Achieving slight distortion with buffer overloading.

[Thecomedian]

I've been playing with this concept for a bit.



at Q4, there is some pretty decent asymmetrical distortion, similar to what a tube might experience, which works like this due to the output resistance being 10k. Increasing resistance to 100k or 1 meg doesn't really change the output voltage level, apart from allowing it to go 99.9% into a complete sine wave.

I "made" (read: slapped it together blindly) this circuit in just under 2 minutes so I could fiddle around with something a little more complex than fuzz face. The Q2-Q3 are supposed to be Darlington pair but I didn't connect the collectors properly.

I've made a "buffer distortion" before using a cross breed jfet/NPN after noticing it's tendency to distort "smoothly" when biased too much for the NPN alone. It had pretty good subtle distortion, but being a buffer it also made "finger sliding" and string touching sounds louder as well. (Maybe some hi-pass filters would clean that up)

Has anyone else tried using the way buffers fail to follow a sine wave if it goes past  <1 gain in order to get distortion? Do the experts think it might have some potential?

[GibsonGM]

I suppose it would depend on a couple of things, no?  Mainly, does the slight distortion it generates provide something of *value* that can't be easier done with fewer components?  Is the sound 'unique', and worth chasing down and modifying/tweaking?

Great 'find', it's awesome to experiment, and you definitely came up with a unique 'topic'!  I'm just curious of it would be 'worth it' to assemble everything, if it could be done simpler.  I once used something like 6 opamps to generate harmonics to make an octave effect...then found it could be done with 2 transistors!!!  )

[Thecomedian]

Yes, I do have to think about that. I just happened on it by chance noticing that I was driving an emitter follower into cutoff, but the other side of the waveform was curving very gracefully while distorted. added a FET and got it to stay more sine wave on both sides with the distortion.

The idea I would suppose is to use the inherent nature of the CC/EF type to cause the output to fail to track the input just enough by "mis"biasing it so it cuts off or saturates very slightly.

I also got this circuit to produce 1.8 volts "clean" sine by fiddling with values, and also tweaking the simulated potentiometer on the end.

I'm sure whatever it does can be done simpler. I don't have the tools for "if it sounds good" tests, really.

Also, I misread it. It's not overloading, it's driving it very momentarily into cutoff.

[GibsonGM]

Right, I was going to say that the 'tube version of a buffer', the cathode follower, exhibits such behavior with one tube.  It's very much a part of the classic Fender tone!! Wouldn't be the same without it.

Really neat to see what circuits do in 'failure mode', when they exceed their design parameters!  

One thing that DOES seem like it COULD lead to something useful....since you have these BJT's going into cutoff less abruptly than they may normally do (? so I've heard)...this may be a way to kind of make transistors a little more tube-like.   Maybe it can be scaled back, use less parts, but lead you to a more 'open' sounding overdrive/distortion?   Just a thought, maybe another step for the idea....

[Thecomedian]

I went off some hunches and scraps of info in search of the "tube-like distortion", I've managed to do a few circuits like this in the last month within spice. The only issue is that this one is very reliant on that input being 20m, so I need to figure out the negative feedback to keep it on the proper Q.



A very simple circuit to achieve a goal that is usually done with a dizzying array of ICs and connections all over the place. Again, I'd still have to test it for "if it sounds good". It's a bit of a pain, but you can pan to the right to see an example of what the Sine of Tube distortion appears like. I also read that the more flattened it becomes, the more harmonics are generated (buzzy bees).



This is the FET helped one. an asymmetrical diode is added to smooth the lower peak, voltage measured from between the simulated 500k pot at R1-R5 250k/250k. Green line node 005. R7 is added just for experimenting how a low ohm trimmer would affect it. 40 ohms produces a much more even spaced wave distortion that just barely hedges the tops and bottoms.

The red line belongs to the circuit below it. The large lower curves are where the current through collector is climbing to it's peak, and to my thinking, failing to "track" the base as it reaches towards the point of .9998+ gain, where it starts to be unable to go further. The idea is taking advantage of the limitations of the Common Collector to cause a gradual decline in amplification factor that is more of an exponential decay similar to the voltage sag of overdriven tube.


This is at least my theory for now..


Here's a "distortion stack"




It's essentially a circuit I found in a book somewhere (when I figure out where I got it I will name the circuit), but it has a specific intended function, though I noticed the clipping going on and thought it'd be great to put in series with a transistor circuit. If the signal is pre-distorted, then you just want to amplify it. Had an issue when I cap coupled it to a working circuit though..

