开云体育

I guess this is why R&S introduced the T-check many years ago as a simple approach to establish quality of measurement

In my own simple words, L calibration establishes the center of the smith chart, O and S define the boundaries so using a fourth known load such as 25 ohm (being substantially different from the 3 calibration loads and easy to create using two calibration loads and a Tee) shows the accuracy in one direction This for sure does not eliminate the amplification of errors when moving away from Z0 but at least gives a understandable indication of quality.

For the nanoVNA this has proven very valuable as the calibration was perfect till 900MHz but the T-check (or a measurement of 25ohm (almost) pure resistance) showed considerable opportunity for improvement.

As the T-check can be done using any impedance it allows any user to establish a sense of measurement accuracy at the used impedance.

It is my strong wish to translate thave this lengthy thread translated into something practical a average nanoVNA user is able to understand. For that reason I was asking about the choice of the 4th load to ensure the % error in the cross ratio has a real meaning. "substantially independent" is not usable for the average user. Is 25 ohm substantially different? A pure capacitor but at what reactance?. It would be of benefit to the community (that has to endure the many mails on this subject) to have simple guidelines that can be applied easily with relevant meaning
--
NanoVNA Wiki: /g/nanovna-users/wiki/home
NanoVNA Files: /g/nanovna-users/files
Erik, PD0EK

On Sat, Jan 4, 2020 at 06:08 AM, Gary O'Neil wrote:

I believe the reference to "absurdness of identicality" was with respect to
the notion that different load values... which, when calibrated, produce
identical results (-1, 0, 1); and those results are expected to yield
perfectly corrected and accurate measurements.

Gary, are you saying that different "characterized" SOL standards, when measured after calibration, produce identical results. (-1,0,1)? If so, this is not true -- they produce the results equivalent to their modeled impedance (a model represented by a cal kit's calibration coefficients).

And therefore, the more accurately one can model this impedance (for example, using a 3rd-order polynomial to describe the fringe capacitance of an "Open" load) , the more accurate will be the measurements.

Are you saying this is not true?

...Through this process, the common user becomes satisfied with the
accuracy of the measurement, unaware of the in inaccuracy introduced by their
attempt at improving it.

Not sure what you are saying, but if you increase the accuracy of the models of the standards you increase the accuracy of the measurement.

I am also puzzled by this quote in your earlier posting...

the also very revealing and important consequence of polynomial error
correction attempts; and my observations... FACUPOV... of their uselessness,
and the propagation of their contributions to uncertainty in the otherwise
straight forward definition of the uncertainty boundary profile.

Which "polynomial error correction attempts" are you referring to as being useless? The fringe capacitance polynomial? Something else?

Thanks for any clarification.

- Jeff, k6jca

I had what seems to be the same problem, but with the last version, updated Windows Vista Professional. The error message was "could not load Qt platform plugin "windows" in "" even though it was found."

I was unable to find a solution, except for the recommendation you also received - use Linux. To which I had the same objections as you have.

Good luck! I really hope there's a solution.

Doug

I find it interesting that some very educated folks like Dr Stebers are quick to critique things but you rarely see them helping with the creation of improved documentation or helpng others on the forum for something like the Nanovna.?

At least there are other very knowledgeable experts on this forum from around the globe who are willing to contribute to the increasing sum of knowledge on this forum with value added tips, techniques, documentation and personalized help in many instances and all the participants are to be heartily commended.?

Cheers to all for 2020!
Looking forward to the new V2 forum (?) Maybe coming soon ?


On Sat, 4 Jan 2020 at 9:49 AM, Nick Kennedy<kennnick@...> wrote: His overall impression seems pretty favorable - like many of us amazed that
such performance could be available at such a price.

He found impedance measurements to be very good in the range of 12 to 200
ohms and acceptable from 5 to 500 ohms.

He cautions potential buyers that clones appearing on the market might be
of varying quality, and that available manuals might be poorly translated
from the original Japanese or Chinese.

Otherwise, " ... if you like to play with hardware and software you may
find this little gem an wonderful toy for experimentation."

73-

Nick, WA5BDU

Search under the concept of de-embedding and on the vna you will use the function electrical delay.
73' Alan

Pentium 4 machines (originally with Win XP) run very nicely with any of the popular Linux distros (like Linux Mint 19.3) ,obviously 32 bit .
With a ( nowadays low cost) SSD the system is fast . Go for it !

