Defeating intermodulation

[Duff]

Ok, does anyone really know how does a fb loop in a distortion´s OPAMP work, & I´m not talkin´ about setting the gain with resistors, I´m talking about how does the signal go in, go round the loop, get´s clipped by the diodes and gets amplified. More precisely, I need to amplify the whole input signal, but only clip the mids & highs to terminate intermodulation (which results in bass loss, treble definition loss, and an awful sound) and to get kicking ass punching mutted power chords, as well as well defined chords and arpeggios. By the way I´m using a simulated tube type dist. with two OPAMP stages from Jack Orman´s website, the first stage has a max gain of 100, this one obviously needs the trick, te second one has a fixxed gain of 10, does this one also need it? Take notice of the unusual fb loop that creates assymetrical clipping.

God bless the one that helps me!

[christian]

Hi. I think I didn´t get that one.. Maybe you should try a marshall tone stack which lets you dial out the treble and bass and enchance the mids.
You´re having problems when two notes(hi-freq) are played at the same time? maybe bending the other and a side-freq comes that changes whle one of the notes is bended?
That´s really common with most of distortions, and I think you can use an EQ that has bass/mid/treble adjust.
And putting treble and bass to 6-8 and blast the mids to 11, you get really sharp and kickass powerchord sound. Listen to Nirvana´s live recordings.

Didn´t asymmetric clipping cause more intermodulation?

[R.G.]

Ok, does anyone really know how does a fb loop in a distortion´s OPAMP work, & I´m not talkin´ about setting the gain with resistors, I´m talking about how does the signal go in, go round the loop, get´s clipped by the diodes and gets amplified.

I'm sure someone knows. I'm sure I don't really understand them, but maybe I can give you a way to look at it that you can use.

Let's take an inverting stage first, positive input grounded. With 0V on the series input impedance, and the + input at 0V, there is zero difference between the + and - inputs, so the opamp gain of infinity minus one times zero is zero, and the output sits at zero. Great, we're all balanced.

If we perturb the input to the series input impedance a little, say by putting a fraction of a volt on it, then the opamp is no longer balanced. The - input is sitting at zero, so a current flows through the input impedance. That current piles up at the - input because the - input is a quite high impedance and the opamp can't eat the current. This causes the - input to rise a fraction of a volt, and the difference between the two inputs is no longer zero. That difference is amplified by the opamp gain of infinity minus one, and the output moves in the opposite direction from the way the - input moved.

However, there is a feedback impedance. The output moving in the opposite polarity to the - input causes a voltage difference between the - input and the output, and that voltage difference sucks a current from the - input to the output. The output can only suck as much current as the input impedance is allowing through (the opamp input cannot either eat or supply current remember), so the output eats exactly as much current from the - input as the input impedance is letting through.

So the inverting opamp works by current balance. The input current through the input impedance is exactly balanced by an equal and opposite current sucked out of the - input through the feedback impedance. The opamp's circuitry maintains this balance exactly from nanosecond to nanosecond. If you look at an inverting opamp circuit under operation, the inverting input remains at ground (well, whatever voltage the + input tells the - input is ground, anyway).  This is why you will hear the summing node at the - input of an opamp referred to as a "virtual ground".

Notice that I've been careful to say "impedance". The impedances leading into the - input and from output to - input do not have to be resistors, and may be nonlinear. The opamp doesn't care. It just balances the voltage on the - and + inputs to be equal.

If the impedances are resistors, everything is easy to see. The current in is Vin/Rin, and the current through the feedback resistor must be exactly equal, to that. So if you have Rf in the feedback loop, the voltage on the output must be Vout = Rf/If = Rf/(Vin/Rin) =Vin*Rf/Rin. A little rearrangement gives the gain: Vout/Vin = Rf/Rin.

If the impedances are not resistors, the current balancing remains the same, but Vout is now whatever it has to be to make the equal and opposite current flow out of the - input and balance the current coming into the inverting input.

If, we have, say, a pair of back-to-back diodes in the feedback loop and the input resistor is 10K, then the output is clipped, not because the diodes clip it, but because they have a variable amount of current that they let through per change in forward voltage. They will let through **amperes** of current with a forward voltage of 0.7V, so to balance any reasonable voltage on that 10K resistor, the output of the opamp only needs to move up or down 0.7V.

Anything in the feedback loop of an opamp is being driven in a constant current mode.

If you put a cap in the feedback loop, things get a little funny. A cap can't possibly pass DC, so if you put a DC current into the summing node, a DC current has to come out, and the only way that a cap can pass that is if the voltage across it is increasing continuously. So a cap in the feedback loop with a DC input shows a ramping output - at least until the opamp runs out of power supply room and quits working correctly. Almost all caps in opamp feedback loops have alternate DC pathways around them so  the opamp can be stable with DC inputs.

