Input Attenuation and Intermodulation Products

Question

I am measuring intermodulation using spectrum analyzer. How does the input attenuation affect the power of intermodulation products?

Answer 1

So, based on your comment, you want to know what happens when you change the "attenuation" setting on your spectrum analyzer's front panel.

Exactly how a spectrum analyzer's front panel controls affect intermod distortion depends on the analyzer's construction. Good analyzers should come with manuals showing their internal arrangement for just this reason.

In general, you can expect a spectrum analyzer to be laid out like so:

(Copied from https://www.keysight.com/main/redirector.jspx?action=ref&lc=fre&cc=FR&nfr=-536900402.536880437.00&ckey=2010290&cname=EDITORIAL)

There's not much between the front panel plug and the mixer, and mixers tend to suffer from intermodulation distortion.

So, in general, if you're seeing a distortion product in the output of your spectrum analyzer, it's either coming from the system your testing, or its being generated in the spectrum analyzer itself. (Or both, just to make your life extra fun).

Also in general, if you decrease the input voltage to the mixer by a factor of \$N\$ (I'm going to try to keep voltage and power separated here), you would expect to see all of the signal component voltages drop by a factor of \$N\$.

If there's \$k^{th}\$ order intermodulation inside the spectrum analyzer, however, then you would expect to see the \$k^{th}\$-order product's voltage to go down by a factor of \$N^k\$ -- so if you dropped the input voltage to the mixer by a factor of 2 (6dB), and there were 2IM products, you'd expect the output voltage of that product to drop by a factor of 4 (12dB).

Take as an example that you're putting what ought to be two signals into your spectrum analyzer; one at 100MHz and one at 101MHz. In the output, you're seeing signals at 1Mhz, 99MHz, 100MHz, 101MHz, 102MHz, 200MHz, 201MHz, and 202MHz.

Six of the signals you're seeing are spurious, and are the result of intermodulation distortion or plain old harmonic generation (the 200MHz and 202MHz signals). The products at 99MHz and 102MHz are the result of 3rd-order intermod; the products at 1MHz and 201MHz are the result of 2nd-order intermod (i.e., plain old mixing), and the products at 200 and 202MHz are simple 2nd-harmonics of the input tones.

If these spurious signals are being generated in the external circuitry, then if you click in 6dB more attenuation, they'll all go down by 6dB.

If the signals at 1, 200, 201 and 202MHz all drop by 12dB, then they were from distortion inside the spectrum analyzer. If the signals at 99MHz and 102MHz drop by 18dB (3 * 6dB = 18dB), then they were also from distortion inside the spectrum analyzer.

Answer 2

I guess your signal is a sum of 2 or more sinusoidal signals. You use that signal as an input for some circuit which is non-linear and causes intermodulation.

The general rule: The intermodulation products generally become stronger as the signal power increases

But that depends on the mathematical form of the nonlinearity. There are some well known theoretical special cases which do not obey the general rule.

1. The circuit is linear as long as the signal amplitude is under certain limit, a small power increase causes no intermodulation, it stays zero

2. The circuit has no linear range, it works like a comparator with treshold at 0V and the input has no DC. The output is the same rectangular pulses at any input level.

3. The circuit has a dead zone around zero like cross-over distortion in badly biased class B audio output stage. There's very heavy intermodulation distortion if the signal is weak, but it doesn't increase if the input signal level is increased. At levels high enough but still so low that no clipping is caused the apparent distortion can be neglible as percents.

Unfortunately we know nothing of your circuit so no actual analysis is possible.

The intermodulation distortion is used as measure of quality for many circuits. Does the circuit fulfill the spec is inspected by inputting a certain test signal and by checking that certain intermodulation products are low enough.

Make a web search for intermodulation and intermodulation distortion measurements.

Answer 3

Following up on the 3 cases given by user287001, there is a 4rth case

During crossover distortion in class AB, there will be an output drive strength glitch, where either

• the pullup and pulldown devices add up to WEAKER than either device alone

or

• the pullup and pulldown devices add up to STRONGER than either device alone

Because of the brief dual_conduction time, we can view this sag/boost in conductivity as an IMPULSE, and impulses will correlate with any transfer function or time response, thus we proclaim "impulses have very broad spectra".

If the electronic system is characterized with SINUSOIDAL stimulus, the error (either a sag or a boost in output drive) becomes a periodic non-linearity, and our modeling tools (and measurement tools) proclaim "some harmonic distortion exists".

If the electronic system is stimulated with multiple tones that are integer-related frequencies, the error (either sag or boost in output drive) shows up at various predictable correlations.

If the electronic system is stimulated with music, our ears "hear a veil over the music" because we have expectations about how music instruments should sound, and the error (either missing or extra energy) is irritating to the brain.

As you start to examine circuits, learn about circuits, think through how the circuits function and manipulate electrons, I recommend highly the book on audio amplifier design by Doug Self.

In one of the later chapters, he repeatedly IMPROVES a complete amplifier, altering the schematic as appropriate, and spends pages explaining what has improved --- a wonderful way to learn.