Symmetrical or asymmetrical diodes for clipping?

[JDoyle]

So, if I am following you correctly, because forward-conducting diodes in the feedback loop dynamically adjust the gain, which in turn changes what sort of feedback signal there IS to feed back, there is a sort of self-regulation in place, not unlike the postgain-feedback arrangement of something like a Dynacomp.  In contrast, with a Dist+ type pair of diodes to ground, the gain is the gain, and the threshold is the threshold, and the two do not interact.  If so, then maybe those folks who are futzing around with this or that diode complement in their Tube Screamers ought to have a switch to select between a pair of diodes in the feedback loop or a pair of diodes in parallel with the cap to ground after than op-amp's output.  Conceivably that would switch between less and more spongy/compressed feels.

The way I think of it is this: in the case of feedback-loop diodes, the diodes along with the op amp and it's gain structure constitute a 'block' whereas with diodes to ground you have the op amp gain circuit as one 'block' coupled to the diodes to ground as another 'block'. This version also allows you to see that with the diodes to ground you get two kinds of clipping, the diodes as well as the op amp itself. With the feedback loop, both are interacting.

I must admit that one of my most glaring holes in understanding comes with what exactly happens in the feedback loop situation. I'm pretty sure that there are some great posts way back in the archives by John Greene and R.G. going into this, but it was way over my head at the time and I never went back to look and learn. As I understand it, in the case of diodes in the feedback loop, when the signal crosses the threshold of the diode drop in either direction, the op amp keeps putting out more current to try to make the math of it's feedback resistors work, but because of the diodes conducting all that current to a virtual ground (the - input of the op amp) it can't, so while the output of the op amp isn't clipping, it is held at whatever the diode drop is, it is driving really hard while the WAVEFORM is clipping; then as the signal swings the other way and the waveform drops below the threshold, the op amp has to reduce it's current proportionally (or increase it when the signal hits the threshold in the other direction) as a result and it is this switching from high to low current drive situations in the op amp, and the interaction between the op amp and the diode at the point of the diode's threshold, that gives each op amp/diode combo it's sound.

For example: Take the case of two different diode types, one like an LED with a very sharp threshold voltage so that before the threshold, nothing but leakage current gets through, but once the voltage at the output of the op amp hits the threshold the diode cranks open allowing all the current through. Then take another type of diode with a much wider turn-on area around the threshold (such as a reverse MOSFET diode) where around the threshold voltage the diode conducts current to varying degrees but still isn't all the way 'on' until a certain amount over the threshold; until then certain voltages cause it to conduct, but in a restricted manner (kind of like a voltage controlled resistor with an extremely small range and nonlinear output). Put differently, the voltage vs. current plot for the reverse MOSFET type diode would have a curve around the threshold voltage, whereas with the LED-type it would be a sharp 'knee' or angle at the threshold voltage. In the first case with the LED-like diode, when the op amp's output reaches the threshold voltage, the diode cranks open and causes the op amp to drive to near it's output current limit trying to make it's inputs match (I believe the current will actually be limited by the parallel resistance of the op amp's - input and the resistance to ground at the - input) and then as the input signal causes the output to fall below the threshold voltage, the diode slams off and the output of the op amp has to drop it's current drastically so that it will fit the math of the resistors in it's feedback loop. In the second case, with the reverse MOSFET type diode, as the output of the op amp nears the threshold voltage, the diode begins to turn on allowing current through it, but not all of the current right away, so the op amp can output a current equal to the task of both supplying the current through the diode at that voltage point as well as the current required to make the math of the resistors in it's feedback loop work. [It helps to think about the input signal as discrete points in time rather than as a waveform sometimes, the opamp does, it just reacts as fast as it can to what is on it's inputs at any given moment; the opamp has no idea you are feeding it a sine or a guitar signal]. What this means for the reverse MOSFET-like case is that the op amp has a relatively easier time then with the first case; instead of having to crank to the output current limit as soon as the threshold voltage is reached, as in the case of the 'sharp knee' LED-type diode, the reverse MOSFET-like diode eases the op amp to maximum current output by turning on gradually, which brings the output slowly (or slower than the first case) up to the maximum once the output voltage is sufficient to turn the diode fully on. It has been mentioned by people much more adept at this stuff than me that it this 'transition zone' in the waveform which contributes a given tonal quality to a given circuit. Note that not just the diodes affect this transistion point, the output current capacity and speed of the op amp itself (not to mention the limiting factors of the current through the diodes mentioned above) contribute significantly.

