Why double diodes in feedback loops?

[FSFX]

Let's  get some things straight, here:  . . . .

Exactly -
This is why I solicited a discussion on clipping beyond the simple view of clipped sinewaves that everyone seems to write about on virtually everything I have seen on the Internet and posted in the pedal community, even from the most respected people like RG and Jack.

The following make for interesting and enlightening reading on this subject.

https://www.gitec-forum-eng.de/wp-content/uploads/2020/09/poteg-8-2-3-pitch-perception.pdf
https://www.gitec-forum-eng.de/wp-content/uploads/2020/09/poteg-8-2-4-partial-grouping.pdf
https://www.gitec-forum-eng.de/wp-content/uploads/2020/09/poteg-8-2-5-inharmonicity.pdf
https://www.gitec-forum-eng.de/wp-content/uploads/2019/02/poteg-10-08-05-distortion-devices.pdf
https://www.gitec-forum-eng.de/wp-content/uploads/2019/02/poteg-10-10-4-distortion-comparison.pdf
https://www.gitec-forum-eng.de/wp-content/uploads/2019/02/poteg-10-10-5-distortion-audibility.pdf

In fact this whole book should be essential reading to anyone involved with guitars, amplifiers and effects pedals:

https://www.gitec-forum-eng.de/the-book/

[antonis]

In fact this whole book should be essential reading to anyone involved with guitars, amplifiers and effects pedals:
https://www.gitec-forum-eng.de/the-book/

From above mentioned book..
(page 10-235)

[teemuk]

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

I don't see the "original signal" in the output. I see a severily distorted signal that just partially conveys some elements of waveform of the "original signal". (The part folding to unity gain
Edit: and the "vertical" part, part amplified with high gain).

Let's flip this around, let's not concentrate on the peak portion. Imagine a brickwall clipped signal, is it - by the above logic - "clean blending original signal" below clipping threshold. No. Way don't say it is.

The "superimposing" is a much more descriptive term of what happens.

[FiveseveN]

Imagine a brickwall clipped signal, is it - by the above logic - "clean blending original signal" below clipping threshold.

I haven't seen anyone claiming that. Maybe a failure to explain what the "above logic" is, though we've been through it a few times and with different examples.

Let's flip this around

Yes, let's!



What happens here, do we not get back the clean signal? Where did it come from if it was no longer there?
Superposition is indeed the correct concept but surely you understand that "mixing" or "blending" describe this same thing.

[teemuk]

Actually my previous post was maybe a bit misleading in its wording. In practice, that vertical slope of the waveform is amplified linearly, it's just not very evident because of super high gain ratio in proportion to gain ratio of amplifying the peak.

In either case we see virtually same result: peak of the waveform is compressed by significant gain reduction. We hear this as distortion. (Are folks just looking at the pictures and thinking "oh, that part looks clean").
There's really not so much difference whether we compress from gain of 100x to gain of 1x, like Tube Screamer, or from gain of 100x to downright brickwall limit; proportionally the waveform peak still compresses significantly and yes, both sound about as much distorted because of that. Both of those waveforms bear a striking similarity to square wave when a clean signal would be a sinusoid. 

No clean signal is heard if you listen signals going through either transfer function.

At lower gain ratios the less steep compression of gain reducing just to unity generates a bit more gradual compression effect but it's still not a clean signal.

In fact, if we talk about this "superimposing" we could also say we have high gain clean superimposed with low gain clean and it still sounds a bit fishy and confusing... The non-linear transfer function where two linear gain curves fold really describes much better how the circuit works. And when we acknowledge that we must also acknowledge that there is very little difference to other nearly similar transfer functions where high gain folds to lower gain.

[FSFX]

Actually my previous post was maybe a bit misleading in its wording.

I am sorry but I find it not just misleading but without merit.

What has been posted by others here is closer to what is actually happening.

The use of the term 'clean signal' does not limit it to a sinusoidal waveform. A plucked string signal into a pedal or even a squarewave may be considered a 'clean signal' as it is the original pre-distorted signal.

There are many other things you say which don't make sense either. Whether this is a lack of understanding or an issue with translation if English is not your native language.

I seriously recommend taking time to read through a lot of the GITEC forum book that I posted a link to before as it is written by a physics professor with an interest in music so he has provided great insight into the whole theory underpinning the subject that this forum is dedicated to.

Another university professor that publishes many in-depth YouTube videos about the whole world of the electronics behind music technology is Aaron Lanterman.
I recommend people to explore some of his fine lectures on YouTube.

https://www.youtube.com/user/abovenyquist
     

[Rob Strand]

In fact this whole book should be essential reading to anyone involved with guitars, amplifiers and effects pedals:
https://www.gitec-forum-eng.de/the-book/

From above mentioned book..
(page 10-235)

More interesting the paragraph with the blue markup answers this thread, especially the part after the blue marked-up text.

My point is this stuff shouldn't have to come from a book (or me).   You should be able to work it out from basic electronics.  That's why I like Physicists you just tell them how an op amp works and they can work out the rest.

[marcelomd]

...  That's why I like Physicists you just tell them how an op amp works and they can work out the rest.

But only point-shaped ideal opamps in a vacuum =)

Bias currents, DC offsets, etc. are dirty things reserved for lowly engineers.

[Rob Strand]

Bias currents, DC offsets, etc. are dirty things reserved for lowly engineers.

True, but if you add that to the model the physicists will get it.   I think a lot of engineers don't get it either.

For heat flow engineering you might find physicists starting on the wrong foot.   They would use differential equations but the engineers would look up the correlation equations in a book with answer already there.   These days people use FEM on computers, which is the physicist's method under the hood.

[bartimaeus]

if something looks and quacks like a duck, surely we can call it a duck?

i tried cancelling out the dry signal from a non-inverting clipper, and compared it to an inverting clipper. they're indistinguishable with a variety of waveforms (triangle, square, saw):



some very interesting reading in this thread!! but i'd like to bring things back to something a little more practical...

increasing the forward voltage of diodes in the feedback loop (either by stacking them or by using LEDs, etc) does two things: it increases the output level of the distortion, and it (as a result) decreases the relative volume of the "dry" signal (i'll say "dry" since "clean" has gotten such a bad name).

if we want no "dry" signal at all, we can use the inverting configuration. in that case, if we use the other half of an opamp to re-invert the distortion output, we can even use that inverting amp to mix in a specific amount of the "dry" signal!

i guess the only drawback is it becomes harder to implement filters without impedance issues? maybe it's no issue. i haven't yet tried an inverting version of the timmy, but now i'm curious...

[Rob Strand]

You are on the right track with everything.  Cancelling is generally a little haphazard.  If you look at reply #7 I've got a non-inverting form that blends.  It's just a series resistor (Circuit 2, R19 4k7).  To kill the blend on that one, just short out the resistor and it's back to the inverting non-blending ckt.  Also have the option of small resistor = less blend.  FWIW, Marshall Blues Breaker pedal uses the non-inverting form with blend (6k8).