The idea is to create correction data for the cheap SMA cal kit by measuring the cal standards after the VNA was calibrated with the Maury kit, and thereby transfer at least some of its precision to the cheap kit. As with most cal kits, the use of correction data is absolutely necessary since the open and short have an offset length, and in this case probably also appreciable losses due to the PTFE dielectric. Moreover, the open is always appreciably non-ideal due to its fringing capacity.
Let's start with the measurements. First, the VNA was configured with adapters to have a male 3.5 mm connector at port 1, and a female 3.5 mm connector at port 2. Then a full one-port calibration was performed for each port, using my Maury cal kit. After some sanity checks of the calibration, the S11 of the six SMA cal standards were measured from 300 kHz to 4 GHz. The results were stored as Touchstone (.s1p) files, each with 1001 data points. Here are plots of the results, directly from the analyzer:
The open and short standards are somewhat lossy, but this is easily addressed by the common error correction models. The match standards are are a bigger problem;surely they are not shoddy and are certainly adequate as a general purpose termination, but are far from what one would expect from a calibration standard. The matches usually will set the limit on the accuracy of a cal kit even after correction with the coefficients method, since open and short are usually very precise after correction, but there is no general parametric model to adequately correct a match. It is interesting to note that between 1 GHz and 2 GHz the return loss of the female match worsens by about 20 dB and remains below only −35 dB, while the male match is much better and remains well below −45 dB almost up to 4 GHz. This despite the fact that the female standards are rated to a much higher frequency!
Another quality feature of a cal kit is the relative phase response of open and short over frequency. The error correction algorithm applied by the VNA is most stable (i.e., the error terms determined during calibration are least sensitive to variations in the measured reflection factors of the cal standards) if the responses of the open, short and match are spaced as far apart as possible on the Smith chart at every frequency in the band of interest. This is the case if, on the Smith chart, open, short and match are located on a straight line at every frequency, with the match at the center.
In the extreme case, if open and short happen to rotate at very different rates around the outer edge of the Smith chart as frequency is increased, they will finally meet. At that frequency, they look alike (except perhaps for small differences in loss), and calibration fails. Thus the phase response of open and short should ideally be 180° apart. In the following two graphs the difference in the phase response of open and short (red trace) for the female kit as well as for the male kit is plotted over frequency. For the female kit, the phase difference remains below 17°, and for the male kit below 6° in magnitude up to 4 GHz. This is actually a very good result.