[GibsonGM]

That is REALLY interesting, Comedian!   You REALLY should try to build a bit of this, and see how you're changing the sound!   The concepts you have going are 100% correct, theoretically....the proof of the pudding is in the tasting, though, right?

The diode pair + resistors is quite familiar...it's almost like you are reverse engineering a differential clipping stage )  Maybe from a limiter of some kind?   I think it's pretty cool that you're building non-linearity into (mostly) linear devices, in a new way!

As you go along, see if you can actually listen to some of these snippets...and I would try to establish a spectrum, and see what's happening with different frequencies!  Try changing caps (esp. those bypass caps!  47u is high for a guitar signal...try 1u, .47, etc).   

Interesting! 

[Thecomedian]

Here is the fet/bjt cross with R7 up to 35 ohm, and the sine wave gets a lot more stable while still distorted/clipped.

the fet was just a fast dirty fix for the low voltage and wavelength distortion.

There is a problem of the "racing" effect which is where the wave takes on sawtooth leaning, but an inverted stage would fix that, I think.

I have the bjt/fet one breadboarded now so I can try it. The cutoff frequency is around 200 hz, so distortion is less down there, but it is consistent and even above that.

I have a second "normal diode distortion" based loosely off of basic diode clipping on a one transistor circuit between collector and power, and I've put on some modifications and taken a few things off to create what passlabs.com would call "second harmonic" distortion. Again, too much input voltage will start clipping it squarely. it is, however, extremely stable in distortion from 100hz to 1000hz and beyond.

The two diode clip design is stable down to 100s hz, cutoff F "knee" being at 76hz.

I figured out that I had produced a "blocking oscillator" by sending the diode/resistor signal through an amp circuit, although it needs an external wave source. Probably fixed with biasing?

I should look into building other public overdrives in spice and on perfboard, to see and hear their differences.


and yes, I believe the diode/resistor was a "limiter stack". There was someplace I saw that had basically a multitude of sets of "basic" intermediary circuits to plug into something bigger.

[GibsonGM]

Well, that's a little less complex, anyway )    You may find the audible differences between the wave you have there and a pure sine aren't too dramatic, Comedian.   A little more 'sparkly' and 'open', perhaps.  Or maybe so little change you just don't hear much difference?  2nd order (octave) harmonics are the root of good drive, and are the characteristics of a moderately overdriven tube tone.   Transistors generally go into odd harmonics, 3rd, 5th etc.  That is 'the land of the flat tops (square waves)".   

So it's cool you've managed to begin clipping smoothly here with the bjt/fet setup.   I'd start here, build it and listen!  It would be especially enlightening if you have access to an O-scope, and could watch the waveform while hearing it AND tweaking components!   

You may have to raise that 100m input voltage to about 1V P-P to more accurately represent the output of a guitar, as well....even 2V!   They vary so much.  So, because all this stuff is analog and variable, it's difficult to REALLY know what's going to happen inside the circuit until you can get in and get your hands dirty!  )

[Thecomedian]

I have to errata the "emitter follower". I thought I was dealing with a PNP, but it's an NPN. This turns out to be even more interesting, but for different reasons. The emitter is having itself "raised" from 0 or ground voltage via biasing, and the voltage bias on th base is what affects the distortion. The distortion occurs as the base tries to reach down to 0 amps, but cant quite get there, and it looks like it's pulled back out of being able to go into cutoff just as it gets there. The reason being that the 0.65v drop between base-emitter is being undervolted momentarily.


More interestingly, leaving R7 at 40Ohm and changing emitter R3 to 60kOhm provides a very nice 5.5v P2P on the output, with an even more stabilized distortion wave. The sawtooth lean is almost completely gone.

1v P2P will push it into cutoff, due the signal completely outranging the biasing DC current. Hmm..

[tca]

Hi, first I would like to welcome you to the world of simulations, I also do a lot of them. There many ways of getting that round knee that your sims shows. Quite general, any BJT "badly" biased will work in a similar way. I've also a made a few sim's with similar circuits, they all sounded good, similar to a "triode" sound but, nevertheless completely different. Let me just put forward some pics:

Fetzer Valve



Fetron type configurations



Jack Orman's Muffer



and my version of Tim Escobedo's bootstrap for gain version of the muffer



Russian schem: http://www.diystompboxes.com/smfforum/index.php?topic=93930.0



And finally the Tube Sound Overdrive



Questions:

Do they sound good?
Yes
Do they sound like a "triode"?
Yes, they all sound like some kind of triode under certain circumstances.

So, what's missing in all of them?
Don' really know, but I know that something is missing

Keep up the good sims.

Cheers.

P.S.
And I've build all of them. You should do the same to get some *real* feeling about the circuits that you sim. One of the topologies that you should sim is the Muff Fuzz (fuzz OD).

