Can Feedback loop soft clipping be done with a discrete transistor stage?

[R.G.]

Soft clipping is one of those things that is easier to talk about in the frequency domain than time domain. My own personal definition of soft clipping is clipping that produces as much of the low order harmonics - up to about #6 - than the higher harmonics, 7 and higher. The higher the harmonic, the harsher the sound.

This was in fact the reason behind the JFET, MOS and Mu Doublers. FETs have the quirk that most of the distortion they produce before noticeable clipping is second harmonic, with a tiny smattering of others. The idea was to cancel out the original signal and leave only the even-order harmonics, primarily the second.

However, you can get an idea about harmonic content from scope traces (or the simulator equivalent). It takes high harmonics to make sharp corners. The sharper the corners of the clipped waveform, the higher the harmonics needed to shape that corner.

"Distortion 101" (http://www.geofex.com/effxfaq/distn101.htm) is my description of this from 1993, updated a couple of times to 2000.

[midwayfair]

Also, would there any tangible benefit to using two MOSFETs in the feedback loop (chained together, so that the body diodes are preventing reverse biasing) instead of LEDs, or will the LEDs (or even silicon) be roughly similar in terms of softness in this situation? Since MOSFETs don't hard clip ever, I wonder if maybe that's even better.

[thelonious]

Also, would there any tangible benefit to using two MOSFETs in the feedback loop (chained together, so that the body diodes are preventing reverse biasing) instead of LEDs, or will the LEDs (or even silicon) be roughly similar in terms of softness in this situation? Since MOSFETs don't hard clip ever, I wonder if maybe that's even better.

If you're talking about the configuration second from the top on Jack's diode clipping page, I've done this in an opamp feedback loop and really liked it. I thought there was a noticeable difference between this and LEDs. As usual, YMMV.

[samhay]

Also, would there any tangible benefit to using two MOSFETs in the feedback loop (chained together, so that the body diodes are preventing reverse biasing) instead of LEDs, or will the LEDs (or even silicon) be roughly similar in terms of softness in this situation? Since MOSFETs don't hard clip ever, I wonder if maybe that's even better.

If you're talking about the configuration second from the top on Jack's diode clipping page, I've done this in an opamp feedback loop and really liked it. I thought there was a noticeable difference between this and LEDs. As usual, YMMV.

+1, but it works much better if you have more than a 9V supply as you need a bit of room to allow the MOSFETs to start clipping.
Also, I am starting to think that 'rubber diodes' (look up the ETI soft clipping thread) can work even better/softer.

[tca]

Soft clipping is one of those things that is easier to talk about in the frequency domain than time domain. My own personal definition of soft clipping is clipping that produces as much of the low order harmonics - up to about #6 - than the higher harmonics, 7 and higher. The higher the harmonic, the harsher the sound.

That is a very clever definition!

[tca]

Since MOSFETs don't hard clip ever...

Sorry but they do! Every device, BJT, JFET, Mosfet,triode, pentode,.., will hard clip if you feed them with a high enough input. Check the introduction of Teemuk's book for reference.

The threshold for hard clipping depends on  the device and on the topology used and bias. Some topologies, for some device, will reveal that hard clipping for higher input voltages, but all can be made to hard clip.

P.S.

Also, would there any tangible benefit to using two MOSFETs in the feedback loop...

The most tangible audible consequence would be the different frequency response due  to the high input capacitance of the mosfet (at these voltages and currents, 9V and mA). I'm not so sure about the "use" of the mosfet characteristics to tailor the clipping.  

[midwayfair]

Also, would there any tangible benefit to using two MOSFETs in the feedback loop (chained together, so that the body diodes are preventing reverse biasing) instead of LEDs, or will the LEDs (or even silicon) be roughly similar in terms of softness in this situation? Since MOSFETs don't hard clip ever, I wonder if maybe that's even better.

If you're talking about the configuration second from the top on Jack's diode clipping page, I've done this in an opamp feedback loop and really liked it. I thought there was a noticeable difference between this and LEDs. As usual, YMMV.

+1, but it works much better if you have more than a 9V supply as you need a bit of room to allow the MOSFETs to start clipping.
Also, I am starting to think that 'rubber diodes' (look up the ETI soft clipping thread) can work even better/softer.

I'm running the circuit I'm working on on 18v, three 2N5457s (no bypass caps) for the preamp. I'm looking at (I think) a ~7V total swing at max gain before I hit a tone control (loses a bit of signal) and then an as-yet undefined "power tube" stage. There's also a mini (mild) compressor involved in the circuit. I was going to run with just the preamp section, but it felt like it was missing some mojo and felt a little stiff, which is why I was looking at some soft clipping so I can keep a lot of the big dynamic swing but enhance the whole compress-y feel of it and amp up the actual distortion. (The preamp doesn't get very dirty ... it's close to the ROG Supreaux Deux at 2:00 on the gain.)

