One of the popular ideas online is that the correction process in the NanoVNA does not correct errors in mismatch at Port 1 and Port 2. This article deals with the first case ONLY, Port 1 mismatch.
An experiment with source VSWR nominally 2:1
A NanoVNA was configured with a SMA tee connected to Port 1 and a good 50Ω termination connected to the branch port, see the pic below. The left hand side of the tee becomes the new Port 1 interface, and by virtue of the additional 50Ω shunt termination, if the native Port 1 was indeed well represented by a Thevenin equivalent circuit with Zs close to 50+j0, the Thevenin source impedance is now closer to half that, Zs close to 25+j0.
Some would calculate this mismatch as causing a mismatch loss of 0.512dB that is additional loss in the s21 path.
Above is the test setup. The NanoVNA was SOLT calibrated with cal parts attached to the left hand side of the tee and the 200mm coax jumper from that point to Port 2.
On completion of the SOLT, the s11 traces for SOL were verified, as was the S21 (0dB) trace with the coax jumper.
A good 10dB attenuator was then connected inline (at the Port 2 connector) and s21 measured.
Above is a screenshot from the NanoVNA, |s21| is -9.97dB, quite close to the expected -10dB.
Above is a capture from NanoVNA-App of the same scenario, again using the stored VNA calibration. The departure is tiny, and does not hint for a moment a source mismatch loss of 0.5dB that was not corrected in the calibration process.
Note that properly calibrated, the NanoVNA will capture the departure of a sample of nominal 50Ω transmission line from exactly 50+j0Ω, and that will show up in |S21| measurements even if Port 2 was exactly 50+j0Ω.
Readers might not be too surprised to find that there is at least one reflection type antenna analyser that has quite a low source impedance, source matching is not an issue with the forward and reflected waves are measured and calculations base on both.
None of this is to suggest that mismatch at Port 2 is corrected in the NanoVNA’s own calibration process. That remains a source of error in measurements that depend on Port 2 match, eg the very popular s21 series through impedance measurement technique.