I know that. As a matter of facts, on this PC I have a dual boot with Debian Buster. And the main HD is an SSD (I have 4 additional disks on this tower).
Yes, I could boot with Debian and run nanovna-saver under it, but all of my ham software (also that written by me) is specific for Windows....
So I would like to be able to run that utility when booted with Windows.
But the reports from others that say that they are able to run it under Windows make me think that maybe there is something to be adjusted on my installation. It is Windows XP Professional, SP3, but now I have the suspect that some Python library must be missing... the programming language of the utility seems to be Python... I have to investigate a bit deeper.

Thanks to all the suggestions.

Alberto

His overall impression seems pretty favorable - like many of us amazed that
such performance could be available at such a price.

He found impedance measurements to be very good in the range of 12 to 200
ohms and acceptable from 5 to 500 ohms.

He cautions potential buyers that clones appearing on the market might be
of varying quality, and that available manuals might be poorly translated
from the original Japanese or Chinese.

Otherwise, " ... if you like to play with hardware and software you may
find this little gem an wonderful toy for experimentation."

73-

Nick, WA5BDU

Thanks , it worked out very well.
Given a 900MHz max frequency I was able to use a 2 inch stripline, find the frequency for its 1/4 wavelength and then measure the reactance at 1/2 the 1/4 wavelength frequency.
It is all exactly the same as when measuring cable impedance thanks to the high upper frequency of the nanoVNA
I could not get unambiguous TDR output at this short length.

--
NanoVNA Wiki: /g/nanovna-users/wiki/home
NanoVNA Files: /g/nanovna-users/files
Erik, PD0EK

Good morning Erik;

@ Erik...

I believe the reference to "absurdness of identicality" was with respect to the notion that different load values... which, when calibrated, produce identical results (-1, 0, 1); and those results are expected to yield perfectly corrected and accurate measurements.

There is some irony to this notion of absurdity as well. After all of the care, attention, and expense taken to ensure the ultimate degree of accuracy of the standards used, the benefits of that effort diminishes rapidly as the distance away from the center of the unit circle is increased. The common user is never made aware of either the absurdity nor the irony, because the biases applied to “accurately” define the standards which are themselves embedded into the calibration, are “accepted” and remembered by the AnyVNA to which those biases are applied with great confidence. That confidence is then rewarded by the verification of calibration success through a re-measurement of those same standards, and the user’s realization that all of the uncertainties embedded in the system (including those introduced by the correction polynomial) have been accounted for and removed from the final measurement. Through this process, the common user becomes satisfied with the accuracy of the measurement, unaware of the in inaccuracy introduced by their attempt at improving it.

Furthermore; the accuracy corrections introduced, while small and often subtle, grow significantly as the measurement is removed from Z0, while at the same time diminishes toward zero in it’s consequence to the magnitude of uncertainty in the measurement itself. If the corrections are not accounted for in the uncertainty model, the uncertainty boundaries become asymmetrical and themselves become an unavoidable uncertainty.

--
73

Gary, N3GO

Pentium 4 machines (originally with Win XP) run very nicely with any of the popular Linux distros (like Linux Mint 19.3) ,obviously 32 bit .
With a ( nowadays low cost) SSD the system is fast . Go for it !

Frank , EI7KS

GIN&PEZ

I failed to include in my summary...

@ Gary O’Neil - /g/nanovna-users/message/9128

the also very revealing and important consequence of polynomial error correction attempts; and my observations... FACUPOV... of their uselessness, and the propagation of their contributions to uncertainty in the otherwise straight forward definition of the uncertainty boundary profile.

@ GIN&PEZ - /g/nanovna-users/message/8505

As regards Erik’s comments...

@ Erik... - /g/nanovna-users/message/9131

I believe his... and my own... percentage world view align quite well; and this shared “view” ... i.e. common to Erik and myself... fits quite well within the uncertainty constraints of your own FACUPOV definition as regards the common user. :-)

I feel the exercise you have walked us through; not only reinforces our common user’s point of view, but it also substantially justifies it.

Thank you once again... and congratulations!

