Crossover distortion

[Mark Hammer]

So, as I understand it, because of the way in which diodes will decline to conduct until the forward voltage criteria are met, if one places a back to back pair in series with the signal path, then any activity in the signal which falls below the forward voltage simply won't show up at the output.  And of course, since the diode will conduct at the forward voltage, you end up with a quasi square wave at those turn-on and turn-off points that has properties of pulse-width modulation, such that there are little gaps between successive peaks but everything above that point where the diode conducts is essentially nicely linear. 

A couple of questions arise from this little prod to my natural curiosity:

• What sorts of diode preferences ought one to have with respect to deliberately producing crossover distortion?  For instance, are fast-recovery diodes better than Ge, Si, or LED for this job?  This assumes one would employ appropriate gain stages preceding any diodes.
• Would it be better to have one series diode stage with a higher threshold (e.g., back to back pair of LEDs, or 2+2 pair of 1N4001), or several cascaded stages?
• Are there some bands where crossover distortion "sounds nicer" than other? For instance might it be preferable to mpose crossover distortion on content below 600hz but something different above that point?  Extending from that, might crossover distortion sound "nicer" if one used different voltage drops for different bands (LED for bass, Si for mids, Schottky for highs)?
• Is there anything to be gained by use of series resistances in combination with series diodes?  For instance, could I effectively produce a PWM-sounding footpedal by sweeping a variable resistance in series with one or both opf a pair of series diodes?

Discuss.

[Joe Kramer]

Hey Mark,

Both the Boss HM-2 and the DOD FX56 (and probably others as well) use crossover distortion diodes, and both use Ge diodes, presumably because too high a turn-on threshold might result in a gating effect.

I'm not sure of your points about the "quasi-squarewave" or of a pulse-width modulation effect.  As you say, the series diodes create a flat spot between the wave peaks, not at the wave tops, which remain essentially round.  In other words, you've got a sort of hybrid signal of AC peaks with dashes of DC in between them.  I guess one could think of a true squarewave as peaks of alternating DC, but in that case the audio signal is produced by the alternating potential, whereas, with the "deadband" distortion, the DC stays essentially at zero, and hence no audio signal.  Would varying the length of these DC dead zones really create a pulse-width modulation effect--that is, a change of the actual harmonic profile (from say odds to evens)?

As for frequency shaping and selective distorting, probably both resistance and capacitance in parallel with the series diodes would be useful, to either moderate the deadband distortion or distort selected frequencies, or both.

Just my 2 cents.

Regards,
Joe

[frank_p]

I've just tried it and at first did not know what you meant in the text Mark.

Than searched a bit on the net and found it on the muzique site:
http://www.muzique.com/lab/sat2.htm

It would be something like the third drawing.

I've tried with some Si 1N4001 : could see the crossover notch on the oscilloscope.
I've tried some Ge diodes with about the same results (escept with one kind of Ge witch did not have any notch).

The more you lower the signal voltage the more the relation of the crossover augment in relation to the height of the sin wave. (PWM ?)

The site say it would be better with Ge or Schottky diodes. (...)

I did not build any kind of effects with this.

[grapefruit]

Hi Mark,

Do you have the Walter G. Jung "Op Amp Cookbook"?
There is a variable dead band circuit in there that you might find interesting. I've been meaning to try it for a while but haven't gotten around to it....

Regards,
Stewart.

[grapefruit]

From repairing solid state amps I know that crossover distortion definately sounds "worse" with some program material, I couldn't tell you how it varies with specific freqencies though. Obviously, the distortion is less noticeable with higher level signals, but it also seemed to sound less annoying with less complex (possibly less high frequency) signals.

Regards,
Stewart.

[DougH]

It's supposed to sound "annoying" to a point, I think. I think it's responsible for the high freq "fizz" you hear in "metal" pedals, but I could be wrong. In any case, if mixed in sparingly it could add some "sizzle" in the right places.

[Mark Hammer]

Following up on these two posts, maybe the "trick" to getting decent sound from crossover distortion is to filter the top end a lot and then re-introduce sizzle via the crossover distortion itself.  Perhaps it is the addition of crossover distortion to a signal that already possesses top end that makes it sound unpleasant.  Just musings, though.  No solid theory underlying it.

[frank_p]

Do you think that there could be a more efficient way of doing it by
- Make two signals out of the input one.
- lowering a first sin wave (opamp) and then cut what is undeneath/near the 0.
- Do the same thing with the "negative" portion.
- Sum back the two signals together.

