Defeating intermodulation

[Transmogrifox]

intermodulation is my friend.  8^)

Yup--> It's cool to take a couple strings and bend close to the same frequency so that you have a low frequency pulsating sound, then slowly bend down until you can hear the low frequency chirp into an audio frequency.  Not ALL intermodulation terms are non-musical.  Some of these cross terms will play a bass line for you if you just happen to pick the correct chords in the right order.

Design a series of digital frequency tracking filters and clip the output of each filter individually to add only harmonic terms with very little intermodulation and then you have a guitar-synth.  Add in harmonics to whatever degree and balance that you like.

[Ben N]

intermodulation is my friend.  8^)

Yeah, mine too, but even friends can get tiresome!

[Brian Marshall]

i dont feel like reading all the posts, so forgive me if im repeating

if you want to clip highs, use a smaller cap to ground in your feedback network.

[brett]

Kinda OT.
Robert Fripp (ex King Crimson) makes some cool comments about intermodulation in Ralph Denyer's Guitar Handbook.

In Jimi H's studio version of Hey Joe, there's some fantastic IM.  Also, Voodoo Chile (slight return), just after the intro.  3rd string 14th fret bent to 2nd string 12th fret (??).  Fantastic.

[Duff]

Well, I think I´m going for the cap in series with the diodes, since this was the first I had in mind. I would love to try splitting the signal in three gain stages(one for lows, one for mids and one for highs) and then mixxing them together, but I already made the pcb and I don´t have enough room in the box.

...Anyway, although its a bit late, I was talkin about the intermodulation that happens when you compress the whole signal with the diodes, and because the low freq wave carries the others, when hard clipping, you get a practically flat line at the + threshold level, then the wave comes down to the - threshold, another flat line and then the wave goes up completing the cicle. So you end up with an almost square wave more than a sine or an audio signal.
(Check the Plate to plate webpage in the Fat Screamer section)
I wasn´t talkin about the IM that happens when you bend a string and simultaneously pick another. Well it depends wich strings you pluck, sometimes it cool, sometimes its disonant, try bendind two strings a full tone, it´s a pain in the ass!  

[R.G.]

I was talkin about the intermodulation that happens when you compress the whole signal with the diodes, and because the low freq wave carries the others, when hard clipping, you get a practically flat line at the + threshold level, then the wave comes down to the - threshold, another flat line and then the wave goes up completing the cicle. So you end up with an almost square wave more than a sine or an audio signal.

That's what I took you as saying initially. You have two choices:
(1) you can decrease the amount of bass by filtering it down before the clipping stage
(2) you can remove the bass with a crossover and then distort bass and treble separately ala the Quadrafuzz.

There is a third choice, I guess. You can use diode connected MOSFETs instead of single-junction diodes to clip the signal. Diode connected MOSFETs have a much bigger roundover "knee" region where they don't clip but just compress the signal so the highs are not necessarily clipped on top of bass signals but just squashed down on the tops. The signals need to be much bigger to do this, as diode connected MOSFETs need to have a series connected junction diode to prevent reverse conduction through the substrate, and they have a large threshold, up to a couple of volts before noticeable compression starts. You won't get as much "fuzz" but you will get noticeable distortion and much less intermodulation because the clipping is less severe.

[WGTP]

I have tried using Jfets (J201/mpf102) and Mosfets (2N7000/BS170) with favorable results.  Does that work like your advocating R.G. or do I need to add another diode in series with the Mosfet.  Using a DMM I measured the different leads of the mosfet until I go a reading similiar to the SI's.  The 2N7000 seems to produce similiar volume and distortion to the std. SI diodes, but sounds a little different.  Right now I have one in the loop of a Muff Fuzz with an LED for the other diode.  I have also fried several mosfets while experimenting.  

[petemoore]

The Crescendo at the end of RUSH "Working Man"///nice thick 'third note' becomes opaque with the two string attack / 1 string bend technique...

[cd]

The Quadrafuzz sounds great when you tweak the knobs right.  I think that's the main problem with it - you can't just crank the gain (attack knob) and expect it to sound good.

[R.G.]

