Designing a "transparent" overdrive/distortion

[JDoyle]

To continue R.G.'s complaint and fill in the bit that will probably cause the most confusion/wonder/'why didn't he EXPLAIN what he said' comment, because it isn't obvious...

if we stick to op-amp based topologies for a second (TS, RAT, etc ...) what we know so far is one tends to produce a more focused hard-clip than the other ...

Translation: Diode's to ground clip harder ('a more focused hard-clip') because they drive the diodes straight through their transfer 'knee' and do not affect the opamp's gain equation like diodes in the feedback loop do. In the feedback loop (see some great posts by John Greene patiently explaining it all to me in a search), the diodes drop the opamp gain to one once their threshold is reached.

what we also know from theory and analysis is that one configuration of the std op-amp gain stage exhibits less harmonic and dynamic distortion than the other (inv versus non-inv) in it's linear region - some hi-fi minded people therefore don't use that topology in their circuits ... it would be silly not to think those facts have repercussions on our design choices ...

Of course you are probably wondering WHICH configuration has less harmonic and dynamic distortion. Right? It would have been nice to have that instead of just the toss off of seemingly 'important' information...

The configuration with less harmonic and dynamic distortion is the inverting amplifier.

This is true because the paracitic input capacitance at each amplifier input changes with the input signal voltage creating second harmonic distortion. But this only occurs in the NONINVERTING configuration because in that situation the voltage on the inputs is changing with the input signal, or in techie speak, the COMMON MODE VOLTAGE is the highest.

In an inverting configuration, because the noninverting input is at a steady DC voltage and the negative input is a 'virtual ground' there is no common mode voltage so the dynamic paracitic capacitance doesn't exsist.

For more information see: http://electronicdesign.com/Articles/Print.cfm?ArticleID=7206

Hope this helps to clear some of the probably questions up...

Regards,

Jay Doyle

[Mark Hammer]

I'm going to drag out my old warhorse concept of "proximity to clip", because it is entirely appropriate here.

Think of diode-based clipping as being a percentage-of-the-time phenomenon.  That is, the greater the average overall amplitude of the signal, the closer the signal will be to getting clipped 100% of the time.  If the signal has minimal amplitude in the first place, and minimal gain applied, and the clipping threshold is high enough, the additional harmonic content may only be added on the initial pick attack and immediately after that things simmer back down to what is "merely" a hot signal with no clipping or coloration.  It is simply too far away from the threshold for appreciable clipping to take place.  As the threshold gets lower and the signal amplitude is boosted through some means, the overall signal comes "closer" to the threshold and clipping takes place over a longer period of the note's lifespan, not just the first couple of milliseconds.  Because the guitar signal decays with time, any attempt to have hard clipping for the entire duration of the note requires doing a couple of things.  First, applying more gain, and second, evening out the resonant peaks in different parts of the spectrum such that nothing ever falls below the clipping threshold...ever.

The corollary of this is that if we wish to, we can move some parts of the spectrum closer to, and further away from, the clipping threshold is we wish.  This is, in fact, the entire rationale behind the design of the Tube Screamer.  The low-end rolloff used keeps the bass notes either well below the clipping threshold, or at no greater advantage (relative to higher notes which usually have less amplitude/energy)  most of the time.  The result is a smoother sounding overdrive.

Of course, the TS is not perfect, and one of the ways in which it is not perfect is that the lowpass filtering required to provide some modest restoration of overall tonal (low vs high) balance creates a noticeable mid-peak.  It could be avoided, but the means to do so would involve appreciably more circuitry.

One way to do so is to use a fixed gain stage somewhat like the Rat, where the low end uses one ground leg network, and the high end uses another.  This provides less overall gain for content below a certain point, but with a relatively flat rolloff below that point (unlike the continuing rolloff in the TS).  Use that to stagger the levels of different ranges, and follow it with a full-bandwidth variable gain stage to set overall drive level before clipping.  If my first stage provide a gain of, say, 4 for lower content, and a gain of 10 for higher content, then following that with a variable gain stage of 10x-30x would get me a cumulative gain of x40-x120 for low stuff, and x100-x300 for the higher stuff.  So, you kleep the low end every step of the way, but you knock its gain down a hitch so that it does not risk clipping quite so hard.

What you do after that is use the de-emphasis approach so standard in Boss and other chorus pedals where there is a different and complementary feedback path for high end in an inverting stage at the output.  This restores tonal balance.

What you should end up with is full-range output signal, with a bottom end that only gets slightly dirty when you dig in, and NO MID-PEAK.  Now this is only in theory.  I leave it to the courageous and diligent to find out if what works in theory works in practice.

