Variable Crossover Distortion

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

SSLtech

Well-known member
Joined
Jun 3, 2004
Messages
5,447
Location
Florida (Previously UK)
Evenin' all...

Someone just asked me if it would be possible to "dial in" a variable amount of crossover distortion...  I reflexively replied that it SHOULD be pretty straightforward... but so far i havent been able to visualize how this would be done.

I'm thinking a symettically-increasing "chopping the base off the mountain" to leave the 'peaks' over the raising (in both poles symmetrically) in hte same places... which would be like a fast on-off "gate" kind of circuit, or a lternatively "lowering" the peaks by subtracting the base.... or possibly asymmetrical operation, or  'forward-phase-dimmer' triac-like behavior, where only the leading crossover 'corner' is chopped.

I'm intrigued now, and I'm wondering what a simple/elegant approach might look like.

PS: How's it going, everyone?

Keef
 
Just make a standard transistor output stage like this:

05328.png


but replace the diodes with a variable bias. A "rubber diode" circuit that is a adjustable would work (wiper to transistor base).

Or get a transistor output transformer (the kind that is like a tube amp output) and make the bias variable.

Or just make a standard tube amp output and adjust the bias.
 
Zvex made a pedal called the "machine" that is described as a "Crossover distortion". Pretty crude design, but very interesting sound.

http://3.bp.blogspot.com/_y4AYtND8Hz8/SGjaGMocQ0I/AAAAAAAAAL4/bh9-1etRUrg/s1600/zvexmachine.jpg
 
I've seen Brimstone's pedal named "Crossover distortion", but that's something completely different...

-It's essentially a 'crossover' (as in an audio spectrum-splitter) and two fuzz-boxes. A coincidence in name, but not what I'm interested in.

Hmmm... I think you may be referring to the "machine distortion"...  I'll check it out.

Still interested in how it would be generated. -I'm currently considering splitting the wave into positive and negative halves, then adding  a controllable -VDC to the upper half, adding a 'matched' +VDC to the lower half,  then precision rectifying a second time before recombining. -not very elegant though...

Keith
 
SSLtech said:
I've seen Brimstone's pedal named "Crossover distortion", but that's something completely different...

-It's essentially a 'crossover' (as in an audio spectrum-splitter) and two fuzz-boxes. A coincidence in name, but not what I'm interested in.

Hmmm... I think you may be referring to the "machine distortion"...  I'll check it out.

Still interested in how it would be generated. -I'm currently considering splitting the wave into positive and negative halves, then adding  a controllable -VDC to the upper half, adding a 'matched' +VDC to the lower half,  then precision rectifying a second time before recombining. -not very elegant though...

Keith

Yes the zvex pedal is different and uses a push pull transistor pair without biasing diodes to produce crossover distortion. Not really what you are looking to do, but a cool design none-the-less.  ;D

Wouldnt a push pull transistor pair with no biasing diodes, instead using a seperate floating variable bias supply acheive exactly what you want to do?  It would be interesting if you used seperate bias supplies for each transistor....asymetrical crossover distortion anyone!?  8)
 
Nice to see you around the internets.

Just to get this out of the way WHY??  :eek:

To vary crossover distortion we first need to define it. Crossover distortion is caused by class AB output stage topology. Any delay or imperfect handoff between the positive and negative portion of the waveform causes the artifact we consider "crossover distortion".

Most design effort is to eliminate CD with the most common/effective strategy class AB where both the up and down drivers conduct a modest current making the transition cleaner because devices are already conducting a little before turning on seriously. 

To make variable CD start by unwinding the circuit efforts to prevent it. Typical basic output stages use diodes in the base driver string to bias up the bases closer to on.

A fairly obvious way to  make this variable would be to replace the diode string with a variable resistor. Perhaps add the variable pot in parallel with the diodes so you don't kill the output transistors by turning both on at the same time hard with too much bias voltage.

It is fairly common to wrap negative feedback around the entire output drivers stage to further reduce CD, so connect any negative feedback before our variable bias CD circuit.

If the base bias is reduced to 0V the input will have to swing from -one diode drop to + one diode drop before fully conducting. As should be obvious this is worse for low voltage signals.

Note: with NF connected around the output stage, CD is most noticeable for low level HF signals due to slew rate limitations for the bases to swing fast enough to keep up.