Does that work like your advocating R.G. or do I need to add another diode in series with the Mosfet. Using a DMM I measured the different leads of the mosfet until I go a reading similiar to the SI's. The 2N7000 seems to produce similiar volume and distortion to the std. SI diodes, but sounds a little different.

It sounds like standard because it is acting standard.

Let's look at an N-channel MOSFET, like the 2N7000. This works in standard mode when the drain is most positive, the source is most negative, and the gate is somewhere between the source voltage and the drain voltage. When the gate is at the same voltage as the source, nothing happens - the channel is off. When the gate is pulled toward the drain, more current flows and (usually) the drain voltage drops. When you connect the gate to the drain, forcing the voltage to be the same, you get the diode-connected MOSFET setup. This starts conducting at the threshold voltage of the MOSFET (0.8 to 3.0V for the 2N7000) and has that nice round knee we talked about.

But what happens if we flip this thing around the other way and make the source more positive than the drain/gate? It ** should** block any current flow if it operates like a normal diode, which is what we want.

But it doesn't. There is an internal "substrate" diode that's an inevitably part of the making of the MOSFET on silicon that acts like a standard silicon diode with its cathode to the drain and its anode to the source. This does not affect normal drain-positive operation at all, but it lets current flow through unimpeded when the voltage is reversed.

If you use two diode-connected MOSFETs back to back, the substrate diodes are also back to back, and the 0.7V substrate diode drop always conducts before the higher-voltage paralleled MOSFET diode can. You **are** using a standard pair of silicon diodes, just doing it the hard way.

This gets fixed if you put another silicon diode in series with the MOSFET so the current can only flow the "correct" way, through the drain toward the source. The external silicon diode then blocks the reverse voltage, and the opposite-polarity MOSFET can do its MOSFET-diode thing in the reverse direction without being shorted out by the substrate diode of the "off" MOSFET.

So - each diode connected MOSFET is really a silicon diode connected in series with the MOSFET with its cathode connected to the MOSFET gate and drain, and the + input of this mess to the anode of the silicon diode, the output being the source of the MOSFET. Take two of those, connect them up back-to-back, and you get a MOSFET diode clipper.

JFETs are different yet. The JFET is a single-junction device. The channel of the JFET is a single trough on the silicon filled with high-conductivity n-type silicon. The gate is a layer of p-type silicon overlaying the n-type ditch. If we ignore the gate, the drain => source is a silicon resistor. Current flows by ohm's law. If we instead put a voltage on the JFET negative with respect to everywhere on the channel, then the reverse bias lets no current flow, but it also depletes a certain cross section of the channel by forcing charge carriers away from the gate overlay. It's exactly like squeezing a garden hose - for the same pressure, less flow happens because the channel/hose is made narrower. By the way, this is why many JFETs have interchangeable drain and source - it doesn't much matter which you call what with this setup.

But if you bias the gate positive with respect to any part of the channel, then a current flows. This is a low-conductivity silicon diode because it's a diode specifically designed to *never* be forward biased, so its conduction characteristics are funny. It has a high (for normal silicon) threshold voltage and a lot of channel resistance in series with it as current flows out whichever end(s) of the channel are more negative. So it has a different sound than, say, a 2N4148. But not all that different, to my ear. You diode-connect a JFET by connecting the gate to either source or drain.

Try the MOSFET/silicon diode thing. It's nice if you like soft clipping. It does not intermod nearly as much as stock silicon diodes.

[WGTP]

Thanks for the explaination, I'll have to think on all that for a while.  I had wondered about using the diode in the Mosfet because I think Aron was using that approach on his SB, but with different mosfets.  The Jfets seem to have more wooley sound.  

[Phorhas]

R.G. ,  Can a Diode connected mosfet clipper be an LED connected mosfet clipper - for greater headroom?

[WGTP]

I was messing with the Muff Fuzz again tonight, and I'm surprised how good the 2 op amps sound by themselves without any diodes.  Adding diodes here and there changes the sound some, but just sort of different timbres.  I'm still not sure about he issues of whether the op amp is distorting with diodes in the FB loop or not.  I know for sure that with the Distortion +, Rat, etc. with diodes to ground that the op amp is clipping at high settings, just disconnect the diodes and you can hear it.
:shock:
Anyway, I still don't know if an op amp distorts as numerous transistors, or just acts like a single one.