The amount of crossover distortion would then be adjustable by moving (adding a constant DC) the two signals before "rectifing".
A simple push-pull with a mis-bias control would be simpler (mind me...).

[The Tone God]

I was not impressed with the sound of series diode crossover distortion. The amount of cut off would be dictated by the voltage drop of the diode so different diodes would create different amounts of crossover. Kinda limiting IMHO

Blade is a really good crossover distortion generator that allows the control over of the crossover point. There are some other tricks one can do with that circuit that I did not include. You can see scope shots of it in the FX-X submission thread it was entered in here:

http://www.diystompboxes.com/smfforum/index.php?topic=51877.msg390996#msg390996

Finish Line does crossover distortion too in a more traditional way using a class A/B setup but its charm is more when you drive it to the point of frequency trippling then the crossover distortion it makes at lower settings.

Andrew

[Roobin]

Would running a parallel clean/unaffected line then mixing the two together work? It could act as a blender type knob, to reduce the effect?

[gez]

Years ago I mixed the outputs of a couple of anti-phase precision half-wave rectifiers.  Hardly anything gets sliced from the signal, but there's audible distortion.  Subtle seems to work best.  I dare say that DC biasing the diodes would yield similar results.

[R.G.]

So, as I understand it, because of the way in which diodes will decline to conduct until the forward voltage criteria are met, if one places a back to back pair in series with the signal path, then any activity in the signal which falls below the forward voltage simply won't show up at the output.  And of course, since the diode will conduct at the forward voltage, you end up with a quasi square wave at those turn-on and turn-off points that has properties of pulse-width modulation, such that there are little gaps between successive peaks but everything above that point where the diode conducts is essentially nicely linear.

I think that's correct, although that's not the way I'd have stated it.

Imagine a guitar signal with + and - excursions approximately equal and no DC offset. If you put a diode in series with it, that diode blocks everything with less voltage than the forward voltage of the diode. With another diode in the opposite direction, the new diode blocks everything at a smaller voltage than its own forward voltage. The output is zero when neither diode is conducting; Aat no time do both diodes conduct; so the output is the sum of the bits that one diode conducts and the stuff the other diode conducts.

It's the conceptual reverse of shunt clipping. Shunt clipping cuts off and throws away everything more positive than +X and also everything more negative than -Y. The condition where X=Y is special. Series diodes effectively give you everything the shunt diodes cut off and threw away - only things more positive than +X and more negative than -Y.

Adding the output of a shunt clipper to the output of a series clipper gives, with perfect devices, the original waveform. Of course, nothing is perfect.

Series "crossover" diode pairs give you the procedural inverse of shunt clipping. This "just give me the peaks" process was used in radio communications in the 1950s and 60s for noise blanking. You set a threshold level, and anything over that level is presumed to be noise; the output is muted for those microseconds. This helps a lot in intelligibility where you're doing voice and things like thunderstorms intrude. Imagine AM radio in a thunderstorm, but with the lightning pulses being periods of silence, not blips.

What sorts of diode preferences ought one to have with respect to deliberately producing crossover distortion?  For instance, are fast-recovery diodes better than Ge, Si, or LED for this job?

Leave it to you to ask the obvious but unstated questions. I have not messed with what Fourier says about this, but -

This assumes one would employ appropriate gain stages preceding any diodes.

You're sniffing around what I consider the great secret of distortion, even though I keep posting it. The characteristic sound of a distortion has more to do with how big the signal is compared to the clipping level than it does with the specific clipping level itself or what kind of device did the deed. So assuming appropriate gain stages preceeding any diodes is vastly subsuming the really crucial process.

This is one of the chapters in the book. 

Would it be better to have one series diode stage with a higher threshold (e.g., back to back pair of LEDs, or 2+2 pair of 1N4001), or several cascaded stages?

Interestingly enough, it makes no difference, this being the inverse of shunt clipping. By doing the first pair, you clip the "core" out of the waveform, and the wave bottoms sit at 0V. The next series clipper subtracts more of the core out. This is precisely the same as if you had used both sets of diodes in one stage. You get to the same answer subtracting 7 from 18 as subtracting 4 from 18, then 3 from 14.

Are there some bands where crossover distortion "sounds nicer" than other? For instance might it be preferable to mpose crossover distortion on content below 600hz but something different above that point?

