Clipping threshold?

[MrGuitardeath]

I'm curious if there is a way to fine tune the clipping threshold of an analog gain circuit.

You know, that would respond to playing peaks so you could get really clean tone that becomes groggy sounding when you play harder and suddenly force it into high levels of overdrive.

I'm getting a similar effect from running a modified tube screamer into a ZooM G1. I figure modding the screamer one last time will give me a full blow tube groggyness when forcing the ZooM into dirty overdrive levels. I definitely have to credit the tube like response to the Zoom and the Screamer equally.

Are there threshold controls of this nature already in use?

Say a resistance threshold that was extremely easy to stay under but when you played harder and jumped across, you go into a nice deep, resonating clipping effect. The results that are very tube like to me tend to be voltage controlled. Something I'd done to manipulate the signal levels and tweaked until I have the equipment running more sensitively.

I guess that is the best way to put it, tweaking the gear until it is more signal/voltage sensitve and you can manipulate it through your playing.
I suspect a voltage sag would even do the trick...

[aron]

This is something I've wanted too. It seems this would be so easy with digital pedals and in fact I've suggested it to companies. I also suggested a "learn" mode where the pedal could optimize itself for the individual player.

[MrGuitardeath]

Yeah it does seem it would be a simple task to pull off with a digital unit. I'm close to it with what I'm doing so they should be able to have figured this out and done it a long time ago.

There is a good chance that this is what the ZooM does and they have picked it up off of people like me and you. I helped someone refine their Boss GT6 rig by ultimately running a Behringer tube mic preamp on the GT6's effects loop. Worked like a charm. The higher end version of the ZooM G1 incoroporate tubes and looks like a greatly improved copy of what the Gt6 was and GT8 probably is. He and I were both very liberal with the information and findings so go figure...

[CS Jones]

R.G. planted the seed of an idea along these lines several years back. Something similar at least. Don't know what he ultimately did with it, if anything. Don't know what became of the thread. Do know that the idea was picked up by a couple of guys who have since moved it on into new directions with varying degrees of success.

[MrGuitardeath]

Anyone have some thoughts to share in regards to using a voltage sag?

I had an old floor processor that would get really tube like when the battery got weak. I eventually found a power pack that gave the same effect and the processor became useless when the power supply finally went.
I figure digital could simulate it but this should pull it off in the analog pedals as well.

[R.G.]

OK, time for some philosophy of clipping.

Let's say we have a silicon diode clipper, back to back pair to ground. We know that the current through the diodes is an exponential function of the voltage across the diode. Put another way, the current through the diode goes up exponentially with the voltage across it. Exponential functions start off quite slowly for small inputs, but at some point turn upwards and head for the sky. A 0.1v signal will not cause significant current to flow in the diode pair. A +/-0.4V signal will start some current flow, and a signal beyond about 0.7V may require amperes of signal to reach. That's clipping. Substantially no signal source can supply enough current to pull the diode voltage any higher than about 0.6 to 0.7V.

So what happens when a signal is just ...barely... into the region where the diode starts conducting? The signal is limited, but not clipped. If we used two frequencies, say,mostly low E at 82Hz and just a whiff of E three octaves up at 656Hz, and run those into this diode limiter, then even when the signal is being limted at the peak, you could see some of the 656Hz on the top of the wave. But make the signal try to get to maybe 2V peak, and the top of the wave is dead flatlined, no more of the 656Hz left in it. It's been well and truly clipped off. The information in that delicate 656Hz signal is literally gone.

This is why the so-called knee of the conduction curve is important. If the knee is abrupt, you get hard clipping at all signals over the clipping threshold. If the knee is rounded, you get progressive squashing, but without clipping off whatever is happening in the signal. Although everyone's ears and preferences are different, most people prefer rounded knee clipping.

So - what defines "roundness"? Two things. First is the change in resistance per unit signal change as a fraction of the clipping threshold. Second is the relative signal height compared to the clipping threshold.