Now, an entirely different, and feasible, way of achieving the same end is to use band-splitting such as in the Quadrafuzz.  That can also be used to create different proximity-to-clip for different parts of the spectrum.  You then blend them together.  This would require some capacity to adjust the gain of any particular band to produce greater or lesser clipping, and some control for setting output level of that band (post-clip) to achieve the desired tone.  One of the weak spots of the Quadrafuzz was that it lacked controls for setting individual band gain.

I had a schematic up for a circuit I called the "Flexidrive".  Unfortunately, they cut off access to my photobucket account at work so I can't give you the link at the moment.  the Flexidrive was a primitive version of what you want.  It splits the signal up into higher and lower ranges, clips them individually, and recombines them.  In an effort to keep control/knob count to a minimum, I used one pot to set the upper/lower gain in complementary fashion (i.e., pot wiper went to ground and each pot leg set gain of one channel), a second pot to set the level balance between the two clipping paths, and a 3rd for overall output level.  The initial drawing overlooked the need for some caps in strategic places, but I think I redrew it with them.  If a person used a pair of pots to independently set gain of each half/path/channel, I think you could pull it off.

[johnadon]

  Leave it to me to post completely off-topic. My posts are a lot like my builds: errors in progress. Sorry!

[Eb7+9]

The configuration with less harmonic and dynamic distortion is the inverting amplifier.

This is true because the paracitic input capacitance at each amplifier input changes with the input signal voltage creating second harmonic distortion. But this only occurs in the NONINVERTING configuration because in that situation the voltage on the inputs is changing with the input signal, or in techie speak, the COMMON MODE VOLTAGE is the highest.  In an inverting configuration, because the noninverting input is at a steady DC voltage and the negative input is a 'virtual ground' there is no common mode voltage so the dynamic paracitic capacitance doesn't exsist.

nice to see a sign of informed competence ...

now, then - after this there's the issue of false triggering on the front-edge of a waveform
which helps keep clip-definition high (cuts down on fizz/noise)

something that Tom Scholz addressed when filtering the incoming signal before a clipper in his OD circuit
... which is inverting in configuration - and then ... DC coupled to the rest of the circuit (limiter) ...

many hints of such thinking having taken place ...

why/how this "clipping edge formation" scenario is different in the two TS/Rat topologies serves as basis for an interesting TRANsient response (real time, large signal) analysis challenge ...

further analyzing the role of the "bad" ground in the TS circuit is another thing one could sink  teeth into in terms of explaining how the clipping is frequency selective at the same time ... or why the circuit "at clipping" sounds different when hooking up the return leg of the op-amp gain/clipper to ground instead of bad-ground ...

many unanswered questions and oversimplified explanations over the years as to what's really going on there ...

[PaulC]

Of course, the TS is not perfect, and one of the ways in which it is not perfect is that the lowpass filtering required to provide some modest restoration of overall tonal (low vs high) balance creates a noticeable mid-peak.  It could be avoided, but the means to do so would involve appreciably more circuitry.

One way to do so is to use a fixed gain stage somewhat like the Rat, where the low end uses one ground leg network, and the high end uses another.  This provides less overall gain for content below a certain point, but with a relatively flat rolloff below that point (unlike the continuing rolloff in the TS).  Use that to stagger the levels of different ranges, and follow it with a full-bandwidth variable gain stage to set overall drive level before clipping.  If my first stage provide a gain of, say, 4 for lower content, and a gain of 10 for higher content, then following that with a variable gain stage of 10x-30x would get me a cumulative gain of x40-x120 for low stuff, and x100-x300 for the higher stuff.  So, you kleep the low end every step of the way, but you knock its gain down a hitch so that it does not risk clipping quite so hard.

What you do after that is use the de-emphasis approach so standard in Boss and other chorus pedals where there is a different and complementary feedback path for high end in an inverting stage at the output.  This restores tonal balance.

What you should end up with is full-range output signal, with a bottom end that only gets slightly dirty when you dig in, and NO MID-PEAK.  Now this is only in theory.  I leave it to the courageous and diligent to find out if what works in theory works in practice.

Yeah - I've been messing around with that very thing for a bit.  There's a draw back to the active de-emp circuit though.  Instead of rolling off the high end like the 1st order filter the screamer uses you're gaining up the low end which can clip the de-emp amp if there's a lot of gain. 

There's a really easy passive way to do it to a ts thing that only takes adding one resistor, and then changing a cap value.  The low end roll off set by the 4k7/47nf is a shelf eq.  If you change the output 1st order to a shelf that's a mirror oposite of the low end circuit you're back to flat.  The bass shelf is effected by the gain a bit, but that also happens in the right style circuit.  Here's a link to some curves showing how it works

http://i58.photobucket.com/albums/g259/ocluap/tscurves.png

The top curve is the low end roll-off.  The middle is the high end roll-off, and the bottom is what you get when you're done.  In the stock TS you can see that 15db 700hz mid bump.  In the second set you can see how changing the 1st order to a shelf flattens it all back out.  Of course it doesn't sound totally flat when clipping because you've got a bunch of new harmonics, but with the gain on zero it's flat, and when you roll off your guitar volume to a clean setting you're back to flat.  You do loose some output gain though (no 15db mid peak), so you might want to reamp it later.

