Why double diodes in feedback loops?

[Mark Hammer]

Of course they sound different. With diodes to ground the opamp clips itself. The differences are subtle though. Especially with same EQ design.

The op-amp CAN clip, itself, if the voltage swing of the chip is exceeded.  Granted, it doesn't take a *lot* of gain to accomplish that, but there is still some room between the point where a diode pair to ground on the output might begin to clip, and the point where what it begins to receive from the op-amp is already clipped.

For instance, if one had a pair of germanium diodes to ground on the output, they would begin to clip when the op-amp output started to hit around +/-300mv.  The voltage swing of the op-amp, using a 9V supply, would be around +/-3500mv.  So, unless the pickups were especially hot, given typical pickup output, one could apply a gain of around 10-20x in the op-amp, before one started to get clipping within the op-amp and reclipping via the diodes.  Of course, given that most op-amp-based overdrives and distortions are going to apply a lot more than a gain of 20x (e.g., the humble MXR Distortion+ is 213x gain at max), the zone between when there is ONLY diode clipping and when there is double-clipping is not all that big; say, the zone between 7:00 and 8:30 on the Gain/Drive control.  But I wanted to note that it DOES exist.

[Rob Strand]

It often seems like quantum mechanics and string theory are better understood than diode clipping by many in the guitar pedal community.

Quantum mechanics and string theory are only attempted by professionals so the majority of opinions come from professional and scientific people.   Even if they turn out to be wrong at least their approach is scientific, and more so, they can see and admit the error of their ways.

Audio is full of quacks and opinions, probably only the top 10% of people in the field actually understand how to apply a scientific approach at all.   A significant fraction just reiterate crap they read and they have no idea of the concepts behind it or how to *prove* the result themselves.

[GibsonGM]

Audio is full of quacks and opinions, probably only the top 10% of people in the field actually understand how to apply a scientific approach at all.   A significant fraction just reiterate crap they read and they have no idea of the concepts behind it or how to *prove* the result themselves.

Of note, amp manufacturers, who recycle old designs...which were not meant to be overdriven, but their legacy structure makes them prone to blocking distortion and other issues because their 'designers' have no clue what's ACTUALLY going on inside, how the amp will be used - they just copied them.  Big bypass and coupling caps, no grid stoppers and so on.  This is at the level of multi-million dollar corporations.

The 'complaint' is absolutely valid, hands down.   But IMO wrong time/place, is all.  I commented because I was concerned that the OP would take the 'general complaint' to be directed at him and his question/conclusions, which are totally appropriate for where he is right now.     

[Rob Strand]

This is at the level of multi-million dollar corporations.

I suspect they hire people that talk the audio talk.   My uncle had a $6k amplifier with silver this and that.  It had low voltage fuses on the 1kV/800V rail.   Oddly it would blow fuses all the time and in one instance the fuses blew but it started a perpetual arc in the enclosed chassis area which totally blackened the innards.

I commented because I was concerned that the OP would take the 'general complaint' to be directed at him and his question/conclusions, which are totally appropriate for where he is right now.     

Definitely not.

[Rob Strand]

This example shows the non-inverting clipper is a clean blend using the frequency domain.

The level of distortion is the same in both circuit but the level of clean of the non-inverting circuit
is exactly 1V (in peak voltage units) higher than the inverting clipper.   The non-inverting clipper
is adding the clean signal to the distorted signal ie. it is a clean blend!

Schematic


FFT in usual dB RMS volts


FFT in peak voltage units
In LTspice the schematics use peak voltage but the FFTs are displayed in RMS.  For this reason the FFT values need to be multiplied by sqrt(2) to convert rms volts to peak volts.   You can see this is correct as the input level of the sine wave is 1V (peak) on the schematic and 1V (peak) on the scaled FFT.

[teemuk]

"Both circuits produce the same level of distortion"

FFT showing additional harmonics of distortion up to 12th....And you still keep calling it a clean blend? 

Why not also post the transfer curve to show how it "folds", like I described, and is nothing but linear. Tubes and even diodes have similar "knee'd" transfer curves, are they "clean blending" too when clipping?

Fit a series resistor to diode chain in an inverting amp. Make it so that you get "unity gain" once diodes conduct. Is this now a "clean blend" (hint: it does exactly the same thing as the non-inverting circuit).

