What audio circuit would you build to intentionally produce high IMD?

[MaxDM]

I was wondering: if someone was to build a circuit to intentionally add heavy IMD, without clipping, what would you start with?

In other words, what kind of circuits are notoriously prone to IMD?

 

[JohnRoberts]

Why... ?

IMD cross modulation is worst in circuits with poor linearity so avoid negative feedback and other linearity improvements.

For TMI the original SMPTE IMD test used 60 Hz and 7 kHz for test signals. Back a long time ago 7 kHz was fast and difficult. In the 70/80s for bench testing phono preamps I modified my Heathkit SMPTE IMD bench unit to use 19 kHz and 20 kHz. This improved IMD test revealed grunge in phono preamps that simple THD testing could not see (the RIAA feedback EQ is effectively a low pass filter so diminished harmonic distortion overtones).

JR

 

[MaxDM]

I'd like to experiment with it as a distortion effect, but I would like to avoid clipping, as much as possible.

 

[JohnRoberts]

The classic fuzz distortion is based on saturating a diode junction. More subtle distortion could be too subtle.

JR

 

[MaxDM]

Well, in a way, I'm looking for subtle. What I'm after is, in part, a masking effect, perhaps one way to describe it is a sort of 'AM Radio' sound, more than a hard distortion.

 

[JohnRoberts]

AM radio is mimicked by narrow bandwidth and limited dynamics, I don't know that distortion is a dominant characteristic.

JR

 

[Bo Deadly]

I was wondering: if someone was to build a circuit to intentionally add heavy IMD, without clipping, what would you start with?

In other words, what kind of circuits are notoriously prone to IMD?

I think this is a worthwhile exercise if only to understand IMD and it's place in audio and in distortion circuits like guitar pedals or synths.

Not that I'm an authority on the topic but my understanding is that IMD is created when a component at one frequency modulates the amplitude of another component at a different frequency by way of a nonlinearity. Unfortunately I don't think I could be more specific than that without dropping down into LTSpice.

But I would start in LTSpice by generating two tones - one at 10kHz and one at 11kHz, mixing them together through to resistors into a simple virtual ground to make a pure signal with 10kHz and 11kHz (I don't think LTSpice has a feature to do this using just a voltage source but I could be wrong). Then I would just run that through a resistor and two anti-parallel diodes to ground and look for components at 1kHz, 21kHz, 9kHz, 12kHz, ...

Note that if by clipping you mean using diodes like I describe, I would not consider that to be clipping per se since it is being used as a simple voltage dependent variable resistor to create a non-linearity. It does not have to be driven to the point where you can see the top of the waveform being clipped. I'm not sure how you can generate IMD without a non-linearity.

UPDATE:

I think this page explains more precisely how one signal modulates another:

Intermodulation Distortion

 

[abbey road d enfer]

I was wondering: if someone was to build a circuit to intentionally add heavy IMD, without clipping, what would you start with?

In other words, what kind of circuits are notoriously prone to IMD?

The closest possible answer is usin an analog multiplier. The transfer function is extremely linear but the output is pure IMD.
However I very much doubt it has anything in common with AM. As JR said, restricted BW and dynamics are key.

 

[Winston OBoogie]

When transistors were expensive and new, weren't portable transistor radios advertised by the number of transistors they had inside?

Build a new fangled "6 Transistor AM Radio!!!".

 

[JohnRoberts]

The closest possible answer is usin an analog multiplier. The transfer function is extremely linear but the output is pure IMD.

But it is unclear what to use as the two factors to multiply. SMPTE is specifically looking for 60Hz modulation of a 7kHz tone because back in the day movie projectors used incandescent bulbs and 60Hz routinely corrupted audio paths. I guess you could use a crude crossover to separate out a LF pass band, and multiply the full range by that LF content.

I make no predictions about how that would sound, it would surely be distorted.

However I very much doubt it has anything in common with AM. As JR said, restricted BW and dynamics are key.

yup... +1

JR

 

[john12ax7]

How technical do you want to get? A previous research field of mine was pre-distortion linearization, if you search for that you should get some info. Essentially it is an intentional generation of IMD that was then used to cancel out (linearize) the system IMD. The concepts could be adapted for audio once you understand the merchanisms, you just ignore the cancellation part.

 

[jacomart]

Normally, to correct a non-linearity distortion, an equal and opposite one is inserted, a linearity corrector can therefore be defined as a non-linearity generator. Precorrectors are widely used to correct intermodulation distortions in RF amplifiers (usually larger than those in audio amplifiers)

Cheers
JM

 

[MaxDM]

How technical do you want to get? A previous research field of mine was pre-distortion linearization, if you search for that you should get some info. Essentially it is an intentional generation of IMD that was then used to cancel out (linearize) the system IMD. The concepts could be adapted for audio once you understand the merchanisms, you just ignore the cancellation part.

well, that sounds interesting!

 

[jacomart]

Normally, to correct a non-linearity distortion, an equal and opposite one is inserted, a linearity corrector can therefore be defined as a non-linearity generator. Precorrectors are widely used to correct intermodulation distortions in RF amplifiers (usually larger than those in audio amplifiers)

Cheers
JM

I see that I have been beaten on time by @john12ax7

Cheers
JM

 

[john12ax7]

Thanks for the paper JM, haven't looked into it too much for audio, so should be a good read.

 

[ruffrecords]

IMD is basically AM modulation where where a high frequency tone is modulated by a lower frequency one and thereby produces sidebands. The simplest way to deliberately create this would be by using a multiplier.

Cheers

Ian

 

[Bo Deadly]

Normally, to correct a non-linearity distortion, an equal and opposite one is inserted, a linearity corrector can therefore be defined as a non-linearity generator. Precorrectors are widely used to correct intermodulation distortions in RF amplifiers (usually larger than those in audio amplifiers)

It looks like that paper is about correcting for non-linearity of transistors in general and not specifically about IMD. Running any signal through a non-linear circuit will generate IMD but it's not clear to me that the principles in that paper will yield IMD any more than just running signal through a trivial resistor / diode circuit.

 

[Bo Deadly]

IMD is basically AM modulation where where a high frequency tone is modulated by a lower frequency one and thereby produces sidebands. The simplest way to deliberately create this would be by using a multiplier.

The simplest way would be to use a diode which is a perfectly effective multiplier for this purpose. Using an analog multiplier IC would be considerably more involved but no doubt you could control things more. Also, as a practical point, analog multiplier ICs tend to be noisy.

 

[abbey road d enfer]

The simplest way would be to use a diode which is a perfectly effective multiplier for this purpose. Using an analog multiplier IC would be considerably more involved but no doubt you could control things more. Also, as a practical point, analog multiplier ICs tend to be noisy.

Well, the OP wanted something that doesn't clip. that's why I proposed a solution that has very linear paths.

 

[jacomart]

the best method is to create a complementary non-linearity to the circuit to be corrected, working on carriers is fine only for CW signals. Often, for RF signals, intermodulation is due to power reflected in the output stage due to poor impedance matching

Cheers
JM