Overdrive designs that treat lower and upper registers separately?

[Vivek]

One of the things to keep in mind is that the different bands that one might split the signal into will likely have different average amplitudes.  That, in turn, means that one either has to:

• pick your bands and bandwidths such that the output levels are approximately equal
• pick your split-points arbitrarily, and apply different amounts of post-split gain to each band in order to bring it up to similar clipping thresholds
• alter the clipping thresholds for each band to compensate for differential amplitude, or
• treat the differential clipping of bands having the same gain applied and clipping threshold as simply "the way the thing sounds"

My own tendency would be to work out whatever I could ahead of time to sidestep endless tweaking from the control panel.  So, nothing adjustable about the split bands themselves, apart from their post-clip output level.

A half-dozen years back, I whipped up something for myself I called the Flexidrive, that simply used a 2-way split: middy-mids and highs, and lower-mids and lows.  The split point was fixed, and I used one pot to adjust the gain of each side in reciprocal fashion, and another pot to adjust the relative balance between channels.  That, and a master output control were all the knobs it had.  Eventually I figured a toggle to set different levels of drive on the input stage would be a good idea.  It had a lotta different sounds available.  Can't speak to the intermodulation distortion, though. 

I made some changes to what you see below, but this was the only drawing I could find and post at the moment.


I think there might be some value in applying a similar strategy to a dual-band circuit, but use crossover distortion for the upper band.  That is, place a back to back diode pair in series with the output of the gain stage.  Personally, I find that x-over distortion likes bridge pickups, so feeding a series diode pair only mids and highs, combined with a clipped-and-filtered low end, might sound decent.

In retrospect, rather than a master drive control and a single pot producing reciprocal gain adjustment of the two clipping stages (i.e., more gain for one means less for the other), it's just smarter to have a drive control for each section, and be done with it.  The nice thing about using those individual pots and a balance/mix control is that you can dial in relatively clean bottom OR top, and blend it with however dirty, or however much, you want of the other band.

Obviously doing that with more than two bands starts to get rather complicated, not least because you'd need two pots (drive and level) for each band, whereas a two-way split permits use of a single blend control: one less knob to find space for.

[Shoeman]

AHA!!  Thanks for hunting that down Vivek.

[Mark Hammer]

How about the Malekko Wolftone Sloika pedal

Not uninteresting, but I don't see much "band-splitting", just differential lowpass filtering of the resulting distortion.

[Vivek]

How about the Malekko Wolftone Sloika pedal

Not uninteresting, but I don't see much "band-splitting", just differential lowpass filtering of the resulting distortion.

and some diode ladder stuff where piecemeal transfer function is also linked to frequency response ?

[bowanderror]

DeadEnd FX has a Infanem Driving Notion clone board called the DILLIGAF. Here is the build doc with a schematic. I seem to remember seeing forum posts saying it was particularly sensitive to biasing & transistor choice, but the build doc is recently updated with info on biasing, so maybe they've figured it out.

[Shoeman]

Question for ya.  When R-C filters is there a best practice for any of the values?  Playing around with the AMZ calc I see that you can make a filter for any typical guitar freq with a wide combination of values.   .1 and a 4k  vs .022 and an 18k for a 400hz low pass for example.  I'd guess this was probably covered in basic audio circuit design 101 but I was absent that day.

[Vivek]

Thats a great question !!

Two filters where R1 x C1 = R2 x C2 will show the same frequency response curves

But their impedance will differ.

Normally there is a "Rule of tenth" or "Rule of tens" (It is just a good design guideline, not a legal requirement. Corners can be cut slightly)

It means: If there are stages that are connected to each other


The input impedance of a stage should be 10 times more than the output impedance of the previous stage


In your example, suppose the low pass filter fed a pot as a volume control

If you use 0.1uf and 4K, the pot could be more than 40K (And whatever is connected to the pot's wiper should have impedance more than 400K)

If you use 0.022 and 18k, the pot could be more than 180K (And whatever is connected to the pot's wiper should have impedance more than 1800K)


Hence, The Product of R and C is decided by the frequency response we need, but the exact value of R and C is decided by the impedance we want.

[amz-fx]

I played around with various clipping ideas a few years ago:

http://www.muzique.com/lab/tclip.htm

Best regards, Jack

[Vivek]

Dear Jack

Thank you for your pioneering work and interesting experiments !

I refer to your web page ever so often.

[Gus]

A post from the past that might be of some help.
I found splitting the signal like a crossover network for a speaker did not sound as good as overlapping the HPF and LPF(this was for a 7 string jazz guitar).
https://www.diystompboxes.com/smfforum/index.php?topic=99697.msg875121#msg875121

[DIY Bass]

I have a bass big muff deluxe that has a switcheable crossover before the fuzz.  I have a feeling though from memory that if you switch in the crossover then only the highs go to the fuzz - the lows go through as clean.

[aron]

Good article Jack!

How did it sound? The problem with most fuzz boxes is that if you let too much bass in, it get's really "messy".
Roll off lows and it cleans up/gets tight but is then too "thin".