[tca]

Actually I know what's missing! It is the transient response of the circuit when you play a note and the way the different harmonics react with the decaying energy of the guitar string, also a transient phenomenon. I guess the most reactive circuit is one that I did not show. Fuzz Face type, "voltage feedback biasing", if you listen carefully to a good fuzz face you will notice what I mean, not only the fuzz but the harmonics game during the attack/decay. Of course the fuzz that you get from that "voltage feedback biasing" is the reason why the harmonics sound the way they do! If you capture that harmonics dynamics with out all that fuzz?!?

One way to do it is to put several circuits in series in order to get not only the round clipping at one part of the signal but also the duty cycle to get the "roundness" on the other part and some of the dynamics. Is there a circuit that does all this? I guess you will need no less than 2 gain stages to perform such a device... maybe 3.

Cheers.

[GibsonGM]

Exactly. With tubes, we bias warm/hot/warm etc.  This leads to clipping pos and neg sides of the wave, but they clip DIFFERENTLY.  You either begin to draw grid current or go into cutoff....1st is less abrupt and more overdrivey-warm, 2nd is more harsh and modern/fuzzy.   If you run multiple stages and play with component values, you can often find an arrangement that gives you a lot of 2nd order harmonic content PLUS the 'fuzz' and drive to go along with it!  Too bad transistors don't react quite as nicely as tubes (even FETs cutoff faster, and the impedance transitions they offer occur rapidly). 

We try to reproduce this tubeness in our pedals, but often fall short.  And you hit the reason why, TCA....trying to do too much with too few devices (as well as the physical differences in those devices).  We need to have several stages, and use them to achieve some non-linear performance, and not too rapidly per stage (which Comedian is coming into with this thread).   Can we get the *real* tube tone?  I don't know; many designs do come close.  Maybe Comedian will come up with the next awesome BSIAB-type pedal!

To me, what we should look at the most is the tone shaping between stages...losing lows before clipping (encouraging 2nd order harmonics and losing the woofiness), and restoring what we want after - that sort of stuff.   The clipping itself is great to understand, but what it does to any particular frequency is more important tone-wise (and yes, esp. with CHANGING, dynamic input!). 

[jishnudg]

Why not extract the duty cycle modulation application from the Carvin amplifiers and use that as a stage that would drive this one (or even mix the two topologies,maybe). From whatever little breadboarding I have done, one consistent result has been that multiple gain stages which gently push the preceding stages give a much rounder, warmer sound than a single IC/transistor trying to do all the work.

[seedlings]

This is a way cool conversation to which I can add nothing of value, but enjoy the dialogue.  is there any way to simulate a decaying signal that starts at 1.5V and decays to 100mV?  as you've discussed, the attack is very important often difficult to manage. I build tube amps about as often as effects, and get frustrated when  a circuit sounds good/bad with single coils, then bad/good with humbuckers.

CHAD

[Thecomedian]

Well sure there is 

The trick is finding it. It's a lot like programming language or other stuff. If you understand the logic behind the behavior, you can manipulate it to do anything you want.

Remember this rule: Voltage rises across the node where resistance is high. When resistance starts dropping, voltage will drop with it as current rises.

You're looking for something which will initially be resistive enough to create high "pressure" (voltage), and then fall down to low voltage (which means that the control of current will determine how fast the fade is, a 1Megohm volume pot seems to make my instrument sing forever from a single stroke).

If i were working with live parts, I'd build a buffer, and then add and remove resistor values after the amplification, maybe change cap sizes, and then record my observations. (warning not to do this while powered and take safety measures when working with high voltages! for all the readers who might be newer..)

[tca]

...is there any way to simulate a decaying signal that starts at 1.5V and decays to 100mV?  as you've discussed, the attack is very important often difficult to manage.

Yes, a simple RLC circuit (check this for the formulas: http://en.wikipedia.org/wiki/RLC_circuit). See the following example:



VG1 is a square wave with .25Hz.

Cheers.

[GibsonGM]

Way cool, TCA  )  Good to know that trick!

Hey - what software are you using there?  Looks pretty neat.   I generally use LTSpice but am always interested in what's out there...

[tca]

I'm using TINA-TI: http://www.ti.com/tool/tina-ti

I've started with LTSPICE... but TINA is very user friendly with a fast learning curve.

[GibsonGM]

Thanks, TCA - I'm still trying to figure out how to model output impedance to accurately reflect the performance of filters (mid scoops etc), and never seem to get it right with LT! 

[Thecomedian]

wow. LTspice is only 13 meg, Tina is 80 meg.