开云体育

If I connect the NanoVNA-H to a length of coax (about 20m of RG-11/U in this case) and perform an SOL calibration through the coax, will the NanoVNA then show me the feed point characteristics of the antenna to which the load end of the coax is connected?

Are there any other special settings that must be made?

Yes, assuming the antenna is connected at exactly the same point as your open, short, load were connected. If there's an additional "pigtail" at the antenna or anything like that, that will have an effect.

I'd say that there is another requirement - range used for SOL calibration and range used for measuring should be identical! More precisely, calibration points and measuring points should be the same. This is less impactful if measuring range is relatively small (narrow) comparing to the "center frequency" (for example 14.000MHz to 14.300MHZ vs 3.5MHz to 29MHz), but still ...

With lengths approaching quarter wave length, things (impedance transformation) become very fast changing, and in my experience, interpolation between points where actual calibration was done done becomes much less accurate.

I would say NO.
Sorry to say, but have worked with VNA's for 30 years and as soon as you come near 1/8 wavelength the measurement errors are taking over.
You know, in the analyzer the matrices that are multiplied together, the error matrice will start to influence significantly. It is very difficult to say that it always will not work when exceeding 1/8 (because you are closing in to 1/4 and the matrice in play will get 0's and infinite in some places) because let's consider you are clever and you measure in a narrow sweep where your cable length is not any close to a 1/4 wave multiplum, then it can actually work to some extend.

Generally if you can't evaluate when you are measuring correct, I recommend you to keep your calibration point as close as possible to the analyzer and respect you might have significant errors if your CAL point exceeding 1/8 wave length away from the port.

People believing they can calibrate out 20 meter of coax at any frequency, does not have any idea what they are talking about.

On Tue, May 19, 2020 at 12:48 PM, <sj@...> wrote:

I would say NO.
Sorry to say, but have worked with VNA's for 30 years and as soon as you come
near 1/8 wavelength the measurement errors are taking over.

I respectfully disagree. My NanoVNA was calibrated from 2 to 50 MHz using NanoVNA Saver. Then I measured a 20M dipole at the feedpoint and saved the results as an S1P file . I added 55 feet of RG58 to the NanoVNA and re-calibrated at the end of the 55 feet of coax. Connected the antenna to the RG58 and saved the results. Then plotted in NanoVNA Saver using both sweeps (one as the reference). The results are in the pdf file below. It is almost a perfect overlay of the two sweep measurements so the 55 feet was successfully calibrated out.

Roger

Here is a file containing the R+jX comparison from the experiment above. Sweep at feedpoint and at end of 55 feet of RG58 overlay on top of each other.

Roger

| On Tue, May 19, 2020 at 09:48 PM, <sj@...> wrote:
=
|
| I would say NO.
| Sorry to say, but have worked with VNA's for 30 years and as soon as you come
| near 1/8 wavelength the measurement errors are taking over.
| You know, in the analyzer the matrices that are multiplied together, the error
I matrice will start to influence significantly. It is very difficult to say
| that it always will not work when exceeding 1/8 (because you are closing in to
| 1/4 and the matrice in play will get 0's and infinite in some places) because
| let's consider you are clever and you measure in a narrow sweep where your
| cable length is not any close to a 1/4 wave multiplum, then it can actually
| work to some extend.
|
| Generally if you can't evaluate when you are measuring correct, I recommend
| you to keep your calibration point as close as possible to the analyzer and
| respect you might have significant errors if your CAL point exceeding 1/8 wave
| length away from the port.
|
| People believing they can calibrate out 20 meter of coax at any frequency,
| does not have any idea what they are talking about.
=
+1

Why are the two measurements not taken in the same display format? One
graph is in SWR one is in RL.

Tony VE3DWI.

On Tue, May 19, 2020 at 17:24 Roger Need via groups.io <sailtamarack=
[email protected]> wrote:

On Tue, May 19, 2020 at 12:48 PM, <sj@...> wrote:

I would say NO.
Sorry to say, but have worked with VNA's for 30 years and as soon as you

come

near 1/8 wavelength the measurement errors are taking over.

I respectfully disagree. My NanoVNA was calibrated from 2 to 50 MHz using
NanoVNA Saver. Then I measured a 20M dipole at the feedpoint and saved the
results as an S1P file . I added 55 feet of RG58 to the NanoVNA and
re-calibrated at the end of the 55 feet of coax. Connected the antenna to
the RG58 and saved the results. Then plotted in NanoVNA Saver using both
sweeps (one as the reference). The results are in the pdf file below. It
is almost a perfect overlay of the two sweep measurements so the 55 feet
was successfully calibrated out.

Roger

On Tue, May 19, 2020 at 05:11 PM, Tony wrote:

Why are the two measurements not taken in the same display format? One
graph is in SWR one is in RL.

