Producing crossover distortion

[Mark Hammer]

When introducing any sort of diode-based distortion, the problem one is invariably faced with is that diodes have fixed absolute forward voltages.  If one was going to attempt some crossover distortion by placing diodes in series with the signal, increments to the signal level would increase the output but have no bearing on how side-clipped the signal would be.  What I'd like to be able to experiment with is essentially "pushing in the sides" of the waveform, such that the circuit behaves as if there were continuously variable forward voltage.

So how does one accomplish this?  Is it a case of feeding both the signal and a DC voltage to the same "entry point" at the start of a back to back pair (e.g., a pair of LEDs) so as to vary the "distance" between the forward voltage and the DC bias, and then sticking a DC-blocking cap after the diodes to subtract the bias voltage?

Being able to vary the sides of thw waveform would yield a tone similar to varyig the duty cycle (i.e., PWM), wouldn't it?  That would make a foot-controlled effect in a wah shell.  The fact that one could independently adjust the amount of amplitude clipping would lend some pleasing variation to the sound.

[JDoyle]

Mark I don't know if I get what you are going for completely, but off the top of my head a 'Vbe Multiplier' or an opamp based 'ideal diode' with your variable voltage as a referece point sound like they may do the trick...

....but the Vbe multiplier would be a bit tricky as the signal would have to bias it on and the additional resistors increase the current demand, maybe too much for a line leve signal...

[gigimarga]

Very interesting Mark...now I makes some tries with the crossover distortion in a New Clipper, as you suggested in a topic about Gretsch Controfuzz and as I found in Boss HM-2/HM-3 schematic (by the way, the crossover diodes where must to be placed: before or after the clipping ones?).
So I am very curious about what you will discover

[alanlan]

I did some experiments some time ago - may help in your quest...

The circuit puts a diode clipper in either the forward path or feedback path of an op-amp (or somewhere in between).  The plots below show the results of extreme settings of R6.  It's the type of circuit where you get a G or 1/G response with the diode clipping circuit being "G" (even though it is non-linear).  I did it for the same reasons i.e. to experiment with cross-over type distortion.  It can be made to sound pretty crappy (I did bread-board it at the time but it didn't find itself in a box).  Feel free to have a play if you think it's worth it.

[Cliff Schecht]

When introducing any sort of diode-based distortion, the problem one is invariably faced with is that diodes have fixed absolute forward voltages.  If one was going to attempt some crossover distortion by placing diodes in series with the signal, increments to the signal level would increase the output but have no bearing on how side-clipped the signal would be.  What I'd like to be able to experiment with is essentially "pushing in the sides" of the waveform, such that the circuit behaves as if there were continuously variable forward voltage.

So how does one accomplish this?  Is it a case of feeding both the signal and a DC voltage to the same "entry point" at the start of a back to back pair (e.g., a pair of LEDs) so as to vary the "distance" between the forward voltage and the DC bias, and then sticking a DC-blocking cap after the diodes to subtract the bias voltage?

Being able to vary the sides of thw waveform would yield a tone similar to varyig the duty cycle (i.e., PWM), wouldn't it?  That would make a foot-controlled effect in a wah shell.  The fact that one could independently adjust the amount of amplitude clipping would lend some pleasing variation to the sound.

I think the trick with what you're trying to do is to set the level going into the diode stage as this would control how much swing you get vs. dead time (crossover time). This is what I got when I simulated two diodes in parallel (one forward, one reverse) in series with a sinusoid. What you could do is have a dual pot controlling an op amp stage before and after the diode pair. You would simply amplify the output of the diodes by the same amount you attenuated previously. Controlling the dead time is another issue entirely and really the only way I can think to do this is to add some DC bias to the signal before the clipper stage. As done in the PAiA 2720 LP filter and Steiner filter (Ken Stone's at least), varying the DC voltage to a diode's anode can control the forward voltage of the diode (before it "saturates" at it's fully-conducting forward voltage). The Steiner filter does this with a very clever differential circuit which you could perhaps take advantage of here.

