Soft clipping with predominantly 3rd order harmonics – possible?

[Rixen]

Here's a conundrum. If you feed a triangle into a CMOS inverter, the soft clip can fold the wave to look very much like a sine wave. As the triangle is all odd harmonics - does that mean that the soft clip has filtered those odd harmonics out?

..It's added odd order harmonics in opposite phase, partially cancelling out the harmonics that make the triangle 'pointy'. Further overdrive will increase these added harmonics further causing flattening of the peaks..

[ashcat_lt]

Bear in mind that if you already have a square wave - such as when a hard clipper drive amp is itself clipping - you can make the clippers as asymmetrical as you like, it will still be a square wave out of them & no even harmonics will be created.

Well, extreme asymmetry can start to look like a change in duty cycle, and make it more like a "rectangle" wave than a pure square.

But if you introduce an RC high pass filter between amp & clippers (just insert cap), the differentiated square wave out of the high pass will be asymmetrical because the different clippers makes the time constant different between +ve and -ve cycles in the filter. You get plenty of even harmonics then, even when the drive amp is clipping. Also you can return the clippers via ground, instead of Vref, thanks to the added cap.
Down side is a thin bottom end and/or it stops making evens when the fundamental frequency is above the filter cut-off and passes the whole square wave.

This is not as simple as it sounds.  "Asymmetrifying" the signal in this way looks a lot like a DC offset, and the cap won't let that stand, especially if it's also got an AC coupling cap on the other end.  Those caps will charge or discharge as necessary to center the signal between the diodes, and you'll be back to symmetrical and odd.

[karbomusic]

Good link Teemuk...
And with reference to that...
I think the easiest S-curve amp has to be the CMOS inverter biased for "linear". Worth experimenting with.

As far as rounded soft-clip, how about electra distortion? Input on top, output on bottom. I don't have the harmonic series handy but can scope it if needed.

[teemuk]

Here's a conundrum. If you feed a triangle into a CMOS inverter, the soft clip can fold the wave to look very much like a sine wave.

You don't even need a CMOS inverter, any soft clipping device or scheme will achieve the same goal.... to degree. Basically sft clipping a triangular wave will convert it to resemble a sinusoidal waveform, but you need a very specific transfer curve to achieve most lowest distortion in the end signal. For that CMOS chips or ovedriven open loop OTA chips tend to work best.

..But in crude form the trick works with just about ANY soft clipping device, including plain diodes, whether solid-state or tube. With incorrect transfer curve the output sinusoidal will just have more distortion in it, meaning its not going to be a "pure" sinusoidal wave. In most applications that is not going to be a big deal. Soft clipping to convert triangular wave to sinusoidal one has been used WAY before anyone even dreamed about CMOS logics and such.

[teemuk]

This is not as simple as it sounds.  "Asymmetrifying" the signal in this way looks a lot like a DC offset, and the cap won't let that stand ...  Those caps will charge or discharge as necessary to center the signal between the diodes ....

This is a good example of the "dynamic shifts" Merlin briefly mentioned: When you capacitively couple an asymmetrically clipped signal to another stage the asymmetry in signal's half waves will gradually slide DC offset of that stage. In turn, that sliding DC offset will affect symmetric vs. asymmetric clipping characteristics of the stage and the resulting spectra of distortion will therefore also change along with the sliding DC offset.

This is what happens in most distorting schemes incorporating at least two capacitively coupled and asymmetric clipping circuits. e.g. A preamp with more than two stages that distort.

You will not get that effect from just a single clipping stage.

In essence the harmonics of distortion vary in interaction with signal amplitude and envelope. Overdrive won't just produce a "static" pattern of harmonics, where amplitude of each harmonic simply increases the more you ovedrive but the pattern as is never changes. It produces a "dancing" pattern of harmonics that varies vividly and interacts with dynamics of playing. Some amps are specifically tuned to morph from asymmetric overdrive to more symmetric overdrive with aid of such DC offset shifts.

This is what makes clipping sound interesting and musical, and what also makes it "touch sensitive". Not a specific pattern of harmonics, nor specific "softness" or "hardness" of the clipping. That effect is also what dominantly separates "tube amps" and alike from simple effects based on single distortion stage, like a pair of anti-parallel diodes. The simple clipping setup will never achieve that interactivity and dynamic DC offset sliding.