Without any NF the crossover distortion from an open loop class AB topology may be worse than expected and unusable, but then again I still haven't heard "WHY" someone would want to do this.  8)

To perhaps make a more realistic demonstration of real world crossover distortion reconnect the NF around the entire output stage but intentionally slow down the op amp slew rate to reveal the slew related CD artifacts. This might be more practical to demo crossover distortion.

JR

PS: Hope you are doing well...
 
> "dial in" a variable amount of crossover distortion...

Trivial as stated.

But if the crossover cut-out is large enough to scream, while the harmonics come up the fundamental goes down (you are cutting out the middle of the wave). IMHO you may want to compensate this. The late-nite hack is to fudge the drive level down for "clean" and let it rise for "dirt". Here's a coarse approximation, calling for a dual-gang pot. There may be a more elegant path.
 

Attachments

  • Vari-crossover-dist.gif
    Vari-crossover-dist.gif
    19.8 KB · Views: 73
Hello Keef,

Glad to see you back.
I trust you are well.

I would like to send you a PM on an unrelated matter, but your inbox is full.
Could you make some space there, or PM me with an alternate contact.


aomahana


 
I have a recipe that works well and I have proven it digitally.

1. Rectify signal, either top or lower half. Make this part switchable.
2. Run into single tunable parametric EQ band. Choice of topology is unimportant, as low as it has sweepable frequency and gain.
3. DC filter
4. Sum back to original uneffected signal. Phase inversion switch is a good idea for the sum if EQ only has boost and no cut.

This gives a nice variable amount of crossover distortion control. Control is simple with the filter sweep and gain obviously. This works especially for lower frequency material - below maybe 300hz - and can work miracles for bass transients. Tuned (ie. hardcoded) in a specific way it's also the very same circuit as some of the "big bottom" sub bass exciters. Can be upgraded to even more interesting things with log() or exp() (tunable saturation) after the DC filter.


For other things it's gimmicky.

Here's a VST plugin where I implemented this with far too much control. It's outdated - does not work on most platforms - but at least you can hear how it works in the demo sound examples.
http://www.michaelkingston.fi/retroband/
 
Kingston said:
I have a recipe that works well and I have proven it digitally.

1. Rectify signal, either top or lower half. Make this part switchable.
2. Run into single tunable parametric EQ band. Choice of topology is unimportant, as low as it has sweepable frequency and gain.
3. DC filter
4. Sum back to original uneffected signal. Phase inversion switch is a good idea for the sum if EQ only has boost and no cut.

This gives a nice variable amount of crossover distortion control. Control is simple with the filter sweep and gain obviously. This works especially for lower frequency material - below maybe 300hz - and can work miracles for bass transients. Tuned (ie. hardcoded) in a specific way it's also the very same circuit as some of the "big bottom" sub bass exciters. Can be upgraded to even more interesting things with log() or exp() (tunable saturation) after the DC filter.


For other things it's gimmicky.

Here's a VST plugin where I implemented this with far too much control. It's outdated - does not work on most platforms - but at least you can hear how it works in the demo sound examples.
http://www.michaelkingston.fi/retroband/
That sounds like a different flavor of nonlinearity, but not same as crossover distortion.

We have yet to learn exactly what keef is actually trying to accomplish

JR
 
JohnRoberts said:
That sounds like a different flavor of nonlinearity, but not same as crossover distortion.

On the scope it looks exactly like cross-over distortion. Except you get full control over the shape of it. Cross-over distortion is a whole family of distortions. There are symmetric and non-symmetric variants, faster vs. slower slewing shapes etc.
 
Kingston said:
On the scope it looks exactly like cross-over distortion. Except you get full control over the shape of it. Cross-over distortion is a whole family of distortions. There are symmetric and non-symmetric variants, faster vs. slower slewing shapes etc.
In my judgement "crossover" distortion is a very specific artifact caused by solid state class AB output stages, related to device finite turn-on times.

I see little point in arguing about our speculations without more input from the OP.  My answer was based in my ASSumption that OP is trying to learn what more severe conventional crossover distortion sounds like, to better identify more subtle crossover distortion. But we know what they say about ASSumptions.  ::)

JR

PS: Tube saturation/overload makes some very interesting looking and sounding waveforms which are likely responsible for their popularity in (lead) guitar amps, compared to solid state saturation.   
 
Thanks all, (and many thanks for the welcome back!)