I've got a couple of odd values because I wanted to show it with the stock 1k resistor.  91 ohm/2.2uf is close enough for real world values.  You're wanting to use the same values as the gain elements (51k/4k7/47nf) just scaled so the 51k  = 1k.

If you make the 92 ohm resistor a pot - say 200ohm then it acts sort of like a bax treb control.  0 is treble cut/ 5 is flat/10 is treble boost.

Later, PaulC
Tim & timmy pedals

[PaulC]

Forgot to add - a flat curve to me works best with a fat/mid rangy amp.  Something like this into a twin will sound like scooped death metal...  You need those mids to fill in that big mid hole those amps have.  A cool thing to do is to change the tone control to a mid peak control to get that back. 

Later, PaulC
Tim & timmy pedals

[YouAre]

Forgot to add - a flat curve to me works best with a fat/mid rangy amp.  Something like this into a twin will sound like scooped death metal...  You need those mids to fill in that big mid hole those amps have.  A cool thing to do is to change the tone control to a mid peak control to get that back. 

Later, PaulC
Tim & timmy pedals

great info!

So the second set of curves...You've replaced the .22uf cap with a 2.4 uf cap, and added a 91 ohm resistor in series before it? I assume that flattens out the high end roll off?

Its a bit unrelated to this topic, but still awesome for me nonetheless because i've been looking to keep the bass cut in my TS (to tighten up the sound) but to reduce the high end roll off so that it doesn't sound so middy.  Also, can you please explain how you got to the last 2 curves?

[PaulC]

great info!

So the second set of curves...You've replaced the .22uf cap with a 2.4 uf cap, and added a 91 ohm resistor in series before it? I assume that flattens out the high end roll off?

Its a bit unrelated to this topic, but still awesome for me nonetheless because i've been looking to keep the bass cut in my TS (to tighten up the sound) but to reduce the high end roll off so that it doesn't sound so middy.  Also, can you please explain how you got to the last 2 curves?

Actually it's a big part of the idea for a transparent type of design.  There's several things people tend to think of when they're talking about "transparent".  One is an eq curve that doesn't change the basic responce of the rig while still being able to shape the curve of the clipping circuit for good OD tones.  The standard ways of pre/post clipping EQ leaves a big mid hump.  People try and flatten this by adding more bass before, and taking out less treble after, but if you do that enough to make things sound flat then you can end up with a woofy bass and shrill high end to the clipping.  The pre/de emp circuits let you keep the pre bass where you want it while still having a somewhat flat responce to then dial in with some tone controls.

In those sets of curves the first ones are taken from the input of the op to it's output showing the pre emphasis.  The second curve is taken from the output across the 1k resistor showing the de-emphasis.  The third curve is taken from the opamp input to the output of the 1k showing the effects of both emphasis circuits.

Later, PaulC
Tim & timmy pedals

[Mark Hammer]

The pre/de-emphasis is everything in achieving the goal of a more "relaxed" tone in an overdrive.  And the need for it is attributable to the basic reality that lower strings and notes always have more amplitude.  As such, they will always be at a greater advantage in terms of adding harmonic content to the output signal unless you compensate.  Additionally, think about how many multiples of a low E are added in an audible manner if that note is clipped, in comparison to how many multiples of the high A on the 17th fret are added, taking into account the way in which the vast majority of speakers roll off content above 6khz. 

The other problem created for the listener (and the musician is also a listener), is that the harmonics of lower notes overlap with, and sometimes conflict with, the fundamentals of higher notes.  So, I think when we talk about "transparent" overdrive, people use the admittedly inappropriate term "transparent" because what they want to be able to do is hear all the notes individually, instead of having some of the notes lost amidst a haze created by other notes.

Fundamentally, the low end is essentially hogging the clipping to itself, so the way to achieve a more relaxed and transparent tone is to use pre-emphasis to create a level playing field.  To some extent, a compressor ahead of such a clipping unit can be a useful part of the solution in terms of keeping all notes the same distance away from (or close to) the clipping point.  On the other hand, people do like to have a little more touch sensitivity, so the ideal solution is not so much to achieve note-to-note balance across the fret-board by imposing a fixed maximum amplitude, but rather by creatively and constructively using pre-EQ in the pedal itself to come closer to the level playing field, whilst still permitting sudden picks to be inserted by the player merely by digging in harder with the pick.