The two MOSFETs have a threshold starting at ~2V all by themselves (that's about 1.2V + .7V for the [DG-S + body diode]), plus the series diode (I'm trying 33K). So that's probably like 3V as the starting threshold for clipping and it goes up a couple volts from there. It might square off above 5V, but I've never actually tried MOSFETs for clipping in an 18V circuit with this much swing, so I really don't know how high they go before they say "@#$! it." To be honest, an amplifier circuit involving feedback loop diodes is probably overkill at that point in the circuit and even a gain of 2x WILL overdrive the transistor ... if I could get the same sort of soft clipping effect and retain a voltage swing north of 2V with something simpler and just follow with a buffer and volume control I'd be a happy dude, but I don't think putting the two mosfets to ground is the solution here. It just doesn't sound right.

It would probably help if I posted a schematic of what I've got so far in a new thread, but I don't want to post something that's really far from sufficiently tested.

This is the ETI thread, right?
http://www.diystompboxes.com/smfforum/index.php?topic=102162.0

I just realized that I HAD seen your rubber diode OD in December, I just didn't pay enough attention to it.

Can something like that still work in the feedback loop of a transistor?

[samhay]

This is the ETI thread, right?
http://www.diystompboxes.com/smfforum/index.php?topic=102162.0
Can something like that still work in the feedback loop of a transistor?

That's the one. Some of teemuk's links make for interesting reading.

Can something like that still work in the feedback loop of a transistor?

I don't see why not. You can use a rubber diode to ground too.


However, One of the key aspects (I think) of the Tube Screamer-type clipping (diodes in the feedback of a non-inverting op-amp), which you initially asked about, is that the gain can never be less than one.
It doesn't work the same way with single common emitter/source transistor stages - which behave more like an inverting op-amp stage - so you do not have quite the same process.

[midwayfair]

Can I still do the diode negative feedback loop with a FET? I've been searching for a bit and I can't find an example. I really don't need much gain at all, I just want a FET's distortion characteristics instead of a BJT.

Does this work?

               10K trim (or whatever)
                  |
   D1/2-33K-Drain--->Output cap etc.
    |
   10uF
    |_______Gate
    |           Source
    100K        |
      |          390R (or whatever)
      |            |
      G           G

(The 100K is small but I think it can still bias; I can't really change it much since it's the tone pot from a SWTC)

[tca]

^ Why not a Mosfet?

P.S.
This might work:

[midwayfair]

^ Why not a Mosfet?

I'm not sure it would make a difference from the BJT, really ... seems like a waste of the transistor. FETs just sound a little different when distorted, that was really my only reason for thinking about using one. But to be honest, the biasing will become a PITA.

This is the schematic I'm working on so that we don't have to rely on my ASCII art, might be a few things that aren't correct, but it's close enough. I've more or less decided to go with the MOSFETs for the clippers, though I did want to try the rubber diode as well. The MOSFETs have "simpler" going for them atm.



I'm not really entirely certain how to figure out the exact signal size coming out of Q3 to figure out where the clipping threshold needs to be set (to avoid just squaring off even the low gain stuff) except to build it and measure it. My best guess is:
-The total dynamic range of a 5457 running on 18V is +/-7.5V
-The 10K+47K divide with the 100K tone pot to knock that down to about 3/5, so 4.5V at max gain, except that that changes when the tone knob is set to max.
-Q4 should have a voltage gain of ... ___? Erm, it's 4x based on the 20K/4.7K, but I don't know how to figure in the base resistor divider. I can tell you that the base is sitting at about 1.5V, the emitter at about 1V, and the collector at 10V with a 2N3904.

Honestly, that amplifier seems like complete overkill to me (and makes the output OMFG LOUD even with the 50% divider before the buffer), but no amplification at all means that the MOSFETs barely clip, and decreasing the collector resistors/increasing the emitter rebiases the transistor closer to cutoff, which I don't really want. I guess the only thing that leaves is the 470K feedback resistor to reduce gain? I would think 2x would be plenty of gain there. I'm not looking for fuzz.

[Gus]

I like BJTs often they are more predictable and they can sound very good with good design

[midwayfair]

I like BJTs often they are more predictable and they can sound very good with good design

Yeah, I just spent a bit swapping around transistors ... the BJT is the right call. It's cleaner and glassier and I don't have to add yet another trim.

The AMZ version of the SWTC didn't work, though. It kills the signal at minimum.

[midwayfair]

I'm still a little unclear on something.



Is R6 doing something OTHER than adjusting the threshold of diode conduction (basically, helping set the total Fv)?