Varying the crossover distortion isn't going to quite give you a PWM type sound. Taking a sinusoidal input and adding crossover distortion to it will increase odd harmonics depending on how much distortion is present. With a relatively small input signal, the distortion will be very pronounced and might sound squarish. Relatively high input levels will conversely sound much less distorted. I don't think this circuit is going to have a large effect on phase however, which is where PWM circuits get their characteristic sound from. The harmonics of a PWM signal (all odd too) move in and out of phase with each other which not only starts canceling out the higher order harmonics but also gives you a pleasant phaser type sound. With a changing pulse width, you get a varying amount of perceived power depending on the duty cycle. The strongest sound is at 50% D.C. and as you start to get closer to 100% or 0% D.C. the sound gets weaker and more nasally, characteristic of out of phase harmonics.

[Quackzed]

variable resistor in series with back to back diodes?

[brett]

Hi Mark
I hope I'm understanding this question...
One example of cross-over control (aimed at minimisation) is found on the secondary winding of the driver transformer in old class B transistor amps.  Basically, it is no good to connect the transformer secondary to the output transistors because they need 0.7 V before they turn on.  What was done was to apply 0.7 V (or 0.3 V for Ge) to the centre tap of the secondary.  This involves a diode from the centre tap to ground and a current limiting resistor to the supply (typically 1k for a 9V supply).  Some designs use an extra resistor parallel to the diode such that the two resistors (without the diode) would form a voltage divider with about 1 V at the junction.  IIRC the pigface uses a 1k2 and 220 ohm resistors in series, with the centre tap and the diode to ground connecting at the junction.

I've used this arrangement a couple of times (in octaves etc), and it greatly reduces gating and crossover distortion, especially if used in conjunction with boosted signals (e.g. a common-source JFET going flat out). 

This design appears in many places including in the Pigface 9V and the famous Deacy amp.  From the Brian May website:

The Deacy Amp circuit is a 1950s audio style Germanium transistor push-pull circuit, utilising in its front end an AC125 and AC126 respectively, with the push-pull Output stage comprising two AC128 Germanium transistors.
It also features a Driver Transformer and an Output Transformer and is similar to several designs of the time, which can be found in the Mullard Reference Manual of Transistor Circuits (see pages 168, 170 and 171).

cheers

[JKowalski]

You could use a inverting op-amp stage, with two antiparallel diodes on the output and a potentiometer in parallel with them - the lug of the pot being the feedback path (of course, with the inv. gain stage resistor in series)

Take the output from the end of the diodes... So it's like a ideal rectifier but the amount of "ideal-ness" can be adjusted. Looks like this works perfectly on my LTspice. I think you should buffer the pot output before the feedback, though...



Works just like an adjustable voltage drop!


Here's five diff potentiometer positions, 100%, 75%, 50%, 25%, and 0%

Signal: 1 khz, 8V P-P

Diodes: LEDs



(Note LTSPICE simnot same as schematic, load resistor decreased to 1k to remove resistive slant on crossover portion, and buffer added to pot middle lug)

[Paul Marossy]

You could use a inverting op-amp stage, with two antiparallel diodes on the output and a potentiometer in parallel with them - the lug of the pot being the feedback path (of course, with the inv. gain stage resistor in series)

Take the output from the end of the diodes... So it's like a ideal rectifier but the amount of "ideal-ness" can be adjusted. Looks like this works perfectly on my LTspice. I think you should buffer the pot output before the feedback, though...



Works just like an adjustable voltage drop!

That's slick! 

[JDoyle]

That's slick! 

Completely agree!!! Great work!

[JKowalski]

Thanks!

I figured I might as well put a better schematic pic up here in case people want to use this in things. Set up for 9V supply with Vref, and lower level signals...

As was mentioned earlier in the thread, when you add crossover distortion you decrease the volume considerably. Cliff's dual pot controlling a op-amp stage sounds like a good solution to this. I managed to incorporate the stage into the stage already part of my circuit. The dotted line shows that the rheostat and the pot are two parts of a dual pot. Wire it up so that as the voltage drop control goes towards the output side of the diodes, the 100k rheostat decreases.