In the non-inverting operation, the input to the - side is held at some static voltage and the + input is waved around instead of being held steady. Since the opamp will cause the - input to go to the same voltage as the + input, as it must for linear operation, then the output must follow the + input; but since the - input is moved around relative to the reference voltage for the whole mess, it looks like there is an additional input to the - side of equal and opposite to the signal at the + input, so the output voltage is equal to the movement of the +





Every nonlinear signal process creates intermodulation distortion when there is more than one frequency present in the signal.

Let me say that again:
Every nonlinear signal process creates intermodulation distortion when there is more than one frequency present in the signal.

This is one of the things that Mother Nature has simply decreed will happen.

You cannot by some diddling with diodes and caps in the feedback of an opamp get free of intermodulating bass with mid and treble. You can only make it better or worse by affecting how much bass, mid or treble you put into the signal-o-matic grinder of the clipping stage.

What you *can* do is to split off the bass from the mids and treble with something like an active speaker crossover (one highpass filter, one lowpass filter with the same cutoff frequency) and process the bass separately from the mid/treble and then mix the whole mess back together. That way you can get no intermod on the bass because it was never clipped, or you can clip it separately from the mid/treble and get no intermod because the two bands were never intermodulated.

Which leads us to Craig Anderton's Quadrafuzz.

[Paul Marossy]

"Which leads us to Craig Anderton's Quadrafuzz."

As I was reading that last part, I was thinking of the Quadrafuzz, and it was the next line!

Anyhow, the concept is good, but it seems like a lot of those who have built it seem to be a little less than enthused with it.  :?
I guess some people might like intermodulation distortion better?!

[Paul Perry (Frostwave)]

Suppose you had a quadrafuzz, plus an extra 'full band' fuzz (no big deal, you already made FOUR fuzzes!) and then you have a mix control for varying between quadrafuzz out & straight fuzz out, then you have an intermodulation control!!

[WGTP]

Some folks have been using bass boost after the distortion lately.  
As I understand it, reducing the bass thru the distortion reduces the IM distortion and then boosting the bass afterwards brings your bass back in, plus more if you want it. By controling the bass roll off in and the bass boost out, the IM gets varied, I'm guessing?   :?

I have noticed one of the things I like about the notch filters I've been using is that adjusting for a fairly high mid-frequency 1.2K of 4db brings out the bass that I moderately rolled off in the first stage of the distortion for a smoother sound.

[Paul Marossy]

That's interesting about the bass boost thing. IM dist. does seem to be most noticeable at those lower frequencies, at least to me.

[R.G.]

I ran out of time typing the earlier reply. Insert this section after the matching quote.

In the non-inverting operation, the input to the - side is held at some static voltage and the + input is waved around instead of being held steady. Since the opamp will cause the - input to go to the same voltage as the + input, as it must for linear operation, then the output must follow the + input; but since the - input is moved around relative to the reference voltage for the whole mess, it looks like there is an additional input to the - side of equal and opposite to the signal at the + input, so the output voltage is equal to the movement of the +

... input plus the gain through the negative side. (I got interrupted at the end of that paragraph). Since the inverting side is effectively moved in the opposite direction from the + input (relative to the + input, that is), then the sign of its input signal is reversed, and the inverting input signal moves *with* the signal from the + input.

So for a non-inverting input, the gain is Vout/Vin = 1 + Zf/Zi, where Zfi is the feedback impedance and Zi is the input impedance from the inverting input to whatever reference voltage the inverting input is referenced to.  The feedback impedance is still run in constant current mode, since the only thing that can keep charge from building up on the - input is a balancing current from the output, so the output moves in a way to keep the net charge on the - input at zero.

Feedback diodes work the same way here, except that the output is equal to the + input no matter what (that "1" in the "1+Zf/Zi") and then the voltage to keep the signal current flowing in the feedback diode is added to the input signal.

The odd way an opamp has of making an output voltage match what the feedback element does for constant current has an "inside out" effect.  A notch filter in a negative feedback opamp loop has the effect of a resonant peaking filter at the output of the opamp. A capacitor from out to input is a lowpass filter, instead of the high pass a series capacitor usually is. An inductor feedback element becomes a highpass filter at the output of the opamp. A resistor-cap to ground-resistor feedback network is a high pass network, not a low pass as it would be normally. A compressor in a feedback loop causes the opamp output to be an expander.

But back at intermodulation. Intermodulation is really not related to a deep understanding of how an opamp operates.

=== then the note picks back up with ====

Every nonlinear signal process creates intermodulation distortion when there is more than one frequency present in the signal.

[jrc4558]

OK, how about this:

You want to clip the highs more than lows: use a 0.01µF capacitor in series with the clipping diode pair.
You still want to clip lows? Use another pair of diodes in the NFB loop, but use a 2.2K resistor in series.

Anybody did that before?

[niftydog]

reducing the bass thru the distortion reduces the IM distortion

Here's my theory;

sum and differenece products make up IM distortion.

The lower frequencies appear to create greater IM dist. because both the sum and difference products are still within the audible frequency range.

High frequency IM sum products are above the audible range, hence you tend to get less overall IM distortion.

