Capacitors In Series with Soft Clipping Diodes

[YouAre]

I've read the AMZ Article, Tone Clipping ( http://www.muzique.com/lab/tclip.htm ), but have yet to experiment with it. Most of the examples given involve hard clipping, and I wanted to experiment with a similar tactic in soft clipping.

What if we took a basic Non-Inverting amplifier with back to back clipping diodes ala the Tubescreamer, and put a capacitor in series with one or both of the caps?

My assumption is that as long as the signal is below the threshold of the diodes, we'll have no tonal change. The moment we start clipping, the capacitor gets introduced into the circuit, and we have an active low pass filter.

Can anyone please tell me if my understanding of this is correct, or point out the flaw? Thank you!

[GibsonGM]

Great idea, and great concept, YouAre!  You're noticing how you can combine things in the feedback loop, and use it to your advantage.

If you take a look at the schematic for the TS-9, you'll see a cap after the diodes that connects to Vref (.047 cap thru 4.7k resistor) - this is doing nearly the same thing you are talking about...altho it's sending (probably harsh) highs to ground rather than using the active properties of the opamp.   Feel free to mess around with this idea on the breadboard - I've never done so, and am curious what kind of results you get!  Try very small (pF) thru very large caps, I bet you'll find some range that is really cool!

[amptramp]

If you have a capacitor in series with a diode, the capacitor will charge up to the applied voltage minus the diode drop in one direction.  This tends to reverse bias the diode to the highest voltage applied so far minus one diode drop to the point where it doesn't appear to be in the circuit anymore until the charge gets depleted through leakage in the capacitor.  A capacitor in series with antiparallel (back to back) diodes is a different story because the discharge path is through the opposite diode to the one that provided the charge.  This is sometimes used to ensure the clipping threshold rises as frequency drops.  This prevents the muddy sound of waveforms being clipped to one height and having none of the original harmonics and partials surviving the clipping process as long as the waveform is over the threshold.

[GibsonGM]

Good point (and good info) - I think the OP meant that he wanted to add a cap after the antiparallel diodes, and not just on "one" of them.  I hope, anyway!

[YouAre]

Great idea, and great concept, YouAre!  You're noticing how you can combine things in the feedback loop, and use it to your advantage.

If you take a look at the schematic for the TS-9, you'll see a cap after the diodes that connects to Vref (.047 cap thru 4.7k resistor) - this is doing nearly the same thing you are talking about...altho it's sending (probably harsh) highs to ground rather than using the active properties of the opamp.   Feel free to mess around with this idea on the breadboard - I've never done so, and am curious what kind of results you get!  Try very small (pF) thru very large caps, I bet you'll find some range that is really cool!

Thank you! I was trying to think of the inverse idea of what the High pass filter is doing that you described. I absolutely want to mess with this on breadboard, but I wanted to make sure my assumptions were valid before I began "hearing what I wanted to hear" after testing this. 

If you have a capacitor in series with a diode, the capacitor will charge up to the applied voltage minus the diode drop in one direction.  This tends to reverse bias the diode to the highest voltage applied so far minus one diode drop to the point where it doesn't appear to be in the circuit anymore until the charge gets depleted through leakage in the capacitor.  A capacitor in series with antiparallel (back to back) diodes is a different story because the discharge path is through the opposite diode to the one that provided the charge.  This is sometimes used to ensure the clipping threshold rises as frequency drops.  This prevents the muddy sound of waveforms being clipped to one height and having none of the original harmonics and partials surviving the clipping process as long as the waveform is over the threshold.

Shoot. I should've applied my understanding of a simple compressors envelope detector (e.g. Hollis Afterlife, Orange Squeezer, etc). I initially did mean a capacitor in series with the diodes individually. I just couldn't decide if they'd be connected to the anode or cathode. My gut says Cathode, because we want the diode to block the charge on the capacitor until the applied voltage exceeds the diode drop.

Any insight onto this would be great.

Good point (and good info) - I think the OP meant that he wanted to add a cap after the antiparallel diodes, and not just on "one" of them.  I hope, anyway!

Kinda 

I figured I'd put the same value cap in series with the individual diodes, to make them part of the anti-parallel network. I figured if I put the cap in series with the anti-parallel network, one diode would always be conducting and the cap would always be connected as part of the feedback loop. My goal is to have the cap(s) in the loop only when both diodes are conducting. This way, we get a roll-off of the highs only when the diodes are clipping.



On a side note, I'm still struggling to figure out what a capacitor in series with the anti-parallel diode network would do. The fact that the cap is connected to the anode of one and the cathode of the other HAS to have some weird effect (charging in one end, blocked in the other?), and I don't think it should be ignored.

[duck_arse]

I would have thought the cap would act as a frequency dependant resistor. as the signal freq increases, the resistance (not the right term) of the cap will drop, allowing the diodes to do their clipping. lower freqs will be less affected by the diode clipping. the big muff pi uses this form of freq dependant soft clipping, no?