Of course, the strategy of mixing some clean signal in is, in many respects, aiming for the same goal.  The clean signal does not have disproportionately more harmonic content of the low notes (well, apart from what you get naturally from the wound string itself), so by bringing in some clean signal and moving the clipped portion back a bit in the within-pedal mix, that typical imbalance in harmonic distortion content across the fretboard does not stand out quite so much.

Always the psychologist, I think the solution/s to these sorts of issues can almost always be identified by asking "What is it that people hear, and how is their attention directed within the content they do hear?"  Once you have some idea of what perceptual analysis and priorities exist for the listener, it becomes easy (or at least easier) to identify the electronic means of achieving it.

[R.G.]

nice to see a sign of informed competence ...

It definitely is. J's a good thinker. I like it when people who post here know what they're talking about and can describe it in non-blathering terms, not using polysyllabic garble to try to produce an impression of knowledge they don't really have.

now, then - after this there's the issue of false triggering on the front-edge of a waveform which helps keep clip-definition high (cuts down on fizz/noise)

Kewl. Can you describe in plain English what false triggering on the front edge of a waveform is in an analog clipper, and how that differs from true triggering, please? Clearly so us simple plow-boys can understand it. 

something that Tom Scholz addressed when filtering the incoming signal before a clipper in his OD circuit
... which is inverting in configuration - and then ... DC coupled to the rest of the circuit (limiter) ...

Hmmmm... (note correct use of ellipsis) do you mean pre-filtering before distortion, as I've advocated many times here and in other places, or the difference between inverting and noninverting clipping and further the difference between feedback and diode-to-ground clipping I've posted about?

And if neither of these, can you explain in plain English what you actually do mean? 

many hints of such thinking having taken place ...

   Yes, such thinking may or may not have taken place somewhere long ago in a galaxy far, far away. The inner mysticism of such thinking pevades the aether, producing a spice-sweet smell of gentle clipping. Ah, how sweet the hints!

why/how this "clipping edge formation" scenario is different in the two TS/Rat topologies serves as basis for an interesting TRANsient response (real time, large signal) analysis challenge ...

So, are you up to that challenge? If so, why not reproduce it here. TRANsient response is indeed an issue in all circuits with nonlinear behavior. But they teach you that in your first circuits course. Or they did in mine. Yours may vary. And 'splain this mystical clipping edge formation to us in plain english, OK? 

further analyzing the role of the "bad" ground in the TS circuit is another thing one could sink  teeth into in terms of explaining how the clipping is frequency selective at the same time ... or why the circuit "at clipping" sounds different when hooking up the return leg of the op-amp gain/clipper to ground instead of bad-ground ...

Even better. Explain to us the difference in good-ground and bad-ground in plain English with no handwaving, OK? I'd love to hear it.

many unanswered questions and oversimplified explanations over the years as to what's really going on there ...

There absolutely are, Obi-Wan. Can you enlighten us, oh mystic one? But in plain English, not innuendo, OK. Some of us are too simple to understand the ineffable subtlties of wafted words.

Ellipsis (plural ellipses; from Greek ἔλλειψις 'omission') in printing and writing refers to the row of three full stops (... or . . . ) or asterisks (***) indicating an intentional omission. This punctuation mark is also called a suspension point, points of ellipsis, periods of ellipsis, or colloquially, dot-dot-dot. An ellipsis is sometimes used to indicate a pause in speech, an unfinished thought or, at the end of a sentence, a trailing off into silence (aposiopesis).
... [note that this use of ellipses is per the surrounding quotation, as I ellipse-d out a jump table.]
The use of ellipses can either mislead or clarify, and the reader must rely on the good intentions of the writer who uses it. An example of this ambiguity is 'She went to... school.' In this sentence, '...' might represent the word 'elementary', or the word 'no'. Omission of part of a quoted sentence without indication by an ellipsis (or bracketed text) (i.e., 'She went to school.' as opposed to 'She went to Broadmoor Elementary school.') is considered misleading. An ellipsis at the end of the sentence which ends with a period (or such a period followed by an ellipsis), appears, therefore, as four dots.

To catch up a bit more:

Quote from: Eb7+9 on March 04, 2008, 05:29:24 PM
if we stick to op-amp based topologies for a second (TS, RAT, etc ...) what we know so far is one tends to produce a more focused hard-clip than the other ...

Translation: Diode's to ground clip harder ('a more focused hard-clip') because they drive the diodes straight through their transfer 'knee' and do not affect the opamp's gain equation like diodes in the feedback loop do. In the feedback loop (see some great posts by John Greene patiently explaining it all to me in a search), the diodes drop the opamp gain to one once their threshold is reached.