BUT...

In the case of a diode-to-ground clipper the op amp doesn't have to deal with those pesky diodes messing with it; it does it's job as it's supposed to do until IT clips because it hits its headroom limits, hell, it doesn't even know or care that the diodes are there. Whether it's output is above a diodes threshold or not doesn't affect the performance of the op amp in the least. It is going to do it's job and then HAND OFF the signal to the diodes via a cap. This takes the interaction between the op amp's output and the diodes in the case of the feedback loop situation, out of the equation. The output of the opamp is the size it needs to be to make the math of the resistors in it's feedback loop work, if that value is above the diode's threshold, the diodes clip the signal, if it is below the threshold the diode don't clip the signal. Case closed.

Almost. It isn't a large leap to start to wonder HOW the current output of the op amp affects the clipping action of the diodes in this second case. It does matter, but unlike in the case of the feedback clippers, the op amp's current output and the diodes don't interact; in this case, the diodes just REACT to the current and voltage that they are fed by the op amp. What affects the current the diodes see? Several things. First off the VALUE of the resistors in the feedback loop determine the current the output has to source to make the math in the feed back loop work. For example, two feedback loops with the same gain of ten are 100k(output to - input)/10k (- input to ground) and 10k/1k. But the current required to drop the same voltage across the 100k/10k combo is one tenth as much as is needed for the 10k/1k combo. Then there is the current capacity and output ability of the op amp itself along with any resistance in series with the signal. Without boring you further, you can see that if we had reverse MOSFET-like diodes to ground following the 100k/10k combo, the current portion of the output SIGNAL (NOT the output of the op amp itself, remember it handed the signal off and isn't going to react to the way the diodes themselves react to the signal) may be small enough to interact with the diode in its wide transition zone, whereas the 10-fold increase in the current for the same signal in the case of the 10k/1k combo could cause the diodes to race right past the transition zone and slam all the way on (despite the fact that diodes turn on at a certain voltage, the current component does affect the action of the diodes as they turn on).

So in short: everything interacts, they just interact in different ways in the two cases. 

If anyone has any corrections or if I need to be set completely straight because I am absolutely wrong, please do so; this is how I understand it to work, but there is every chance that I could be wrong.

Regards,

Jay Doyle

PS - I called the diodes in the feedback loop an 'absolute value' circuit above, that is incorrect. It is a 'logarithmic amplifier'.

[Mark Hammer]

So, um, we should expect something a wee bit different if we have diodes for one half-cycle in the feedback loop, and diodes for the other half-cycle after the op-amp?


BTW, thanks for that investment of time.  It warrants printing for a more focussed read-through on the bus.

[petemoore]

  So, um, we should expect something a wee bit different if we have diodes for one half-cycle in the feedback loop, and diodes for the other half-cycle after the op-amp?
  Or in a BMP, have GE / Si clipping one half of the waveform, and Si/Ge clipping the other half, provided the stage following the Ge clip has sufficient gain to then make the Si clip.
  The only clipping diodes I'm currently using are DIST+'s BTB Si diodes, not measured to match.
  Often times I'll use a pre-boost, and I don't know how symmetrical it's output is, or how the 741 then swings +/-, perhaps assymetry of the voltage peaks at the diodes produces noticably assymetrical clipping sound, I don't have an Oscope to look check that.
   

[brett]

In the case of a diode-to-ground clipper the op amp doesn't have to deal with those pesky diodes messing with it; it does it's job as it's supposed to do until IT clips because it hits its headroom limits, hell, it doesn't even know or care that the diodes are there.

Is this correct?  I've always assumed that there's a change in input impedance of the diodes when they clip, and that the op-amp sees that.  Maybe ??

[WGTP]

My simple (caveman) perspective is that the TS (diode in loop) type circuit has 1 stage and the Dist+ (diode to ground) type has 2.  Is that correct?   

Someone pointed out that with the Rat shunting to ground at the feedback loop thru a 47 ohm resistor and 2.2uf cap, that the diodes placed there are almost used both ways. 

[frokost]

Interesting read, JDoyle. I think I understand more of my Rat now.

I have very limited experience with experimenting with diodes, but I thought I should share this. I set out to make a rat with switchable diodes to ground. I chose to switch between LEDs and MOSFETs, because I had never tried MOSFETs but likes the Turbo Rat very much. First, I wired the MOSFETs with the Ge diode as shown in the schematic of the FKR of Beavis Audio. This resulted in a huge volume drop when switching from LEDs to the MOSFET clippers. And here I thought that the clipping threshold was supposed to be very high! So after searching around here a bit I flipped the Ge diodes around, and got the results I wanted. Almost.

