soft clipping the eti way

[duck_arse]

I have a drawing of a circuit, marked "solid-state valve sound, eti, april 1990", and I've had it so long, I don't know where I got it from. the thing is, I can't remember seeing the same clipping method used anywhere else. I'm sure the old(er) hands will be able to come up with a slew of circuits using similar, but I can't.



I haven't tried it either, but knowing how many other schemes for op-amp diode feedback clipping networks there are, another one, and variable at that, surely can't hurt.

[rutabaga bob]

Looks interesting.  Now...who's going to try it first? 
Could you sub a 3904/3906 pair for the trannies?

[duck_arse]

eti used to use the 549/559 as their "low noise/high gain" transistors. I've got them, so I never use the 2N series you blokes are stuck with. perhaps the 2N5089 and its complement would be a better fit.

[slacker]

Funnily enough, I found this Muscman amp schematic recently that has a similar thing http://www.drtube.com/schematics/musicman/gb2.gif. I haven't tried building it but I simmed it and it looked quite nice, much softer than traditional hard clipping, more Tube Screamerish but the tops of peaks still looked like a sine wave. I guess that's where the "valve sound" comes in might give that compressed cleanish valve sound.

[effection]

I'm trying this out as soon as I get home tonight. Will report the results.

[duck_arse]

effection, you'll need to add some buffering of some sort in front of this before it will work good.

and I found the item was still in the same april 1990 issue of eti where I first found it. one of the last issues before they folded, I've only kept it because it has some valve basics in it.

[Paul Marossy]

So has anyone tried building this? I did. I think it's really cool, sounds pretty natural. The only weird thing with it is that when it's boxed as a pedal with a bypass switch, when the 1M pot is towards maximum clipping it takes about 1/2 second for the effect to "turn on". The signal almost cuts out and then comes on the way you would expect it to. When towards less clipping, it works just like I would expect it to.

Any ideas what the heck is up with that?

The only thing I have done differently is added a 1M pulldown resistor to the input & output because it POPPED when switching and added a 100uF cap on the power supply and a 22uF cap on the Vref supply. I didn't try an input buffer, and I don't want to add one as it's quite responsive to picking dynamics as it is now. Oh, and I didn't use high gain transistors, just some low gain gen purpose transistors. Maybe that's what my problem is?

EDIT: It helps if you don't forget to make the connection between the emitters of the transistors and the input.  
So now the gain control also works like I would have thought, although the way I had it miswired also was interesting, but not as useful. Anyway, I still like this circuit.

[Electron Tornado]

Just for kicks, try decreasing R2 and R3, disconnect the emitters, and see how it sounds.

[Paul Marossy]

Just for kicks, try decreasing R2 and R3, disconnect the emitters, and see how it sounds.

I might try that as a seperate build sometime. As far as the stock circuit goes, looking at it on the scope with a simple sine wave, I don't really see much of a difference between one extreme of the gain pot and the other. But when playing guitar thru it they sound a little different from eachother, one sounds a little bit warmer than the other. This is with lower gain transistors, will try some higher gain ones sometime.

[tca]

I'll do a sim on this.

[Paul Marossy]

I like mismatching the transistors, one low gain one (240 Hfe) and a high gain one (550 Hfe)

[samhay]

Interesting idea.
I ran a quick SPICE sim using 2N3904/2N3906 and a TL072. It makes a nice enough looking soft-clip waveform. The pot subtly and progressively squares up the waveform and gives a mild extra boost at 12 o'clock. Over all, it has a gain of about 5.
It doesn't seem to be picky about matching transistors.
If you put a buffer before it (which I would recommend) you can change the waveform quite dramatically by varying the current into the op-amp by changing the value of the resistor to the (-) input. A 1M pot in series with a 1-10k resistor will give a fairly large variation in waveform and gain - low values will obviously have higher gain and the op-amp will start clipping somewhere below 10k).

[duck_arse]

I'm glad this has proved of some use. I don't think I would be able to stop myself from doing something stupid with the base connections, though. it looks to me as though they want to be driven externally, by a filter or something ...... you get the idea.

[Paul Marossy]

I simmed it with Circuitmaker Student Version and it doesn't look anything like what I see on my scope.

