Move a NanoVNA Reference Plane Before You Match

A NanoVNA calibration is at the connector you screwed the standards onto. The antenna pad is not that connector. On the synthetic $2.45\,\mathrm{GHz}$ demo antenna the pad is $10.64 + j5.56\,\Omega$. Fifty picoseconds of lossless $50\,\Omega$ line still in front of it, $15.0\,\mathrm{mm}$ at velocity $c$, is what the SMA sees: $24.6 + j54.8\,\Omega$. Match that number and you have matched the connector.

The Smith chart calculator can slide the plane along one uniform line before it solves. Positive length moves toward the load and removes a feed. Negative length adds line. The same $50\,\mathrm{ps}$, entered positive, with $Z_0 = 50\,\Omega$ and $\varepsilon_{\mathrm{eff}} = 1$, puts $24.6 + j54.8\,\Omega$ back on $10.64 + j5.56\,\Omega$.

Port extension and electrical delay

Port extension is that positive length. You are telling the chart the calibration plane is too far from the pad, and the feed in between is one line.

Electrical delay is the same slide when you know a time instead of a ruler length. Type it in picoseconds or degrees. Both are the one-way delay at the design frequency. The reflection coefficient rotates by twice that, because the wave goes to the pad and comes back. Fifty picoseconds one-way is about $100\,\mathrm{ps}$ of reflection delay. Enter $50$, not $100$.

If the VNA manual’s electrical delay is already applied in the instrument, leave the length at zero and export the file. Applying it twice moves the plane past the pad.

Millimetres and mils do not depend on the design frequency. Degrees and picoseconds are converted once, at $f_0$, and that physical line is then used at the other sweep frequencies. A loss in $\mathrm{dB/in}$ is optional. De-embedding a lossy line increases $|\Gamma|$. Adding lossy line decreases it.

Where the line numbers come from

$Z_0$ defaults to $50\,\Omega$, which is the usual NanoVNA cable. A microstrip feed is not that cable. Open Trace Studio, then use “Use the open Trace Studio line” on the match form. That copies $Z_0$, $\varepsilon_{\mathrm{eff}}$, and loss at the match frequency. A fab preset does the same for a line locked to $50\,\Omega$. If the open Trace Studio design is a differential pair, the copied impedance is $Z_{\mathrm{diff}}$ and the label says so. Edit it if the measurement was single-ended.

This is one line of constant $Z_0$. It does not remove a pad, a via, a width step, or two different dielectrics. Those still belong in the measurement, or they stay inside the answer.

The gray Smith trace is the calibration plane. The dark trace is after the shift, and that is the load the match uses. Share links keep the shift. An old link with no shift keys still opens on the file plane.

The network .s2p, the matched .s1p, and the BOM download without an account. The .s1p is $S_{11}$ of the network on the shifted load.

References

• D. M. Pozar, Microwave Engineering, the input impedance of a terminated line.
• P. H. Smith, “Transmission Line Calculator,” Electronics, vol. 12, January 1939. Length rotates Γ.
• The $10.64 + j5.56\,\Omega$ point is the synthetic demo in IndieRF Match, not a vendor antenna and not a NanoVNA capture.

Related

• Match a 2.4 GHz chip antenna to 50 Ω
• How to read a Smith chart
• Smith chart calculator

FAQ

Why is a NanoVNA calibration at the SMA and not at the antenna?

The standards screw onto the connector, so the reference plane is that SMA. The antenna pad is further along the board. On the synthetic 2.45 GHz antenna, 50 ps of lossless 50 Ω line (15.0 mm at velocity c) moves 10.64 + j5.56 Ω at the pad to 24.6 + j54.8 Ω at the connector.

What length do I type for port extension?

The one-way delay, with a positive sign, to remove the feed. Positive 50 ps, Z0 50 Ω, and εeff 1 puts the connector impedance back on the pad. The reflection phase is twice that delay, about 100 ps round trip. Type 50, not 100. If the VNA already applied the extension, leave the length at 0.

When should I use electrical delay instead of a measured length?

When the instrument quotes a one-way electrical delay and you have not applied it yet. Degrees and picoseconds in Match convert to a physical length at the design frequency, then that line is used across the sweep. A length in millimetres or mils does not depend on the design frequency.

What does this shift not remove?

A pad, a via, a change of width, or two dielectrics in series. It is one uniform line. Copy Z0, εeff, and loss from the open Trace Studio line, or from a fab preset. A differential pair copies Zdiff and says so.