Why double diodes in feedback loops?

[bartimaeus]

A lot of really famous overdrive pedals double up the 1N914 diodes in the feedback loop. Gainster, Red Snapper, Timmy, Xotic BB, Wampler Ecstasy/Euphoria, Barber LTD, Barber Gain Changer, etc.

What is the reason? I understand why you'd use two diodes one on side for asymmetrical clipping. But why two on the same side?

I know it increases the clipping threshold. But once you turn up the gain, won't it sound the same (as long as you tune the feedback cap to get the same treble response)? And in that case you're boosting noise too, just to get the same amount of clipping.

Does it give better control of the gain knob for low gain? Couldn't you simply limit the max gain to get that result?

Does the "softness" of the diodes double up in some way? Extra soft clipping if you stack a pair of Mosfets?

Or is it just to keep the output volume higher, so you don’t need an output boost to compensate like you might on a Distortion+?

EDIT: for example:

[Rob Strand]

There's a few reasons:

- It sounds different.
  For a non-inverting clipper the clean signal leaks through with a gain of 1.
  With two diodes the clip voltage is 0.6V and is effectively blended with the clean signal.
  With four diodes the clip voltage is 1.2V the distorted signal is now 2 times higher
  but the clean signal is the same - so more diode sound.

- In some cases the two diodes are just use as a clipper to prevent the opamp from clipping.
  That means the sound largely comes form the diodes and not the opamp.

IMHO the asymmetrical argument isn't the primary reason.   The 2 + 1 diode format is more like an in between
of the two above cases.   If you listen to the asymmetric case it sounds in-between the two symmetric diode
cases, it doesn't possess any magic second harmonic tone.

[PRR]

> But once you turn up the gain, won't it sound the same

If you have to turn it up, it must be weak. Am I right?

You NEVER want to be the "weakest" pedal in the showroom.

[marcelomd]

  With four diodes the clip voltage is 1.2V the distorted signal is now 2 times higher
  but the clean signal is the same - so more diode sound.

More _undistorted sound that is proportionately less clipped_... which is more similar to the original,so LESS diode sound, no?

[Rob Strand]

More _undistorted sound that is proportionately less clipped_... which is more similar to the original,so LESS diode sound, no?

It's a little tricky to construct an apples to apples comparison.   One angle is you want to compare the clippers with same input level but also the same diode current (since the diodes are then active the same in both cases).   That means the feedback resistor needs to increase by a factor of 2.   Of course the output level will go up but then you can adjust the levels with an output attenuator.

If you think about comparing a 20 diode clipper to a 2 diode clipper clearly you need to make some changes to have any hope of comparing the underlying *tone* without the levels entering the equation.  (Assuming the opamp won't clip, but on 9V it will  ).

For things like Timmy and the OCD the clip voltage has an impact on choosing the gain of the second stage.   The second stage can clip or not clip, it's a design choice.

[bartimaeus]

> But once you turn up the gain, won't it sound the same

If you have to turn it up, it must be weak. Am I right?

You NEVER want to be the "weakest" pedal in the showroom.

Fair enough haha! Louder sounds better!

  For a non-inverting clipper the clean signal leaks through with a gain of 1.
  With two diodes the clip voltage is 0.6V and is effectively blended with the clean signal.
  With four diodes the clip voltage is 1.2V the distorted signal is now 2 times higher
  but the clean signal is the same - so more diode sound.

Hadn't thought of the ratio of the 1x clean and the clipping, interesting!

That's at least something beyond the simple "it's louder" because of the higher threshold haha!

I think I'll try some level-matched examples in spice to see if that comes through...

[bartimaeus]

ok, here's the result of a test clipping a sine wave:





in the double diode circuit, i halved the value of the feedback cap, doubled the value of the feedback resistor, and reduced the volume of the output.

i had to reduce the volume by different amounts for the frequency graph and the clipping graph.

i didn't expect it, but looks like the single diodes have a more linear knee after the clipping threshold is reached?

[Rob Strand]

It can be a bit of an illusion due to adding different amounts of clean!

If you plot the voltage across the 4xdiode/2 and the voltage across the 2xdiodes you get an identical waveform meaning the diode part is doing the same thing except double the voltage.

From that we know if we double the output of the 2xdiodes, that the diode voltages must agree with the 4xdiode.   The theory is the 2xdiodes case has one clean blended, and the  4xdiodes has one clean blended.  If we double the 2xdiode output signal we get 2xdiode + 2xclean blended, and if we then subtract one clean we expect it should match the 4xdiode+one clean blend.