The change in resistance one is obvious. If you have a magical diode that starts rounding off at, say, 1V, and does not reach full clipping until 2V, gradually increasing along the way, it would be soft compared to silicon. Silicon starts conducting noticeably at ..., oh, just pick a place, 0.5V. It is in full conduction somewhere around 0.6 to 0.7V depending on the junction. The transition region is 0.15 to 0.2V wide compared to a 0.5V threshold. That's a much smaller range of variable conduction to threshold, and it's a sharper clip.

A high feedback opamp running out of power supply room will run the full range of the power supply right up to its limit, many volts of signal room, and then flat top over a few microvolts. That is hard, sharp clipping.

The second thing is obvious if you think about it. If you have a silicon diode, it will clip with some degree of compression for signals less than the full conduction voltage. Perhaps 0.5V peak will leave some information on top of the signal that's not clipped off. What happens when you feed the same diode a signal that would otherwise go to 10V peak? Yep, it shoots right through any silly little conduction knee so fast that the corner looks like the clipping corner of an opamp. So the magnitude of the applied signal compared to the conduction threshold matters.

If I have an arbitrary signal, and I want to clip it, I have some options. I can diode clip, but then I'm limited to the voltage of a junction. I can use Vbe multipliers and construct artificially-amplified "diodes" that are a magnified version of the single junction. I can use an amplifier that will power supply limit and vary the power supply to the amplifier.

Multiplied junctions have bigger knees as well, and there may be something there. Changing the power supply to an amplifier may well have something in it because the turn-on and turn-off of the amplifier is not affected as much as its linear region, so a low- or no-feedback amplifier could be modulated by its power supply and change the ratio of threshold to knee size, to good advantage.

Finally there are the analog calculation circuits, where opamps have diodes or transistors in their feedback loop hooked to voltage thresholds. When the signal exceeds the threshold, the conduction of the diode/transistor connects in another resistance and changes the opamp's linear gain. These work well, but need big signals to approximate really rounded knees, and you have to do it for both positive and negative signals.

So in thinking about clipping thresholds think about
- how big is the non-clipping region?
- how big is the knee compared to the non-clipping region?
- how flat is the area after the knee?
- how big is the signal compared to the non-clipping region and knee?

They all matter.

[aron]

What I am talking about is a programmable curve for the response of the distortion circuit. Very common with sophisticated MIDI modules. You manipulate a curve for different response over signal levels. This would make an extremely dynamic pedal. The Anatek breath controller module made you blow into the unit the first time it was turned on to set the curve.

[MartyMart]

RG, fantastic information .... as usual I've learned something .. again !
Dont we have that "kind" of control in various circuits ( in a way ) ?
The humble TS-9 will be a touch "cleaner" with less "strummed/picked" attack
from the gtr, even my simple "Mr Drive" circuit, respondes to picking "dynamics"
quite well  ! .  ie: More attack = more crunch = diodes being driven harder ...

aron, I know what you mean, my Peavey master keyboard has a function to
"learn" a velocity "curve" by striking the keys at two points at different velocities, to give the
lowest/highest threashold of your "curve" and also has various versions
available, exponential/linear  etc etc.
Setting it strangely can produce an almost "compressed" result !

Whilst not "programmable" I'm sure that careful choice of circuit, with
diode's in a feedback loop/ to ground can be made more "responsive"
and could possible be made variable with a couple of switches to change
some resistors - affecting the "feel" if you like, even your "softness" control
from the FAQ is a very useful  device 

Marty.

[Khas Evets]

It sounds like you want an expander before the clipping stage.

[aron]

aron, I know what you mean, my Peavey master keyboard has a function to
"learn" a velocity "curve" by striking the keys at two points at different velocities, to give the
lowest/highest threashold of your "curve" and also has various versions
available, exponential/linear  etc etc.

Yep, that's what I am talking about although I mean a visual response curve that can be either manipulated on the computer with an editor or a "learn mode" which could analyze the input/dynamics from the player.

Now these response curves could be applied to different parameters of a pedal. Drive, tone control, volume, filter resonance etc... could be affected +- from the input response.

It think it would be sooo coool.