I couldn't tell the difference between LED and MOSFET clipping. Not at all. Not in sound, not in volume, anything. I thought this could have something to do with what Mark Hammer calls "proximity to clip" and deduced that the MOSFET characteristics wasn't able to shine when shunting to ground. So I changed the switch and made it switch between LEDs to ground and MOSFETs in the feedback loop. Much better! The MOSFETs have a softer character character to them than the LEDs to ground. Still, I'm not sure if the differences justify all those extra parts. When using LEDs to ground I can ALMOST dial in a sound adjusting the filter knob that sounds like MOSFETs in the feedback loop. But of course I tend to have the switch set at MOSFET all the time because that is a much cooler and boutique way to do your clipping .

[gutsofgold]

I can grasp certain parts of JDoyle's post but in simpler terms could someone lay out the differences in asymmetrical and symmetrical? Specifically in a tube screamer using silicon diodes.

[Caferacernoc]

"Has anyone implemented this (is it even possible) in the same fashion as Anderton's Quadrafuzz where the signal is split into multiple differently eq'd bands and then clipped differently and mixed back together?  Even if it wasn't necessarily into as many bands as the QF, a Hi-Lo split with different clipping sounds like it might be cool in theory. "

    I think the easiest way to do this would be with seperated diode clipping to ground for highs and lows using a Big Muff tone circuit. As shown here:

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

[Gus]

Not for fuzz but I have built a box for a friend that plays jazz.  It was not a cross over but more of an overlap box.  This might be something to keep in mind.  We thought about open and fretted string sound changes with open and closed back amps, signal inversion and other things.  That was the starting point.  I found it better not to think "standard" active speaker crossover setups.  A "standard type " active crossover did not sound good

The idea was a closed back solid state bass amp for the lows and and an open back tube amp .  It worked out OK

[DougH]

Now the real thing to do would the be feed a stock TS sine waves  at say  720Hz 360Hz and 1440Hz at different input levels and gain settings(low middle max) and diode settings( a 1 and 2  1n4148 setup).  They do a FFT and look at the harmonics.  Been meaning to do stuff like this just got to make time.

I've done this sort of thing a lot with circuit simulators. It can tell you a few things but doesn't tell you everything. In the end, the bottom line always comes down to what my ears hear. For that reason I don't have any "preference" for this sort of thing. It all depends...

I did tweak Vr a little for more pleasing harmonics in a pedal I designed. I usually don't use diodes too much.

[JDoyle]

In the case of a diode-to-ground clipper the op amp doesn't have to deal with those pesky diodes messing with it; it does it's job as it's supposed to do until IT clips because it hits its headroom limits, hell, it doesn't even know or care that the diodes are there.

Is this correct?  I've always assumed that there's a change in input impedance of the diodes when they clip, and that the op-amp sees that.  Maybe ??

I thought about this too last night and it is a good point. While I'm sure that the diodes affect the following input impedence that the op amp sees, and the op amp will have to react to that, I think it still is different.

With the feedback loop case, you have the opamp and diodes interacting at the same time to produce a clipped wave; the diode's operation works to affect DIRECTLY the action of the op amp as it tries to balance it's two inputs and make the math in the feedback path work, with the diodes there, it just can't make the math work when the diodes conduct, it can't balance the value of the voltage on it's inputs even though it does it's damndest to do so. In the case of diodes to ground, the opamp is able to do it's job (although it still may clip) and output a signal that conforms to the math in it's feedback path. THEN the signal is sent to the diodes to ground via a cap. When that signal is large enough to cause the diodes to conduct, they will drop the value of the impedence the op amp sees, and therefore require it to output a larger current to still make all of the math in the feedback loop work, but it can do that. In the diodes to ground case the op amp CAN do it's job, with the feedback diodes it can't. That is the difference. First, the opamp is still able to do it's job and isn't working as hard as it can but never getting there, as in the case of the feedback example. Second, there is the cap handing off the signal from the op amp to the diodes. The cap does it's job by charging and discharging as the output of the op amp moves positive and negative and while it does this fast it still take time to do this (that is what happens when the phase changes). So when the signal is clipped by the diodes as it's magnitude crosses the threshold of the diodes, it is the CAP, not the op amp, that supplies the initial current that is clipped and sent to ground around the diode's threshold voltage. What this does is effectively take the situation in the feedback clipper where the diode's act of conducting robbed the op amp's output of the current it needed to make the math of it's feedback network work, out of the equation. By the time the op amp has to react to the change in the impedence it sees at it's output from the diodes conducting, the cap has already brought the signal past the diodes conduction voltage and the diodes are on. There is none of the interaction at the threshold voltage of the diodes between the output of the op amp and the diodes; the op amp suddenly sees a lower impedence at it's output, which forces it to output more current to make the math in it's feedback path work, but it can do that. The interaction that occurs around the threshold voltage in the feedback case doesn't apply because at that point the op amp isn't turning on the diodes, the cap is.

