How to measure input impedance of matching network (nanovna)

[spacehen]

Hi all,

I would like to measure the input impedance of a matching network I have designed using the nanovna. There is some confusion with regards to the calibration technique required for an accurate measurement.
The input impedance as seen through port 1 is around 5-5j ohms which is why I am using the shunt-through impedance measurement method (S21 impedance).
The DUT consists of a 15cm PCB, with the matching network positioned in the middle and two SMA connectors at either end.

Now as I understand it I can't simply perform SOLT calibration on the cables and make a direct measurement because this does not take into account test fixture (DUT) effects, i.e there should be some phase error.

So my question is how do I setup port extensions in this case? Do I create additional PCBs with the matching network removed (open) and shorted and then add electrical delay until I see a single point on the smith chart?

EDIT: Frequency of interest is 500MHz

[antenna]

I am probably going to make a fool of myself here, but I'll take a guess that the nanoVNA only uses the OSL for S11 calibration and isolation/through for S21, and that the calibration will normalize the S21 logmag to 0dB and phase to 0° if you had a through fixture of the same design/geometry as the matching network board.  If I am wrong, at least this topic is getting a bump lol.

Not sure what nano you have, but mine will not show the S21 shunt impedance values directly.  The best mine can do is show the polar values or the logmag in dB and the phase angle.  So, what I ended up doing while trying this is setting mine to display S21 polar values and punch that complex value into a formula for the shunt through device impedance.  But I think in your case, you would do that math backwards and calculate the expected polar values from the impedance you need  to match so you can watch for that while adjusting.  For example, I am fairly sure that 5-5i would translate to a S21 polar plot at 0.19-0.14i on my nanoVNA.

Edit:  Table 1 of the article linked below shows the formulas for converting between polar s-parameters and the device impedance.  The only issue is that where you see Zdut, Zref and S11 or S21, those are all complex numbers and doing that math without a calculator that does complex math (like a TI-89, a phone app like HyperCalc, or even MS Excel), it gets a bit messy. 
https://www.edn.com/accurately-measure-ceramic-capacitors-by-extending-vna-range/

Edit 2: I have read that shunt-through measurements are very susceptible to common-mode currents.   That means sheilding or baluns if you want to get serious.  My nano does fine with S11 until it gets below about 2Ω, so depending on your situation and your nano, it may be easier to trust the S11 reading.

[LM21]

Can you short the track just at matching network. At least that would help to find out what time delay would be.   What I understand, calibrating only fixes reactances, If your matching network gives 50R+-j0R, that is still pure resistive 50 ohms at the connectors.

[radiolistener]

Just load your matching network output with a proper dummy load and measure matching network input S11 with NanoVNA, that's it

[spacehen]

Just load your matching network output with a proper dummy load and measure matching network input S11 with NanoVNA, that's it

Hi radiolistener, I just have one question then. Is the s11 reflection method, in general, accurate down to 2-5 ohms for both R and j?
What is the minimum useful impedance that you can measure with this method?

I've seen many different figures quoted, i.e 0.5 ohms, 1 ohm, etc

[radiolistener]

Hi radiolistener, I just have one question then. Is the s11 reflection method, in general, accurate down to 2-5 ohms for both R and j?
What is the minimum useful impedance that you can measure with this method?

Measurement uncertainty depends on how close impedance to the port base impedance (50 Ω for NanoVNA). If it's close to 50 Ω then uncertainty is low, if it's too low or too high uncertainty will be high.

I think for 2 Ω you can get acceptable results with NanoVNA, but for lower impedance the result may be worse.