There's another way to look at it as well. I've posted this a number of times. Diodes to ground are driven through their nominal exponential forward voltage by a voltage source (the opamp output) through a resistor, which does not matter until current flows in the diodes. As I've posted a number of times, the sharpness of clipping is entirely dependent on how fast the waveform moves through the conduction knee of the diode. In a diodes to ground setup, this is maximized.

Diodes in a feedback loop are driven with a constant current proportional to the voltage across the inverting-input impedance to ground, perhaps mixed with any signal voltage which drives the common mode voltage on the + input. As such, the transition across the knee region is much more controlled than diodes to ground. The impedance driving the diodes into clipping is much higher, approaching an ideal current source. It makes for a much less abrupt transition

Good explanations for diode clipping will explain what happens on the coming-out-of-clipping edge, too. I'm glad to see you putting some constructive thought into this J.

[YouAre]

i'd be lying if i said half the stuff in this thread isn't over my head

[Mark Hammer]

And if neither of these, can you explain in plain English what you actually do mean? 

   Yes, such thinking may or may not have taken place somewhere long ago in a galaxy far, far away. The inner mysticism of such thinking pevades the aether, producing a spice-sweet smell of gentle clipping. Ah, how sweet the hints!

Even better. Explain to us the difference in good-ground and bad-ground in plain English with no handwaving, OK? I'd love to hear it.

There absolutely are, Obi-Wan. Can you enlighten us, oh mystic one? But in plain English, not innuendo, OK. Some of us are too simple to understand the ineffable subtlties of wafted words.

Now, now boys...play NICE!  JC isn't the only person on the face of the planet (or on this forum) who posted a note hastily without editting for clarity.  A simple "Sorry, but I didn't get it " will do.

Honestly, guy, you know I luv ya, but you'd simply never fit in with the Gulf Island crowd.   When you wake up with the orcas and the eagles, Strunk & White is probably the last thing on your mind.

[JDoyle]

I'm assuming by 'good' and 'bad' ground he is talking about an 'actual' ground like the - battery lead vs. a 1/2V+ voltage reference.

There's another way to look at it as well. I've posted this a number of times. Diodes to ground are driven through their nominal exponential forward voltage by a voltage source (the op amp output) through a resistor, which does not matter until current flows in the diodes. As I've posted a number of times, the sharpness of clipping is entirely dependent on how fast the waveform moves through the conduction knee of the diode. In a diodes to ground setup, this is maximized.

Diodes in a feedback loop are driven with a constant current proportional to the voltage across the inverting-input impedance to ground, perhaps mixed with any signal voltage which drives the common mode voltage on the + input. As such, the transition across the knee region is much more controlled than diodes to ground. The impedance driving the diodes into clipping is much higher, approaching an ideal current source. It makes for a much less abrupt transition.

So basically with the diodes to ground setup you are using a voltage to clip the signal which happens a lot faster as a small voltage change cause an enormous current change through the diode, whereas with a feedback loop setup you are driving the diodes with a current and it takes a large current change to cause even a little change in the voltage drop across the diode. It's all about the impedance of the source...

...and what the op amp is driving...

...for example (I think that was the 'colloquial' use of ellipses ), in Paul's suggestion above I can see an issue - the added resistance/impedence to ground of about 1k is now, along with the 1k in series with it, in parallel with the impedance on the - input when the diodes are clipping, which reduces the effective load impedance on the output for AC signals from 4.7k to a little under 1k because the output sees the filter in parallel with the 4.7k/.47 uF when the diodes are clipping.

This will change the way the circuit clips and make it clip harder. As the signal approaches the diodes threshold, the current through the feedback loop is determined by the setting of the gain pot, the impedance to ground on the - input and the load attached to the output. But when the diodes conduct the gain pot is removed from the equation (sort of, as the signal begins to clip and then as it leaves the diodes clipping range, the diode and the pot sort of 'hand off' the feedback impedance duties, but this happens REALLY fast as the diode goes from being an open circuit to a short in under 0.7 V) leaving only the impedance on the - input in parallel with the load resistance. It is this parallel resistance that determines how hard the op amp output has to drive to satisfy it's requirement that both inputs be at the same voltage when above the diode's threshold (because this current out, dropped across the impedence on the - input, is the signal on the - input and is what the output is trying to make the same as the voltage on it's + input); the smaller the resistance, the more current required from the op amp to drop the same amount of voltage, which means you drive the diode further into the conduction curve when the diodes are clipping with a smaller parallel impedence (- input impedence || load impedence).