OK, while we are spilling the guts, how about this? You know how the #1 problem with distortion pedals is that the tone controls don't work "in the right place" for a particular amp? OK, as a start, you put the tone controls in a "set state" and what they do is sweep the center freq. of the knob, then you "unset them" and you can sweep in a different range. Yes, it's basically a parametric filter but I think it would be so cool to be able to quickly set the center frequency quickly on any piece of gear and start affecting the tone from there.

These things are so easy to do but have never been done.

I wish I knew DSP - just went to school a little too early. 

[MrGuitardeath]

Well, since we have said this much about it, they'll probably do it. I've got a brand spanking new floor unit processing at 32 bit and it has a good tube like response as does the pirated version of the Tech 21 amp I own. I still am stuck modifying and adding gear to get it just right though...

What about simply coming up with a power source that responds to the input load??
Seems like a really easy solution and you wouldn't out date everyones favorite gear. I'm still under the impression a voltage sag would do the trick with plenty of analog pedals and amps, particularly those that are tube emulating.

I've noticed lights fade and intensify in certain pedals due to my playing, obviously some gear is sensitive enough to respond to the input signal it is processing. I'm betting one of you guys could figure out some sort of voltage sagging power supply or an input buffer that controls votage draw off the power supply.

[petemoore]

  Ok...there's current coming from the clipping diodes to ground when they clip, what about putting something between the diodes that either slows the current's skyward travel soon after it begins, or use the current to trigger a stage before the clipping section to drop gain.
  Sounds wierd, but has anyone tried putting a row of large caps under clipping diodes? not that it would work...I'm just wondering if they could have small bleed resistor[s so they'd maybe have to get charged up by current through the diodes, and the diodes would be busy filling them up enough to soften their work on the clipping knee.
  Maybe even have the clipping diodes supply current to some dummy active component? Of course you might want to increase output the dioedes 'see, a little bit.
  [Just fodder of speculation?].

[petemoore]

   Here's one you may like...
  Take a speaker of high ohmage, and connect it between diodes and Gnd.
  Then take same speaker and set it where the speaker of your amp is 'driving' it [like a microphone].
  I feel this is almost certain to do "Something Wierd or Cool [amplitude sensative at the least] then, invert the 'mic-speaker' in relation to the drive speaker.
  I started thinking along the tracks that the speaker would move on the output of the small signal device alone [maybe a larger Sm Signal device and sensative speaker  ], but then decided to have a drive speaker [since everybody has one of these and amp to go with] provide voltage/current to the diodes as well !!!
  As potentially untested *ideas go, I think it's a good enough [and probably easy enough to do to get preliminaries from] to be worthy of trying. There are certainly driver types very cheaply available...I pulled four from the curb today...parallel/seriesing more than one of them would of course be one of the first mods to try.
  Perhaps have a diode or two in one direction through it's own speaker[s facing one way, and the other two opposing clipping diodes through the other set of speakers 1/4 turn or more from out of phase.

[phaeton]

I'm pretty sure someone has designed artificial voltage sag into a pedal.... superficially it seems so easy to do:

1) Make the diodes LEDS, and use an LDR upheave either transistor biasing or other wise open/close a transistor that controls signal strength.  This is not unlike your standard compressor.

2) If LEDs require more voltage than your circuit has, then using transistor junctions as diodes would be the next step- depending on how you do it, the transistors could 'switch on' when forward biased and do the same as above.  Setting transistor bases "behind" diodes is another approach.  Lemme think, do we have to use an NPN/PNP pair?  Voltage forward-biases the diode and brings the base of a transistor away from ground.  Surely some capacitors are involved in this too.

Anyways... I've got tons of other crazy thoughts too, but i dunno if you want to hear them, heh....

As always, a great read, R.G.  Aron- did you go to school in an EE field?  You can probably go back for a few classes regarding DSP, without having to do it all over again, jaknow...

[petemoore]

  I have no idea whether such a thing exists, but a DVD DSP learning set
 

[Alex C]

Does anyone else copy and save almost all of RG's posts?

Thank you sir, both for your wealth of information and your generosity.  This is fantastic.