This gives you a no-amplitude-loss voltage drop adjustable crossover distortion.  

EDIT: There is a weird issue that I just found out, the way the output wave with crossover distortion relates to the input wave's amplitude is not linear. A exponentially decaying sine wave input outputs a steeper exponentially decaying crossover distorted wave. You get a quick decay, basically. Interesting.  



This is with a 0.5V P-P signal input, with the exact schematic above. (same 0%, 25%, 50%, 75%, 100% pot location graph thing as I did before) You see that the amplitude matches perfectly with the signal input at all levels of crossover distortion. EDIT: However, when I adjust the amplitude, this matching goes away. Hence, the fast decay that I was talking about.



Here is the decay:





Okay, now I gotta b-board this. Be right back with my opinions. 

[petemoore]

  Split the signal.
  Send output A to a Dist+ which has a photocell in series with the clipping diodes, very low R at max bright is probably a good photoresistor choice.
  Send output B to an envelope detector / LED lighting circuit for the photocell.
  SHould make harder distortion when attacking, then lifting the clipping threshold would let more output pass the diodes as the notes fade, a Dist-Comp.

[JKowalski]

Interesting...

On the b-board, I upped with signal strength by x3 to get the best sounds out of it. The circuit worked just as in the simulation. (I used a dual polarity bench supply to get DC-coupled signals on my scope, they were just like in the pics I posted for LTspice)

Ill try to describe the sounds:

100% crossover:
Very sputtery, have to play really hard to get anything decent. Decays super quick, makes almost like a background "bell" sound as you hit a note. "duowwwnnnnggggg"

50% crossover:
Pretty neat, the sputter goes away mostly. Buzzing sound in the background of note, like you are playing guitar and a quiet synth together.... The timbre of the synth-y sound changes as the note decays (because the crossover width gets less with smaller signals!) so you get a kind of bzzzeeeeooooowwwwwwwwww..w.ww...... sound in the background. The more guitar-y sound with it is distorted slightly.

5% crossover (can't get any less):
Mostly original note, but the buzzy sound is there! Not as strong but interesting nonetheless, slight distortion.

NOTE: I don't have to fret about the dual pot gain circuitry, that worked perfectly. Stayed almost the same apparent loudness at every angle of the pot, compared to the original signal!

A FFT analysis showed that:

- 100% crossover, odd harmonics were up, amplitudes exponentially decaying, even harmonics half as high, same decay.

- 50% crossover, odd harmonics up, linear decay, even harmonics half as high, linear decay.

- 5%, 3rd harmonic up and that was it.


It ocassionally tickled my fancy, but I think this little circuit is more suited to people who like weird fuzz pedals and noise stuff. It's not classic distortion... well, at all.

[petemoore]

  Interesting approach to achieving the latest Fuzztones !

[Cliff Schecht]

Interesting...

On the b-board, I upped with signal strength by x3 to get the best sounds out of it. The circuit worked just as in the simulation. (I used a dual polarity bench supply to get DC-coupled signals on my scope, they were just like in the pics I posted for LTspice)

Ill try to describe the sounds:

100% crossover:
Very sputtery, have to play really hard to get anything decent. Decays super quick, makes almost like a background "bell" sound as you hit a note. "duowwwnnnnggggg"

50% crossover:
Pretty neat, the sputter goes away mostly. Buzzing sound in the background of note, like you are playing guitar and a quiet synth together.... The timbre of the synth-y sound changes as the note decays (because the crossover width gets less with smaller signals!) so you get a kind of bzzzeeeeooooowwwwwwwwww..w.ww...... sound in the background. The more guitar-y sound with it is distorted slightly.

5% crossover (can't get any less):
Mostly original note, but the buzzy sound is there! Not as strong but interesting nonetheless, slight distortion.

