Re: errors of "error" models
#109 : reply to a newcomer
| From: "******* ********" <***********@gmail.com>
| Sent: Thursday, September 03, 2020 08:43
| Subject: Re: S-Parameter Uncertainties in Network Analyzer Measurements
| no, I am not aware of this discussion. anyway thanks for share that link.
errors of "error" models : https://groups.io/g/nanovna-users/topic/34237712
| I would like to understand the One-Port Error Model and Calibration
| (One-Port, 3-Term Error Model).
| I`m using the attached file
[ Network Analyzer Error Models and Calibration Methods by Doug Rytting ]
| Do you know how it works? see page 11/43 (pdf).
| To: "******* ********" <***********@gmail.com>
| Sent: Thursday, September 03, 2020 15:51
Dear Mr. ********,
Yes, of course we are aware in our group of Doug's papers.
Well, it is a pity that you don't want to follow the suggested
discussion. Because, it was only after the intensive thinking
enforced by the demands of our participation in this very
discussion that our point of view, regarding the kind of these
measurements, was finally made absolutely clear.
Anyway, we could theoretically conclude from this discussion
the following essentials:
- (c) gin&pez@arg (cc-by-4.0) 2019 : start - - - - - - - - - -
(0) We can adopt a so-called by us "virtual"-real two-port
S-parameter "error" model, that is one consisting of a
Virtual-Measurement "Port": one that you can't see it as
really existing around, in order to connect a cable to it, and
the other, indeed a really existing one, just because you can
see it and connect a cable to it: the familiar (Under)Test Port.
(1) To arrive at this model you have * u n a v o i d a b l y *
begin with the real four-port consisting of the familiar Ports:
Input, Incident, Reflected, and (Under) Test - once again:
"real", because you can really see its four real ports and
connect cables to them.
(2) As usual, you have to write down the familiar four linear
S-parameter equations for this four-port and then -after some,
rather lengthy indeed, mathematical manipulation- to form
just one equation relating those familiar g and G ratios of
signal-samples of Reflected-to-Incident waves -g for the
Virtual-Measurement Port and G for the Real-(Under)Test Port.
(3) This equation involves 3 parameters and express the
g-Measurement in terms of G-(Under)Test, but since it is also
an invertible equation, you can also express the desired
G-(Under)Test, that is of an Unknown Load, in terms of
g-Measurement that is of the Known Network Analyzer
Readings you can see, also write down and/or collect,
as G = G(g).
(4) The involved 3 parameters, which are complicated
expressions of the four-port S-parameters mentioned above,
are those well-known (HP) "errors" - although unfortunately
enough: widely-non-understandable until now.
(5) Well, after all that said, not only here but especially in
the aforementioned discussion -currently with 292 messages-
it is also an unavoidable conclusion that for many years until now,
Doug and his colleagues at HP, also produced, under various
additional assumptions, more-or-less approximate equations
G ~= G(g) for the multi-ports of more than four ports they
- end : (c) gin&pez@arg (cc-by-4.0) 2019 - - - - - - - - - - -
That is all.
ARG IAOI NFI
Antennas Research Group-Informal Association of Individuals-No Finance Involved-Austria-EU