Hmmm, why is everyone suddenly assuming that I want to have a detailed look at the connectors?
That would make very little sense given the fact, that BNC connections are pretty crappy in the world of VNAs. I am just trying to improve my measurement skills (which works well due to the tons of useful information everyone is generously dropping in this thread) and to optimize my measurement setup around the cables to be reasonably good for getting as much as possible out of the limited capabilities of the liteVNA.
At the current point I am able to take fairly reproducible TDR measurements of my cables. The results, do not allow me to derive any reliable conclusions about the absolute cable impedance yet, partially, because I do not have sufficiently characterized calibrations standards. However, I can detect significant deviations from the target of 50Ω and I can see which cables have large local or global impedance variations. In addition to that I discovered that I can even detect quality issues of my BNC crimp connectors resulting from crimping quality variations by comparing the respective TDR peaks, which is also nice.
On the non-TDR side, I can now measure the quality of my BNC terminations and do repeatable return loss and insertion loss measurements. At the moment, return loss and insertion loss measurements for my cables do not line up with what I would expect yet and I still have to determine if that is due to measurement errors on my side, insufficient equipment or simply my DUTs not being comparable with the reference measurements I am comparing them with.
With respect to getting the absolute cable impedance values as good as possible I am currently investigating the software side of things, since I have observed large differences between software solutions that transform my S parameter measurements into the TDR domain. I found that on the time/length scale of the TDR graphs, the Solver64 software seems to align very well with reality when using the correct velocity factors for my cables, one thing where the VNA Tools seem to be completely off. The impedance values on the other hand appear to be more consistent in the VNA Tools than in Solver64 which I would base on the previously described behavior when switching the filters. Since I am not sure what is causing the observed differences between the software solutions it is possible that one software has an issue, both have an issue or none of both have an issue and the issue is only hiding in my brain. Maybe I should try to implement a minimal version of the time domain conversion on my own to get an impression what is happening there and to have some code I can play with and get direct feedback what changes to the code have which effects.
Once I have sorted this out, I will try to optimize my calibration gear, by getting a better kit, or finding an opportunity to characterize one using well known standards.