Article

Reading a nanoVNA measurement into a Smith chart

A nanoVNA gives you a Touchstone file. QuickSmith reads it as the load, which means you can design a match against what your antenna actually does across the band, rather than against one impedance you typed in.

Get the file out of the analyser

Every VNA exports Touchstone, and for a one-port measurement you want a .s1p. On a nanoVNA it is Save, then the s1p option. On the PC software it is usually File, Export. What lands on the SD card looks like this:

The line beginning with # is the option line: frequency in MHz, S-parameters, magnitude and angle, referenced to 50 ohms. Everything on it is optional and the defaults are GHz, S, MA, 50 ohms, which catches people out when a file that looks fine reads back a thousand times off.

If your analyser wrote the file against 75 ohms and you work in 50, do not edit it by hand. QuickSmith renormalises on import.

Read it in

Data → Import Touchstone, pick the file, and S11 becomes the load. Not one impedance: the whole measured curve, interpolated between the points you measured, so the load changes as the frequency changes. The sweep range is set to the range of the file.

This dipole is not in good shape at the bottom of the band:

Match a band, not a point

This is where a measured load changes how you work. You cannot get VSWR 1.0000 across a band, and chasing it at one frequency will usually make another frequency worse. The goal is different: get the whole band inside a circle.

So set the VSWR circle overlay to 2, run Sweep → Generate, and look at where the swept trace sits relative to that circle. You are no longer aiming at a point, you are herding a curve.

For this antenna, two components do it: 32 nH in series to drag the curve towards the middle, then a 25 ohm shorted stub 95 degrees long to fold it back on itself. Folding is the trick with wideband matching, because the frequency dependence of the stub partly cancels the frequency dependence of the antenna.

From 3.9:1 down to 1.84:1 across a 2:1 frequency range, with two parts. In mismatch loss terms that is about 1.9 dB down to about 0.4 dB at the worst point in the band.

That link carries the measurement itself, which is worth pausing on. The share link holds the twelve measured points as well as the network, so what opens is not an idealised model of the antenna, it is your antenna.

Send the answer back out

Data → Export Touchstone writes the matched network back out as a .s1p, so whatever you use next reads it. If you have not run a sweep it runs one first, so the file describes the circuit rather than whatever happened to be on the chart.

Data → Export Sweep (CSV) gives you the same thing as a spreadsheet, one row per point, with impedance, VSWR and return loss worked out for you.

Things that go wrong

• The trace looks like noise. Check the option line. A file with no # line defaults to GHz, so a 145 MHz measurement is read as 145 GHz.
• The impedances look wrong but plausible. The file was probably taken against a different reference impedance. QuickSmith renormalises what it imports, but if you edited the numbers by hand you may have half-converted them.
• The match is beautiful and the antenna still does not work. A matching network makes the transmitter happy. It does not make a resistive loss into radiation. A dummy load is a perfect match too.
• Too few points. The load is interpolated between what you measured, so a coarse sweep through a sharp resonance will lie to you. Twelve points across an octave, as here, is enough for a broad dipole and nothing like enough for a high-Q loop.

There is a sample .s1p in the touchstone folder if you want to try the flow before you go and measure something. It is the same dipole used here, and the same measurement as Example 4.