[YouAre]

I would have thought the cap would act as a frequency dependant resistor. as the signal freq increases, the resistance (not the right term) of the cap will drop, allowing the diodes to do their clipping. lower freqs will be less affected by the diode clipping. the big muff pi uses this form of freq dependant soft clipping, no?

Hmm...I don't think I would be counting on the impedance of the capacitor. I suppose the implementation that you're suggesting gives us frequency dependent clipping, but not in the same way that the big muff gain stage does it. I believe that the Big Muff's capacitor/diode network has a high pass filter effect. My understanding is that this is a low pass.

Rather than frequency dependent clipping, I'm trying to achieve clipping dependent filtering. I hope that makes sense.

[duck_arse]

now I'm lost.

the anti-parallel diodes are in series with a cap, in the feedback loop of an opamp, yes? so the signal reaching the diodes is blocked by/dependant on the reactance (might be better) of the cap. Xc falls with rising freq. would mean a high pass in the clipping section = highs clipped more, with the low freqs getting the gain of the feedback network sans diodes.

to the breadboard!

[YouAre]

now I'm lost.

Welcome to the club. Please join. 

the anti-parallel diodes are in series with a cap, in the feedback loop of an opamp, yes? so the signal reaching the diodes is blocked by/dependant on the reactance (might be better) of the cap.

That is frequency dependent clipping. I won't contend that. However, I want the opposite effect. A filter that is only active after the onset of clipping. It's the capacitor inside the loop I want to affect.

Xc falls with rising freq. would mean a high pass in the clipping section = highs clipped more, with the low freqs getting the gain of the feedback network sans diodes.

The capacitor in the loop is part of the low pass network. So in your setup (capacitor in series with anti-parallel diodes), would that capacitor that normally filters out high end in the low pass (in the absence of the diodes), now simply just clip the highs instead of filtering them out? Interesting.

[GibsonGM]

to the breadboard!

+1, ha ha!   

1) draw this up so we can see what you intend

2) breadboard it, and tell us what it sounds like with small and then larger values of capacitance.  I'm trying to picture what you intend, but it's a little vague still.    We're ALL hitting on some of the possible outcomes, but are all 'around it' and not focused.

So, we need to either simulate what you're meaning and see the response, and then build the darn thing since it's so easy to do!   Only your EARS will tell you if it's a good thing.   Might do exactly what you think, but sound out of phase and ugly...or it might sound like heaven on earth )

[YouAre]

+1, ha ha!   

1) draw this up so we can see what you intend

2) breadboard it, and tell us what it sounds like with small and then larger values of capacitance.  I'm trying to picture what you intend, but it's a little vague still.    We're ALL hitting on some of the possible outcomes, but are all 'around it' and not focused.

So, we need to either simulate what you're meaning and see the response, and then build the darn thing since it's so easy to do!   Only your EARS will tell you if it's a good thing.   Might do exactly what you think, but sound out of phase and ugly...or it might sound like heaven on earth )

You're right. I will simulate it first.

I was hoping someone could chime in on some of the basic engineering principles of this, like what's getting clipped and what's filtered.

[GibsonGM]

Gotta see it first, my man  )

My gut tells me that if you wire it the way I think you're going to, it's gonna clip, THEN filter, as no current will flow until the diodes conduct...

[Quackzed]

i think the highs get full clipped and the lows pass unclipped and in between frequencies go from very soft  clipped at the low end to soft to slightly less flat/hard clipped at near the cap corner freq.
stated more clearly, the highs would be hard clipped and then more and more soflty clipped as frequency goes down till the lows are barely clipped at all...
because its a series string, frequencies have to pass both the cap and the diodes to get to ground... so they have to be high feq AND above the threshold to get full clipped.
thats my guess anyway... 

[GibsonGM]

Gotta see it first, my man  )

My gut tells me that if you wire it the way I think you're going to, it's gonna clip, THEN filter, as no current will flow until the diodes conduct...

Yup.  "Filter" covers the 'high freq' portion of that.    Let's see what he draws up/sims....

[duck_arse]

yer, I'll wait for the drawing too. ohh, wait, it's not me supposed to be drawing, is it?

@quack - that's not a bad name there, q.

[YouAre]

i think the highs get full clipped and the lows pass unclipped and in between frequencies go from very soft  clipped at the low end to soft to slightly less flat/hard clipped at near the cap corner freq.
stated more clearly, the highs would be hard clipped and then more and more soflty clipped as frequency goes down till the lows are barely clipped at all...
because its a series string, frequencies have to pass both the cap and the diodes to get to ground... so they have to be high feq AND above the threshold to get full clipped.
thats my guess anyway... 

I agree. I think the highs may get clipped, but filtered out anyway. The spice simulation hopefully offer some insight.

I will do the sim, but it's unfortunately the lady friend's birthday. I will definitely be tied up this weekend.

[GibsonGM]

I tried it (simply, and set up as:   source >> parallel diodes >> cap   and source >> para diodes cap in each leg).    I get a very linear rolloff...pretty interesting and worth exploring further when I have time...