This set up shows they are in fact the same,



Back in July 2022 I posted these sims showing explicitly that the non-inverting clipper is the same as a clean blend.
(differences only due to finite opamp bandwidth.)

[marcelomd]

Rob, I understand just adding a pair of diodes (and everything else equal) to get a different sound (signal clipped at different point) was the whole point.

Unless I'm misunderstanding.

[ElectricDruid]

Or is it just to keep the output volume higher, so you don't need an output boost to compensate like you might on a Distortion+?

Mostly this, I'd say. A 0.6V signal isn't that loud. If you've got anything that does any sort of boosting ahead of the drive, the distorted signal might even finish up *quieter* than it went in, especially if it's an inverting clipper. Doubling up the diodes gives you more level to play with.

Yes, it affects the clipping level, so you need more gain to get the same amount of clipping, but that's not such a big deal as all that.

[bartimaeus]

Rob, I understand just adding a pair of diodes (and everything else equal) to get a different sound (signal clipped at different point) was the whole point.

Unless I'm misunderstanding.

But if it's ONLY clipping at a different point, it's not a different sound. It's just louder. If that's the only difference, then you can just increase the input level and it'll sound the same.

Maybe louder is reason enough for some designers.

But I want to learn if the sound changes in any way besides the clipping voltage. And as Rob explains, it seems that the amount of clean blend has a noticeable effect!

It can be a bit of an illusion due to adding different amounts of clean!

If you plot the voltage across the 4xdiode/2 and the voltage across the 2xdiodes you get an identical waveform meaning the diode part is doing the same thing except double the voltage.

From that we know if we double the output of the 2xdiodes, that the diode voltages must agree with the 4xdiode.   The theory is the 2xdiodes case has one clean blended, and the  4xdiodes has one clean blended.  If we double the 2xdiode output signal we get 2xdiode + 2xclean blended, and if we then subtract one clean we expect it should match the 4xdiode+one clean blend.

Wow, it really was an illusion! I replicated your idea, and you're right: the signals look identical if you mix in an inverted clean signal to cancel it out.

So I it seems the only differences are output level and ratio of clean blend.

But I have to wonder, how does the different versions feel once they are level-matched? Does the lower ratio of clean signal feel like a different knee into clipping? Or does it just feel like there's more overtones? I guess I need to breadboard these to compare...

[Rob Strand]

But if it's ONLY clipping at a different point, it's not a different sound. It's just louder. If that's the only difference, then you can just increase the input level and it'll sound the same.

Maybe louder is reason enough for some designers.

But I want to learn if the sound changes in any way besides the clipping voltage. And as Rob explains, it seems that the amount of clean blend has a noticeable effect!

You got it, spot on!

An interesting experiment is to go the other way, ie. have more clean.   As a thought experiment that would be using half a diode (0.3V clipping).  In practice you don't have an exact half a 1N914, you can use a germanium or Schottky but we know they don't sound quite like a 1N914 so it blurs the result.   So what I did in the early 90's was place a gain of 2 (or more) before the clipper and use a single pair of 1N914's - IIRC technically need to increase the resistor and decrease the cap to ground for the same diode currents.   To me lots of clean with a heavily clipped diode pairs sounds very unnatural.

When you switch through the three cases you can hear a difference.   You can also find what amount of clean sounds right to you.   From the theory (or spice sims) we know the difference is largely caused by the amount of clean blend.

There's some finer points that with too many diodes and a hot input signal the clipper stage opamp can itself clip, since the swing is the diodes + the input signal.

But I have to wonder, how does the different versions feel once they are level-matched?

They sound different that's for sure, but it's subtle.   You really need to experiment.   Over time I came to like the 4 diodes (2+2).   A friend of mine had a TS9, for a few years he loved it, then he got to the point where he was going to bin it.  I think I put a 1+2 diode clipper in there and he was happy with it ever since.   The point is it's a subtle thing and it can take time to work out what point suits you.

Does the lower ratio of clean signal feel like a different knee into clipping? Or does it just feel like there's more overtones? I guess I need to breadboard these to compare..

Yes, you need to do the experiment.   If you are someone who likes to ride the clean/overdrive point with how hard you hit the strings etc, you will find one of the diode combinations just seems to work better, kind of a more natural transition.   For more cranked settings less clean is better (I think that's another reason I like 2+2 diodes, usable over a wider range of overdrive settings).   After hundred of hours playing with these things that's about the only things I can say ... that's life with overdrive pedals.   Then there's using different diodes ...

[Steben]

With an opamp that has an infinite treshold, the higher the diode treshold the more the sound will be distortion rather than overdrive. For all the reasons mentioned.
Asymmetrical set up does add some gradual clipping.

