Intermodulation Distortion

[FredB]

Is there any easy way to guesstimate how much IM distortion is being generated in a circuit from the FFT graph and the magnitude relationship between the fundamental and the first side bands?

http://s19.postimage.org/kk1zffzxv/im_distortion.jpg

[Steve Conner]

Well, your two biggest sidebands are 60dB down on the fundamental. So you have about -57dBc worth of IMD. I don't have a calculator handy, or I'd work out what that is in percent. I'll guess 0.3%.

If you're working in audio, you might want to weight the sidebands perceptually, and that could get messy.

[FredB]

This is just to determine which circuit configurations have the lowest IMD, so perceptual weighting would be over kill.

I think I got it, Just find the RMS total power for the fundamental and how ever many sidebands, then subtract the fundamental, then divide that by the total got get a relative percentage. That should work well enough.

[J M Fahey]

Yes, that's about it.
In fact we are looking at an incredibly clean signal.
You are modulating the 7KHz with 60Hz, aren't you?
Where did you get the original signals from?

[Steve Conner]

I think I got it, Just find the RMS total power for the fundamental and how ever many sidebands, then subtract the fundamental, then divide that by the total got get a relative percentage. That should work well enough.

Kind of. Because of the way decibel quantities sum, you may be able to ignore all but the two biggest sidebands, as I did.

The equation would be distortion power divided by total power. I simplified that as the power in the two biggest sidebands divided by the power in the carrier.

If you want to express it as a percentage, remember that it's a voltage ratio. So, -57dB is nearly -60dB, which is one part in a thousand, 0.1%. The 3dB difference makes it up to 0.14%.

Not bad for a vinyl record, but not good enough for the front end of a shortwave receiver either.

[FredB]

Yes, it was 60Hz and 7kHz,

It was generated by the SPICE simulation of my clean booster circuit.

[J M Fahey]

Well, that's one of the cleanest boosters I have ever seen.
What signal levels are we talking about?

[FredB]

60Hz@0.3Vpk and 7kHz@0.1Vpk

The booster is currently setup for 20db boost, so the 60Hz output with the 7kHz ripple is right about 2Vpk.

[Steve Conner]

Well, that's one of the cleanest boosters I have ever seen.
What signal levels are we talking about?

Have you ever seen a TL071? Any modern op-amp should beat this IMD performance easily.

That was why I mentioned perceptual weighting earlier. Nowadays op-amps are so good performing and so cheap, that we only use discrete circuits in situations where we want euphonic distortion. So, designing a discrete small-signal circuit for low measured distortion is pointless. (Unless it's a power amp to handle voltages higher than op-amps can cope with.)

I'm not sure how perceptual weighting works with IMD, but I expect it would imply more low-order products and less high-order ones, just the same as it does with THD.

[FredB]

I specifically went discreet with this booster to be able to shape content of the harmonic distortion before clipping, and the manor of the clipping itself. The booster sounds clean and smooth with plenty of dynamic headroom. It does sound less warm than say RG's basic MOSFET booster, although it still sounds warmer than an op-amp.

[kg]

clean is no fun.

dirty is where it's at.

[FredB]

That's a matter of perspective. This clean booster project arose out of my latest amp design based on the Fender 5E3. I had originally set out to use one half of the stock 12AY7 as an in amp booster for 12db gain.

After talking with a few 5E3 gurus, I decided to keep the input section of the amp in the original configuration for two channels, and just use a external booster. I wanted a clean boost as to avoid coloring the natural sound of over-dirving the boost stage for the PI.

An op-amp probably would do fine, I just wanted to warm the boost tone a little, although mostly I waned to be able to get at least 12db boost without clipping in the booster. I've bread-boarded this circuit, which sports a JFET and two BJTs, and it sounds wonderful, just what I wanted. It only clips when the boost is turned all the way up and the guitar strings are picked extremely hard, harder then normal playing would entail.

In simulation with a 250mV 1kHz test signal the booster has 0.2% THD, and that's with 20db of gain. The JFET input allows the boost to be attuneated to 0db with out loosing any of the frequency reponse on the high end. I feel this achieves the goal for this project.

As a variant I worked out the component changed to increasing the THD to something like 5%or more which would be mostly 2nd, 3rd, 4th, and 5th harmonics to add in warmth. We'll see.