Tony VE3DWI.

They are all taken in the same format! All 4 graphs have both measurements on them. The results are so close that one is directly overlaid on the other and you can't tell one from the other except at a few frequencies. Take a close look at this plot - I zoomed in to show the 2 plots.

Roger

Thank you. I see it now. My old eyes need some magnification hi... hi

Tony VE3DWI.

On Tue, May 19, 2020 at 20:59 Roger Need via groups.io <sailtamarack=
[email protected]> wrote:

On Tue, May 19, 2020 at 05:11 PM, Tony wrote:

Why are the two measurements not taken in the same display format? One
graph is in SWR one is in RL.

Tony VE3DWI.

They are all taken in the same format! All 4 graphs have both
measurements on them. The results are so close that one is directly
overlaid on the other and you can't tell one from the other except at a few
frequencies. Take a close look at this plot - I zoomed in to show the 2
plots.

Roger

On Tue, 19 May 2020 at 22:36, pez@arg <petrosez7@...> wrote:

| On Tue, May 19, 2020 at 09:48 PM, <sj@...> wrote:
=
|
| I would say NO.
| Sorry to say, but have worked with VNA's for 30 years and as soon as you
come
| near 1/8 wavelength the measurement errors are taking over.
| You know, in the analyzer the matrices that are multiplied together, the
error
I matrice will start to influence significantly. It is very difficult to
say
| that it always will not work when exceeding 1/8 (because you are closing
in to
| 1/4 and the matrice in play will get 0's and infinite in some places)
because
| let's consider you are clever and you measure in a narrow sweep where
your
| cable length is not any close to a 1/4 wave multiplum, then it can
actually
| work to some extend.
|
| Generally if you can't evaluate when you are measuring correct, I
recommend
| you to keep your calibration point as close as possible to the analyzer
and
| respect you might have significant errors if your CAL point exceeding
1/8 wave
| length away from the port.
|
| People believing they can calibrate out 20 meter of coax at any
frequency,
| does not have any idea what they are talking about.

The standard Agilent cables for my HP 8720D VNA are 24" long, which is
about 600 mm. The maximum frequency of the VNA is 20 GHz, so a wavelength
of approximately 15 mm. Even ignoring the velocity factor of the cables,
that makes them 40 wavelengths long.

Maybe you mean something different to what I think you mean.

Dave

Calibrating to the end of a cable was SOP in the lab I ran and every
time some one would tell me it introduces error I'd give them what ever
time they needed to prove it. They all failed in their assertions
as did I when I had that assumption in my head many years before.

Your data is as good as the cal to the plane of the cal. Even with
45ft of RG400 at 400mhz it worked.

The only yabut of it is if the losses are so great and the cable
long enough it will eventually self terminate. At that point its
not useful to do a SOL cal at the far end as the reflections are
absorbed before they reach the instrument.

Allison
-----------------
No direct email, it goes to bit bucket due address harvesting in groups.IO

One caveat to measuring at the end of a long cable is that the calibration and measurement may need to have the sweep speed reduced. A round trip reflection cal or measurement with a 100 M transmission line with a velocity factor of .66 will be roughly a microsecond.
This can be controlled on a commercial VNA - I'm not sure if the nanovna / saver has that capability.

Bob

NanoVNA need ~1-2ms for measure 1 point
Pause 0.5ms between measure points for wait si5351 freq lock (can be larger on band change)

On Tue, May 19, 2020 at 12:48 PM, <sj@...> wrote:

I would say NO.
Sorry to say, but have worked with VNA's for 30 years and as soon as you come
near 1/8 wavelength the measurement errors are taking over. ...
People believing they can calibrate out 20 meter of coax at any frequency,
does not have any idea what they are talking about.

If that were true, nobody would ever use a VNA above DC, or some frequency very close to it. Like others here, I also have taken measurements at the end of a long coax, with good results.

On Wed, 20 May 2020 at 02:58, aparent1/kb1gmx <kb1gmx@...> wrote:

Calibrating to the end of a cable was SOP in the lab I ran and every
time some one would tell me it introduces error I'd give them what ever
time they needed to prove it. They all failed in their assertions
as did I when I had that assumption in my head many years before.

Your data is as good as the cal to the plane of the cal. Even with
45ft of RG400 at 400mhz it worked.

The only yabut of it is if the losses are so great and the cable
long enough it will eventually self terminate. At that point its
not useful to do a SOL cal at the far end as the reflections are
absorbed before they reach the instrument.

Allison

There is another thing to be careful of, if using some of the older
professional VNAs, like the 8753. The VNA transmits a signal continuously
during the sweep, changing the frequency during the sweep. If the cables
are long, the transit time can be such that the received signal will be on
a significantly different frequency to that transmitted. It can then fall
outside the passband of the filter. This is well documented in application
notes from HP/Agilent/Keysight, and in the manuals for the VNAs. One just
increases the sweep time.

