Intermodulation: How can we avoid or control it?

[Findeton]

I was designing my first distortion pedal when I realized that intermodulation distortion is one of the main things that marks the difference between a good distortion pedal and a bad one. I am designing a distortion that gives mostly second harmonics... but intermodulation is too problematic. So I've been searching how to manage it, because what we all want is to obtain a distortion that sounds well when we strike 2 strings together!


One way of avoiding intermodulation that i've read about is  through frequency filtering. Basically, intermodulation happens when you soft/hard clip a signal containing 2 frequencies (for example f1 and f2). Harmonic distortion would produce frequencies like 2*f1 or 3*f2 (harmonic distortion is generally good), but intermodulation creates frequencies like f1+f2 or f2-f1, which sound extremely harsh.  So, if we separate the frequencies into two or more parallel channels with some band pass filters, and we soft/hard clip each channel, we could theoretically avoid all the intermodulation: that would happen if each frequency of the input signal goes to a different channel. Finally all those channels would be added to the final output stage of the pedal/preamp.

But, normally, we don't use different bandpass filters feeding different channels, do we? Some people just do some pre-EQ to boost the frequencies we want to overdrive/distort and then we use a post-EQ to get rid of some of the intermodulation products we've created.

Another way to reduce intermodulation is to create the distortion through many stages in series, each one giving a little amount of soft distortion. What I think this mechanism achieves is to get a progressive reduction of gain versus input amplitude, this can somewhat lower intermodulation products because it manages to retain a good amount of the original tone. Of course you can combine this mechanism with frequency filters so each stage soft-clips one part of the audible bandwidth.

And, what a tubescreamer does (among other things) is to mix the clipped signal with an attenuated version of the input. This way, it doesn't matter how much you overdrive/clip the signal, because part of it will come to the end, so the output will retain part of it's original tone, and while it doesn't actually reduce the energy of intermodulation products, it increases the signal/intermodulation ratio.

Well, those are the options I've come to reading here and there, but, if anyone knows of other methods of avoiding intermodulation, please tell us!

[Gus]

Are you sure IM is a problem?

I think of the distortion products as harmonic partials that are added.

http://en.wikipedia.org/wiki/Harmonic_series_%28music%29#Terminology

[FiveseveN]

Yes it seems you've mixed up a couple of phenomena but try this:
Hexaphonic pickup -> distort each string individually -> mix (and maybe filter)
It's been done before; there was a page with sound clips somewhere but I can't find it.

[Findeton]

Are you sure IM is a problem?

I think of the distortion products as harmonic partials that are added.

http://en.wikipedia.org/wiki/Harmonic_series_%28music%29#Terminology

IM is always a problem! IM always sounds harsh to the ear. There's good and bad distortion (for electric guitar sounds) and IM is almost always on the second group, afaik.

Yes it seems you've mixed up a couple of phenomena but try this:
Hexaphonic pickup -> distort each string individually -> mix (and maybe filter)
It's been done before; there was a page with sound clips somewhere but I can't find it.

Well, I just tried to summarize the different ways of avoiding IM that I know about. Hexaphonic pickups sound interesting, although I wouldn't use a different circuit to overdrive each string (too many circuits indeed). I think that a good way of avoiding IM while getting lots of distortion could be joining into two groups the 6 separate pickups of an hexaphonic pickup. For example, we put strings 1,3 and 5 in one group and send it  to one stompbox and strings 2,4 and 6 on the other group to another identical stompbox. That way, if we only play power chords, the only IM products we could get would actually sound musical.

But, this is not what tubes normally do, and, still, they achieve to get lots of distortion and low IM products...

[earthtonesaudio]

Play fewer notes.

[FiveseveN]

But, this is not what tubes normally do, and, still, they achieve to get lots of distortion and low IM products...

Sorry but that's just not true
Funny, I was just thinking while I wrote the last reply about the same kind of setup you've suggested. Yes, it would be better than nothing but people (hopefuly) usually play more than 2 strings at once. And there are plenty of cool simple distortion circuits. The pickup and cable is the tricky part IMO.
Dammit... I knew I should've picked up drums instead of guitar!

[earthtonesaudio]

If you ever hear a hex guitar pickup with fuzz, you'll know instantly that it does NOT sound like a typical distorted guitar.  It sounds more like an organ.  Actually, it sounds a lot like an organ. 

