Feedback Loops & Buffers

[Antero]

So, I built this feedback/true-bypass loop.  It's pretty cool.

But.

It also sucks tone and volume; I guess that makes sense, really, because the feedback knob is controlling how much of the signal is shunted out through the loop, so it's splitting the signal... so I figure, I need a buffer.

What would be the right sort of buffer to use in such a device?  Is it just a clean boost, or is there more magic to the buffer than that?

[Elektrojänis]

Most clean boost circuits will probably work. If you dont need the gain you could use a simple buffer like one of the circuits on this article from AMZ: http://www.muzique.com/lab/buffers.htm

The difference between booster and buffer is thet a buffer does not have voltage gain. It simply presents a high impedance load to the input signal and has a low impedance output.

The bipolar transistor based stuff probably has a bit low input impedance if you are going to connect an electric guitar (with passive pickups) to the input. I would probably use a fet.

Depending on the configuration of your feedback loop you might need to use some resistors to mix the buffered signal to the feedback signal.

[Antero]

Thanks!

Hmm.  The feedback loop presents an entirely new set of problems... maybe I need a pre-gain and post-gain?  Just dropping a buffer in could, it seems, kick up the feedback signal in ways that could potentially be problematic.

I used NaBo's layout design:

http://aronnelson.com/gallery/NaBos-Layouts/NaboLoop

[Elektrojänis]

Hmmm... I have an idea how to use a buffer in there in the most efficient way... or at least how I think it would be most efficient. It took me a while to get my head around that picture as I'm more comfortabel with "old style" schematics than with wiring diagrams.

I'll try to draw a mod on that picture and put it somewhere temporarily so you can see it. I hope the original designer won't mind (i'll leave the original info in there and document what I changed).

[Elektrojänis]

Ok. Here goes: http://koti.welho.com/pjunno/temp/loopermod.gif

This is untested. I think it will work and this is how I would try it, but I can't be sure.

I hope the drawer/designer of the original will not mind. If anyone finds me posting this inapropriate, please reply here and I will remove the image as soon as possible. I won't keep it on the web for long anyway.

[Antero]

Hmm, innnnteresting!  Thanks!

[The Tone God]

You shouldn't just wave your hand and say use any buffer (impedance ? phase ?). Mixing with resistors may not always work depending on what effect you are hooking up too and the characteristics of the buffer you choose. I would think it would be better mix before the buffer then after or use another stage as a unity gain mixer.

Just a few of my thoughts.

Andrew

[Elektrojänis]

You shouldn't just wave your hand and say use any buffer (impedance ? phase ?).

Umm... Yeah. That's true. You are right. It should be non-inverting buffer with high input impedance. The basic JFET-buffer should work.

Actually... Where I suggested to put the buffer, the phase should not matter too much. It's only for the input signal and not in the feedback loop.

Mixing with resistors may not always work depending on what effect you are hooking up too and the characteristics of the buffer you choose. I would think it would be better mix before the buffer then after or use another stage as a unity gain mixer.

I assumed that it will be used with guitar FX. Most of them have input impedance of 100k or more as you probably know. Mixing with 10k resistors should not present too much signal loss if the next stage has an output impedance that high. Some fuzzes might have lower input impedance, but as the impedances at this point are mostly resistive, it would cause mostly volume loss but the basic tonality would not change. 500kohm feedback pot might actually be a problem... Lower value could probably be used if it works weirdly.

If you mix before the buffer, the output impedance (that is usually low compared to inputs) of the effect can load the pickups down if you turn up the feedback. I think this might have been the original problem. If not, then there should not be any problem (when using an effect that has high input impedance).

This is just how I think it works though. I still have not tried and I have some other projects to do before I can. And no... I'm not pro by any means, so if my reasoning is somehow faulty, I would like to know ofcource.

Another buffer after the mixing could be added if one wants to eliminate the potential problems from the low input impedance effects.

[The Tone God]

Umm... Yeah. That's true. You are right. It should be non-inverting buffer with high input impedance. The basic JFET-buffer should work.

Agreed.

Actually... Where I suggested to put the buffer, the phase should not matter too much. It's only for the input signal and not in the feedback loop.

Agreed. I made that statement to deal with various buffer options like the buffer and mixer option I suggested later in the post where it could make a difference. Just for others to keep it in mind.

I assumed that it will be used with guitar FX. Most of them have input impedance of 100k or more as you probably know. Mixing with 10k resistors should not present too much signal loss if the next stage has an output impedance that high. Some fuzzes might have lower input impedance, but as the impedances at this point are mostly resistive, it would cause mostly volume loss but the basic tonality would not change. 500kohm feedback pot might actually be a problem... Lower value could probably be used if it works weirdly.

If you are designning for yourself you can control the external conditions under which this circuit is being used but if you are planning on releasing this to the public then unfortunately making assumptions can be...well lets say bad.

I was not concern about the value of the mixing resistors but what the impedence of the driving load was going to be. I tend to design assuming the user is going to put the circuit into less then ideal situations so my thought was that you can mix the signal under your own internal conditions before sending it out to the real world. This would reduce some problems that could come back later on.

