Negative Feedback in Pedals (e.g. Honey Bee Overdrive)

[YouAre]

I've been dipping my toes into amp design, and have made my way to Negative Feedback. I don't see it being used all that often in pedals (specifically across multiple stages).

For reference, below is the Madbean Yellow Shark project. I've used the Yellow Shark/Honey Bee because it utilizes 2 different kinds of feedback that we can analyze.

http://www.madbeanpedals.com/projects/YellowShark/docs/YellowShark.pdf

Looking at R7 & C7, it seems to feedback the hard clipped signal back into the negative input on the op amp. After that, C8 & R9 feeds back the signal after it's been filtered by R8/C9 in the same way is the previous stage. Is this an example of negative feedback? My gut is telling me that it is. I'm guessing the effect is probably consistent with what you'd expect from negative feedback (increased bandwidth and smoothed out "eq spikes" at the expense of some output level). The Yellow Shark is a fairly low output pedal, so that could make sense. I'm very likely over simplifying, so I'd love if anyone could comment on this feedback.

The second type of feedback is the "Nature" control. On the surface, it seems to simultaneously adjust the transistors bypass cap (affecting the gain/frequency response of the output stage) and the opamp's bass response. While this is happening, we've always got 50k connecting the non-inverting output of the transistor to the inverting input of the opamp. This seems like it could be a case of (global) negative feedback. What inclines me to question that assumption is that the middle lug goes to ground. I'm not sure if I should treat the potentiometer as 2 separate adjustable resistors, or as 2 adjustable resistors and 1 static resistor. Can anyone offer an insight to this?

Thanks!

[GibsonGM]

What I'm seeing (correct me if I'm wrong) is a non-inverting output of the opamp, followed by a non-inverting output from the JFET.   Fed back to the inverting input, yes, this would oppose the signal in a NFB loop.    There's a little bit of filtering going on, too, so you'll have emphasis on some frequencies over others...and the Nature control seems to be able to ground out the NFB, and cancel it?

I don't THINK the transistor has a bypass cap as such (not grounded thru most of the pot's rotation), but when the "Nature" control IS grounded...then it does!  Hmm, kind of cool little loop going on, for sure!  It's doing at least three things, from where I'm sitting. 

[YouAre]

What I'm seeing (correct me if I'm wrong) is a non-inverting output of the opamp, followed by a non-inverting output from the JFET.   Fed back to the inverting input, yes, this would oppose the signal in a NFB loop.    There's a little bit of filtering going on, too, so you'll have emphasis on some frequencies over others...and the Nature control seems to be able to ground out the NFB, and cancel it?

I don't THINK the transistor has a bypass cap as such (not grounded thru most of the pot's rotation), but when the "Nature" control IS grounded...then it does!  Hmm, kind of cool little loop going on, for sure!  It's doing at least three things, from where I'm sitting. 

Seems like you agree with analysis. But, I think the transistor is bypassed. If you're familiar with the Orman MOSFET boost, it's gain control is set by adjusting the series resistance of the bypass capacitor. The nature control seems to do just that at one end of its rotation. At the opposite end of it's run, it almost performs a similar function to the Proco RAT Ruetz mod, maybe? Clever stuff.

Can anyone else comment on the feedback aspect?

[teemuk]

I've been dipping my toes into amp design, and have made my way to Negative Feedback. I don't see it being used all that often in pedals (specifically across multiple stages).

Actually, you would have hard time finding a circuit that doesn't employ negative feedback. Yes, GLOBAL feedback might be rare but LOCAL feedback can be found from almost any active circuit. ...at least in some form.

Looking at R7 & C7, it seems to feedback the hard clipped signal back into the negative input on the op amp. After that, C8 & R9 feeds back the signal after it's been filtered by R8/C9 in the same way is the previous stage. Is this an example of negative feedback?

Yes, but both are very complex examples. A simpler example is just the resistive divider between output and inverting input.

I'm guessing the effect is probably consistent with what you'd expect from negative feedback (increased bandwidth and smoothed out "eq spikes" at the expense of some output level)

The circuit uses very complex feedback schemes, which have several other purposes besides archetypal negative feedback properties (increased badwidth, lower output impedance). Mostly it's all about gain controlling, but this control is made amplitude dependent with two different diode+NFB configurations (clipping both peaks and crossover regions of the signal) as well as frequency dependent with RC filters. Mostly utilising feedback is all about that (gain control).