So, with the TS circuit, as the signal 'goes into' clipping, the rate through the diode's knee is determined by the difference between the current required to drive the feedback loop (this depends on the setting of the gain pot, mostly), in parallel with the load impedance, while the diodes aren't clipping, and the current required to drive just the impedence on the - input in parallel with the load impedance when the diodes are clipping. It is this difference in current levels and the speed the opamp can and/or has to change between those two levels that determines the way the signal transverses the diode's 'knee'.

Another seperate issue that I think would change the overall response of Paul's setup is the fact that for AC, the op amp is now being asked to drive a heavy load, about 2k in series with a large cap. I don't know much about driving capacitive loads, but I do know that I've read some warnings about doing it and also that 2k is getting awfully low of a load for most opamps to be able to effectively drive without adding its own separate source of distortion.

None of this is to suggest that Paul's circuit is in any way 'wrong' or would sound bad, only that by doing it, more than the frequency response is changed.

Regards,

Jay Doyle

[R.G.]

You know, Mark, you're absolutely right. 

Let me try again. I'll be nice.

now, then - after this there's the issue of false triggering on the front-edge of a waveform which helps keep clip-definition high (cuts down on fizz/noise)

Wow! That's deep. I never knew there that an analog clipper had triggering, false or otherwise, on its leading or trailing edge, JC. Can you explain this triggering, and how false triggering differs from true triggering, and what difference that makes in the resulting sound, either in generated harmonics or the frequency response?

Please be thorough, because this is an area of which I have no knowledge, and I'd really like to learn about this. Since I'm new to this, please explain in small words because there may be new terms with which I'm unfamiliar. Triggering has a specific electronic meaning in all the teaching I've ever had, this doesn't seem to match those, so there's pretty sure to be something new (to me at least) there.

something that Tom Scholz addressed when filtering the incoming signal before a clipper in his OD circuit
... which is inverting in configuration - and then ... DC coupled to the rest of the circuit (limiter) ...

Do you mean pre-filtering before distortion, as I've advocated many times here and in other places, or the difference between inverting and noninverting clipping and further the difference between feedback and diode-to-ground clipping I've posted about?

If I've missed the point entirely and it's neither of these, can you explain in more detail? Frankly, it seems like the Rockman distortion is pretty ordinary, as most decent distortions have some pre-distortion filtering of some kind. And is there a nuance there about inverting ODs that's different from my comments about using inverting clippers?   

By the way, since lowpass filtering before clipping slows down the (transient!) rate of change of the incoming waveform to a clipper, how does that affect edge triggering?

many hints of such thinking having taken place ...

That's pretty cool. Can you point me to a number of those hints so I can learn about them? Perhaps annotate a few so I can understand better what you mean?

why/how this "clipping edge formation" scenario is different in the two TS/Rat topologies serves as basis for an interesting TRANsient response (real time, large signal) analysis challenge ...

That sounds very interesting. Could you describe this clipping edge formation scenario a bit more clearly, and also describe how the scenario is different between the TS and Rat. Also, since I realize that transient response is indeed an issue in all circuits with nonlinear behavior, could you describe why this is a challenge (presumably to model? Is that what you meant?) and maybe tell us some of the results of your work on the challenge. I'm always eager to learn.

further analyzing the role of the "bad" ground in the TS circuit is another thing one could sink  teeth into in terms of explaining how the clipping is frequency selective at the same time ... or why the circuit "at clipping" sounds different when hooking up the return leg of the op-amp gain/clipper to ground instead of bad-ground ...

I'd very much like to see your analysis of the differences in good-ground and bad-ground. Are these good-bad differences based only on the relative impedances of the grounds, or is there some other factor of which I'm not aware?

many unanswered questions and oversimplified explanations over the years as to what's really going on there ...

Can you list the unanswered questions you see, the oversimplified explanations, and your correct explanations for what's really going on? I'd very much like to add any new technical views of clipping situations to what little I know. Please tell me more, and remember that if you introduce new terms or concepts, it would be helpful for you to explain them as you go along so I can understand.

[Eb7+9]

I'm assuming by 'good' and 'bad' ground he is talking about an 'actual' ground like the - battery lead vs. a 1/2V+ voltage reference.

that's right, a resistive/capped 1/2+ voltage reference will work fine on the non-inverting input of an op-amp because that node doesn't draw any current but on the return path of a feedback loop it's an entirely different story, there's an interaction that's identifiably exploited in the TS circuit ... hence the need for op-amp buffered references in some one-rail circuits as described in the well-known Linear app-note ... tie the return path to ground in the TS circuit and hear the difference in "fizz" when clipping ...

again, I don't know who's qualified to determine who's a tinkerer here or not but having some IC design and univ. TA experience under my belt it might be a little hard for me to express myself at a level of plain language and plain understanding the resident snowjobber routinely uses ... I'm very sorry if incorporating std. elements of current EE design and analysis techniques is inappropriate in this forum ... why we should pretend to be above/below that is beyond me - other than to keep other people ignorant so one can hold dominion over them ??

