Op amp clipping theory

[WGTP]

This probably doesn't have a simple answer, but it's a simple question.  Does the op amp clip in a Tube Screamer type circuit?  I'm wondering because I'm experimenting with using Mosfet clippers in the loop and wonder how much distortion is the clippers, and how much is the op amp.  The clipping threshold of the Mosfet/Schottkey combo I'm using is around 1.5v.  Do rail to rail op amps minimize the op amp distortion?  Are we talking current or voltage?  What's the deal?  Thanks.   

[JimRayden]

http://www.geofex.com/Article_Folders/TStech/tsxfram.htm

Have fun!

------------
Jimbo

[WGTP]

Ok, I re-read that again and found, "While the clipping opamp is not driven into overload, I theorize that the sudden change of gain every time the diodes go into and out of conduction can cause a similar although briefer recovery period at each "corner" of the clipped waveform," R.G.

On down I found, "using a CMOS opamp for the relative softness of the clipping that CMOS based amps give." R.G.  This would seem to imply that the op amp will be clipping.   

[johngreene]

This probably doesn't have a simple answer, but it's a simple question.  Does the op amp clip in a Tube Screamer type circuit?  I'm wondering because I'm experimenting with using Mosfet clippers in the loop and wonder how much distortion is the clippers, and how much is the op amp.  The clipping threshold of the Mosfet/Schottkey combo I'm using is around 1.5v.  Do rail to rail op amps minimize the op amp distortion?  Are we talking current or voltage?  What's the deal?  Thanks.   

Once the 'diodes' in the feedback loop conduct, the opamp is basically a voltage follower. So the opamp will not clip unless the input is driven with enough level to drive the opamp against the rail. You would need a 9V p-p input signal (approx.) to do this. But, as RG says, the opamp can 'impart' it's own characteristic in other ways.

--john

[WGTP]

Cool, that means it's almost all Mosfet clipping.   

[Paul Marossy]

Hey, that's a cool idea! A MOSFET Tube Screamer... 

But, how does it sound? How much of a change does it make in the sound?

[WGTP]

So far I would have to say that it is smoother and more organic sounding with less treble content, I assume due to fewer high order odd harmonics.  Seems to compress more before distorting.  All as advertised.   

[Paul Marossy]

Cool. That sounds like an interesting thing to try. 

[johngreene]

What you'll find is that the sound you get will vary as much as the gm varies with MOSFETs. A high-gain MOSFET will sound not much different that a silicon diode. Low-threshold MOSFETs are usually high-gain and every one I've ever tried sounded very 'silicony'. Finding the best sounding MOSFET in this configuration can be a loooooong process.

The major drawback: very, very sensitive to battery voltage. Once the battery starts to sag the distortion gets very garbley and harsh sounding.

What has always puzzled me is that with the soft clip of the MOSFET, the low frequencies don't totally 'squelch' out the highs when they clip yet the exact same circuit (Tubescreamer for example) will sound much more bassy with MOSFETs than with diodes. They sound great in the bedroom but don't always cut through the mix with a band.

--john

[Transmogrifox]

Ok, I re-read that again and found, "While the clipping opamp is not driven into overload, I theorize that the sudden change of gain every time the diodes go into and out of conduction can cause a similar although briefer recovery period at each "corner" of the clipped waveform," R.G.

On down I found, "using a CMOS opamp for the relative softness of the clipping that CMOS based amps give." R.G.  This would seem to imply that the op amp will be clipping.  

RG is speculating.  You would really have to get a data storage oscilloscope so you could record a couple hundred frames of the waveform right at the output of the clipping op amp to see if this actually happens.

What he's referring to, is that the effective resistance of the diode decreases at an exponential rate.  At the "knee" of the exponent there is a rapid change in feedback resistance.  That parallel cap in the feedback loop combined with the reactive elements internal to that opamp may prevent the opamp from smoothly reacting to the gain change.  It would be like a delay, and the signal would blip to a high amplitude for a short period of time during the transition.  During that time, the transient would actually cause the op amp to clip, and the different nonlinear effect on the transient RG hypothesizes may make an audible difference in the overall tone.

I'm honestly a little skeptical, but I don't rule it out completely because I know that closed loop high-gain feedback amplifiers don't always respond nicely to sudden changes in feedback impedance.  It's very difficult to figure such a nonlinear event into the feedback loop equation and manipulate it in a way that reveals such effects as RG has presented.  Again, the only practical way to find out is to record it with a storage oscilloscope.  If it's a significant effect, you could probably even put a function generator or audio test oscillator on the input and record it into your sound card.  You could even play your guitar through it and record it in the sound card.  The problem with going into the sound card is that the antialiasing filter on the front-end of the A/D converter may "hide" some of these anomalies.  My thinking is that if it makes a distinct audible difference, it would occur for a long enough period of time that you would see *something* strange at the corners of the clipped signal.  There would be a sigificant spike resulting from something that actually clipped at the rail of the op amp.  Through the antialiasing filter, this would still look like a "bump" of some sort--though the phase response of the filter may be enough to disguise it.  This is the reason you would want to come right off the output of the clipping stage, and not through the tone control stage.

[TimF]

The opamp doesn't clip its self as stated in an earlier reply. I think about it as how the opamp "reacts" to the clipping created by the diode/capacitor combination in thr FB loop.  Every opamps design is going to produce a different harmonic structure. You just have to figure out what you like best. I find experimenting with different combos fun.