What effect does using imbalaced diode pairs to get more asymetrical distortion have on IM? :roll:

I'm also not sure the effect all of this has on IM, but I think this is a cool thread, so I'm pushing it back to the top.  

[R.G.]

Can a Diode connected mosfet clipper be an LED connected mosfet clipper - for greater headroom?

Not exactly sure of your question, but some possible answers are:
(a) yes, a diode-connected MOSFET can be connected in series with an LED for more headroom, although a diode connected MOSFET already has a forward voltage of 2-4V, so adding an LED in series would make the thing clip at about 4-6V, way too big for a single 9V supply, and all you'd hear is the basic circuit clipping, not the diodes
(b) no, there is no way to connect a MOSFET as "an LED connected MOSFET"; a MOSFET when diode-connected does what it does, and there's no clear way to emulate other diodes with it.
(c) yes, a diode-connected MOSFET clipper can be substituted for an LED clipper

I was messing with the Muff Fuzz again tonight, and I'm surprised how good the 2 op amps sound by themselves without any diodes.

An opamp all by itself produces very flat topped buzzy distortion, and some people like that. Craig Anderton did one like this - his Optimum Fuzz Adapter.

Adding diodes here and there changes the sound some, but just sort of different timbres.

That's true of any circuit. Diodes are a fairly brutal way of directing current, and they usually - will - change things.

I'm still not sure about he issues of whether the op amp is distorting with diodes in the FB loop or not. I know for sure that with the Distortion +, Rat, etc. with diodes to ground that the op amp is clipping at high settings, just disconnect the diodes and you can hear it.

There's an easy way to tell. If the opamp *can* respond linearly with the
power supplies, feedback network, and signals it's fed, it will. The trick to understanding whether the opamp is clipping or the diodes in the feedback loop are clipping is to think about how big a signal you're asking the opamp to produce. An opamp running from 9V and biased in the middle thinks it's running from +/-4.5V. Common opamps can't product outputs all the way to the power supply rails, often only coming within a volt or two, so something like a 741 or TL07x opamp will only do maybe +/- 3V of signal. That's all the opamp can do, and when it's asked by the signal and feedback loop to put out more than that, it clips. If you have a signal of 100mv and a gain of 30, it clips - for example. Different opamps have different output voltage ranges, so they will all have different closenesses to the power supplies that they can come. Rail to rail opamps can get within millivolts of the power supply voltages. But at some point, if you ask the opamp to go beyond its output voltage range, it clips.

Enter the diodes. If I have an opamp with a gain of 100 (Rf=100K, Ri = 1K) and I parallel the Rf with a pair of clipping diodes what is the gain and signal size out? We know that with the diodes, it's going to change.

At 1mv in, the gain is 100, so the opamp tries to put out 100mv - and succeeds because 100mv is within its power supplies and does not turn on the diodes. So there is no clipping, and the opamp is running linearly.
At (about) 6mv of input, the output starts trying to put out 600mv of signal, and the diodes start turning on. When the diodes turn on, they look to the opamp like almost a short circuit from output to input, and the effective opamp gain drops to 1. The opamp is fine with +/- 600mv on its output, still linear, but the diodes in the feedback loop are telling it to have a gain of 1 whenever they turn on, so the feedback diodes are doing the clipping and we hear the diodes.

What happens then depends on whether we are driving the opamp through the (-) input or the (+) input. If we're putting signal in through that 1K input resistor, then we're done - the diodes will clamp that signal to +/- a diode drop forever, because the diodes can actually get low enough in resistance to duck the gain below 1, and the opamp is still running linearly, but the diodes have converted the feedback loop into an active attenuator, and the output will be adjusted to suit the diodes right up until the opamp can't supply enough current through the diodes to keep the (-) input happy. What happens next is usually destructive, and it's hard to get to, so we don't need to worry about it now.