Another briefly mentioned aspect is EQ. That also makes ton of difference! When you think about it, post and pre-distortion EQ:ing is really effective in controlling what frequencies will be intermodulated most. Typical pre-distortion EQ reduces low end response by great deal to minimize IM at low frequencies. Does marvels for "note separation" and for turning that distortion tone from farty 190's fuzz realm towards modern, "tighter" distortion sounds. Sometimes high end is cutted too pre-distortion because clipping distortion will artifically recreate most of those frequencies back. If you study it, you see that manufacturers have various schemes to treat mid-range, which is the most dominant range of guitar. For example, many amps incorporate some sort of "mid-range notch" filter for overall voicing. At what frequency that "notch" is centered at can have a great effect on overall amp tone. Sliding that notch frequency up or down can effectively have much bigger impact on your amp's tone than slight differences in transfer curves of distorting stages.

Highest order harmonics sound "fizzy" so its customary to cut them post clipping. Most guitar cabs, for example, already work aslow pass filters, cutting pretty much everything above 5 kHz and it does great benefit for overdriven tones. Ever tried how your O/D tone sounds through a cabinet with a tweeter? If you ever listened to pure output signal of your amp you will quickly realize why people fuzz about all those "cab sims", which are effectively low-pass filters.

Good overdrive tone is sum of many different characteristics, including those dynamics and EQ:ing at various points of the circuit. It's not just about the mere clipping process. For example, in certain Carvin schematics tone shaping filters are BLOCKED from view, because Carvin's trade secrets - and what actually defines the "tone" of their amps - actually lie in those "voicing" circuits, not in the distorting gain stages that are nearly or totally identical with those of many other amps.

[Rixen]

Yes. The tube screamer circuit is a symmetrical clipper, and relatively soft. I have seen a spectrum of the output which shows predominantly odd order harmonics. BTW I just visited the Wikipedia page on the Tube Screamer, and see some BS about how it works has crept in. It is not a crossover distortion.

Just for the record it is more like a logarithmic amp than crossover distortion.

I have corrected the Wikipedia page now 

[lion]

Thanks to all who contributed so far, I'm really learning a lot here.

Teemuk - great info about tape saturation. Also the waveform distortion article, very helpful.

I created a few diode clipping models in Excel last year based on iterative solutions to the Shockley model. They actually line up very well with scope observations for simple circuits. I can either upload or forward them to you if you want to play around with them.

PBE6 - not sure I'd know how to use it, but I'm very interested.

I'm liking what I get with a common source N channel pre-amp followed by an inverter stage, all from a 4007 chip at the moment.

Jim - would you mind sharing the circuit details?

As far as rounded soft-clip, how about electra distortion? Input on top, output on bottom. I don't have the harmonic series handy but can scope it if needed.

Kary - with not many harmonic plots around, if it's not too much trouble I'd like to see it.

I'm struggling to understand the parts about "dynamic shifts" and sliding DC offset - and other details as well - but I'm working on it

Erik

[robthequiet]

http://www.runoffgroove.com/thunderbird.html

ROG has purposelly modeled the dc shift Teemuk and Merlin described, if you have not already seen this.

[anotherjim]

I've shown the 4007 parts from individual mosfet's so spot the pin numbers of the chip.
From left you have an N channel amp followed by a tone control (which loses quite a bit of signal), then a P channel source follower, finally an inverter. With R15 at 1M the inverter gain is wide open, but the S transfer curve means it's hard to measure what that is! It stopped sounding louder at about 680k, but 1M seems fine to me.
The whole thing is an amp sim type affair & still under development. Bypassing the tone control network results in a very thick sound. With the tone, I think it's a nice, well rounded "clean" tone from "woody" to "chiming" - meaning not really clean in the hi-fi sense at all, but definitely no fizz. Increase C9 if you want more bass.
Note the unused mosfets are all tied up, which is what I think the datasheet says you should do.

[lion]

I'll look into the ROG/Thunderbird - haven't noticed it before, thanks Rob.

Jim - thanks for sharing the 4007 circuit, appreciated.

Erik