--
73

Gary, N3GO

On Sat, Jan 4, 2020 at 12:53 AM, gin&pez@arg wrote:

but which they have in fact different values from these (-1, 0, 1), simply
because it is

* C o m p l e t e l y * A b s u r d *

the belief that all these (S, L, O) standards in this World are completely
Identical
and equal to (-1, 0, 1), not to mention their unavoidable Uncertainty, that is
of values
the VNA * d e m a n d s * of them to have

@gin&pez

On this group we have people that live in the percentage world (me) and people that live in the "per millionth" world (such as Dr. David Kirby)
For those that live in the percentage world (e.g. they are happy if their measurements are correct within some percent) the "standard" S,O and L loads are easy to manufacture to be within these required tolerances and their position within the smith chart helps to establish a calibration that will also satisfy their need to (limited) accuracy.

So when making statements about "absurdness of identicality" or "substantially independent" it would help the scientific discussion if you provide a non-ambiguous qualification of the world in which you make this statements.
Identical within a percent or identical within a millionth (of even more)?

For people like me that live in the % world all nanoVNA calibration standards are "perfect" and I gladly use them to calibrate all my VNA's. You may call this "absurd" but I am perfectly happy with it.

--
NanoVNA Wiki: /g/nanovna-users/wiki/home
NanoVNA Files: /g/nanovna-users/files
Erik, PD0EK

#105: On the Absurdness of the Identicalness of All Standard Loads in this World...

@Gary O'Neil , N3GO - 4 January 2020

Dear Gary,

Thank you for your time you spend all these months to comment on our work;
thank you very much, indeed !

First, regarding your remark about "the definition of G which is constrained",
you are right, of course. All the involved passive loads S, L, O, G have by
construction values which are complex ratios, with modulus within the closed
interval [0, 1]. The same is expected to be true for their VNA measurements
s, l, o, g, respectively. But, once again, you are right : we had to emphasize
this very point.

Next, regarding your remark about the Smith Chart, allow us, please, to inform
you that we found very interesting the following fact :

- - - - - - - - - - - - - (c) gin&pez@arg (cc-by-4.0) 2019 : start - - - - - - - - - - -

the cross-ratio of the Standard Loads (-1, 0, 1), in the inverse form of the one
we already gave, that is as:

(G - S)*(L - O)
- - - - - - - - - - - =: CR' := 1/CR
(G - O)*(L - S)

results in (1+G)/(1-G), which is equal to (Z/Zo) indeed - or to (Y/Yo) from CR.

Well, after all that, we are forced to amplify even more our conclusions regarding
The 4th Load Trick, as follows:

It seems that any VNA - the NanoVNA anyway - implements by default only this very
one of all possible Moebius transformations, that is the one of Standard Loads
(-1, 0, 1). This finding simply means that The Common User, who connects a triplet
of loads, with values he likes to think - regardless of the variety of reasons for that
likeness - as the standard ones (-1, 0, 1)
-
but which they have in fact different values from these (-1, 0, 1), simply because it is

* C o m p l e t e l y * A b s u r d *

the belief that all these (S, L, O) standards in this World are completely Identical
and equal to (-1, 0, 1), not to mention their unavoidable Uncertainty, that is of values
the VNA * d e m a n d s * of them to have
-
plus one load G that is definitely unknown to him, he then always gets from his VNA
the measurement for that G as a * c o m p u t e d * result based on this very simple
calculation :

(g - s)*(l - o)
- - - - - - - - - =: CR'
(g - o)*(l - s)

(1 + G)
- - - - - = CR' =: (Z/Zo) immediately on a Smith Chart
(1 - G)

while, by applying the well-known property of ratios:

a/b = c/d => (a-b)/(a+b) = (c-d)/(c+d)

to the above ratios, The Common User gets back his most valuable G-mini
formula - from which all this adventure began some months before:

(g - l)*(o - s)
- - - - - - - - - - - - - - - - - - - - = G
(g - o)*(s - l) - (g - s)*(l - o)

That's all.

Just a Trick - as usual.

- - - - - - - - - - - - - finish : (c) gin&pez@arg (cc-by-4.0) 2019 - - - - - - - - - - -

Kind regards,

gin&pez@arg

:105#

On 1/3/20 11:37 AM, Alberto I2PHD wrote:

It cannot be upgraded, not even to Windows 7, as its Athlon CPU lacks some features Microsoft deems absolutely necessary for Windows 7 and 10.

???? It can be upgraded to Linux, under which NanoVNA-saver runs just fine.

Hello GIN&PEZ;

@ GIN&PEZ - /g/nanovna-users/message/9108

#103 : The Fourth Load *

“It seems that the ratio of these two cross-ratios, the one of the loads and the
other one of their corresponding measurements, is equal to the constant (1,0). “

I agree and find this statement to be true FACUPOV.