For it to sound like a distorted guitar, it must have intermodulation distortion.

[PRR]

> Are you sure IM is a problem? ....wiki/Harmonic_series....

Different things, the harmonic series, the single-tone distortion, and the multi-tone distortion.

> IM always sounds harsh to the ear.

So does odd-order (to a lesser degree), yet that's "essential" to modern guitar sound.

I have also observed that people don't "think it is loud" until they get significant IM. In dance studios, speakers with higher IM (2-way) get played at lower acoustic levels... they turn-up until IM is 5%-20%. Give them a cleaner (more-filtered, 3-way) speaker, they use a lot more Watts and annoy many more neighbors.

> normally, we don't use different bandpass filters feeding different channels, do we?

Nearly all fuzzes filter. Many-band filtering is too much work. So some of the differences are in how they filter versus how they handle multiple notes at once.

I think you have the physics in hand. I'm not sure that taking several small bites gives lower apparent IM than one good bite: it may have a lower peak but there are more artifacts at more frequencies, more "haze". We may also wonder how much filtering the ear really wants. On full-range material at high level, through modest-size speakers where IM is intrinsic, I don't like 2-way (3KHz crossover) because bass-vocal ranges beat together, true 3-way (800Hz/5KHz) speakers are so much better that I'd rarely want more filtering just to clean the midrange (the next step would be 150Hz crossover so the midbass is not muddied with the deep-bass). I dunno what guitar "wants".

Don't overlook slope. Boost the highs, clip, cut the highs. Many fuzzes do this, even if incidental to empirical tweaking. I have built a wideband differentiator, clipper, and integrator: when handing speech in GROSS clipping, it was "unnatural" but intelligible and not offensive.

[Mark Hammer]

For this discussion to go anywhere, I think we need to distinguish between IM produced by the circuit, and sideband products produced by the string itself, which are then amplified by the pedal.

Plug into almost any octave-up fuzz, and do a double bend.  You will hear a tone moving down even though the bent strings are moving up in pitch.  That's a sideband product produced by the circuit, but it begins with a very small sideband product produced by the string itsel, which is then magnified by the circuit.  It happens because, unless you use ultra-heavy gauge flatwound strings, the pitch of the string is not fixed.  The string is bendable and when you pick then bend a string, you have asked the string to vibrate in one way, but then are quickly making the string vibrate in other ways, in real time.  The string is a physical body, not a circuit, so the physics dictate that it will produce vibratory components as a result of the various interference patterns created by the pick-n-bend action.

I  honestly can't think of a single player that would use a distortion and NOT bend strings.  Maybe somewhere out there in the experimental field, but nobody who tries to play 4-minute songs would be like that.  So the presence of string-originated intermodulation and sideband products is rarely, if ever, going to be something that goes away, even with the most cleverly-designed circuit.

Having said all of that, the $64,000 question is how much of the objectionable IM we hear comes from the string alone, how much from the circuit alone, and how much from the circuit making what the string does worse.

[Findeton]

If you ever hear a hex guitar pickup with fuzz, you'll know instantly that it does NOT sound like a typical distorted guitar.  It sounds more like an organ.  Actually, it sounds a lot like an organ. 

Something like this video, I guess. Yes, it actually sounds  well but in an organ-like way, not a normal guitar distortion.

For it to sound like a distorted guitar, it must have intermodulation distortion.

This actually is a relief, because is difficult to control IM when overdriving the guitar signal.

> IM always sounds harsh to the ear.

So does odd-order (to a lesser degree), yet that's "essential" to modern guitar sound.

Well, afaik, tubescreamers create only odd harmonics and they sound well.

Also, if it's not because of IM, why is it good to use many soft-clipping stages, appart than for frequency shaping in between them?

[Ben N]

IMD is why double stops and, as Mark pointed out, double bends, are so effective as "exclamation points" in distorted leads. Having said that, you already alluded to one possible way short of a hex pickup to reduce, but not eliminate, IM, and that is by seperately fuzzing fairly narrow frequency bands. This was one of the promised benefits of the Anderton Quadrafuzz, IIRC. As Mark had commented efore, the Quadrafuzz is an interesting idea that could stand some tweaking, ut, for example, reducing clipping on the lo-pass band would not interfere with IMD reduction at all.