Musical Soft Clipping Circuits - Tri-to-Sine Waveshapers & the Differential Pair

[bowanderror]

(Starting this thread as a continuation of the conversation in the LG1 Gain Block thread.)

Some of the most musical sounding soft clipping follows the hyperbolic tangent (tanhx) transfer function, which is also desirable for triangle-to-sine waveshapers in synthesizers. There are a variety of methods for sine waveshaping, including the overdriven open-loop differential pair, voltage-controlled amplifier (often diff pair), OTA, Middlebrook & Richer, and piecewise linear approximation.

Although quite common in synth circuits for both waveshaping & distortion, I haven't seen many overdrive pedals that utilize these circuits. The Hagerman Distort is based on this old dynamic gamma correction patent, which uses a differential pair w/o feedback to mimic various nonlinear transfer curves. The device that originally got me interested in this is the Zvukofor Tanhx (user manual with more details), a stereo soft-clipping studio preamp. Based on some posts I saw of his Colour Module, I believe Zvukofor's design uses a JFET differential pair, but I have yet to trace it.

ElectricDruid has a great example of the Differential Pair Tri-to-Sine Waveshaper in his article on Sub-Oscillators, based on a schematic from Thomas Henry's excellent "Making Music with the 566":



I've played around with this circuit as well, and it sounded really nice on the breadboard, but still needs to have the level dialed in as the diff pair can go from no clipping to too much with a very small change in signal level:



I'd like to keep working on a tanhx-style design like this, and have a pretty massive library of reference articles & example designs on the topic, so let me know if anyone else is interested!

[bowanderror]

If anyone wants to play around with that Thomas Henry circuit in LTSpice, I've posted the .asc file here.

You'll need a TL072 spice model & the standard potentiometer model from the LTSpice groups.io Library (used to be Yahoo Groups). If you're not a member, I've re-hosted it here.

[julungas]

The overdrive side of germanium 4 big muff

[bowanderror]

The overdrive side of germanium 4 big muff

That's an interesting one! How does it sound?

I've played around with using Ge transistors in diff pairs, but as matching is important, the variability of Ge's is doubly problematic. I'll have to sim it to see how they deal with that. Any idea what type of PNP Ge's it uses?

[bowanderror]

If you're a little confused as to what tanh(x) is, it describes a very musical soft clipping sound. Tanh(x) is the name of the mathematical function relating the input signal to the output signal & is arguably the most musical sigmoid transfer function. This relationship is present in a wide variety of electronic circuit elements, such as the open-loop (no feedback) differential pair and is crucial in converting a synthesizer triangle wave to a near-perfect sine wave. For additional musical pedigree, it's a good model of the response of differential transistor pairs in the Moog Transistor Ladder Filter.



I'd like to find a way to apply some of these circuit elements to guitar, but there is more than one way to skin the tanh(x) cat.

I've uploaded that EHX Ge Overdrive file to a shared Drive folder with collected schematics, articles, & LTSpice sim files for a tanhx/sine waveshaping/differential clipping overdrive here.

There are folders for each method of sine shaping/tanhx clipping:

• Differential Amplifier (BJT)
• Differential Amplifier (JFET)
• Middlebrook & Richer (JFET)
• OTA
• VCA
• MOSFET
• Transistor Ladder Filter (Moog)
• Zvukofor Tanhx
• Books & Articles

[Mark Hammer]

Look into the SDS subcircuit of the Roland Funny Cat.  A terrific soft distortion effect that is very different.  It uses envelope ripple to produce the "distortion".  Over the years we've had many posts from people whose autowahs, noise gates, and compressors had what they described as "distortion".  In all cases, it was actually envelope ripple from a poorly smoothed half-wave rectified sidechain.  A tip of the hat to Roland for putting that ripple to good use!

[bowanderror]

"Funny Cat" is the most stereotypical Japanese effect name I've heard so far! The SDS section alone does sound quite nice in this demo video, although once the filter switches on it gives me a quite queasy stomach.

Here is the SDS section only of the Funny Cat (you can see how it connects to the sidechain and rest of the circuit here):


Definitely an interesting setup with the FET in the feedback loop.

The opamp labelled as "Module A" is a discrete opamp that was previously traced here. I've corrected the pinout from the post on this schem:


It looks like someone had luck subbing in 2N5088 & 2N5087 on a breadboard, so I may have to give it a try.

[Mark Hammer]

The FET serves as the ground connection of a non-inverting op-amp, whose drain-source resistance is varied by the half-wave rectified output of the op-amp being fed back to the Q1 gate via D1.  C3 is a fairly small/low value so it provides very little smoothing.  The result is that the gain-setting resistance of Q1 "wobbles" at an audio frequency.  Not sufficiently to produce sum and difference sideband products, but enough to produce a "burr" to the signal.  Very different from everything else out there.