[Rob Strand]

"Both circuits produce the same level of distortion"

FFT showing additional harmonics of distortion up to 12th....And you still keep calling it a clean blend? 

I don't know how you can be so smug with your opinion when it's a simple matter of addition of two signals in the circuit domain.  I think most 1st year Engineering students would get the idea of the addition of two signals, whatever those signals are.

If I increase the input level to 2V peak the only thing that differs in the spectra is the level of the fundamental.  The clean signal component  is effectively passing through and adding to the distorted signal.





I've given three pieces of evidence that supports the idea that the non-inverting clipper blends the clean input with the distorted signal.

What evidence can you provide that contradicts that?

[FSFX]

. . . . clipper blends the clean input with the distorted signal.

I think that what you are saying here can be explained by considering the forward slope characteristics of a typical diode.
By this I mean that we are dealing here with real world diodes rather than ideal diodes and these have a exponential type of 'knee' followed by a slope that is begins to approach a more a linear shape. The effective resistance of this slope in combination with the other feedback components provides some forward gain for the clean signal. This appears as a rounding of top of the resultant clipped wave and increases the relative level of the clean signal (or the fundamental in the case of a simple sine wave) to the harmonics and intermodulation products produced by the clipping of the diode.

Although the following is not for soft clipping diode configuration, the principles are similar.     

[Rob Strand]

I think that what you are saying here can be explained by considering the forward slope characteristics of a typical diode.
By this I mean that we are dealing here with real world diodes rather than ideal diodes and these have a exponential type of 'knee' followed by a slope that is begins to approach a more a linear shape. The effective resistance of this slope in combination with the other feedback components provides some forward gain for the clean signal. This appears as a rounding of top of the resultant clipped wave and increases the relative level of the clean signal (or the fundamental in the case of a simple sine wave) to the harmonics and intermodulation products produced by the clipping of the diode.

No I'm saying more than that.   I'm saying the non-inverting feedback configuration always adds the input signal, regardless if the circuit is non-linear or not.

It's a very basic idea,  you don't even care what the blocks are (other than B cannot be a short circuit, or, in practice, A cannot be an open circuit),

[FSFX]

. . . you don't even care what the blocks are

What about the hypothetical case where A is in fact a pair of 'ideal' diodes.

[Rob Strand]

What about the hypothetical case where A is in fact a pair of 'ideal' diodes.

It still works.

Suppose B passes current IB = Vin / ZB.     The voltage across VA is determined by the diode equation, since ID = IB,

                    ID = IS (exp(VD/(n*Vt)) - 1)   ; where Vt = kT/q

It won't take much current before the diode voltage is some defined value.

If B is open then IB = 0 and ID = 0 then the diode will have VD = 0.  So it still works.

If you use opamps with BJT input stages you have to expect some deviation from (simple) theory because the input bias currents need to flow through the diodes - it will result in a DC offset at the output.

It's easiest to think of A and B as being passive but they can be linear or non-linear.  (I don't want to get off topic with special cases like opamps in A and B.)

[Rob Strand]

You can also build a diode divider.   The circuit produces a gain of 2 with essentially no distortion,

[teemuk]

What evidence can you provide that contradicts that?

Plot the transfer function. It is not straight line but two lines, "folding". Even the very lack of overall linearity should point out that this circuit will amplitude distort the signal.

I have explained  how the gain "folding" to a decreased ratio - (in this circuits case to unity gain) - creates amplitude distortion. The same explanation demonstrates that, other than below diode Vf, you do not get a clean output signal from this circuit. It is an impossibility.

There is no "clean blending" anywhere; different parts of the waveform are just being amplified by different gain ratios.

I have stated the rudimentary fact that "clean" signal contains no extraneous harmonics and that a signal is distorted if it does. Blending the two is fallacious concept, they can not co-exist, and "clean blend to distortion" is an oxymoron. Even your FFT plot shows a great deal of these extraneous harmonics.

Now it is your turn to show evidence of the contrary. IMO, the OP came hear to learn and you shouldn't confuse him with false theories even though you happen to hold dear this "clean blend" concept.

[Rob Strand]

All you have done it reiterate the same words with no quantitative results.

Even the very lack of overall linearity should point out that this circuit will amplitude distort the signal.