[Duff]

Almost all caps in opamp feedback loops have alternate DC pathways around them so the opamp can be stable with DC inputs.

So how do I make the alternate DC pathway without disabling the cap´s ability to filter freqs?

[puretube]

OK, how about this:

You want to clip the highs more than lows: use a 0.01µF capacitor in series with the clipping diode pair.
You still want to clip lows? Use another pair of diodes in the NFB loop, but use a 2.2K resistor in series.

Anybody did that before?

that`s the way to go...

[puretube]

Almost all caps in opamp feedback loops have alternate DC pathways around them so the opamp can be stable with DC inputs.

So how do I make the alternate DC pathway without disabling the cap´s ability to filter freqs?

big R parallell to any other FB stuff...

[gez]

OK, how about this:

You want to clip the highs more than lows: use a 0.01µF capacitor in series with the clipping diode pair.
You still want to clip lows? Use another pair of diodes in the NFB loop, but use a 2.2K resistor in series.

Anybody did that before?

On the old forum I posted a mod for diode clippers which involved a cap in series with the diodes.  You have to set the amp up to limit bass as gain is increased (like the MXR dist. +) and the cap is set up to have slight reactance to the very lowest frequencies so that it acts as a high-pass filter to the diodes and low-pass to the output.  In this way a small percentage of lows are passed on to the output 'un-clipped' as the gain knob is turned, which compensates for the thinning effect but intermodulation is kept low.  

If you get the balance right (and I posted ideas on how to make this variable), there should be no discernable increase/decrease in low end as gain/distortion/whatever knob is turned.  Also, because some of the original signal is mixed in there, you retain more of the natural tone of your instrument.  You do get slightly less distortion on the low strings, but in practise this works out ok - you get a huge sound minus all the mud.

I used to harp on about this from time to time but, to my knowledge, nobody's ever bothered to try it.  Considering the amount of people here who post schematics for ABCs (another bloody clipper), I find this truly amazing...

[zachary vex]

intermodulation is my friend.  8^)

[GM Arts]

Best approach for me has been suggested above - roll off bass before the overdrive (like the TS design), and boost it afterwards.

Another approach that DOES work is to use a synth hexaphonic pickup, and feed each string through it's own overdrive channel - that way you can customise each channel to suit the string and pickup position.  You'll need to find a way to control 6 drive levels from one control (and 6 tones if you want to get that fussy).

You can try this on some of the older Roland guitar synths (I'm not sure about the newer ones).  Has to be heard to be believed - you can even play jazz chords at over-the-top fuzz levels.  Having said all of that, don't
expect to hear this on Metallica's next release - I think they actaully want the extra bass notes produced by overdriven 5th chords ;-)

GM Arts

[jrc4558]

Zachary, I always wanted to ask you this:
What are your sympathies in music? I would imagine that these would somehow influence the choise of sounds you will be aiming your designs after. Having heard some of your marvellous pedals, I didn't come to any conclusion as to what music you may be listening. What inspired you to design The Machine for example?
Sorry for the off-topic.

[WGTP]

Welcome to the Machine - Pink Floyd   :twisted:  You played a mean guitar...

I've wondered about using diodes in series with a cap and whether or not it boosted the bass since the bass would be forced to go thru the resistor instead.  I guess its a matter of balance (like most things in life) as to whether the cap/diode or resistor offers the path of least resistance (usually my favorite) for the bass.  

The dual channel idea sounds cool.  

Combined with Tim's TMK's circuit and you could waste your life away.  :shock:

[Mark Hammer]

The thing everyone has been ignoring here is that the guitar is a string-tension-based instrument with far too much string compliance built in.  What that means is that pitch stability is more imposed mentally by the listener than actually present.  

Last year when JC Maillet and I were noodling around with octave boxes in my basement, we realized that it was common to just about all of them that IM would be more evident with string bending, and in some cases highly obvious.  The most obvious case would be the Green Ringer which people have a hard time deciding if its a ring modulator or an octave box ("It's a dessert topping!"  "It's a floor wax!" "Stop, you're both right!"  New Shimmer is a dessert topping AND a floor wax." - Little SNL nostalgia there).

Okay, although the source of harmonics is slightly different in your average distortion box than in a FWR or other sort of octaver, the fact remains that, by definition, any distortion box is putting out harmonics that crash into each other in the face of both multiple strings exitting the same jack, and strings that vary in pitch as we fret.

I think the bottom line is this: there are things you can do to reduce IM but if you really want to get rid of it, try using heavy gauge flat wound strings, strung tight enough so that they are unbendable.

[WGTP]

Mark is that a "Steel Guitar"?  I tried it, but it sounded like a Gong.  

I remember seeing a strat style pickup that had a coil wound around each magnet individually and then 6 outputs. (And another that had them mounted on a slide so you could reposition each string, plus use different magnets and coils for each, can't find the link)

Low E Black Fire
     A Sweet 16
     D Vulcan
     G TS
     B  Rat
Hi   E  Tone Bender

6 Channel Mixer with parametric EQ on each.