[R.G.]

RG is speculating.

He for sure is!!

What he's referring to, is that the effective resistance of the diode decreases at an exponential rate.  At the "knee" of the exponent there is a rapid change in feedback resistance.  That parallel cap in the feedback loop combined with the reactive elements internal to that opamp may prevent the opamp from smoothly reacting to the gain change.

So far, so good.

It would be like a delay, and the signal would blip to a high amplitude for a short period of time during the transition.  During that time, the transient would actually cause the op amp to clip, and the different nonlinear effect on the transient RG hypothesizes may make an audible difference in the overall tone.

Actually, that's way more than I said, or intended to imply.

Opamps vary a lot in their recovery from overdrive - being driven to clipping. Some take this gracefully, recovering softly and cleanly, and others have funny shelved edges and non-symmetrical slewing in the recovery from overdrive. The TS clipping stage *does not clip* - it responds to the change of conduction of the diodes, and I don't think, nor did I mean to imply that there is a transient hiding in there. There could be, but I don't have the equipment to see it, and have not looked.

What I have noted is that there is a correlation in the smoothness of the sound produced in a feedback-diode clipping stage, like the TS, and the smoothness of the opamp's recovery from overdrive. The JRC4558 happens to be very nicely behaved for a bipolar opamp of that era.  But it's not the only one. When people started popping in with their own results on substitution, I noticed that the results mirrored mine, and that the only predictor I could find in any data was that the cleaner the overdrive recovery, the better the opamp seemed to work in a clipper.

It may be that there's an internal transient overload of the input stage or something that never gets out of the opamp, but still leaves the internal transistors in recovery for a few microseconds. It is speculation - but it seems to fit the observed data.

I'm honestly a little skeptical, but I don't rule it out completely because I know that closed loop high-gain feedback amplifiers don't always respond nicely to sudden changes in feedback impedance.  It's very difficult to figure such a nonlinear event into the feedback loop equation and manipulate it in a way that reveals such effects as RG has presented.

I'm a lot skeptical - but I haven't found any other predictor for what might sound good there. And, as you note, there are strange things happening in high feedback amplifiers inside the feedback loop.

I think about it as how the opamp "reacts" to the clipping created by the diode/capacitor combination in thr FB loop.  Every opamps design is going to produce a different harmonic structure.

And that's a good way to look at it.

[TELEFUNKON]

anybody took the nominal slewrates of various opamps into account
when comparing their "sound"?

[Transmogrifox]

Actually, that's way more than I said, or intended to imply.

you can put my name on that theory    Now I see what you were getting at, and that seems a little more down-to-earth than my eisegetical interpretation of what you were indicating.

[brett]

Hi.

That parallel cap in the feedback loop combined with the reactive elements internal to that opamp may prevent the opamp from smoothly reacting to the gain change.

Interesting theory.  The frequencies at which the gain is affected by the 51pF cap in the tubescreamer are very high (but as low as about 50kHz at maximum gain if I recall correctly).  However, there might be some subtle effects.  And the diode has some capacitance, too.
Anybody use a much larger or smaller cap than the stock 47/51pF?

[R.G.]

you can put my name on that theory

No need for embarassment - I was not very clear there.

The frequencies at which the gain is affected by the 51pF cap in the tubescreamer are very high (but as low as about 50kHz at maximum gain if I recall correctly).

I make it about 6.2kHz. F = 1/(2*pi*500K*51E-12) = 6244Hz if I punched the buttons correctly. Of course, that varies wildly as the diodes come into conduction.

[WGTP]

Excellent discussion.  Thanks to all.  I appreciate your assistance in helping me understand this stuff, without actually having to be an EE. 

It appears that MOST the distortion I'm hearing is from the MOSFET clippers.  The op amp currently on the bread board is the a CMOS TI with the lowest slew rate I could find - TLV2372.  I'll have to try some others ... again.   

[Mark Hammer]

1) Slew rate in op-amps, unless particularly low, and the gain attempted particularly high, is generally moot when it comes to guitar stuff. 1v/usec is probably more than adequate enough for what we do.  Many of the "better" op-amps do have noteworthy slew rates like 13v/usec, but that isn't why we use those op-amps.  If you're trying to get gains of 400 to mic up cymbals and a violin section, then you'll likely want something that is capable of responding a little faster than that.  Repeat this mantra: processing one limited-bandwidth instrument and processing many instruments with full 20-20khz bandwidth are two different things.

2) Country/blues slide player Lee Roy Parnell apparently swears by his Ibanez MT-10 Mostortion pedal.  I won't say it is "just" a Tube Screamer but it isn't too far off either.  A major difference is the use of a CA3260 dual MOSFET op-amp.  Never tried one myself, and haven't spent much time listening to Mr. Parnell other than an ACL appearance some 5 years ago that impressed me.  I'm just repeating what he said, and what I know about the pedal.

[WGTP]

Whoops, shouldn't have said that.    It's also one of the cheapest CMOS.  The CA3260 is one of the 2 dozen different op amps I have, I'll try that along with the Hi Mojo 4558's...  IIHO the diodes are where it's at.   But a cap value change is easily detectable too.    I'm not EVEN going to get into cap type.

I haven't been using a cap in the loop because noise and excess treble aren't an issue, but it appears that it potentially has more effect than that.  Where is that silver mica...    The things I go thru just to try and get the most wonderful magnificant thick creamy distortion sound in the galaxy.