[MrGuitardeath]

Can I get away with just swapping the clipping diodes in my tube screamer with LED's and get a sag effect?

[aron]

did you go to school in an EE field?  You can probably go back for a few classes regarding DSP, without having to do it all over again

I wonder if they have anything here in Hawaii? I received a Masters in Electronic Music composition from San Jose State University. We were doing some DSP stuff, but nothing like now. Then there is the issue of time......

BTW, years ago I made the Shaka HV and even put a voltage regulator in it for sag. It didn't affect the pedal as much as you might think for certain things.

[petemoore]

  I know I only read bottom two or so posts, and sometimes miss an interesting/silly page because it just happened to have another 'fall' on it right away...so I'm bumping the thread here, and duplicating this: [is this idea nuts or what...
  Here's one you may like...
  Take a speaker of high ohmage, and connect it between diodes and Gnd.
  Then take same speaker and set it where the speaker of your amp is 'driving' it [like a microphone].
  I feel this is almost certain to do "Something Wierd or Cool [amplitude sensative at the least] then, invert the 'mic-speaker' in relation to the drive speaker.
  I started thinking along the tracks that the speaker would move on the output of the small signal device alone [maybe a larger Sm Signal device and sensative speaker  ], but then decided to have a drive speaker [since everybody has one of these and amp to go with] provide voltage/current to the diodes as well !!!
  As potentially untested *ideas go, I think it's a good enough [and probably easy enough to do to get preliminaries from] to be worthy of trying. There are certainly driver types very cheaply available...I pulled four from the curb today...parallel/seriesing more than one of them would of course be one of the first mods to try.
  Perhaps have a diode or two in one direction through it's own speaker[s facing one way, and the other two opposing clipping diodes through the other set of speakers 1/4 turn or more from out of phase.

[Mark Hammer]

One of the things that is often neglected in discussions of clipping is alluded to (but not beaten to death with a stick) near the beginning of Brother RG's post, and that is the manner in which concurrent frequencies contained in the signal SUM together to produce the actual signal amplitude at that particular moment.  If you were to scope the signal, you would see that not all peaks produced in such a manner cross the magical line where clipping begins.

As a result, it is important to think about what proportion of such peaks are going to be supra-threshold (getting clipped), and what proportion of signal content will be sub-threshold.  Just as important, it is useful to think about how the proportion of such clippable peaks changes over the life of the note.  This is, of course, why many distortion pedals can seem to sound richer when preceded by a compressor.  The compressor assures that a larger proportion of peaks will be supra-threshold for a longer period of time.  It's also why you would NOT want to use an archtop with a floating wooden bridge (i.e., a "classic" jazz guitar) with a distortion pedal; because the rapid decay and dynamic contrast within notes makes such peaks disappear too quickly, producing a ratty sputtery sound.  I'm not saying such a tone is not useful musically, but it is generally not the sort of tone most are looking for.

This is where my allegiance (and that of millions of amp and Big Muff Pi users) to double clipping comes from.  One of the things that double clipping (two or more cascaded stages, each of which produces some clipping) does is reduce the contrast between those mini-peaks, and the average signal level on the first stage, so that more generally consistent clipping can be imposed on subsequent stages.  Of course the manner and desirability of how this gets done would vary with the gain structure of the device in question.  

In the case of the Big Muff Pi, we have three gain stages, a passive (lossy) tone control and a gain-recovery stage.  The first stage assures that the signal can make it up to the clipping threshold set by the diodes in the first clipping stage, and the "Sustain" control determines how frequently that will happen (i.e., how many peaks will be in the critical zone).  The first stage evens out the contrast between peaks and average amplitude somewhat, and clamps the maximum output level a bit in doing so.  The signal then passes to a second identical clipper stage where differences between peaks and troughs are smoothed out even more and the more effective and consistent clipping across the lifespan of any note yields the smooth sustain people like this pedal for.

Multi-stage preamps, of the sort that Randall Smith drew musicians' attention to with Mesa Boogie amps in the early 70's, use a similar sort of strategy.