I've had a bit to play around and listen more. First, I've switched to a MOSFET in the test circuit for a few reasons. First, I couldn't get a transistor to stay as clean despite the higher voltage (my guess is because an emitter sits much lower than the MOSFET's source), even though it was definitely much cleaner than a FET (the voltage coming out of previous stages was too high even for a 2N3819). The MOSFET seems to retain treble better. Also, it pretty much dictated a larger R2 (which is 1M, along with R1), which I thought might give me more leeway in adjusting R6.

I've noticed that adjusting R6 works almost more like a clean blend than simply raising the threshold at which clipping occurs. If the resistor is very high, it sounds more like the distortion is in parallel with the clean. It's neat on some level, but maybe not exactly what I wanted.

So I was wondering if R6 were, say ... another pair of diodes, would that actually be different than a particular adjustment in R6?

In other words ...
Drain -> 10uF -> Rx -> DG-S > S-GD -> Gate

vs.

Drain -> 10uF -> D1 -> DG-S  -> Gate
                  |<- D2 <- S-GD |

(D1 and D2 could be different Fv for just the right amount of asymmetrical-ness, but for this question, let's pretend they're the same)

[thelonious]

Is R6 doing something OTHER than adjusting the threshold of diode conduction (basically, helping set the total Fv)?

I believe it is also setting the amount of negative feedback that occurs once the diodes conduct.

I've noticed that adjusting R6 works almost more like a clean blend than simply raising the threshold at which clipping occurs. If the resistor is very high, it sounds more like the distortion is in parallel with the clean.

That seems to be similar to what R.G. and others say in this thread, although they're referring to clipping to ground: http://www.diystompboxes.com/smfforum/index.php?topic=82764.msg687084#msg687084

[midwayfair]

Is R6 doing something OTHER than adjusting the threshold of diode conduction (basically, helping set the total Fv)?

I believe it is also setting the amount of negative feedback that occurs once the diodes conduct.

Forgive my ignorance, but what exactly does that mean? The diodes conduct, the signal clips, and then ... Is it then like having a 100K in parallel with the 1M? Which drops the gain so things don't clip as hard overall?

[thelonious]

Forgive my ignorance, but what exactly does that mean? The diodes conduct, the signal clips, and then ... Is it then like having a 100K in parallel with the 1M? Which drops the gain so things don't clip as hard overall?

Yes, as far as I know, that is how it works. When diode clippers in a feedback loop like this conduct, they appear in parallel with the feedback resistor, allowing more negative feedback through, which effectively lowers the stage's gain. Sounds pretty similar to an op amp feedback loop to me, although it's probably more complicated and nuanced than I'm making it out to be.
____
Edit: Btw, you could also probably create the asymmetry you were talking about by placing a different value resistor in series with each diode. Or another interesting option would be a ladder configuration where you put two sets of diodes in series and then tap another resistor off the junction of the two sets of diodes. One set conducts and places a resistor in parallel with the feedback resistor, then the second set conducts and places a lower value resistor in parallel. Might be softer than you want and/or require higher B+ than you want, though.

[pinkjimiphoton]

jon,
have ya tried using crummy germanium transistors as a clipper like i did here?

http://www.diystompboxes.com/smfforum/index.php?topic=106353.0

very tubey and responsive, with a sweet edge only ge seems to get. i don't see why it wouldn't work in a transistor fuzz, too. will definitely be "softer" than mosfets i bet.

[midwayfair]

jon,
have ya tried using crummy germanium transistors as a clipper like i did here?

http://www.diystompboxes.com/smfforum/index.php?topic=106353.0

very tubey and responsive, with a sweet edge only ge seems to get. i don't see why it wouldn't work in a transistor fuzz, too. will definitely be "softer" than mosfets i bet.

Your project was actually one reason I started this thread, because I had it sitting on the breadboard and wanted to see if it would work in a transistor. But it didn't really act the right (or just same?) way in a transistor's loop. It definitely sounded different than your demo, more like it was just using the transistor as a single diode, or maybe there just wasn't as much voltage gain in the transistor. Maybe I didn't fiddle with the values enough, though, so if you get it working, maybe there can be yet another project using it. I also didn't just want to just duplicate your project, and as a final blow it would have been extra awkward to fit it on the layout I did after settling on the MOSFETs (I basically had half a layout for something that I was adding a final stage to).

Just knowing that the resistor affects the amount of negative feedback was a big help. I'm pretty much ready to post the project I started on Monday with the intention of just tossing something together in an afternoon.

[Zoom]

Yes, as far as I know, that is how it works. When diode clippers in a feedback loop like this conduct, they appear in parallel with the feedback resistor, allowing more negative feedback through, which effectively lowers the stage's gain.

Huh, so can some compressor-like action be achieved from this arrangement? It would be nice to have some real simple, even if dirty, compression stage option for a drive pedal.