Anyway, it works nicely in conjunction with another overdrive pedal. Not so great as a standalone device, but I'm sure it could be tweaked into something that could be.

[samhay]

The thread's been cold a while, but I finally got round to playing with this on the breadboard and figured there was little point starting a new discussion.

I have made some changes, and below is a snippet that can be dropped into e.g. a TS in place of the clipping stage. This version is non-inverting, so the input impedance is high, and I have scalled the resistors down 10x to keep the noise down.
While it will work as drawn, like any distortion stage, it will sound a lot better with some filtering after it, so I would suggest you use the other half of the TL072 (or whatever op-amp you care to chose) to add a touch of LP filtering.

Perhaps more signficantly, I have added series diodes (D1,2) to stop current flowing back through the B-C junction; I borrowed this idea from the mucicman schematic that slacker posted. The addition of the diodes makes the 'distortion' pot more useful. Now, when it is fully ccw (B-Es shored) the transistors are effectively out of the circuit and it cleans up nicely. IMO, this makes the distortion pot a lot more useful.

Edit - I should mention that I quite like the sound of this thing. Definitely has fairly soft clipping with a hint of tube/valve sound to it.

[Mark Hammer]

I take it that the transistors are being used more or less as diodes?  In which case maybe using Germaniums might be interesting?

[samhay]

My take on it is that the transistors are essentially acting as switches (I may be wrong):
The emitter is at virtual ground and the 'distortion' pot biases the base voltage between that of the emitter and collector (+ diode drop). When the signal voltage is large enough to pull the switch open, current flows through the C-E junction via the 10k resistor and diode. This increases the feedback (10k||100k vs 100k of distortion pot) and reduces the gain, thus giving 'soft' clipping.

Edit - I was wrong. Rubber diode it is.

[teemuk]

The transistor is configured as discrete zener. The circuit is also known as "rubber diode" or "rubber zener" because one can adjust the zener voltage with the resistor network.

With two transistors two rubber zeners can be simplified to circuit posted by OP. One can use devices of same or different polarity, the choice only matters in which orientation the devices are connected. The circuit can also be outside a feedback loop and work just as equivalent diodes would work.

http://www.st-andrews.ac.uk/~www_pa/Scots_Guide/audio/part2/page3.html

the thing is, I can't remember seeing the same clipping method used anywhere else.

Aside from this circuit being today's standard in biasing solid-state power amplifiers I have seen it numerously used in clipping purposes too: NAD HiFi amps used it as soft clipper. Music Man amps used it. Fender amps used it. Crate amps used it. Peavey amps used it. Genz Benz amps used it. Otis PA amps used it. Dirk Baldinger (maker of Dualdrive pedals) has a patent of a refined version where the clipping threshold of both transistors is completely adjustable and also made frequency dependent. In addition the new patented ADA "T-Torsion" clipping cell is also based on this idea.. and so on. It's a very, very common circuit.

[samhay]

^The circuit is also known as "rubber diode" or "rubber zener"
Thanks, I figured this must a catchy name like that from antiquity.

The schem the OP posted works fine, but the 1M 'distortion' pot isn't terribly intuative as you get the cleanest output at noon. I added the diodes in an attempt to make it more of an 'effect' by allowing you to dial in the effect of the transistors/rubber diode as you turn the pot.

^One can use devices of same or different polarity, the choice only matters in which orientation the devices are connected.
Right - the musicman schem that slacker posted uses two NPNs oriented opposite to each other.

[samhay]

Aside from this circuit being today's standard in biasing solid-state power amplifiers I have seen it numerously used in clipping purposes too: NAD HiFi amps used it as soft clipper. Music Man amps used it. Fender amps used it. Crate amps used it. Peavey amps used it. Genz Benz amps used it. Otis PA amps used it. Dirk Baldinger (maker of Dualdrive pedals) has a patent of a refined version where the clipping threshold of both transistors is completely adjustable and also made frequency dependent. In addition the new patented ADA "T-Torsion" clipping cell is also based on this idea.. and so on. It's a very, very common circuit.

Cool - looks like it was worth reviving this old thread. Thanks teemuk.