If I feel like it, I will continue to critique "blather" when it does not qualify as a substitute for sound reasoning ...

[R.G.]

that's right, a resistive/capped 1/2+ voltage reference will work fine on the non-inverting input of an op-amp because that node doesn't draw any current but on the return path of a feedback loop it's an entirely different story, there's an interaction that's identifiably exploited in the TS circuit

Can you describe how that causes a difference in the TS circuit and how you identify the exploit?

again, I don't know who's qualified to determine who's a tinkerer here or not but having some IC design and univ. TA experience under my belt it might be a little hard for me to express myself at a level of plain language and plain understanding the resident snowjobber routinely uses

Oh, wow! Who's the resident snowjobber? I'd like to know who to ignore.

... I'm very sorry if incorporating std. elements of current EE design and analysis techniques is inappropriate in this forum ...

Why would standard items of current EE design and analysis techniques ever be inappropriate here? I think they should be explained in detail so everyone can benefit. That's what a lot of my last post was about. Please 'splain this to me.

why we should pretend to be above/below that is beyond me - other than to keep other people ignorant so one can hold dominion over them ??

So, who's these ignorant ones and who's holding dominion? Or was that a rhetorical question/assertion/statement/innuendo, without a real basis?

If I feel like it, I will continue to critique "blather" when it does not qualify as a substitute for sound reasoning ...

Well, it's pretty much open for you to say whatever you like. I like it very much when a new way of looking at things is found that fits into the way natural laws as expressed in technical analysis can be found. It rounds out my, and I'm sure everyone else's understanding. That's why I was asking for clarification. I was taught that I should not make assertions for which I could not produce facts to support. But that's just me, and the teaching may be different in other places. You go right ahead and say whatever you like, whenever you like it, and for whatever reasons. Furthermore, I think that if one has something to say and has a factual basis behind it, that makes whatever one is saying logically unassailable.

And I fully support your determination to critique "blather" when it does not qualify as a substitute for sound reasoning. Matter of fact, I'll do the same. Perhaps between the two of us, we can bring more clarity to technical issues which are so often passed over, improperly described or mystically mojo supported in the effects advertising world. It really irritates me when pedal "designers" have no clue what they're doing and attribute circuit operation to mystical or mythical views of the technology. Mother Nature is a strict controller. Her Rules are what happens no matter what people say.

As my friends in the design community used to say, you can't B.S. an electron. It will always do what Mother Nature's Rules say to.

[MartyMart]

I'm assuming by 'good' and 'bad' ground he is talking about an 'actual' ground like the - battery lead vs. a 1/2V+ voltage reference.

that's right, a resistive/capped 1/2+ voltage reference will work fine on the non-inverting input of an op-amp because that node doesn't draw any current but on the return path of a feedback loop it's an entirely different story, there's an interaction that's identifiably exploited in the TS circuit ... hence the need for op-amp buffered references in some one-rail circuits as described in the well-known Linear app-note ... tie the return path to ground in the TS circuit and hear the difference in "fizz" when clipping ...

JC, reading this, is there any benefit from running the op-amp as a "true" bipolar +4.5v "0" -4.5v rather than the "false" 4.5v or V-ref ??
I've only tried this once or twice and the results were quite decent IMO.
Could this be applied to the TS "type" circuit ?
Marty.

[R.G.]

What JC's trying not to say is that impedance in series with any point that is nominally "ground" has to be taken into account, or Nature will take it into account for you.

Most of this is covered in my article on designing Vbias networks, at http://geofex.com/circuits/Biasnet.htm at least in passing. The ideal bias voltage would have an impedance of zero. We can't do that, and we approximate it with either two resistors and a cap to ground, or an opamp buffering this resistors-and-cap point, or some other means to get low impedance.

Low impedance is what matters when you have current flowing into and out of a nominally bias node. What JC calls a "bad ground" is an AC ground point with some impedance in series with it. Unfortunately, it's impossible to make a zero-impedance point. You can only make the impedance so low that it vanishes by comparison to the other circuit parts.

Let's look at the AC impedance of a resistor/resistor/cap Vbias. Notice that I always refer to this as "Vbias", not "AC ground" as gets bandied about. There's a reason for that bit of insistence, as you'll see. If your power supply is a perfect AC short circuit (it's not, but we'll come back to that) then the resistive part of the impedance of a Vbias node is the two resistors in parallel, as the high and low ends are "shorted" by the DC power supply. Two 10K's for instance, produce a resistive part of this impedance of 5K (within the tolerances of the resistors). The capacitor appears in parallel with this point in the path to what we have to call ground for lack of a better place. So the impedance of Vbias for the simple case is 5K paralleled with 1/(2*pi*F*C), which reduces to (1/F)*(1/2*pi*C) = (1/F)*15.9K.  That's a far cry from zero. We might be tempted to use 22uF to make it (1/F)*(7.2K), 47uF to make it 1/f*3.4K, etc.

