I have written online and in many many emails that a very common failure of VNA measurements of components is the test fixture, and the standout problem is most often the length of connecting wires.
This article works a couple of theoretical designs based on a validated model and experience of building and measuring many baluns of similar or identical design. We will then look at extracts from a Youtube video by ferrite manufacturer Fair-rite and appraise the results.
Validated theoretical choke designs
FT240-43
It is possible to calculate a pretty good estimate of the impedance of a common mode choke wound on a #43 material ferrite core over 1-30MHz. Measurement of a real choke suggests an equivalent shunt capacitance to calibrate the model to measurement. Whilst I have given the generic name to this core, it is based on Fair-rite’s 5943003801 and Fair-rite’s published 2020 #43 mix characteristics. There are imposters, and they may be significantly different.
Let us take a practical example design and calculate the expected choke impedance and from that, the expected |s21|dB in a VNA series through measurement setup.
Above is a SimNEC model of a FT240-43 with 11t winding and 2.5pF equivalent shunt capacitance to calibrate the self resonant frequency. The model calculates and plots choke impedance, and |s21|dB in the series through measurement configuration shown.
Above is a closeup of the graph. Choke impedance is shown as Zcm.R (resistance) and Zcm.I (reactance). This is pretty typical of a good HF common mode choke.
Also shown is the calculated |s21|dB, note the shape of the curve, it dips to a minimum at the frequency where the impedance peaks, the choke’s self parallel resonance at 9.3MHz.
These three curves are typical of a good HF common mode choke, and the design based on #43 is sound and quite popular.
FT240-31
It is possible to calculate a pretty good estimate of the impedance of a common mode choke wound on a #31 material ferrite core over 1-30MHz. Measurement of a real choke suggests an equivalent shunt capacitance to calibrate the model to measurement.
Let us take a practical example design and calculate the expected choke impedance and from that, the expected |s21|dB in a VNA series through measurement setup.
Above is a SimNEC model of a FT240-31 with 11t winding and 3pF equivalent shunt capacitance to calibrate the self resonant frequency. The model calculates and plots choke impedance, and |s21|dB in the series through measurement configuration shown.
Above is a closeup of the graph. Choke impedance is shown as Zcm.R (resistance) and Zcm.I (reactance). This is pretty typical of a good HF common mode choke.
Also shown is the calculated |s21|dB, note the shape of the curve, it dips to a minimum at the frequency where the impedance peaks, the choke’s self parallel resonance at 9.3MHz.
These three curves are typical of a good HF common mode choke. This design based on #31 is popular for reasons that are not always sound.
Fair-rite’s measurement
Fair-rite published a Youtube video showing how to measure a HF common mode choke.
Above is the DUT, it appears to be 11t on a FT240-31 core.
No information is given on instrument calibration, and the location of the reference plane… it is probably at the instrument jacks.
Note the connections, try to work out where the inner and outer conductors are connected and the lengths of them. Though shielded cable is used (coax) the shielding appears to be quite ineffective.
Above is the measured result. Compare it to the theoretical cases above.
The measurement relies solely upon the |s21|dB metric, and it is not appropriate to baluns in antenna systems. The metric is often called CMRR, quite a misnomer in this application.
If they could be relied upon, the results do not indicate a good common mode choke.
The measurements are essentially worthless due to the length and type of connecting wires.
Conclusions
You could not use this measurement setup to make a meaningful comparison of the #43 and #31 based designs.
If you shorten the connections and it changes the results, the connections were too long, and they may still be too long… do it iteratively until the results are not significantly effected by the connection lengths.
This is not to criticise Fair-rite alone, there are a huge numbers of videos like this on Youtube, probably hundreds, possibly thousands, and they are very popular. If you believe that popularity determines fact, you might accept that nonsense.