If we're driving the opamp through the (+) input, the gain is actually 1+Rf/Ri; the "1+" doesn't matter when Rf/Ri is 100, but when diodes change the gain down it does. When the diodes conduct enough to get the gain down to 1 in the feedback loop, there is still that "1+", so from there on you get the sum of the input signal at a gain of 1 and the diode voltage drops, so you get half signal and half diode clipping. When the input signal is increased still further, the "1+" lets the opamp output increase more and eventually it will bump into the power supplies and you'll hear the opamp limit against the power supply.

Clear as mud, right? It gets worse. Now what happens when we put a resistor in series with the diodes?

Same thing, but now the diodes can't reduce the gain as much as just the diodes by themselves.

Anyway, I still don't know if an op amp distorts as numerous transistors, or just acts like a single one.

You can't really think of it that way. The forward gain is so high and feedback so dominates the opamp's response that the idea of it acting like one or more transistors doesn't apply. That's like asking if the stock market acts like one person buying and selling stock or numerous ones. It doesn't do either - it acts like a mob, which is different yet from one or more people.

What effect does using imbalaced diode pairs to get more asymetrical distortion have on IM?

It makes the results hard to predict. IM is the result of passing
**any two or more sine wave signals**
through
**any nonlinear process**
what you get is not only the nonlinear process on each pure tone, but the nonlinear process times the integer multiples of sums and differences of the input pure tones.  Having mismatched diodes causes the answer of that calculation to be different from + to - signal polarities. You get a different slew of intermodulation products that would be predictable if you knew very clearly the differences between diodes and had a loooooong time to do the math. There may be magic mismatches in there that sound GREAT, but it would be hard to reproduce them

[Phorhas]

Posted: Sat Oct 09, 2004 1:47 pm    Post subject:  

--------------------------------------------------------------------------------

Quote:
Can a Diode connected mosfet clipper be an LED connected mosfet clipper - for greater headroom?

Not exactly sure of your question, but some possible answers are:
(a) yes, a diode-connected MOSFET can be connected in series with an LED for more headroom, although a diode connected MOSFET already has a forward voltage of 2-4V, so adding an LED in series would make the thing clip at about 4-6V, way too big for a single 9V supply, and all you'd hear is the basic circuit clipping, not the diodes  

Code Select Expand



Yeap... that's what I was asking, and I do itnend to use this with a supply  greater then +9V, or may be before any gain stage like th BD2 - can't tell why exactly, but I'll have to give it a shot.

Thanx R.G. - great article... confusing but that just means I have so much more to learn

:)

I'll post if I find anything worth mentioning :)

[WGTP]

Thanks the additional explaination.

It sounds like plugging the Mosfet/Diode clipper into the Muff Fuzz isn't going to have as much effect as even LED's. if I understand correctly.  It appears that the op-amp is likely to do more clipping on it's own rather than the diodes because of the high threshold.  I guess that may or may not reduce IM for the same level of "perceived" distortion.  It seems that the "richer" the harmonic content, the greater the potential for IM???

It also sounds like the fancy Rail to Rail Op amps will distort less, I guess that is the idea. :roll:  May also explain why different op amps sound different.  The "Ratio" of clipping by diode vs. clipping by op amp :shock:

So do Op-amp circuits have more IM than decrete circuits, is that were we started?  

[Phorhas]

So - each diode connected MOSFET is really a silicon diode connected in series with the MOSFET with its cathode connected to the MOSFET gate and drain, and the + input of this mess to the anode of the silicon diode, the output being the source of the MOSFET. Take two of those, connect them up back-to-back, and you get a MOSFET diode clipper.

Hi R.G.
In Jack Orman's Shaka 5 the diodes are connected to the gate and source of the mosfets and the drain is the output
http://www.muzique.com/schem/shaka5.gif

Is he wrong or is it to produce a differant outcome (BTW he use Ge diodes with the MOSFETs too) ?

[WGTP]

Jack may just be using the diodes in the MOSFET's, in addition to the GE diodes.

Since "most" op amp clippers use diodes, and some folks don't like op amp distortion, does that mean that Op-amp distortion without diodes has more IM.  Or, are they not liked for other reasons.