Viz.

AnyVNA is physically bound by the limits of its sensitivity (noise floor) and its maximum full scale signal magnitude, and those physical boundaries are calibrated, scaled, and transformed mathematically to precisely define a unit circle of magnitude G = infinity centered on a user defined load of any arbitrary value needed/desired/chosen by the user to be used as the reference Z0 (characteristic impedance) of the measurement system.

Without regard to accuracy, uncertainty, resolution, or absolute value of the chosen Z0, all VNA's can be considered equal on these points, and any measurements made of an identical and arbitrary DUT would be expected to map to precisely the same identical coordinates, when calibrated/characterized with the identical set of reference standards, in an identical test environment.

@ GIN&PEZ - /g/nanovna-users/message/9073

“So, for each triple of (S, L, O) values we think, believe, accept,
or whatever else we like to be consider as "true", this equation
produces a * d i f f e r e n t * value of G.

I do not think this statement fits the definition of G which is constrained to within the limits of 0 and 1 by definition (Gamma = E-reflected / E-forward).

FACUPOV, the statement would be precisely true for the mobius transformation of G to Z (or Y)... i.e. the Smith Chart.

“The Smith chart provides a graphical representation of Γ that permits the determination of quantities such as the VSWR or the terminating impedance of a device under test (DUT). It uses a bilinear Moebius transformation, projecting the complex impedance plane onto the complex Γ plane:. “ (1)


@ GIN&PEZ - [#96]:The Ultimate Criterion for Endless Reliable VNA Comparisons
28 December 2019 - /g/nanovna-users/message/8749

As a criterion for endless reliable VNA comparisons, I think the ratio of cross-ratios falls short of being the Ultimate test, and is probably better described as a fundamental criteria of AnyVNA.

Alternatively; this may be a potentially valid tool for comparison of a known (or assumed) healthy VNA to another of questionable performance. To that end, one would interpret cross- ratio results uncomfortably removed from (1, 0 ) as a confirmation of defective hardware; e.g. gross instability, signal saturation/compression, non- uniformity in frequency step size, incorrectly compensated system errors during calibration, etc.

The cross-ratio results do appear to confirm that:

1) The process referred to as calibration does in fact mitigate all but but the uncertainty of accuracy in the measurements.
2) Accuracy is defined by the known precision of the standards used for calibration, and precision of control over the position of the measurement reference plane.
3) Sources of unavoidable uncertainty in the measurements are exposed and revealed; resolution (number of bits used to quantize the measurement), noise floor (minimum number of bits toggled), and the linear range of the measurement; all of which are measured and remembered through calibration.
4) The Mobius transformation process compresses all values from 0 to infinity into the unit circle, and by so doing, amplifies the accuracy uncertainty exponentially as G approaches zero and infinity. In addition; the resolvable accuracy of the load used and defined as Z0 establishes/defines/bounds, the “Ultimate” accuracy achievable by any appropriately and correctly calibrated AnyVNA measurement system.

I have attempted to summarize the findings of this discussion thread to the best of my understanding, and only from this common user's point of view. It is important to point out that this summary is limited in scope. It does not include any consideration or influence of the DERDEI model (2) used to graphically reveal the upper and lower boundaries of uncertainty defined at the onset of this thread. This is not to suggest, much less conclude, that the boundaries of uncertainty are not important.

I do perceive the findings as evidence that the uncertainties are an independent artifact of the measurement, and not an artifact of the measurement itself nor of the process. In light of that, I also perceive that understanding of the DERDEI models as proposed by GIN&PEZ are very much worthy of scrutiny, validation, and understanding. FACUPOV and amateur radio users in general, they may well be of minimal value or concern. Product developers, and researchers however, would be well served to pursue it with great interest.

72

Gary, N3GO


Ref: (1)

(2) @ GIN&PEZ - /g/nanovna-users/message/2984


73

Gary, N3GO

I was wondering a similar thing, more generically how do you test an antenna with an UN-removable length of cable.... ?

