Clipping diodes exposed

[seedlings]

If I had only two sine waves, 100hz and 1Khz,

This is rudimentary, but gives an idea of what combining frequencies looks like.  First pic is a snap of 1000Hz, 100Hz and 1000+100Hz sine waves:




Now a pic of the combined wave clipped:





CHAD

[R.G.]

I had this realization some years ago, and I think I've posted it here before. If not, here it comes again.   

Imagine a pair of silicon clipping diodes, which start conducting faintly at about 0.5V and are fully conducting at about 0.7V. Below 0.5V, not much effect at all on the signal being clipped. Above 0.7V, everything is flat. Well, not completely, because there is still SOME slope, but pretty flat. It's what happens  between 0.5V and 0.7V (in this made-up example) that matters, right? Curve of the knee and all that?

Well, the input signal matters too, and what matters most is what percentage of the signal's cycle time is spent in the knee curve. Obviously, for signals that spend 0% of their time in the knee, there's (nearly) zero effect. The opposite is true too. If a signal is large compared to the knee, it's a nearly square wave, and doesn't vary much with amplitude changes. It stays square.

Likewise with mixed frequencies, you lose the smaller-amplitude stuff entirely when the bigger amplitude signal pushes everything over the top-side clipping limit.

It's only the time spent inside the knee that makes varying, touch sensitive things happen that are not just massively clipped square waves. While there is a market for massively clpped square waves, most people want the smoother, touch sensitive stuff. And that all happens in the knee. It would seem that the critical stuff is the relative size of the signal to the knee, and the portion of the signal's cycle time spent in the knee, not above or below it. There's a lot of finessing to be done within this range, of course, but amplifying or attenuating your signal to spend a lot of its time in the knee is going to be fruitful work, regardless of the exact voltages or shape of the knee.

[teemuk]

While there is a market for massively clpped square waves, most people want the smoother, touch sensitive stuff.

Most people?

Don't be too quick with generalisations. Hi-gain stuff has a huge market, not niche one, and it has had it since what? 1980's? 1970's? 1960's? And "hi-gain" is mostly about those "massively clipped square waves" because the mild soft clipping simply doesn't achieve that same effect.

Then again, you can get touch sensitivity with square wave-ish clipping too; For example, most tube amps (and in fact, many modern solid-state amps / effects too) don't just clip to "square waves" but also bias shift in interaction with the magnitude of overdrive and the harmonic output changes quite a lot due to variance in clipping symmetry or asymmetry. The effect is controlled with magnitude of overdrive so it's a "touch" thing too. Another touch sensitivty effect is "sagging", which practically reduces the clipping threshold in interaction with signal envelope.

IMO, there's actually surprisingly little difference between clipping waveform peaks off with soft knees versus clipping them off at same magnitude but with harder knees. Like in case of Si vs. Ge the most drastically perceived differences do not come from different "slopes" of nonlinearity but from overall magnitude of overdrive applied.

It kinda makes sense if you think about it.

But I think this wanders somewhat off-topic.

[ashcat_lt]

RG, you talk about this "staying in the knee" thing relatively regularly.  At least, I've seen it (and been inspired by it!) a number of times since I started hanging out around here.  One of the biggest issues with diodes is that knee is so small in relation to the linear section before it.  Your example sort of illustrates this.  If you set the peak voltage to just hit 0.7V, then about two thirds of (that side of) the signal swing is still below the 0.5V where things start to get curvy.

It occurs to me just now that what you're describing might be better fit by a very poorly designed compressor.  Usually, we want to filter the sidechain path in a compressor so that the gain reduction does not follow the instantaneous voltage of the signal.  If it's not filtered well enough, then it causes a sort of distortion that we call "ripple distortion", and generally don't want in a compressor.  But if we're looking for distortion...

[merlinb]

One of the biggest issues with diodes is that knee is so small in relation to the linear section before it.  

+1. I disagree with RG that the knee has anything to do with much. Unless you're dealing with valves where the knee is sometimes very broad, or not even a knee at all but a wide curve.

