Measurement / evaluation of an RF filter response with NanoVNA gave an example measurement of a low pass filter.
Above is a plot of |s11| and |s21| from the article.
This article remeasures the filter with and without Enhanced Response Correction (ERC) in the VNA.
Let’s talk about the VNA for a moment…
The measurements were made with NanoVNA-H4 v4.3 and Dislord’s NanoVNA-D firmware v1.2.30.
At An experiment with NanoVNA and series through impedance measurement and An experiment with NanoVNA and series through impedance measurement… more I discussed inaccuracy of the combination for that type of measurement.
The problem raised was demonstrated with series though impedance measurement of a known 200Ω resistor, and the error was quite significant. Similar error might be expected for series though impedance measurement of a common mode choke.
The problem is due to uncorrected source mismatch error, and will have most effect when the DUT presents a significant mismatch to Port 1. The series though impedance measurement and shunt though impedance measurement will both be susceptible, as will DUT such as filters in the reject regions where InsertionVSWR is high, but less so in the pass region where InsertionVSWR is low.
An issue was raised as IMHO it was not a very difficult thing to implement (I had prototyped it in NanoVNA-App). Within days Dislord had coded a solution which implemented ERC, and it works very well, it is in NanoVNA-D firmware v1.2.32.
Impact of source mismatch error
The series though impedance measurement and shunt though impedance measurement will both be susceptible, as will DUT such as filters in the reject regions where InsertionVSWR is high, but less so in the pass region where InsertionVSWR is low, or any other DUT that presents high InsertionVSWR to Port 1.
200Ω
A s21 series through measurement of the 200Ω SM 1% resistor was conducted and results were very close to the measured value of the resistor.
A subtle point lost to those who rate series through impedance measurement over s11 reflection measurement is that ERC depends on corrected measured s11.
Low pass
This repeats the measurement of the filter at Measurement / evaluation of an RF filter response with NanoVNA, freshly recalibrated and measured (the filter is susceptible to temperature change).
Above is a comparison of measurements (corrected in the NanoVNA) with and without ERC, green is with and magenta is without. At 14.09MHz, the ERC trace shows approximately 1dB lower |s21|, ie 1dB more rejection, 80% of the power output.
Load mismatch error
Load mismatch error is at this time uncorrected, so it remains important that the load is very close to 50+j0Ω for good s21 accuracy.
Some of the NanoVNAs have quite high ReturnLoss on Port 2, some are mediocre and warrant a good external attenuator to improve it.
A common way in which people compromise Port 2 ReturnLoss is the measurement fixture… measure it through the fixture to check.
Conclusions
Whilst ERC might seem insignificant in some scenarios, in others it might be quite significant. The series through measurement techniques warrant ERC.