[Radio Wrangler]

To measure third order intercept points you first need a pair of very good signal generators.

The issue is any signal from one making it to the other, backwards into its output connector will cause the intermod products you are looking for in the output amp of that signal generator... and this works both ways.

So you run the sig gen at a high level so you can have plenty of attenuation before the combiner so that there is plenty of loss from one sig gen to the other. In some cases I've added minicircuits medium power linear amplifier modules to the outputs of a pair of HP 8662s.

THEN you need a combiner which provides as deep a null as you can get in the source-to-source path. Such couplers tend to be transformer-based and you need to find good ones so that they don't create the intermod products you are after. THEN you want some more attenuation so that the balanced transformer bridge type combiner sees a good load impedance to help it null cross-coupling, and to bring the level down to what you want to apply to the equipment under test.

If you're looking at RF or IF of the front end, then you need a good spectrum analyser, something with about 100dB dynamic range.

If this sounds terribly expensive and professional, then it's what's needed to measure the sort of RF front ends that hobbyists build.

Stewart of Reading have a number of 8662s in stock and were asking £750 each when I bought a couple for work-work. The minicircuits amps, combiner and attenuators came to about the same in total. Spectrum analysers were in the £2500 region. This is what it takes to build a rig capable of measuring the TOI of a reasonable RF amp or mixer.

The same gear will measure compression and blocking etc as well, so you get those for free.

Noise figure is easier, you need a calibrated noise source, and a very linear receiver/detector.

These are non-trivial measurements that stretch test gear to the limit. Get it wrong, and you still get a result, and unless you are suspicious and take precautions, there is nothing to tell you that the result is badly wrong. The internet has a lot of people who've mislead theirselves and don't know it.

David