I made myself a Funny Cat clone (thanks RG) but I think one of these days I need to make an SDS unit on its own, and play with the values.

[bowanderror]

How important do you think the discrete opamp is to that sound? The original used 2SC1000 NPNs & 2SA640 PNPs, which seem to be low-noise with high current gain. Would BC549C/BC559C be good modern subs?

[Rob Strand]

The opamp labelled as "Module A" is a discrete opamp that was previously traced here. I've corrected the pinout from the post on this schem:

FYI,   there was an earlier thread back in Jan 2003, where we back engineer that pedal.   That lead to RG's schematic.

There is a hard wired feedback resistor on each module,

Module A:   470k across pin 3 (out) and pin 4 (in-)
2xModule B: 100k across pin 3 (out) and pin 4 (in-)

As measured by moosapotamus,
https://www.diystompboxes.com/smfforum/index.php?topic=7416.msg44159#msg44159

[Vivek]

https://www.desmos.com/calculator/2857hjenmf


Tanh function with gain control

[Steben]

Reminds me a lot of AC30 power amp curves.
AKA: non feedback (EL84) push pull.

[teemuk]

Well, that's no surprise. Pentode transfer curve isn't all too different from that of bipolar junction transistors or FETs, which likewise have logarithmic transfer functions. In both push-pull and differential circuits the transfer function is also a composite of two (device) transfer curves, now enforcing symmetry. (Two moderately asymmetric curves interweave to a single symmetric curve).

AC30 output is a pentode push-pull circuit driven by a differential circuit, also in open loop. Therefore it provides transfer characteristics that are symmetric and highly non-linear, yet in "smooth" fashion. And yes, probably too smooth for all applications.

Ironically, or perhaps because aforementioned, one rarely sees any differential or push-pull tube circuits exploited as distortion circuits in tube amps (excluding genuine overdriven power amps and PPIMV schemes). The Laney Klipp circuit is likely the most famous example of using differential in a preamp and then there are small handful of "power amp emulation" -style circuits (Jim Kelley "Line Amp" as likely most famous of them, though definitely not the only example). Crate employed differential circuit distortion in some amp models (AFAIK only two of them), and some amps feature remotedly similar SRPP circuits (namely only Marshall used them for distortion, and not in any of their popular amps). But for some reason overall these circuits have never gained a status of seeing more frequent use in tube amps although they should be the closest option to generating "power amp like" distortion.

[julungas]

The overdrive side of germanium 4 big muff

That's an interesting one! How does it sound?

I've played around with using Ge transistors in diff pairs, but as matching is important, the variability of Ge's is doubly problematic. I'll have to sim it to see how they deal with that. Any idea what type of PNP Ge's it uses?

Sounds something like this https://www.youtube.com/watch?v=oWm1yVhLOTQ&t=538s
It uses NKT275, but I used 2SB475, it sounded very similar.

[Groovenut]

Does the bias offset at pin5, 6 of the summing opamp cause any additional clipping behavior?

The bias of the summing amp will change depending on where you set the symmetry R12

I would think you'd want a cap between R6,R7 and pin 6,5

[Vivek]

Can one hear the differences in sound between all these different transfer functions that look approximately same ?

[mdcmdcmdc]

I'm curious - for the purpose of this discussion, what are your metrics for "musical"?
For example, I really like the sound of a BJT differential pair arrangement like the one Dan Coggins used in the Tube Bender:



But it sounds like you're after something different?

[Steben]

Can one hear the differences in sound between all these different transfer functions that look approximately same ?

Not in full gain.
I made a sketch of what you may call "intuitive suggestion" of the similarity between drive based on transfer curve only, not EQ.
It clearly shows two basepoints: no to low gain, almost no harmonics and high gain (where the driven opamp circuits comes in)
This graph would be a flat line in the case of a perfect sim of a transfer curve.


Note : a non inverting feedback loop soft clipper will never reach 100% compared to an amp

[ElectricDruid]

Can you explain what this curve describes a bit more, please?

I don't understand what I'm looking at, sorry. "similairity"? similarity to what?

[Steben]

Can you explain what this curve describes a bit more, please?

I don't understand what I'm looking at, sorry. "similairity"? similarity to what?

I think it is clearer reversed in a difference graph



In the lowest region of the headroom many transfer curves are alike = quite linear. In high gain, with the output reaching the rails quite easily, the output is quite similar = as good as square waved. Mid gain is where the difference reigns. And it is why opamp clippers can be used as an easy high gain device.