Vlad Dvorkin on a clever way to measure IP3 and non-linearity

So last Tuesday I called an RF aficionados meeting. I was sure to ask Earl McCune, author of Practical Digital Wireless Signals. I also asked James Long, an RF consultant and Vietnam vet. James told me he was prouder of his extra class Ham license than he was of being a professor at Cal Tech. Another go-to guy for RF is Linear Technology’s Vlad Dvorkin. Vlad is a hard-core systems engineer that helps LT customers design cell phone base stations and other RF hardware, like military and high-performance medical.

So the four of us are enjoying lunch, telling stories and discussing Spice, RF design tools, and 100 other things. Towards the end, we were talking about how the parts are getting so good it is getting hard to measure the distortion products and non-linearity of amplifiers and such. Vlad mentioned he had to use a Rohde & Schwarz spectrum analyzer for some to get that last bit of noise floor that high-dollar equipment can provide. Then he mentioned he had a neat trick for measuring distortion. You measure distortion by feeding your amp two tones that are nearly spaced in frequency. Any output frequency spurs are mixing products that non-linearity in the amplifier creates. Like I said, the parts are getting so good it is getting hard to measure the little spurs. One problem is that you don’t want to send the original two tones to the spectrum analyzer, it will saturate its front end and prevent you from seeing “down in the grass” as RF folks talk about signals just above the noise floor. So to take out the original two tones, Vlad whips out a napkin and sketches a test setup.

• Vlad Dvorkin nulls out the fundamental tones in the linearly evaluation setup.

I let the busboy pick up Vlad’s napkin, so I re-drew it here. He sends the two tones into a 180-degree phase splitter. He uses two identical amplifiers, and two identical attenuators. The trick is, on the amp he wants to measure, he hammers the input with big signals to get a full-scale output, and then attenuates it so he won’t burn out the spectrum analyzer. On the other phase, he puts the identical attenuator first. That amplifier is driven much less so it has hardly any distortion. After that amp, Vlad puts a phase tweaker and an adjustable attenuator, just so he can perfectly match the fundamental tones in the top path, only at 180 degrees out of phase. Then he adds the signals together in a 0-degree combiner and Voilà, the bottom path helps to minimize the two fundamental tones by 30 dB or so, and your spectrum analyzer is not over-driven. As a result you can really see the OIM3 (3rd order output inter-modulation distortion product)  -100 dBm down.

I assume other RF engineers do this, Vlad never said it was his invention, it has been widely used in feed-forward RF power amplifiers, only that it was a really good way to measure IP3 when you have really good parts to evaluate, like Linear Tech makes. I am trying to get Vlad to write up these RF tricks for EDN, I will keep nagging him, this one was pretty sweet.