Roger, very interesting!
Some questions and comments...
First, what is the attenuator (how much attenuation, its input Z, and how is it terminated)?
The 8753 and the Kirby reference are very similar at low frequencies (although there is a small offset that is interesting), but then diverge as F increases. It¡¯s possible this divergence is due to the (assumed) 50 femtofarads of intrinsic capacitance in the SMA open. One way to test this theory is to use your external calibration routine. But rather than assume that the Open reference¡¯s ideal S11 matrix (I.e. its ¡°A¡± matrix) equals ¡°1+j0¡± for all entries, create a new A matrix by calculating gamma for each frequency assuming the 50fF cap.
In other words, create a new A ¡°ideal¡± matrix for the open by making the following calculation for each of the 101 frequencies:
A(i) = s11(i) = (1 - 50*(j*6.28*freq(i)*(50e-15))) / (1 + 50*(j*6.28*freq(i)*(50e-15)))
Where i is the index for a ¡°for¡± loop and Zo equals 50 ohms.
And then run the compensation routine with this new ¡°ideal¡± matrix. (I think this should work).
I¡¯m returning home tomorrow afternoon and the first thing I want to do is write a Matlab script to perform the compensation with both the Matlab equation I posted earlier and also with your method using the ideal A matrices that I can then manipulate to include a C0 term of 50 femtofarads.
As for the NanoVNA¡¯s 0.3 dB delta, that¡¯s weird. And maybe it explains the fairly course delta I first posted. Be nice to find the calibration code to see what is happening.
- Jeff
