes the cut-off down.   The low-pass effect is pushed upward.  However when the TS-9 clips the loop gain is reduced due to the clipping and and high-pass effect moves back up and the low-pass effect back down, so it behave more like the original pedal.   There are some overdrive pedals which use this feedback effect.  (A slight complication is the TS-9 adds gain to the forward loop and is affected by the level control.)

For flangers and phasers you will see similar behaviour to when normal feedback is added.  However, it's a little different because you have the clean signal getting fed back as well.    As a rough guide it will affect response similar to EQ.   In these cases you have to deal with feedback being in and out of phase so that is likely to limit how much feedback you can apply.

Hard to cover all the cases but that probably covers the main points.

[Matthew Sanford]

Rather a simplified drawing, that was. I was almost thinking to draw 1M feed back resistors, but left it plain to start... then was also thinking making them inverting instead of non-inverting, possibly variable gain for tweaking that response as well. Do you think inverting would have any real difference to it, since both with negative feedback, both ends would be same...?

I was also thinking that if it's taking a full chain of effects, there should be interaction between all of them, right? The delayed signal from a delay would then create havoc in a later part of the song, flanger may make the distortion have filter additions, etc.

I suppose it would be a simple thing, only way to really know the outcome is to try it! Maybe this weekend. IF it is a bit of something but a destroyer, then maybe just a momentary switch to add at specific moments...

[Rob Strand]

Rather a simplified drawing, that was. I was almost thinking to draw 1M feed back resistors, but left it plain to start... then was also thinking making them inverting instead of non-inverting, possibly variable gain for tweaking that response as well. Do you think inverting would have any real difference to it, since both with negative feedback, both ends would be same...?

I suspect so.  You can see even with a simple phaser the feedback sign has a significant effect,
https://www.diystompboxes.com/smfforum/index.php?topic=132500.msg1290670#msg1290670

I was also thinking that if it's taking a full chain of effects, there should be interaction between all of them, right? The delayed signal from a delay would then create havoc in a later part of the song, flanger may make the distortion have filter additions, etc.

It's definitely different, especially with things like a delay with distortion.   I'd be thinking are the differences we hear just due to changes in filter/voicing cause by the overall feedback and the secondary effects are just complications in our mental picture of the signal flow.

A phaser or flanger with it's own feedback and the global feedback are very similar.  The difference is the dry signal is part of the feedback loop so that means the global feedback case will mess with the dry level.  However you can do that by changing the amount of dry signal mixed at the output of the basic pedal.    So the overall feedback isn't really adding much more than normal feedback; with normal feedback you can choose the best tap and polarity to feedback!  What it might do is when you adjust the overall feedback level it controls the dry level in a usable or appealing way.

I suppose it would be a simple thing, only way to really know the outcome is to try it! Maybe this weekend. IF it is a bit of something but a destroyer, then maybe just a momentary switch to add at specific moments...

Yes definitely.   Lots of variables to play with.  Especially things like the actual level setting of the overdrive.   That's going to have a different effect to a level control post the feedback point at the output.   If the loop goes unstable it might be possible to add filter or compensation to the feedback loop to put it on track.

[Matthew Sanford]

I had taken avid interest in that phaser post, seemingly an even number of stages being the sensible way. For the current conundrum, I realized it doesn't have to wrap the full chain, or possibly two set to different points (beginning coming back from 3rd pedal, final 5th feeding back to second, which adds the buffers between different effects too).

I guess it deserves a breadboarding, figure pots for all feedback resistors to best see what's what. And I could put an extra non-inverting after everything and mix the input dry signal to the inverting input to remove some dry signal, couldn't I?

[R.G.]

What Rob said.
The perversity of feedback is that any time the sum of phase shifts in the forward path (everything other than the feedback) hits a phase that is 0 degrees, 360 degrees, or any integer multiple of 360, and the total gain from the input, through the signal chain, and back around the feedback loop is greater than 1, it oscillates. As Rob says, this limits how much gain you can stand in the forward path before oscillation; and it does odd things to frequency filters - narrowing pass bands, introducing notches and peaks and so on, even if the overall effect is still stable. 

I'll have to think a bit about what a high gain distortion pedal does in this kind of setup. For signals beneath clipping - like hiss and rumble - the clipper doesn't limit the signal, so hiss and noise can build up at the forward gain of the distortion until they do hit the threshold. Since you're only mentioning a signal chain, does that imply that you might let yourself swap in any arbitrary pedal in the forward chain?

[Matthew Sanford]

Yes, any pedal. But we will see what works and not, distortions may not be cut out for it.

[R.G.]

Go for it! Let us know how it works!

[Rob Strand]

I had taken avid interest in that phaser post, seemingly an even number of stages being the sensible way. For the current conundrum, I realized it doesn't have to wrap the full chain, or possibly two set to different points (beginning coming back from 3rd pedal, final 5th feeding back to second, which adds the buffers between different effects too).

No doubt there's going to be groups which work together.

For phasers we usually see even numbers stages.  That's so the low frequencies and high frequencies have the same phase, typically some boosting.   You can use odd numbers of stages, it does work.  With the common all-pass stage it inverters at low frequencies but not at high frequencies.  With an odd number of stages it cuts the lows and boosts the highs, giving it a brighter voicing.  However if we also add an inverter to the all pass path the low frequencies are now in phase and it will boost the lows and cut the highs giving it a darker (maybe vintage) voicing.   With even numbers of stages it doesn't matter what the highs and lows so phase-wise because the even number of stages make the phasing the same at each frequency extreme.

I guess it deserves a breadboarding, figure pots for all feedback resistors to best see what's what. And I could put an extra non-inverting after everything and mix the input dry signal to the inverting input to remove some dry signal, couldn't I?

Sure.   When you start with these types of experiments it's best to have a lot of options then tinker around to see what works and what doesn't.   When I did this with phasers years ago I built a VST plug-in to explore all the options and the conclusion was only a few configurations had something to offer.

[Matthew Sanford]

Sure.   When you start with these types of experiments it's best to have a lot of options then tinker around to see what works and what doesn't.   When I did this with phasers years ago I built a VST plug-in to explore all the options and the conclusion was only a few configurations had something to offer.

I kind of figure that will happen, I'm just hoping that indeed there will be a few - like when I get to seeing pickups in all different series/parallel/phase options, I'm sure much will be trash. But this idea seems a bit madcap, which makes it very interesting to me, just have to know, you know. I figure if it makes any kind of sense then I may want to build it, but have an option for it to send the end of chain signal either as feedback to the first or out to another chain (while still having the output from the effects, so 2 outs)... and maybe feed something else in through the negative feedback. A sort of inside/outside switch.

So it goes on the mental list, hoping I get through the rest well in time to try it this weekend! (shouldn't be hard, just a dual op amp, four jacks, a SPDT, and some pots...)