But with SS the knee is always too sharp. If you're "in it" then the distortion just isn't great enough to be a big deal, and beyond it, it no longer matters what sort of diodes you used to get there. What matters more is EQ, and IMO what matters even MORE is the self rectification effect. In other words, the ability of a circuit's headroom to follow the signal envelope, instead of there being a rigid headroom limit like you get in most SS clipping circuits. Valve amps do this naturally - you overdrive them and they continue to be overdriven as the note decays, instead of it suddenly dropping below the clipping threshold and becoming clean again. That's what makes touch sensitivity, more than any knee effect. Most pedals don't do this, so they get boring and predictable after the initial excitement.

[R.G.]

Well, it is *only* my opinion, which is worth what you paid for it.

[seedlings]

Most pedals don't do this, so they get boring and predictable after the initial excitement.

And so we wait for the upcoming Exciting Dirt Drive.  Seriously.  Someone better design one for us hacks.  I'll buy the first PCB!

CHAD

[ashcat_lt]

Well, it is *only* my opinion, which is worth what you paid for it.

Well, you know, it's the carefully considered opinion of one of our most respected and senior members.  I figure that should count for something...

I've been playing a bit with different non-linear processes in the digital world lately.  It's kinda fun, but I can't say that I've got too much in the way of meaningful results to report at this time.  I can that something like a sin function, which is by definition always curved and only hits a true 1:1 slope right at the 0 crossing, is a pretty subtle distortion even when the top of the swing actually hits pi/2 point.  It gets kinda interesting when it goes past that point and the peaks start to fold back down, but that's a specialty sound.

I have also played a bit with digital compressors with attack, release, and RMS time set to 0.  Again, it's mostly just been like plug in and turn the knobs and see how it sounds, but there is quite a bit of room for exploration there.  Different thresholds, different ratios, different "knee size"...  I think this is very much what merlinb is calling "self rectification".  I've heard (or rather, read) other people talk about "side-chain" distortion, which I think works out very much the same.  Different processing on the detector path of a compressor can lead to all kinds of different effects.

Edit - OK, that was all kinda wishy-washy.  I think my experience has shown that the "gentler" curves which keep the signal "in the knee" are not particularly satisfying for guitar distortion.  It really isn't until the signal gets pretty far around the bend that we have any noticeable added harmonic content.  These things are maybe better suited for adding a little bit of "warmth" in a mix context, or maybe for a "not quite clean" feel, but won't generally give us the kind of "crunch" that we usually want when we step on an OD pedal.

[Tall Steve]

So if all diodes clip relatively the same, excepting forward voltage, what exactly does account for the tonal variations between diodes in the same circuit?  Does the dB rise to THD have anything to do with it?

[merlinb]

So if all diodes clip relatively the same, excepting forward voltage, what exactly does account for the tonal variations between diodes in the same circuit?

Well if you change nothing else, then at the very least the diodes will change the headroom.

[ashcat_lt]

Here's something else, might not really make any sense, but I'll throw it out there:

Consider the diode not as a switch, but as a variable resistance which changes based on the voltage across it.  Now consider that this is never alone in the circuit.  With a diode-to-ground arrangement, there is often a capacitor leading into it, which makes the arrangement look a whole lot like an RC high-pass, but since the R is always changing, so is the cutoff frequency, no?  So, in a strange sort of way we have a voltage controlled filter following the instantaneous signal voltage.  How the signal hits the knee, how far "in" it gets", and how long it spends there will have a noticeable impact on the harmonic content coming out.  I guess...maybe...

[pappasmurfsharem]

So if all diodes clip relatively the same, excepting forward voltage, what exactly does account for the tonal variations between diodes in the same circuit?

Like Merlin said.

Imagine running a pedal into a perfect compressor that had infinite clean headroom from the input.
If you had two of the exact same pedals one with LEDS one with 4148s
The pedals would sound the same. on the 4148s with the gained turned down a little and the LEDS Turned up a little to match the clipping ammount.

In the real world the LED pedal would more than likely have higher output volume.

[samhay]

It occurs to me just now that what you're describing might be better fit by a very poorly designed compressor.  Usually, we want to filter the sidechain path in a compressor so that the gain reduction does not follow the instantaneous voltage of the signal.  If it's not filtered well enough, then it causes a sort of distortion that we call "ripple distortion", and generally don't want in a compressor.  But if we're looking for distortion...