I hope this makes sense. The differences are subtle, but they are different.

Jay Doyle

[JDoyle]

I can grasp certain parts of JDoyle's post but in simpler terms could someone lay out the differences in asymmetrical and symmetrical? Specifically in a tube screamer using silicon diodes.

In a stock tube screamer you have symmetrical clipping because there are two of the same diode in either direction so the signal clips when it goes over .6V positive and under .6V negative. If you were to add another diode in series with one of the existing diodes, you would have assymeterical clipping because you now would have a threshold of .6V on one side and 1.2V on the other (two .6V drops in series).

There are A LOT of other things (like two randomly selected diodes of the same type rarely have the same threshold voltage so how close to 'true' symmetrical do you have to be to be considered symmetrical?) but this is the gist of the idea.

Jay Doyle

[Caferacernoc]

"I can grasp certain parts of JDoyle's post but in simpler terms could someone lay out the differences in asymmetrical and symmetrical? Specifically in a tube screamer using silicon diodes. "

Are you are asking the difference in sound between assymmetrical and symmetrical clipping? The simple answer is that symmetrical clipping contains almost no even order harmonics, only odd. Assymmetrical clipping has even and odd order harmonics. Like a single ended or class A tube amp.

[Gus]

But note the output of the distortion is direct coupled to the tone section then a cap to the rest of the circuit.

It might be offset into the tone section with mismatched diodes is a part of the mismatched sound.  To test opamps in a TS circuit I set the gain to max and turn the tone to max treble.

Take a TS  and look at the signal after the clipping section after the before and after the output cap with a scope turn the tone knob.  Look at the peak to peak after the cap with the center line of the scope set to 0 V also look at the shape.  Not a sim but a real build.

[aron]

>In the end, the bottom line always comes down to what my ears hear. For that reason I don't have any "preference" for this sort of thing. It all depends...

Very true. In the case of symmetrical vs. asymmetrical, it's also a touch thing too for me. I'm thinking I like a _little_ asymmetry, but very little. Too much and it starts getting that ring mod type of sound with two notes playing. (minor sixth ring mod sound). That's my test, play a minor sixth (A and F) together and hear how much asymmetry you have.

[ultar]

i like asymmetrical more often than not because the sound seems more dynamic and sometimes more powerful. right now in my RAT i have the choice of 2-1n34a/1-914si or 2-914si/1-blue LED. the first has an awesome fuzziness and the LED/si is loud and brutal.

for a while i had a keeley modded metal zone pedal and in that the asymmetrical clipping sounded cool with chords, but awfully thin and weird with single notes. i think the type of circuit has a lot to do with what can sound good. my rat sounds good with a large variety of clipping options (and i've changed them around quite a bit) while in my big muff only a few types of diodes sound acceptable.

[Brian Marshall]

i dont mind asymetrical clipping, but i dont usually like it when it's done with diodes.  it usually sounds gated or sort of splattery... probably wouldnt be so bad if the diodes werent all the same type.  say if you used a regular si diode in one direction and a si and a ge in the other.

I especially hate asymmetrical clipping in feedback loops.  if its ground clipping it doesnt sound nearly as bad, but i suspect that is probably because the center voltage moves toward the middle after the initial attack.

[WGTP]

http://www.aronnelson.com/gallery/v/WGTP/FeedbackLoop.JPG.html?g2_imageViewsIndex=1

Does the signal path with the diodes constitute a feedback loop?   

[Mark Hammer]

Nope.

If you redrew it with the 4u7/47R path going to ground on its own, and the diode pair going to ground on its own, you'd smack your forehead and realize it was pretty much the standard Dist+ topography.

Funny how a drawing style can confuse, isn't it?

[WGTP]

Yes, I drew it that way on purpose to get a different perspective on things.  Wasn't sure if the diodes would effect the op amp "thru" the 47 ohm/4.7uf cap or not.  I guess the ground offers way less resistance.