Re: errors of "error" models

Gary O'Neil

Hello Yza;

I admire and appreciate your persistence, as well your desire to stay on topic. I suspect this discussion is far above most on this list... myself included. There may be a few whose attention you have not yet captured, but whose support and insight may prove worthwhile when they realize the extent and focus of your work.

Your posts have been difficult to follow, but not impossible, and I’ve tried to dig a bit into your questions to see if there are any I can answer.

Some questions first:

1) Are you proposing your DERDEI software (algorithms) with “uncharacterized” standards as an alternative to using error coefficient polynomials that rely on “characterized” standards?


2) Are you proposing your DERDEI software (algorithms) with “characterized” standards as an enhancement to existing calibration processes as a means to establish boundaries on residual post calibration uncertainties?


3) Are you proposing a tertiary algorithm to be used as a means of fairly comparing the hardware performance of two independent VNA platforms; such as when making like measurements of an identical device on two completely independent VNA’s but under otherwise exacting conditions?

3) Are you in search of Beta testers of your DERDEI program using NanoVNA raw data, corrected post calibration data using uncharacterized standards, or corrected post calibration data using traceable characterized standards?

4) In your example INPUT.TXT file, the fourth input (AUT) is assumed to be Antenna Under Test. Are these the measurements of the UHF ground plane antenna in the described anechoic chamber environment described?

5) Is the purpose of the INPUT.TXT file intended for our evaluation and comments, or as a test file to verify your software is correctly installed and functional?

6) What do you want/need from users on this list in our response to you on this software? Most on this list lack VNA experience, and most DO NOT have characterized SOL or SOLT standards.

To answer your early question as to what is used in the NanoVNA, my “guess” would confirm your suspicion of missing terms in the calibration routine. There are no provisions for the input of calibration error coefficients of a user’s unique set of standards. Therefore; calibration terms embedded in the firmware of the NanoVNA would at best have no meaning, and at worst introduce error uncertainties of their own. I do not however have any knowledge of how calibration is achieved, or uncertainties defined... other than to say they are declared to be zero as this is generally acceptable to most amateur enthusiasts.

There are however two sources you may want to review if you haven’t done so already.

The hardware and software is open source, and derived from the efforts of edy555 in Japan. The NanoVNA is feature rich, but the product is restricted to the processors on chip resources, and thus memory limited...likely insufficient to accommodate supporting more sophisticated error correction algorithms.

NanoVNA Saver is an Open Source GUI supporting a PC connected NanoVNA that has made provisions for incorporating calibration coefficients for the standards in use in the calibration routine employed by that program, presumably using the NanoVNA’s raw data. It is written in Python and should be straight forward to analyze. It’s author Rune Broberg would the source to question the algorithms he employs.

You commented: “Well, we don't mean to offend you but, in our humble opinion, the whole story of
hp "error" models is just a myth serving the powerful advertisement.” While not intended to sound offensive, this likely did raise a few hackles. :-) Your use of the term myth suggests you may be slightly mis-informed.

HP, R & S, and others have a competing interest in growing the RF industry. There is no question that marketing among them is highly competitive as well. To their market however; capability, repeatability, stability, and accuracy are unquenchable demands, and the key parameters used to differentiate one vendor from another. Mythical error models would yield failed products by their customer base, and as such their business wouldn’t thrive. That said; the quest for absolute accuracy will continue until everybody is certain, and in unanimous agreement, that absolute accuracy has been achieved. :-)

I wish I could help you more, but this is likely my best effort toward your cause. Perhaps it will seed further discussion of your proposal, or the comments of others; as it appears to have merit for consideration.

You stated your efforts are Open Source, and have provided good resource links to launch and evaluate your software. Is there an online forum corroborating your efforts?


Gary, N3GO

Join to automatically receive all group messages.