Ways to distort

[Dylfish]

Hey guys, what methods are used by pedal makers to distort a signal. From what I've got so far there is.

1) Amplify the original signal with an Opamp or Transistor and then clip it with a set of Diodes.
2) "Overdriving" a transistor or Opamp

What other ways are there that are used by pedal makers around the place?
Cheers

[R.G.]

See Distortion 101 (http://geofex.com/effxfaq/distn101.htm) from The Guitar Effects FAQ at GEOFEX. That's update 3, from over 12 years ago. The first version of that was about '94.

[Kesh]

Split your signal into +ve and -ve portions. Invert the -ve. Feed both through logarithmic amplifiers. Attenuate each with a potential divider(each different for asymmetrical effects). Feed each through an exponential amp. Invert the originally negative back to negative. Combine.

I have no idea if this would work in practice, or even sound good, it came to me in a dream. The math is sound though.

[R.G.]

It's right. In fact, the voltage out of an ideal clipping diode is actually the logarithm of the applied current. P-N junctions are used in logarithmic amps for doing exactly this.

A back-to-back diode pair is really doing what you describe: one diode acts for each polarity and produces the logarithm of the applied current. Opamps and other feedback amplifiers do a much more abrupt limiting in that the feedback is causing much more nearly ideal limiting when the amp runs out of power supply room to amplify.

This school of thought leads you rapidly to thinking about the curve of the "knee" of the limiting function as being one of the critical issues of a clipper.

[Kesh]

Yeah. But normal clipping doesn't exponentiate.

What my circuit actually does is take roots. Log, half, then exponential gives the square root of a voltage. Log and square root curves look rather similar so it may sound good.

[R.G.]

Yeah. But normal clipping doesn't exponentiate.

No, it doesn't. It takes the logarithm of the applied current to some base, expresses that as a voltage.

What my circuit actually does is take roots. Log, half, then exponential gives the square root of a voltage. Log and square root curves look rather similar so it may sound good.

True. There is a fair amount of work on analog computing that's been largely forgotten; your circuit is one of them. The math, as you say, is correct. Square root is quite similar to logarithmic in shape if you get the scaling right. Not exact, but quite similar. I believe I remember an analog log/anti-log amplifer setup from an old National Semiconductor app note that did the + and - sides properly without separating them and logging/anti-logging in two separate units.

The thing about doing the math in analog computation blocks instead of the raw nature of devices is that you *can* do the math and get the curves to be what you want. A junction logarithm has scaling and offset errors as well as temperature drift that makes it somewhat error prone.

I had to do a signal squarer with an old Philbrick tube analog computer back in in school as a lab project. That took a while to get right. 

[Kesh]

The nice thing about splitting, if anyone is crazy enough to build this, is asymmetric clipping, or rooting. Sqrt on one side, cube root on the other. And if it is ever built it has to called The Rooter.

Trivial to implement in DSP though.

[FiveseveN]

Trivial to implement in DSP though.

Maybe someone familiar with Max/MSP or SynthEdit or such voodoo could provide a shorter path to audio samples than dealing with the idiosyncrasies of an analog implementation. That way people could get an idea of what it would sound like and decide if it's worth pursuing for the purposes of musical production and not just for the hell of it (which is a perfectly good reason).
Come to think of it, this methodology could be generalised into a useful tool. I know it's been done with feeding .wavs into SPICE and such but I'm talking less granular building blocks and realtime simulation (something akin to ReValver's in-depth ampsim editing or the upcoming Head Case).
My, my, haven't we got a bit off track? 

[pinkjimiphoton]

isn't splitting the signal, and distorting each half separately, the basis of the klon centaur?

[WaveshapeIllusions]

Trivial to implement in DSP though.

Maybe someone familiar with Max/MSP or SynthEdit or such voodoo could provide a shorter path to audio samples than dealing with the idiosyncrasies of an analog implementation. That way people could get an idea of what it would sound like and decide if it's worth pursuing for the purposes of musical production and not just for the hell of it (which is a perfectly good reason).
Come to think of it, this methodology could be generalised into a useful tool. I know it's been done with feeding .wavs into SPICE and such but I'm talking less granular building blocks and realtime simulation (something akin to ReValver's in-depth ampsim editing or the upcoming Head Case).
My, my, haven't we got a bit off track? 

I've heard this is how Line 6 does things. They digitally model the circuit that effects the signal. It is a great idea though. More accurate digital effects and you can do things that don't exist in reality, like a perfect current source or other ideal components.

On topic, here are a few weird distortion makers. Build a discrete opamp for slew rate limiting. Make a low power AB gain stage and mis-bias it into crossover distortion. Use a compressor set to high ratio (basically a limiter); I think one small builder uses that one.