Good question. No good answer right now.

Extending from that, might crossover distortion sound "nicer" if one used different voltage drops for different bands (LED for bass, Si for mids, Schottky for highs)?

I think this is modestly immaterial in some ways. Of course there will be differences, but the differences will be those of comparison of the signal magnitude in the band with the voltages clipped out. By filtering first, you select out a frequency band with some signal level. The frequency domain transformation is on that signal level, and is dependent on the relative size of the signal and the diodes. Then you add back to the rest. What it does do well is remove intermodulation with other frequency bands, along the lines of the Quadrafuzz. That probably improves things overall, but it's something that's always good, any distortion style.

Is there anything to be gained by use of series resistances in combination with series diodes?  For instance, could I effectively produce a PWM-sounding footpedal by sweeping a variable resistance in series with one or both opf a pair of series diodes?

And here's the winner! A resistor of 0 ohms in parallel with the diodes completely eliminates the distortion. A resistor of effectively infinity compared to the source and load impedances lets the diodes control the sound. In between it's, well, in between; with the dead flat place between + and - pips filled in with tilted places. What is actually happening is that this resistor is adding back in the "clipped out" shunt clipping portion that the series diodes had removed, but in varying proportion.

[frank_p]

Would running a parallel clean/unaffected line then mixing the two together work? It could act as a blender type knob, to reduce the effect?

This what Mark asked before in a kind of way.  What I would see more is putting it in a circuit such as the harmonic enhancer:
http://www.mif.pg.gda.pl/homepages/tom/files/harm.jpg

Finish Line does crossover distortion too in a more traditional way using a class A/B setup but its charm is more when you drive it to the point of frequency trippling then the crossover distortion it makes at lower settings.

I have trouble at putting a "pictures" on these words Andrew.  What could I do to understand that ?

[The Tone God]

I have trouble at putting a "pictures" on these words Andrew.  What could I do to understand that ?

It hard to describe without scope shots but when the class A/B setup is driven with more input signal the output will "fold" through the transition point and create a rough recreation of the original signal at three times the input frequency.

Andrew

[frank_p]

It hard to describe without scope shots but when the class A/B setup is driven with more input signal the output will "fold" through the transition point and create a rough recreation of the original signal at three times the input frequency.

I suppose it's not possible to do it with that diode pair?   You really nead a class A/B stage ?
Is that it ?

[puretube]

There are other ways...

[Mark Hammer]

Like the Autoctave and the Tri-fold.

The X-over distortion scope-movies you have remind me a lot of the many waveform shapers from those old Elektor circuits in the various summer and X-mas circuit collections.  I'll need to take a look at them again this evening.  I have no doubt there will be useful content there.  The only problem is that most of them are assuming a robust steady (and simple) VCO output to work with, and obviously that's not what a guitar provides.

[frank_p]

There are other ways...

Thanks Puretube !  It gives me some "light".

Ermh...  Will I drill those holes int those enclosures today ?

I guess having homemade enclosures helps motivation to get to that pressdrill...

[puretube]

and then some other (with sound):
cleany:
http://www.youtube.com/watch?v=z2fRXR-V_rw
gainier:
http://www.youtube.com/watch?v=Re2aKbVCeTc

[kvb]

although I can not comment using theory I am going to mention some observations made while working on a distortion behemoth.

I had a switch that engaged either clipping diodes to ground or GE diodes in series. The clipping diodes are an Si and a red LED.
I read the X-over distortion info at AMZ and took a good look at The Tone God's "Blade". Looking at the scope pics it seemed like the wave was more tweeked when the controls were not at center. This made me think that two different types of diodes could give a more noticeable result.

So now the switch takes the Si diode and the red LED and places them either to ground or in series with the signal.(signal goes 'through' the LED' - and it lights up.
The diode circuit is Jack's, and it is the one with a pot that can blend clippers with clean signal.

Driving this is a two stage Jfet set-up with a gain control in-between. Following that is an op-amp that is engaged along with the clipping diode switch. The op-amp is set up like a typical gain stage (I don't remember what value pot is in the feedback loop).
Lets just say though, that there is a ton of gain/volume here. It gets really loud and will oscillate.

The Observation is that the Si LED pair gives a very noticeable effect and the large signal being thrown at it makes it all happen.



Augh, I'm getting kicked off this computer . . . gotta go!