It's not the fact it distorts that's in question.  Both the non-inverting and inverting forms distort.  The point is the difference between the two outputs *is* the input signal, which is clean.

There is no "clean blending" anywhere; different parts of the waveform are just being amplified by different gain ratios

This alone shows the output signal is the sum of two components, one of them clean.

I have stated the rudimentary fact that "clean" signal contains no extraneous harmonics and that a signal is distorted if it does. Blending the two is fallacious concept, they can not co-exist, and "clean blend to distortion" is an oxymoron. Even your FFT plot shows a great deal of these extraneous harmonics.

Well I can take a distorted signal and add it to a clean signal.   It is a simple process of mixing.  You can buy pedals that do it.

The point is the non-inverting form does this mixing by nature.

Even your FFT plot shows a great deal of these extraneous harmonics.

They are not extraneous they *are* the distorted signal.   The whole point was to make the harmonics the same in the two circuits.

The only difference between the non-inverting and inverting forms *is* the fundamental.   In the last example 2V of fundamental is there in plain sight.

Now it is your turn to show evidence of the contrary. IMO, the OP came hear to learn and you shouldn't confuse him with false theories even though you happen to hold dear this "clean blend" concept.

What's wrong is your concept of clean *blending* altogether.    By definition it is adding a clean signal to a distorted signal.


There's a difference between theories and a demonstration.  Theories are only of use if they explain an observation.  I've provided a number of demonstrations where it's clear the observation is the the clean signal adds to the distorted signal.   You have provided no demonstrations at all.



The theory that is explains the observations is this,

[FiveseveN]

they can not co-exist, and "clean blend to distortion" is an oxymoron

Now I'm intrigued to find out what you mean.



You guys familiar with dualities? Or maybe you're not even arguing two equivalent models but about the particular words to describe one.

IMO, the OP came hear to learn and you shouldn't confuse him

Perhaps a thread split is in order.

[antonis]

Perhaps a thread split is in order.

With thermionic gas diodes inside the NFB loop...

[merlinb]

Rob is right, in a non-inverting clipper the original signal is effectively blended (summed) with a pure clipped signal, ala:

[FSFX]

Rob is right, in a non-inverting clipper the original signal is effectively blended (summed) with a pure clipped signal, ala:

This is also what happens when resistance is added in series to hard clipping diodes or naturally with the various forward transfer slopes of different diodes. This is what makes the difference in sound between Schottky, germanium. silicon, gallium arsenide (LED) and so on.

[FSFX]

Perhaps a thread split is in order.

What Rob posted above makes a lot of sense regarding the case of the non-inverting soft clipping op amp stage.
I would personally welcome a sensible discussion regarding the many aspects of diode clipping that get overlooked or taken for granted.
One example is just how diode clipping actually produces a 'fuzz' sound as just the effect of adding harmonics to a single tone is inconsistent with the sound or spectral content of fuzz.
Virtually everything that has been written in the pedal community avoids discussing this in favour of just showing simple clipped sine waves.

[Mark Hammer]

Let's  get some things straight, here:

1) Guitar signals, produced by picking/plucking/strumming a string, are dynamic.  This means their amplitude changes over time (generally very quickly, unless artificially held constant), and their inherent harmonic content changes over time.  Any additional harmonic content produced by a circuit is ultimately a function of what that circuit is fed, and what it is fed changes quickly.  In short, a guitar string is not a signal generator, producing a waveform of constant shape and amplitude.

2) Let us distinguish between what is sent to our ears, and what we perceive.  I keep harping on about "auditory scene analysis", but I'm dead serious about it.  Here's a nice, profusely illustrated, explanation/summary of what it means: https://www.phon.ucl.ac.uk/courses/spsci/AUDL4007/Scene_analysis.pdf  Since we actively (though generally unconsciously) "group" auditory content, and since the harmonic content produced by feeding a guitar signal into any sort of clipping circuit will change over time, and generally much faster than the note fundamental does, we are likely to mentally group that auditory content into the basic string signal and "other stuff".  In other words, unless the added harmonic content and original signal last the same length of time and at similar levels, we will "hear" them separately, the same way you are currently "hearing" various differentiable sound sources (as well as seeing different letters and screen windows), even though it's all landing on two hapless eardrums.

Do not confuse what is on a scope screen with how you perceive sound.