antenna-theory.com

[Sekutma]

I know the answer to some of my questions will depend on the analyzer but I'm more interested in knowing if anybody has done this. So I was using the average feature (takes the last 16 measurements and averages) during some measurements and I had to re-cal for another range but didn't turn it off. Turns out you can calibrate using the average feature and my resulting signal looked significantly cleaner (even compared to doing a normal cal and then turning averaging on).

1) Is this a valid way to calibrate my analyzer and trust the results or could this be screwing anything up?
2) Has anybody else done this and what where your results?
3) Anything else I should know about averaging in general? Is it a bad idea to do this regularly? I feel like seeing the signal go up and down can be a sign of instability but with averaging you don't see that which is why I would like some more information by somebody who knows more.

[helix]

Averaging ONLY reduces the effect of noise (random error) on a measurement, or a cal, and has a similar effect as reducing the IFBW. Calibrations normally do not have SNR issues (what kind of cal are you doing?), so averaging generally does nothing except slow the cal process down. However it should do no harm to the end result data. There are a few special cases where you may run into SNR issues with a cal (e.g., if you include port isolation). But if your cal or measurement traces are "moving around" then there is likely something else wrong with your setup (e.g., bad cable, connector, or cal standard), and averaging isn't really removing the error (which is systematic, but time-varying).

Agilent/Keysight had many good VNA app notes. Look for one that is called "Ten hints for better VNA measurements" it may be helpful. R&S has a really nice text book on VNAs and their use as well.

[Sekutma]

It's just a standard OSL cal using the physical calibration kit.

What you're saying makes sense. I'm basically tricking myself into thinking it's "better" just because it looks smoother.

What about switching to an eCal? Any advantage to using that? My machine is badly in need of a hard calibration however I can't exactly afford that.

[helix]

eCal generally produces less accurate calibrations than mechanical standards that are within spec (ie, not damaged).

I suggest looking into ways to verify your calibration (I'm sure there are many). Given the calibration coefficients for your mechanical open standard, you should be able to calculate the S11 that the VNA should report with the open standard connected AFTER you cal. If the measured differs from the calculated much then you have a problem. This would be a fairly rigorous test. Also checking the RL of your load standard., which should be more than 20-30dB. Simply connecting the two test cables together should produce zero phase/delay and 0dB IL. Other things I can think of are to sweep a known attenuator or filter.

If there's something wrong with the VNA then calibration (systematic measurement error correction) of the kind we've been discussing probably isn't going to fix your problem.