I had this thought some time ago. After some considerable breadboard time, I came to the conclusion that really soft clipping is quite hard to hear in most musical contexts.

[Tall Steve]

If you had two of the exact same pedals one with LEDS one with 4148s
The pedals would sound the same. on the 4148s with the gained turned down a little and the LEDS Turned up a little to match the clipping ammount.

In the real world the LED pedal would more than likely have higher output volume.

I understand that, but what about the difference between 4148s and 4001s?  They both have a FVD of around 0.7V, right?  Yet in my own experimentations with a boutique TS circuit I built, I found the 4001s to have a much more mellow sound.  Now going back to Paul R's post, he found that 4148s went from 1% to 10% THD with 7dB of input rise, while the 4006s went from 1% to 10% THD with only 5dB of input rise.   Does this have anything to do with the differing tonalities between theses diodes?

I'm building a Rat with selectable clipping for my EE100 final project, and I'd like to really explain what's going on beyond a snapshot of an oscilloscope.  This thread has been very enlightening .

[pappasmurfsharem]

If you had two of the exact same pedals one with LEDS one with 4148s
The pedals would sound the same. on the 4148s with the gained turned down a little and the LEDS Turned up a little to match the clipping ammount.

In the real world the LED pedal would more than likely have higher output volume.

I understand that, but what about the difference between 4148s and 4001s?  They both have a FVD of around 0.7V, right?  Yet in my own experimentations with a boutique TS circuit I built, I found the 4001s to have a much more mellow sound.  Now going back to Paul R's post, he found that 4148s went from 1% to 10% THD with 7dB of input rise, while the 4006s went from 1% to 10% THD with only 5dB of input rise.   Does this have anything to do with the differing tonalities between theses diodes?

Doesn't that basically mean that with 7db increase in input that the clipping will be the same as with 5 db input increase? AKA both at 10%.

If I understand that correctly.

[Quackzed]

     i experimented at LENGTH (oh yes) with diode ladders and came to the conclusion that you dont have to rely on the smallish diode knee. you can make your own knee out off strings of diodes and resistances. VERY soft clipping knees if you want em. people seemed to worry about intermodulation distortion, but i gotta tell ya, intermod is inherent in soft clipping. the bigger the knee, the more apparent it is.
   this route is the only one i've found other than ts-type feedback loop clipping and really small signals 'barely' hitting diodes to ground type clipping that does this (ok tubes but thats cheating)  both of those are 'soft' type clipping but the ts style is limited by the diodes and -gain of one- above the threshold and the other is limited by the diode threshold and noise due to the signal needing pre attenuation...
      the diode ladder or better series strings of diodes/resistances to ground gives you alot more control over how steeply you clip the signal at each higher threshold... imagine just 5% dist for .6v 10% for .8v 20% for 1v 40% for 1.2v etc... a consecutive clipping slope. it can be done. you need a bit of extra headroom to avoid hitting the rails but nothing crazy... you can even tailor in separate odd/even order slopes. sounds like bs i know, and its not perfect, but you can do it.

[Quackzed]

sorry if i derailed there a bit, just wanted to mention a way of creating your own knee curve with multiple diodes and resistors... sort of on topic but another branch...
ok back on topic

he found that 4148s went from 1% to 10% THD with 7dB of input rise, while the 4006s went from 1% to 10% THD with only 5dB of input rise.   Does this have anything to do with the differing tonalities between theses diodes?

it may be that with a lower threshold the scale of 1% to 10% is smaller just as the scaled down signal would be smaller... so it evens out in the end?

[merlinb]

I understand that, but what about the difference between 4148s and 4001s?  They both have a FVD of around 0.7V, right?  Yet in my own experimentations with a boutique TS circuit I built, I found the 4001s to have a much more mellow sound.

The 4001s will have more junction capacitance, which might make a different if you have a 1Meg feedback resistor and no other capacitor there.

[aron]

> Use diodes (if you must- haha) to set headroom but use EQ to set the "goodness" of the sound.

I think that's what the simple mods have been saying for years. Glad to see some data to back it up.
http://www.diystompboxes.com/wiki/index.php?title=Simple_mods

[pinkjimiphoton]

merlinb,
if ya wanna pm me your snail mail, i will send ya some ge diodes to mess with, i'd love to see the results!!!