NOTE: I don't have to fret about the dual pot gain circuitry, that worked perfectly. Stayed almost the same apparent loudness at every angle of the pot, compared to the original signal!

A FFT analysis showed that:

- 100% crossover, odd harmonics were up, amplitudes exponentially decaying, even harmonics half as high, same decay.

- 50% crossover, odd harmonics up, linear decay, even harmonics half as high, linear decay.

- 5%, 3rd harmonic up and that was it.


It ocassionally tickled my fancy, but I think this little circuit is more suited to people who like weird fuzz pedals and noise stuff. It's not classic distortion... well, at all.

A quick look at your simulation results says that the distortion is going to have a nasty gating effect. The only way to negate this is to hit the diodes with a lot of signal as you did. Really, what you are going to get is something like the sound of a mis-biased amplifier being pushed into distortion - very gatey and sputtery like you describe.

Your FFT notes also make sense. With 100% crossover distortion you get the longest trail of odd harmonics. The "exponential" decay makes sense because your waveform is really looking something like a triangle wave with a bit of lowpass filtering applied. A triangle wave has a squared term in the denominator of the gain constant that gets multiplied by the sin(wt) term (looks like 1sin(wt), -1/3sin(3wt), 1/9sin(5wt), -1/27sin(7wt), etc..). This matches your results (I think). At 50% pot rotation, your higher order harmonics are going to dissapear and leave you with something that looks more linear although I suspect it's still "exponential" (really a squared term), you just can't see enough harmonics to tell.

Perhaps you could post pictures of your FFT results for us to look at, this is all a guess based on your description .

[JKowalski]

Perhaps you could post pictures of your FFT results for us to look at, this is all a guess based on your description .

I'll have to take some pics of my scope tomorrow then, and maybe record some audio.

We'll see how terrible my FFT description is 

[cpm]

back to mark's first post...

i think the idea is to make the "crossover flatness" proportional to signal amplitude, and not a fixed value determined by the voltage drop of the diodes.

my take would be:
1) separate both positive and negative halves, (with opamp precission rectifier, for exaple)
2) take each half wave, centered around a stated Vref and rectify again, keeping the upper  half again (or lower in each case)
3) put together the halves of firt halves

tune the second splitting bias to get different amount of crossover drop than 1/2

[cpm]

another idea

what about using the lambda-diode idea in place of the standar clipping diodes, the Vf graph suggest it would do some interesting quirks to the wave:

[Johan]

I didnt read all the earlier replies, so perhaps someone allready suggested this....you want the X-over dist to be dynamic, otherwise it wont sound realistic...take an bixonic expandora type circuit, but take the FET section and strap it across two back to back diodes in series with the signal...no/weak signal: no/weary weak x-over, stronger signal: x-over distortion...
j

[Mark Hammer]

First, thank you all so much for your responses.  Some really interesting and very different ideas.  Thanks to Jay for what he sent me off-line (it's a complicated diagram, so give me a bit to wrap my head around it ).

What I initially posted the thread about was not in search of a "dynamic" effect.  Rather, I simply wanted to be able to continuously vary the degree of crossover distortion with a control knob in the same way that I can easily increase gain to produce clipping using the standard diode format.  Where a pair of diodes is placed in series, simply upping the gain will do nothing for how the transition between half-cycles occurs.  It will simply make the result louder.

JKowalski appears to be zeroing in on the sort of change I'm looking for.  And if I could understand Jay's circuit, maybe that one would too!

Certainly one of the things that I hadn't considered before, which has been brought to my attention quite clearly by the scope-sim picturs (which I unfortunately can't see here at work), is the extent to which more crossover distortion and "side-flattening" necessarily reduces amplitude.  I had forgotten that full diode action will subtract the diode's full forward voltage from the signal.  So, if the series diode pair are red LEDs, one might conceivably subtract as much as +/-1.5v from the resulting output signal.  That, of course, would necessitate both a substantial signal amplitude hitting those diodes, not to mention a fairly respectable recovery stage after them, if the signal leaving one's output jack is to have any meaningful output level.