I just built another SD9 and it sounds like there's too much lows going through the overdrive. Very buzzy sounding.

[Mark Hammer]

Again, I'll note that differential-clipping can be pursued to different ends.  One, as Aron alludes to, is to yield a more pleasing distribution of harmonic content, by either allowing for different bands/ranges of harmonic content to be adjusted in their representation.  This is part of the strategy underlying what the TS-9 and derivatives do, as well as the approach adopted by the Rat and Bluesbreaker derivatives, with dual ground legs providing more gain for some parts of the spectrum than others.  All of them take note of the fact that the brunt of the signal "lives in the basement", as I like to say, resulting in more clipping for lower strings/notes, unless one compensates by reducing the amplitude of lower notes, prior to clipping.

Jack's approach, that he linked to, does not alter how much a given part of the spectrum clips, but rather how much a part of the additional harmonic content generated ends up in the final audio output.  That's not "lesser" in any way, just a particular approach to use of harmonic content; more like harmonics-specific EQ-ing.

The approach adopted by the Anderton Quadrafuzz and the Yamaha MBD-100 is different in that there is no separate post-clip filtering of bands, only tonal adjustment of the aggregated outputs of the individual clipping bands (although Yamaha's Balance control only adjusted for the upper two bands, and allowed the lowest band to be mixed in without any attenuation-control).  I don't know what Yamaha's intention was, but Craig states explicitly that a primary objective was to reduce intermodulation.  Now, just how many bands are truly required to accomplish that is pretty much an unknown, but Craig settled on 4.  I suppose people can feel free to explore more bands, if they wish.  I will simply note that pursuing individual clipping within narrower bands also obliges some means of adjusting proximity-to-clip, since individual bands have differing amounts of acoustic energy, requiring adjustment of gain to reach whatever forward voltage is needed.

With the Flexidrive, I stuck with two bands, chiefly because it was simpler and smaller, and allowed for the reciprocal gain/drive pot to be implemented (something that 3 or more bands would not permit).  But I also adopted filtering both before and after the clipping point.  This provides a sort of blend of the approach adopted by Anderton and Yamaha, and what Jack proposed.  I won't say it is "better", just different, and available for folks to explore, and expand on.

We have an entire history and legacy of processing guitar as a sort of monolithic signal, when it comes to adding harmonic content, with occasional digressions.  One example that comes to mind is the placement of Uni-Vibes ahead of distorting pedals, such that large segments of the guitar signal are moved away from, and closer to, the clipping threshold, as the pedal modulates.  This suggests that there are opportunities for "interesting" sounds via amplitude modulation of individual bands, having their own clipping circuitry.

[ElectricDruid]

Question for ya.  When R-C filters is there a best practice for any of the values?  Playing around with the AMZ calc I see that you can make a filter for any typical guitar freq with a wide combination of values.   .1 and a 4k  vs .022 and an 18k for a 400hz low pass for example.  I'd guess this was probably covered in basic audio circuit design 101 but I was absent that day.

+1 what Vivek said about impedances.

Aside from that, I'd add that it's more difficult to get caps in many values (E6 or E12 series if you're lucky) whereas getting resistors in E24 series or better is simple. So it makes sense to start with a cap value in the right ballpark and then adjust the resistor to taste.
Caps have poor tolerance, and PCBs have parasitic capacitance (of several pF or more) so tiny cap value <100pF are definitely better avoided. Large value caps finish up being electrolytics or expensive or large (or all three) so you need to keep the cap values below <1uF so you're in the common range of the sorts of film caps that are good for audio.
For resistors, you've got more scope perhaps, but again, there are practical limits. Large resistors >470K start to produce appreciable amounts of noise, so you want to avoid huge R values too.

Luckily for us, the mid-value-range around nF caps with KOhm resistors produces the audio range we're commonly after, so meeting these practical guidelines is rarely onerous.

HTH

[Steben]

This may be a little more complicated than it needs to be.  The Big Muff overdrives use capacitive coupling through antiparallel diodes in a transistor feedback stage to get isolation of low-frequency signals from the diodes to a greater extent than high-frequency signals.

For another approach, you could use a state variable filter with high, low and midrange frequency outputs and only apply distortion to the midrange.  The lows would get too muddy if they were clipped and the highs would get additional unnecessary harmonic content, so if you just apply distortion to the mids, the world becomes a better place.

The Big muff stuff can be used on any opamp (loop) clipper but it almost never is used. I wonder why. flat response at low gain and bigger low response at high gain surely is a dynamic effect.
A quality rather MUFFled under the typical scooped Big Muff tone control.
One of those interesting paths to work on ...if one has time.

[Vivek]

The Big Muff overdrives use capacitive coupling through antiparallel diodes in a transistor feedback stage to get isolation of low-frequency signals from the diodes to a greater extent than high-frequency signals.

Here is a graph of the current in the caps before the antiparallel diodes (BMP Triangle)

[aron]

I need to try this. I have that tube screamer clone I modded with sockets for the diodes. Perfect for this.