One thing to be aware of is the phase stability of the cables. Good quality
VNA test port cables are expensive - over $5000 for my VNA.

G8WRB.

Hello.

Let us ask, please, two "rhetorical" questions, which are addressed to all
readers of this thread:

- Is there anyone, ever, wondered, in any similar to this, under discussion,
case - that is both "near by"- and "far away" from- a VNA:

= (1) What is the behavior of the three "Traditional" measurement "errors"
(ED, EM, ER) or (D, M, R) ?

- We had; and found the following results [1][*]:
-

= (2) How much "the measurement errors are taking over" ?

- We had; and found the following results [2][*]:
-

for two substantially different loads under the two, most typical,
telecommunication situations, respectively, that is of:

- a load "near by" a VNA - in fact, a DC-50-Ohm Resistor - and

- an Antenna "far away" from that VNA - in fact, a Ground-Plane one.

Well, from the last figure, any reader can easily see how much the
Error of Measurement is, as:

(a) A "Traditional", Absolute, Real Non-Negative, Error: Delta-S11,
as it is calculated by using the technique introduced by HP and thus
resulted as a couple of circles on the Complex Plane of this figure,
as well as,

(b) An Estimation of the Complex Core Uncertainty of the Measurement
Error Delta-Rho, computed and drawn by using two methods [3]:

a differential and a numerical one, and thus resulted as a continuous line
and a cluster of points, respectively, on the Complex Plane of this figure.

CONCLUSION

Although, in comparison, these two Estimations of the Measurement Error
have quite different appearance, because the Traditional Error is arbitrarily
less or more than the Core Uncertainty Error, "near by"- and "far away" from- that
VNA, respectively, the final judgement of which one of these two estimations
represents more accurately the Measurement Error is left to the reader.

Sincerely,

gin&pez@arg

REFERENCES

[1][*] : Fig. 2 & Fig. 4
[2][*] : Fig. 13
[*] : [researchgate] :

[3] : gin&pez@arg : [errors of "error" models"] : a thread, beginning with
#2,770 : 2019-09-21: /g/nanovna-users/message/2770

REF: /g/nanovna-users/message/13681

Hello, again.

Well, after the Special Interest shown here, we would like to inform you
that, in direct contrast with what Gary O'Neil, N3GO believed on 2009-09-30
-
in the "Errors of error models. Who is following this and who is confused"
thread, at his message: /g/nanovna-users/message/3492
-
that is that: "They appear to have done limited testing of their math model",
there are plenty Measured Results by a VNA under these Most Interesting
Extreme Conditions created by many Open- and Short- circuited loads, which
were produced by purposely constructed for that reason Special Antennas,
and we plan to present some of these results, as soon as possible, within
the frame of discussion: "errors of "error" models".

Sincerely,

gin&pez@arg

DR Kirkby,

That would be a corner case and a very very long cable.
Typical cables used ran maybe as long as 100M, so we
haven't reached a radar mile yet. Usually we have no
reason to make sudden frequency range changes but
have in the past. Most of the expensive VNA/PNAs
the dwell times, numbers of points are programmable
and that also dictates the sweep rates. However,
as HP/Agilent/Keysight has noted you can tell the machine
to do ill advised things and get the deserved nebulous results.

More typical would be a piece of RG400 10M long. We
used a lot of specially made cables that were ferrite loaded.

Bench cable are typical 1M commercial made like minicircuits
ULC-1M-NMNN (1 meter cables). or ULC-6FT-NMNN (6ft).
Costly but quality when using 6ghz PNA in the 5700mhz range.
The payback is quality commercial cables have good useful
lifetimes. Often they can be found used or surplus in good
condition.

Allison
-----------------
No direct email, it goes to bit bucket due address harvesting in groups.IO

Dr. Kirkby is right, as I can confirm from my own experience. Last year I wanted to do a RL measurement o find the losses of several lengths (up to 25 m) of leftover coaxial cables on all amateur bands from 2 m up to 13 cms IN ONE GO. So I set my HP8753C to a span of something like 100 to 2500 MHz and measured the RL of each cable on one end while leaving the other end open (or shorting it), with default settings of the 8753. Or maybe I used 401 iso 201 points. Anyway, I then noticed unexplainable steps along the way in the measured RL curves. This turned out to be simply a matter of the default sweep time being too short. I haven't tried a similar thing with my NanoVNA-F up to 1.5 GHz, and maybe I will just for curiosity. But you can't change the sweep time of a NanoVNA anyway, to my early knowledge of the device.

73's Hans PA3AZA