GIN & PEZ;
Up to the presentation of your [LeastVNA] formula: G=(g-l)(o-s)/[(g-o)(s-l)-(g-s)(l-o)], and provided in executable form as BBC BASIC source code, I interpret as an alternative calibration algorithm.
As currently being performed in the traditional manner, it requires 3 reference measurements (SOL), which are combined mathematically with the measurements of an unknown (DUT) in order to compute accurately the reflection coefficient of the DUT FACUPOV. It there a comparison to make that will differentiate and otherwise highlight the virtues of using your algorithm?
Your 36 character long formula is misleading and imprecise as presented in, A3 of #62 : On the Most Reasonable Questions with the Most Clear Answers.
It is a compact presentation of 36 typed characters, but the variables are neither defined nor identified as complex. This may be obvious to knowledgeable readers, and those who dissect your source listing.
The NanoVNA outputs its raw results represent the uncorrected measured reflection coefficient in cartesian form. Is this a condition of your BBC BASIC implementation, or is the reader to assume the measurements can be in either rectangular or polar form?
The output of [AnyVNA] is generally understood to be S-Parameters. S11, S22,...Snn, are in fact reflection coefficients. Each of the input variables in your formula are reflection coefficients of the form S11, S22,... Snn, etc.
"No errors" implies absolute accuracy. That's going to be a tough sell.
... better yet... You might be better served if you purge the long list at the end of A3 altogether... It discloses your passion, but the content is confusing and does nothing to support or defend your claims.
My questions and comments here need not interrupt the progress of your report, but are instead offered for your consideration whilst the pdf version of your final report is being composed.
--
73
Gary, N3GO