But it is frequency dependent. So at the bottom guitar note, the 10uF is 15.9K/82 =194 ohms. It hits 19.4 ohms at 820Hz, and 1.94 ohms at 8.2kHz. As long as the things which are connected to this bias voltage are much larger impedances, you're modestly justified in ignoring the impedance of the Vbias point. Notice I said "modestly" and that the Vbias article talks about the imperfection.

If you absolutely, positively have impedances connected to this point which are so small that 5K paralleled with 194 ohms capacitive matter, then you have to go to some other kind of Vbias. In effect, to hear a difference 20db down, you need an impedance that is less than 10x the Vbias impedance, or around 2K in series with the Vbias point.

There is no mystery about this effect. Mother Nature is very picky that every one of Her Rules get followed, whether you personally know about it or not, or whether or not you have taken it into account. So can you hear a difference in a TS circuit between the inverting input of the clipper stage being tied to a 10K/10K/10uF Vbias, an opamp buffered Vbias ( with an internal impedance in the few-ohms range, about a 10:1 change for the better) and it being connected to "real" ground? Maybe. It kinda depends on what else is connected between that "real" ground and the power supply point that feeds the opamp that uses the power supply to push current through the ground node. It all matters. Of course.

Some more imperfections that Mother Nature will insist on: the power supply is not a perfect AC short circuit. It has at least the DC impedance of the battery, tens to hundreds of resistive ohms, paralleled by the impedance of any power supply filtering (actually bypass) cap that's being used. Every bypass cap will have an ESR and an ESL to be taken into account, as well.

A good-ground/bad-ground distinction blurs what's really going on. You need to know the impedance that's in series with your power and ground, and design accordingly. There is no mystery, just the need for understanding, completeness and mental hard work.

Now back to your question. Will running the opamps from a "true" bipolar supply make a difference?

The answer is a solid maybe. It depends on (a) the impedance of that "ground" in the bipolar supply and (b) what you have tied to it as well as (c) whether you look for microscopic differences. When you know the impedances involved, a little calculation gives an answer.

Notice that there is no "transient" issue. The composite network of power supply, ground, R-network and cap on a bias network behaves the same way as if you put a lumped impedance in series with whatever you are testing that is equal to the combined impedance of the network.

You have a chance to hear a difference depending on the impedances you connect up.

[PaulC]

Another seperate issue that I think would change the overall response of Paul's setup is the fact that for AC, the op amp is now being asked to drive a heavy load, about 2k in series with a large cap. I don't know much about driving capacitive loads, but I do know that I've read some warnings about doing it and also that 2k is getting awfully low of a load for most opamps to be able to effectively drive without adding its own separate source of distortion.

None of this is to suggest that Paul's circuit is in any way 'wrong' or would sound bad, only that by doing it, more than the frequency response is changed.

Regards,

Jay Doyle

The reason I posted it as such is to show the curves, and how the pre/de emp circuit Mark talked about would look, and how you would go about trying it in something like a TS design.  It's not a perfect circuit when used with a TS, but the idea of mirror opposite curves that noise reduction uses when applied to overdrive circuits can give some interesting results.  I have built many variations on this idea for several years, and they are something to consider.   I even said that other things would be going on with this set up, and the curves are gain dependent so they are not perfect.  


It's not something that just goes with non/inv opamp diode clippers.  You can use the idea with pretty much anything, and you can use the idea using passive or active (or combinations of both) emp filters.  For me the idea is to sort of "undo" the eq curves you had to make for good clipping so you can then design some tone controls to really nail it.

Having said that you're right about the loading issues, and i always felt it was a bad choice to go with that 1k resistor that the TS uses.  When I've built TS type of designs I used values more like what you'd find in the graphic eq units for the reasons you stated, plus the fact you can get a more linear dB responce across  the range of the tone control than you do with the values used in the TS.  

What I was trying to do was put some pictures behind what Mark said to show how the idea differs from what happens in a ts.

Later, PaulC
Tim & timmy pedals

[Mark Hammer]

And for that, sir, I thank you appreciatively.

I'd be curious to hear from players that use strings or guitar types that provide note-to-note difference in amplitude across the fingerboard that are highly discrepant with what something like a TS or other pre-emph/de-emph overdrive design would normally anticipate.  For example, if you play baritone, does a TS variant sound limp on your lowest 2 or 3 strings?  If you play skinny bottom, heavy top does the pedal sound "off"? somehow.