I think you still have plenty of reading left concerning the topic, though.

IMO, you chewed a little bit too much with this example and explaining the different effects of different feedback schemes found from that circuit would be a pretty extensive topic in its own right.

What is "smoothing out EQ spikes" BTW? That's not a generic characteristic of negative feedback in any way....

The second type of feedback is the "Nature" control.... While this is happening, we've always got 50k connecting the non-inverting output of the transistor to the inverting input of the opamp.

No we don't. The potentiometer's wiper is grounded so there is no connection between JFET source and opamp inverting input. That is not a loop.

I'm not sure if I should treat the potentiometer as 2 separate adjustable resistors, or as 2 adjustable resistors and 1 static resistor.

It's two adjustable resistances, one always increases while the other decreases and vice versa. The other end of both resistances always terminates to ground so there is no loop for global feedback. The resistance at inverting input of the opamp alters the ratio of how much output signal gets fed back, the resistance at JFET's source alters how effectively the capacitance bypasses the source resistor.

By the way, the source resistor is also form of applying negative feedback because it generates a voltage drop and thus "mirrors" the signal at FET's gate to its source. Because gate is an inverting input and source a non-inverting input this voltage potential creates local negative feedback. People often call it "degenerative feedback" but it's negative feedback allright. Capacitively bypassing the source resistance naturally lowers the source resistance at AC signals and consequently also decreases the amount of negative feedback at such.

Since these all are more or less RC circuits (resistors and capacitors) they are essentially filters and thus gain variations from feedback are largely frequency dependent. (e.g. capacitor has higher impedance at lower frequencies and lower impedance at higher frequencies).

[PRR]

> we've always got 50k connecting the non-inverting output of the transistor to the inverting input of the opamp

Not if the wiper is grounded. Which it is. (as others noted while I was thinking)

If the ground is perfect (or good-enough), it is TWO resistors. Maybe 25K and 25K, maybe 1K and 49K, or 49K and 1K. But not "global" feedback.

One way it extends the bass of IC1. The other way it jacks-up the treble of Q1. (There's a similar dingus in my brother's 1941 Plymouth radio.)

> R7 & C7... expect from negative feedback (increased bandwidth

Work-out the reactance of C7 at various frequencies. By eye, it is "large" at 80Hz and "small" at 8KHz. I think it rolls-OFF treble--- makes frequency response "worse". (But gooder for our uses: naked diode clipping can be ear-searing.)

CA3130?? Man, flashbacks of the 1970s.

[YouAre]

Wow! Thank you for the intensely comprehensive answer!

::cracks knuckles and pulls out keyboard::

Actually, you would have hard time finding a circuit that doesn't employ negative feedback. Yes, GLOBAL feedback might be rare but LOCAL feedback can be found from almost any active circuit. ...at least in some form.

I didn't explain myself well, I apologize. I wanted to look at the global feedback between stages with their own local feedback (opamps/transistors)

The circuit uses very complex feedback schemes, which have several other purposes besides archetypal negative feedback properties (increased badwidth, lower output impedance). Mostly it's all about gain controlling, but this control is made amplitude dependent with two different diode+NFB configurations (clipping both peaks and crossover regions of the signal) as well as frequency dependent with RC filters. Mostly utilising feedback is all about that (gain control).

Can you explain more about the gain control being amplitude dependent? We've got hard clipping before the signal is tapped for feedback, so my inclination is to believe that the fed back signal's amplitude is fixed (dependent on diode chosen).

I think you still have plenty of reading left concerning the topic, though.

Ouch. But I'd love to be pointed in the right direction. Most of my reading, as I'd mentioned, is with regards to amp design. If you can point me towards anything more comprehensive, I'd be extremely grateful.

IMO, you chewed a little bit too much with this example and explaining the different effects of different feedback schemes found from that circuit would be a pretty extensive topic in its own right.

Gluttony is my second favorite Deadly Sin. That was my purpose of this thread. I picked this pedal because it has multiple schemes.

What is "smoothing out EQ spikes" BTW? That's not a generic characteristic of negative feedback in any way....

It's very typical for the Global NFB in a guitar amp's poweramp section, which I should've clarified. As the NFB is reduced (series resistance is increased, or shunt resistance is reduced) you'll notice more "peaks." The bandwidth is reduced, but we barely notice it because of the heavy filtering of the guitar amp speaker. We will mainly perceive a significant increase in volume, because the frequency response becomes considerably less even and we get the very loud "peaks." 

