Opamp-based distortions - The frequency-selective network to ground

[Labaris]

Hi everyone,

In this article ("Clipping Stage" section) Mr. Keen describes how Zi (wich is formed by the series resistor and capacitor from the inverting input to ground) is frequency selective.

This idea is used in other distortion circuits, but I have a question:

¿Why do they always cut the low frequencies and not the highs? ¿"Muddiness" is always undesirable?
Boss DS1 for example has a "drive cut-off frequency" that gets higher as gain gets bigger. So, with max gain setting the low frequencies don't get distorted.

I'm designing a modification for this kind of networks to use it "the other way", to get more bass as more gain is set.
Will this be a useless "muddiness-machine"?

I'd like to learn how to use the good idea in a good-sounding way.


Thanks! 

[teemuk]

Why do they always cut the low frequencies and not the highs?

Think about this scenario:

The clipping of the fairly high-amplitude low frequency component pretty much eliminates all the high-frequency components "riding a top" the low ones. Decreasing the amplitude of low-frequencies will help to retain note clarity and the resulting clipping signal is less muddy farting and buzzing.

So I think you pretty much answered your own questions.

Will this be a useless "muddiness-machine"?

In my opinion, Yes.

But you can always try it out to make up your own mind.

[Labaris]

It's very clear to me now, thank you!

[Earthscum]

I've actually done this, but in a completely different way.

I always liked the way my bass sounds through an MPSA13... nice clean highs. I also love the way the inverter distortions sound, so I made a phase splitter (10k to V+, 10K to ground, .1uF output caps from the collecter AND emitter, base biased with 1M's to pos and grnd). I use the non-inverting side to drive the distortion, running straight in to an inverter with no feedback, then shape the distortion after that. I used a 27k(.022)27k t-filter between the first and second inverter stages and fed the bottom of the cap straight to the output of that stage (feed-forward). 33k-330k on the third stage, then fed that to the output. The other side of the phase splitter got fed to an inverting buffer (4th stage of CMOS) and mixed via a 1M pot.

Worked well, nice gritty low end, but my bass was still shining through with all the punch. This mixed a bunch better than other clean/distortion mixes I've tried. I'm guessing, from info I've read by Mark and R.G. that it has to do with impedances. The high impedance of running straight into a CMOS inverter doesn't cause any change on the impedance at the inverting output of the transistor stage. Since I'm also using a high-impedance buffer (although significantly lower because of feedback biasing) on the inverted output, it doesn't change the distortion side. Both outputs like seeing a higher impedance signal, so the 1M pot worked good... tried a 500k, which seemed to work better, but it was audio taper, so hard for me to tell. If I remember right, though, I determined that an audio taper worked better, but would be best if it was a half taper instead. The distortion drives alot harder than the clean signal.

[Labaris]

Thanks for for sharing your results, Earthscum.


Now I have another question, about a specific circuit: Boss DS-1.

In the schematic you can see clearly the "Zi" network formed by a 4k7 resistor in series with a 1uF capacitor (80's version).
Zi varies as the distortion potentiometer does, so the cut-off frequency of the signal being clipped also changes.
The thing is that the maximum frequency of the Zi network is about 33Hz (and the minimum is less than 2Hz). What's the effect then? At any setting, all guitar frequencies will be amplified en distorted, with the associated "muddiness"...
I'd like to change the maximum "drive-frequency" as I call it, to a higher value. Maybe 4 octaves up or more, by changing the capacitor.

Any opinions?

[teemuk]

The thing is that the maximum frequency of the Zi network is about 33Hz (and the minimum is less than 2Hz). What's the effect then? At any setting, all guitar frequencies will be amplified en distorted, with the associated "muddiness"...

But you are ignoring the effect of the earlier stages: the RC filter formed by C3 and the total impedance presented by base circuit of Q2 rolls those low frequencies off very heavily. Thus there's a drastic bass reduction before the signal even hits any of the gain stages. There was perhaps very little point in implementing more of it to the gain control circuit.

However, there is a "frequency selective network" In DS-1's gain control circuit. In this it case affects high frequencies, not the low ones.

[Earthscum]

basically, the cutoff frequency at NO gain is ~34Hz. Since this is in the feedback (obviously what we're talking about here), then it boosts all signals above 33hZ. Now, when you have the distortion UP all the way, the frequency cutoff is 100k+4.7k, ~2Hz. This is different than the TubeScreamer, which boosts high frequencies only. Only thing without losing the DS-1 sound that you could do is put a larger cap in there, making it "flubbier" possibly, or drop that cap down which would make thin distortion tones and even thinner clean tones.

Basically, the DS-1 ensures that all audible frequencies that have entered the input are boosted at the output.

[WGTP]

A commonly used technique is to reduce the bass thru the distortion, and boost it back afterward, either using passive or active filters.  IMHO this gets a more ampish distortion and sounds less like a fuzz.   

[Mark Hammer]

The single common challenge that so many filtering schemes attempt to address is that:

1) diodes have a fixed forward voltage (clipping point)

2) the signal produced by thicker strings, especially at the lower end of the fretboard, is much higher amplitude.

If you want to apply the same degree of clipping across all notes, you can't easily make the diodes behave differentially, according to frequency.  All you can easily do is attenuate the higher-amplitude portions of the spectrum so that they are roughly the same proximity to the clipping threshold as other portions of the spectrum.