John,  a colleague is looking for a particular 'effect' for a particular (single) use. His description (and he's not design-minded, so I'm having to infer from his description) rather made me think that lower level signals being 'punished' more than higher-level signals might suggest that crossover distortion might be something to investigate.

In simple terms, he wants something that -on long tones or chords- would become 'dirtier' as it decayed... and also for example on dynamic, rhythmic playing, the 'stronger' strikes might 'punch-through' more cleanly, and lighter strikes be more 'lost' in the effect...

I knot that in audio post-production, distortion-related effects are generally referred to as 'futz' units... for example the LaFont Telephone Simulator... These are invariably combinations of clippers and EQs/filters, and people tend to try and 'futz' up a signal using the few approaches they have available to them... the limited selection of which tends to lead to a limited 'palette' of results.

I love approaches like the BBC's ring modulator use, which led to the 'Dalek' voice; different tools producing different results... but I've really grown up in an age when the typical maturity of design has made it so simple to almost eliminate various distortions to the point where I'm honestly not truly familiar with their characteristics, and -on a philosophical level- perhaps I should be...

Things like clipping I -and I'm sure most every reader here- am pretty familiar with, both in terms of how it happens, and what it sounds like, both in static sine waves, and dynamic complex signals... I'm curious to know if it might be instructive or enlightening for me to make something which could demonstrate the nature of crossover distortion, making it obvious by unsubtle exaggeration, then  dialing it back -once identified- to show how it is in more 'subtle' manifestations.

It may -or may not- produce the effect which my colleague was envisioning, but it set me pondering.

Keith
 
PRR said:
But if the crossover cut-out is large enough to scream, while the harmonics come up the fundamental goes down (you are cutting out the middle of the wave). IMHO you may want to compensate this. The late-nite hack is to fudge the drive level down for "clean" and let it rise for "dirt". Here's a coarse approximation, calling for a dual-gang pot. There may be a more elegant path.
Brilliantly simple. -I love the approach!

I was thinking in terms of 'addition' and 'negative addition', and completely overlooked the 'just block it a bit' idea!

I'll probably start by building something like this and seeing what it sounds like.

Many thanks!
 
SSLtech said:
Thanks all, (and many thanks for the welcome back!)

John,  a colleague is looking for a particular 'effect' for a particular (single) use. His description (and he's not design-minded, so I'm having to infer from his description) rather made me think that lower level signals being 'punished' more than higher-level signals might suggest that crossover distortion might be something to investigate.
yes...  another characteristic of typical crossover distortion is also worse or increasing with HF due to speed relationship...

A variable dead band around zero crossing might avoid the frequency sensitivity.
In simple terms, he wants something that -on long tones or chords- would become 'dirtier' as it decayed... and also for example on dynamic, rhythmic playing, the 'stronger' strikes might 'punch-through' more cleanly, and lighter strikes be more 'lost' in the effect...
If you think about digital quantization distortion it cleans up for loud signals and gets dirtier the lower the level...

It is against my training to make distortion on purpose, but intentionally trashing the LSBs of a digital stream would cause the level dependent smutz. 
I knot that in audio post-production, distortion-related effects are generally referred to as 'futz' units... for example the LaFont Telephone Simulator... These are invariably combinations of clippers and EQs/filters, and people tend to try and 'futz' up a signal using the few approaches they have available to them... the limited selection of which tends to lead to a limited 'palette' of results.

I love approaches like the BBC's ring modulator use, which led to the 'Dalek' voice; different tools producing different results... but I've really grown up in an age when the typical maturity of design has made it so simple to almost eliminate various distortions to the point where I'm honestly not truly familiar with their characteristics, and -on a philosophical level- perhaps I should be...

Things like clipping I -and I'm sure most every reader here- am pretty familiar with, both in terms of how it happens, and what it sounds like, both in static sine waves, and dynamic complex signals... I'm curious to know if it might be instructive or enlightening for me to make something which could demonstrate the nature of crossover distortion, making it obvious by unsubtle exaggeration, then  dialing it back -once identified- to show how it is in more 'subtle' manifestations.

It may -or may not- produce the effect which my colleague was envisioning, but it set me pondering.

Keith
Have fun...

JR
 
JohnRoberts said:
In my judgement "crossover" distortion is a very specific artifact caused by solid state class AB output stages, related to device finite turn-on times.

Yeah. Well model can do this very limited view of crossover distortion, too.
 
Back
Top