No we don't. The potentiometer's wiper is grounded so there is no connection between JFET source and opamp inverting input. That is not a loop.

It's two adjustable resistances, one always increases while the other decreases and vice versa. The other end of both resistances always terminates to ground so there is no loop for global feedback. The resistance at inverting input of the opamp alters the ratio of how much output signal gets fed back, the resistance at JFET's source alters how effectively the capacitance bypasses the source resistor.

Man! The 50k connection has always seemed funny to me! I thought that the grounded wiper effectively breaks the resistance when I looked at it. But reading the Madbean document, it says "Additionally, C5 and C4 are connected through a constant 50k resistance regardless of the position of the control" under the Nature control's description. That is where the incorrect assumption came from, and what through the monkey wrench in my system. It seemed very fishy to me, and I'm glad I asked for you guys to check my assumptions.

By the way, the source resistor is also form of applying negative feedback because it generates a voltage drop and thus "mirrors" the signal at FET's gate to its source. Because gate is an inverting input and source a non-inverting input this voltage potential creates local negative feedback. People often call it "degenerative feedback" but it's negative feedback allright. Capacitively bypassing the source resistance naturally lowers the source resistance at AC signals and consequently also decreases the amount of negative feedback at such.

Since these all are more or less RC circuits (resistors and capacitors) they are essentially filters and thus gain variations from feedback are largely frequency dependent. (e.g. capacitor has higher impedance at lower frequencies and lower impedance at higher frequencies).

I understand that the FET is utilizing it's own form of negative feedback. This ties back into my ignoring the effects of local NFB for the purpose of this exercise.

I also understand that these applications of NFB are essentially filtering the signal, somewhat similar to how an amp's Presence control would. Not entirely the same, but similar in concept I suppose.


I would also like to take this time to thank you again, and to take a break to say that none of what I typed in my initial post was a statement of fact. I wanted to lay out my understanding on the table plainly, and have everyone whittle at it to remove the fallacies. You've done this marvelously. Thank you.

[YouAre]

Not if the wiper is grounded. Which it is. (as others noted while I was thinking)

If the ground is perfect (or good-enough), it is TWO resistors. Maybe 25K and 25K, maybe 1K and 49K, or 49K and 1K. But not "global" feedback.

Yes, thank you. Again, I'm very happy that I questioned the statement (or at least my understanding of the statement) from the original document.

One way it extends the bass of IC1. The other way it jacks-up the treble of Q1. (There's a similar dingus in my brother's 1941 Plymouth radio.)

It's damn clever. It's a pretty badass "tone" control, that relies on the internal filtering of stages rather than applying any additional filtering.

Work-out the reactance of C7 at various frequencies. By eye, it is "large" at 80Hz and "small" at 8KHz. I think it rolls-OFF treble--- makes frequency response "worse". (But gooder for our uses: naked diode clipping can be ear-searing.)

Any increase in bandwidth isn't likely to be noticeable because of the intense filtering of guitar amp speakers, correct?

Interesting note about the effects of C7 (and possible C8?). How would we perform the opposite, bass roll off?

CA3130?? Man, flashbacks of the 1970s.

It's a fun little opamp. Does it have a low slew rate? I'm trying to understand why this opamp was chosen specifically over something like the lm308.


Thank you for the help!

[teemuk]

It's very typical for the Global NFB in a guitar amp's poweramp section, which I should've clarified. As the NFB is reduced (series resistance is increased, or shunt resistance is reduced) you'll notice more "peaks."

I think I understand what you are trying to imply, but that particular characteristic actually is an effect of negative feedback decreasing amp's output impedance.

That decrease leads to increasement of amp's damping factor and that is basically another way to state that the amp has more linear frequency response to reactive loads like loudspeakers. With more linear response the "peaking" at ascending and descending edges of the waveform is reduced.

Ever seen what, for example, a hi-pass filter does to a clipped waveform? It makes the edge of clipping more "peaky", transient. The low-pass is simply reducing the low frequency components within the "flat" top of the clipped wave and from waveforms perspective it looks like a transient on the rising edge. This is what the "peaking" you mention is about. It's basically just an issue of frequency response and how flat or unflat it is. Make the frequency response flatter and the transients disappear or at least decrease in amplitude.