[Elektrojänis]

With an opamp that has an infinite treshold, the higher the diode treshold the more the sound will be distortion rather than overdrive. For all the reasons mentioned.
Asymmetrical set up does add some gradual clipping.

But do you consider diodes in feedback of an inverting opamp stage just a distortion like diodes to ground? (Inverting stage does not have that +1.)

Its' been a long time since I experimented with opamps and clipping, but I remember even inverting stages with diodes in feedback to sound a bit different from just diodes to ground after the stage. It also might be that my circuits using those were just too different in other ways.

[Mark Hammer]

A very simple answer to the original question: more dynamics.  When the clipping threshold is low, the resulting sound is more likely to hit the ceiling most of the time; the principal reason why people describe the TS-9 as sounding "compressed",  When the clipping threshold is raised by some means (and more diodes in series, or use of diodes with higher forward voltage, is a common strategy) nuances in pick attack produce more graduated degrees of clipping.  Yes, you CAN do this with a simple 1+1 diode pair as well.  Heck, you could even stick a pair of Schottky diodes in a TS and do it.  But the degree of restraint and fine motor control required of the player to achieve nuanced clipping in that context is well beyond the majority of players.  The easiest situation is to simply raise the threshold via diodes, so that the player's wrist can do what it does without having to think too much about it.

A second aspect is that creating more gain, in a clipping stage that uses more diodes, allows for use of tonestacks that introduce passive signal loss, without having to add gain 'makeup' stages to compensate.

[Steben]

With an opamp that has an infinite treshold, the higher the diode treshold the more the sound will be distortion rather than overdrive. For all the reasons mentioned.
Asymmetrical set up does add some gradual clipping.

But do you consider diodes in feedback of an inverting opamp stage just a distortion like diodes to ground? (Inverting stage does not have that +1.)

Its' been a long time since I experimented with opamps and clipping, but I remember even inverting stages with diodes in feedback to sound a bit different from just diodes to ground after the stage. It also might be that my circuits using those were just too different in other ways.

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

[teemuk]

For a non-inverting clipper the clean signal leaks through with a gain of 1.

Sigh. Here we go again with the "clean blend" idea.


Now that you have the SPICE, input the circuit triangular waveform. The linear ramping up or downwards is plotted in output as the transfer curve of that distortion circuit.

I bet you get a "folding" curve, a "piece wise" transfer function, where the waveform ramps up in steep fashion at input levels below diode threshold, and in less steep fashion ("unity gain") above the diode threshold.

If a completely straight curve in output is a linear, distortionless, "clean" output, then does that "folded" curve look like a straight curve to you? That's a rhetorical question...

Next, plot the harmonic spectrum with a sine wave input. If a distortionless, "clean" signal shows only a single, "fundamental" harmonic does the folding transfer curve produce that? Another rhetoric question....

If by definition amplitude distortion introduces harmonics to the waveform not present in the input signal then what do you think happens when you "blend" harmonics to fundamental frequency or "clean signal"? Another rhetoric question.

Could we please bury this fallacious concept already?

[FSFX]

Could we please bury this fallacious concept already?

Thank you.

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

[bartimaeus]

We can spend all day trying to be 100% accurate on this stuff. But if something SOUNDS like a dry blend, and the end goal is SOUND, I see no reason we can't call it that for convenience. Sure, you can say that the "clean" blend on all effects is fake, because once you add some effect signal it's no longer clean. But it sure does sound like mixing between clean and effect, and at the end of the day that's all I care about.

[GibsonGM]

This is a DIY site, not an advanced audio engineering forum.  Thus, MANY people here aren't as committed to the deep level of learning required to FULLY understand things like the non-linear behaviors of certain components.  For most all applications, that level of knowledge just isn't required to 'do the thing', which is to make good-sounding pedals, mods, and maybe hash together some stuff of your own.   Some people simply don't have the time in their lives to take that journey, esp on their own with no formal education behind them.  Some might just have poor math skills, a young family, a busy job...    It's learn as you go for very many.

It's great when those who DO have that much depth behind them try to share and increase understanding. TEACHING is very appreciated.  It's great when someone explores a topic, and presents something that us lesser-informed hobbyists otherwise would never read about! 

Analogy: Rather than saying "so many in the beginning guitar community" simply lack the understanding to achieve a certain end, I personally prefer to try to show the one with the question how to do something, or at least send them to some resources that might help them grow.  Can't build your own arpeggios if you don't know what scales are, or positions over the neck, y'know.   They don't care what I know (often it goes right over their head for lack of context), they want an answer to their problem put in a way they can understand. 

My 2 cents, YMMV.