The point I'm trying to make is that the -112deg phase measurement doesn't
really change if you could magically alter the Q of the 470pF cap across a Q
of 300 to 3000 as it was being measured on the VNA.
Well, it does change! I just did the maths. But I have to concede that the phase change seems to be less important than the amplitude change.
The nanoVNA just has to be
able to measure something like -112deg at 10MHz to let you know that it is a
470pF capacitor at 10MHz.
No, it also needs the amplitude! When the amplitude changes while the phase stays constant, the part has a different capacitance and of course also a different resistance.
We really always need to consider the vectorial value - that's why the thing is called a VECTOR network analyzer, after all! I was horrified a while ago when I looked for online reflection coefficient calculators, and found several that consider impedance to be a scalar value!!!! Fortunately there are also some that correctly take impedance as a vector, and allow to enter both dimensions of it.
And then of course the NanoVNA needs the calibration values, to remove the line length effect and other minor things. After all, a bare bones pure capacitor causes a -90¡ã reflection, but the line length can turn this into anything, such as your -112¡ã.
What the NAnoVNA does, for S11 based measurements, is placing the part under test plus some line length in a 50? resistance bridge, feeding a test signal into it, running two direct conversion receivers, one for the driving signal and one for the bridge imbalance signal, digitize the two audio outputs from the receivers, and then process that digital data. It can basically measure three things: Reference amplitude, imbalance amplitude, and the phase difference between the two signals. Everything has to be derived from this. I trust that the maths used in the software are correct, but the results are bound to the linearity, stability and noise of those circuits. When we connect a capacitor, the imbalance of the bridge is very large, so both the reference and the imbalance signals are large, and should be easy to measure. But the phase difference, which contains not just the noise of the mixers and ADC but also the phase noise of the oscillators, might be harder to measure accurately. That's what I meant in my post Whether or not I'm right, is of course a different matter... I cannot be the judge on that.
Anyway, regardless of whether the phase or the amplitude is more critical in any given case, I think we agree on the fact that the NanoVNA's accuracy and/or resolution in measuring them is insufficient for directly measuring the Q of high quality capacitors. Which is a pity, but fully expected, and shared with many much more expensive instruments. When I bought the NanoVNA, I expected far less measuring range than it actually has, so I'm still happy with it, even if it cannot plot the Q of a good capacitor over frequency, with reasonable accuracy!
Manfred