--
Regards,
Chris

#104': On the Criterion of Cross-Ratios - Errata

Section 1 - In 3rd line from below, instead of the wrong :
"Thus there is by any chance possible", the correct :
"So, there is not by any chance possible"

Section 2 - In the 5th line from the bottom, instead of the wrong :
"we do not publish t", the correct : "we do not publish it"

Section 2 - In the 3rd line from the bottom, instead of the wrong:
"there is absolutely nothing to do in such actually existing 2-port
cases to eliminate such a problem", the correct :
"there is absolutely nothing that we could do in such actually existing
2-port cases, in order to eliminate such a problem"

Section 3 - In the 1st line from below, instead the wrong: "[3]-[5]"
the correct : "[3]-[5], in order to find if and what we could do"

gin&pez@arg

@Gary O'Neil , N3GO - 3 January 2020
/g/nanovna-users/message/9087

Dear Gary,

Yes, it is.

Happy New Year 2020 !

gin&pez@arg

#104: On the Criterion of Cross-Ratios

@Erik, PD0EK - 2 January 2020 - /g/nanovna-users/message/9074

Dear Erik,

Thank you, very much indeed, for your time spent on these questions.

1
Well, both cross-ratios are suffering errors in either side of their equation, the
one of measurements because of the unavoidably limited precision or accuracy
in their 4 readings, and the other of loads because of the existing uncertainties
in the values of 3 standards. However, since the linear S-parameter equations
do * N O T * formulate errors of either kind
-
* may we have your attention, please, to this most crucial point ? *
-
the same holds for * E V E R Y * other consequence of them, that is for this very
equation too, as well as of the totality of its consequences. Thus, we have to
confront any error problems within this available S-parameter frame. Thus there
is by any chance possible a "substantial "amplification" of uncertainty" due to
"an unfortunate choice of loads" - see, please, "The Fourth Load" trick [0].

2
Anyway, we search the whole thread for "sufficiently independent" and we found
nothing such that was said by us. Instead, we call "substantially independent" our
experimental results we selected to present "On our Comparison of our [NanoVNA]
with our [HP 8505A] Using the Ultimate Criterion of Cross-Ratios" [1], because we
have in hand other results using a loaded 2-port, which are not appropriate at all to
call its experimental results "substantially independent" from them
-
and although we found the required explanation for that fact, we do not publish t
until we complete our experimentation with another 2-port - see, please, our reply
to Gary O'Neil , N3GO [2]. However, we can assure you right now that there is
absolutely nothing to do in such actually existing 2-port cases to eliminate such
a problem.

3
Finally, although we are not sure at all what you mean by characterizing the SOL
loads as "traditional" ones, regarding your -very interesting, indeed- claim for an
"independence criterion", as well as for the related to it of "maximum independence",
"best comparison" facilitation, and "better set of 3 loads", it may be obvious from
all that said above that such a condition can not be set in anything else different
than the single requirement of their distinctness. However, we already look at the
direction imposed in our three articles "On the Results of Measuring Two-Ports
Using Only [NanoVNA] Channel [CH0] or a [LeastVNA]" [3]-[5].

Best regards,

gin&pez@arg

REFERENCES

[0] #103 : The Fourth Load
3 January 2020 - /g/nanovna-users/message/9108

[1] #97 : "On our Comparison of our [NanoVNA] with our [HP 8505A]
Using the Ultimate Criterion of Cross-Ratios"
29 December 2019 - /g/nanovna-users/message/8763

[2] @Gary O'Neil , N3GO
17 December 2019 - /g/nanovna-users/message/8293

[3] #86: On the Results of Measuring Two-Ports Using Only [NanoVNA] Channel [CH0]
16 December 2019 - /g/nanovna-users/message/8275

[4] #86': UPGRADE : On the Results of Measuring Two-Ports Using Only [NanoVNA] Channel
[CH0] or a [LeastVNA] - 16 December 2019 - /g/nanovna-users/message/8278

[5] #86": Two Notes On #86': UPGRADE : On the Results of Measuring Two-Ports Using Only
[NanoVNA] Channel [CH0] or a [LeastVNA]
19 December 2019 - /g/nanovna-users/message/8400

:104#

On Fri, Jan 3, 2020 at 02:56 PM, Nick Kennedy wrote:

" … I should have mentioned that the QEX article was probably pretty out-of-date by publication time …"
============================================================================
I was just wondering the same thing as QRP's most recent firmware has the frequency resolution down to 1 or 10 Hz, and using NanoVNA-saver's increased measurement point feature, crystal testing should be in-line with the VNWA 3 below 300 MHz. With 101 measurement points, characterizing a crystal in the NanoVNA standalone mode would not be my first choice.

I still haven't received my first 2020 QEX issue. Overall what was Dr, Stebers impression regarding the NanoVNA?

- Herb