Another buffer after the mixing could be added if one wants to eliminate the potential problems from the low input impedance effects.

That was what I was referring to with "unity gain mixer".

Andrew

[Elektrojänis]

If you are designning for yourself you can control the external conditions under which this circuit is being used but if you are planning on releasing this to the public then unfortunately making assumptions can be...well lets say bad.

True. I have similar goals. This however was not meant to be a "finalized design". It was meant an idea/suggestion. I tried to point that out from the start. And I hoped someone would join in the discussion... And that hapened. Thanks!
 

I was not concern about the value of the mixing resistors but what the impedence of the driving load was going to be. I tend to design assuming the user is going to put the circuit into less then ideal situations so my thought was that you can mix the signal under your own internal conditions before sending it out to the real world. This would reduce some problems that could come back later on.

Yep... Thats good. However there is always a point where you just have to make at least some assumptions adout what is going to be interfaced. It's a balance between considering too much of untypical uses and assuming that the operating conditions are too easy. For example it would definitely be overkill to assume that any output of a device could be connected to a speaker cab. (Someone might try that.)

I actually thought that the hardest case would be something like a Fuzz Face. If you place a 10k resistor in series with the input, it works. It eats a bit signal but it still works quite good.

Other thing I considered is that with that one buffer it probably works quite nicely with more things than without.

That was what I was referring to with "unity gain mixer".

A simple opamp mixing stage (summing amplifier) would be even better otherwise, but they have one problem in this. Yes... They are inverting.

[The Tone God]

True. I have similar goals. This however was not meant to be a "finalized design". It was meant an idea/suggestion. I tried to point that out from the start. And I hoped someone would join in the discussion... And that hapened. Thanks!

And I fell for it. Damn you!!!

Yep... Thats good. However there is always a point where you just have to make at least some assumptions adout what is going to be interfaced. It's a balance between considering too much of untypical uses and assuming that the operating conditions are too easy. For example it would definitely be overkill to assume that any output of a device could be connected to a speaker cab. (Someone might try that.)

Well I could also design on the basis that someone is going to use this circuit to prop up the couch with the broken leg in the OT-Lounge but I won't. I'm not suggesting to design for every possible situation but it is not unreasonable that someone would using this with an effect that has less then either desirable or optimal operating conditions. The Fuzz Face is an example.

A simple opamp mixing stage (summing amplifier) would be even better otherwise, but they have one problem in this. Yes... They are inverting.

Yes the simplest opamp mixer stage is the inverting one so what can we do about it ? How about put an simple inverter stage afterwards to flip the signal again. Here's an idea, that inverter stage can be made a attenuation/amplifier stage so that issue with the pot can be resolved.

Andrew

[Elektrojänis]

And I fell for it. Damn you!!!

Sorry... 

Well I could also design on the basis that someone is going to use this circuit to prop up the couch with the broken leg in the OT-Lounge but I won't.

Well that's good princible for selecting enclosures. Err... Sorry for OT...

Yes the simplest opamp mixer stage is the inverting one so what can we do about it ? How about put an simple inverter stage afterwards to flip the signal again. Here's an idea, that inverter stage can be made a attenuation/amplifier stage so that issue with the pot can be resolved.

Yes... All true. But it will make the thing a lot more complex if you think that the original is practically a passive design (as far as the signal goes).

Another problem in putting opamps or other active stuff inside the feedback loop is that if they get driven hard they distort and change the sound. If the feedback is used to make things oscillate (like I quess it will) then the opamp might be driven to distortion. Depends on what you put in the loop. The change might be to better too though. This is also something to consider when trying to make it "foolproof".

Argh... This got me intrested about the feedback loop stuff again, and it will probably take a few months before I have time to build it.

[The Tone God]

Well it sounds like you'll need to do some playing around to see what works for you.

Have fun.

Andrew

[Antero]

This is extremely useful stuff for me.  I'm afraid I don't, in my general newbieness, understand quite what the resistors are doing in the diagram, or what the effects of different values would be.

[Elektrojänis]

The resistors balance the blending of the two signals so that one will not totally overpower the other. I'm not the best to explain why and how it works. It's about the output impedances of the circuits that are feeding the resistor mixer and about the input impedance that the resistor mixer is feeding. The resistors are kind of used for balancing the impedances of the outputs that are to be mixed.

Hmmm... My description is not so good... But anyway... Those resistors are how it is usually done.

Tha value of the resistors depends on the stuff arond them. 10kohm is an estimate only. It was based on assumption that most guitar effects have an output impedance of 1kohm or lower and input impedance of 100kohm or more. 10kohm is nicely in between those.

[birt]

i've noticed the same problem with a feedback loop. the pot has less resistance than teh effect in the loop so the signal doesn't go back trough the pot to the input of the effect but just goes trough the pot and bypasses the effect resulting in volume loss.

isn't it best to just put a buffer IN the feedback path so signal only goes in the opposite direction of the signal in the effect? or am i overlooking something here?

[stobiepole]

How about using the stereo micro-amp board from Tonepad and adding the feedback pot to take the signal from the second amp and putting it back into the first? I'm going to have to give it a go myself...

Chris