If you have a waveform with transients in it then introducing some NFB doesn't magically remove them. That is what I'm meaning. If you have an amp with high output impedance, which - because of that - has a very non-linear response to reactive loads (like loudspeakers) and hence outputs waveforms that have transient edges THEN negative feedback can be used to decrease output impedance, which then leads to more linear response to reactive loads like loudspeakers and consequently to less transient content in the signal.

[YouAre]

I think I understand what you are trying to imply, but that particular characteristic actually is an effect of negative feedback decreasing amp's output impedance.

That decrease leads to increasement of amp's damping factor and that is basically another way to state that the amp has more linear frequency response to reactive loads like loudspeakers. With more linear response the "peaking" at ascending and descending edges of the waveform is reduced.

Ever seen what, for example, a hi-pass filter does to a clipped waveform? It makes the edge of clipping more "peaky", transient. The low-pass is simply reducing the low frequency components within the "flat" top of the clipped wave and from waveforms perspective it looks like a transient on the rising edge. This is what the "peaking" you mention is about. It's basically just an issue of frequency response and how flat or unflat it is. Make the frequency response flatter and the transients disappear or at least decrease in amplitude.

If you have a waveform with transients in it then introducing some NFB doesn't magically remove them. That is what I'm meaning. If you have an amp with high output impedance, which - because of that - has a very non-linear response to reactive loads (like loudspeakers) and hence outputs waveforms that have transient edges THEN negative feedback can be used to decrease output impedance, which then leads to more linear response to reactive loads like loudspeakers and consequently to less transient content in the signal.

Bah! You're absolutely right! I completely ignored the effect of speakers! I was foolish to make assumptions and remove variables. Thank you for correcting that really fallacious line of thinking and oversimplification.


Going back to what you mentioned earlier, the complex examples of the feedback from C7/R7 and C8/R9, are you able to comment on their effects on the signal coming from the opamp?

Thanks again!

[PRR]

> As the NFB is reduced (series resistance is increased, or shunt resistance is reduced) you'll notice more "peaks."

Most audio amps with simple loads:

A) Don't have peaks, -or-
B) Have deliberate peaks which may-or-may-not change with gain (NFB).

Large change in NFB *will* change frequency response (unless compensated). Here's an extreme case: amp gain can vary from 0dB to +60dB. At high gain, the bass/treble limits move in. From 1Hz-1MHz inward to 50Hz-50KHz.



That was supposed to be Full Range, at least up to fairly high gain (~~45dB). There are fuzzes which start with less bass+treble, and DO narrow-up to a semi-peak at 1KHz. This design would too, if taken to 90dB(!!) gain.

> actually is an effect of negative feedback decreasing amp's output impedance. ... ... loudspeakers.

True; but this YellowShark isn't driving a loudspeaker directly. Yes, voltage-feedback reduces the variation of voltage seen at the loudspeaker; that does not seem to be this thread's question.

> C7/R7 and C8/R9, are you able to comment on their effects on the signal coming from the opamp?

No, not without more time than I care to put in. Don't forget R8 C9. And R5 C2, not to mention C4 and the half-pot. It's complicated.

[YouAre]

> As the NFB is reduced (series resistance is increased, or shunt resistance is reduced) you'll notice more "peaks."

Most audio amps with simple loads:

A) Don't have peaks, -or-
B) Have deliberate peaks which may-or-may-not change with gain (NFB).


> actually is an effect of negative feedback decreasing amp's output impedance. ... ... loudspeakers.

True; but this YellowShark isn't driving a loudspeaker directly. Yes, voltage-feedback reduces the variation of voltage seen at the loudspeaker; that does not seem to be this thread's question.

I understand that I mistakenly believed that the effects of Global NFB would in guitar amps and their interaction with speakers would apply directly to pedals. I remember that the effects are a result of changed output impedance. It was a silly mistake and had to do with the fact that I simply forgot some of the key concepts.

> C7/R7 and C8/R9, are you able to comment on their effects on the signal coming from the opamp?

No, not without more time than I care to put in. Don't forget R8 C9. And R5 C2, not to mention C4 and the half-pot. It's complicated.

Understood. It does seem like it would be fruitless to explain a complex example to someone with a flawed understanding of the principle concepts, but I would greatly appreciate if you could point me in a general direction of what concepts to look into. I plan on testing the design by switching the feedback in and out, viewing the results on a scope, and listening to the results as well. Unfortunately, I would only get a qualitative understanding of the differences at best. I would love if I could read about the underlying concepts, so that I might actually understanding the reasoning behind what my eyes see and what my ears hear.

Thanks again for all your help.