Hi,

I am just starting to use my NanoVNA Pro to measure a four port hybrid coupler I am making. I noticed that after calibrating the unit with the included cables, one on each port, when I connect the cables with the thru, the VNA says the input impedance of port 2 is 53.4 Ohms with about 635pf of capacitance. The narrow band hybrid I am making is quite sensitive to the impedances on its ports. So, what happens is, if I do a one port measurement with the other three ports of the hybrid terminated with 50 ohms, I get a S11 curve with a minimum say at 150 MHz. When I connect the second port of the VNA to port 2 of my hybrid the 90 deg port, the minimum of the S11 curve shifts  up about 151 MHz. And when I connect the VNA  to the port 3 of my hybrid the 0 deg port the S11 minimum shifts down about 149 MHz from the one port measurement. Of course port 1 of the hybrid is being driven by the VNA and the other two hybrid ports are terminated with 50 ohm loads when doing the two port measurements.

So, I searched for some old messages on this and found something that said the port two of the VNA has a frequency dependent return loss. This is bad for what I am trying to do. So, Does anybody know of a workaround for the lack of a stable 50 ohm port 2 input impedance on the VNA?

Hi,

I am just starting to use my NanoVNA Pro to measure a four port hybrid coupler I am making. I noticed that after calibrating the unit with the included cables, one on each port, when I connect the cables with the thru, the VNA says the input impedance of port 2 is 53.4 Ohms with about 635pf of capacitance. The narrow band hybrid I am making is quite sensitive to the impedances on its ports. So, what happens is, if I do a one port measurement with the other three ports of the hybrid terminated with 50 ohms, I get a S11 curve with a minimum say at 150 MHz. When I connect the second port of the VNA to port 2 of my hybrid the 90 deg port, the minimum of the S11 curve shifts  up about 151 MHz. And when I connect the VNA  to the port 3 of my hybrid the 0 deg port the S11 minimum shifts down about 149 MHz from the one port measurement. Of course port 1 of the hybrid is being driven by the VNA and the other two hybrid ports are terminated with 50 ohm loads when doing the two port measurements.

So, I searched for some old messages on this and found something that said the port two of the VNA has a frequency dependent return loss. This is bad for what I am trying to do. So, Does anybody know of a workaround for the lack of a stable 50 ohm port 2 input impedance on the VNA?

Hello Frederick,



I‘d simply put an attenuator on port 2 and thereby ‚force’ the return loss to
a higher level. Something like 3 to 6 dB should do the job.



You may lose some dynamic range though.



Best regards



Dieter



Von meinem iPhone gesendet



> Am 03.12.2023 um 09:56 schrieb frederick.brown@gmail.com:
>
>

> Hi,
>
> I am just starting to use my NanoVNA Pro to measure a four port hybrid
coupler I am making. I noticed that after calibrating the unit with the
included cables, one on each port, when I connect the cables with the thru,
the VNA says the input impedance of port 2 is 53.4 Ohms with about 635pf of
capacitance. The narrow band hybrid I am making is quite sensitive to the
impedances on its ports. So, what happens is, if I do a one port measurement
with the other three ports of the hybrid terminated with 50 ohms, I get a S11
curve with a minimum say at 150 MHz. When I connect the second port of the VNA
to port 2 of my hybrid the 90 deg port, the minimum of the S11 curve shifts
up about 151 MHz. And when I connect the VNA to the port 3 of my hybrid the 0
deg port the S11 minimum shifts down about 149 MHz from the one port
measurement. Of course port 1 of the hybrid is being driven by the VNA and the
other two hybrid ports are terminated with 50 ohm loads when doing the two
port measurements.
>
> So, I searched for some old messages on this and found something that said
the port two of the VNA has a frequency dependent return loss. This is bad for
what I am trying to do. So, Does anybody know of a workaround for the lack of
a stable 50 ohm port 2 input impedance on the VNA?

_._,_._,_

* * *

First, be sure some of it is not aggravated by your cables.  I have to
be careful to use real 50-ohm cables on the test bench. It isn't
uncommon to find 55-ohm cable sold as 50-ohm.

You could use a good 5 or 10dB pad at port two, and calibrate S12 or S21
with that pad.

You could also correct port 2 a bit with a small surface mount shunt
resistor. The 635pF series equivalent capacitance is negligible at 150 MHz.

73 Tom


On 12/3/2023 3:56 AM, frederick.brown@gmail.com wrote:

> Hi,
>
> I am just starting to use my NanoVNA Pro to measure a four port hybrid
> coupler I am making. I noticed that after calibrating the unit with
> the included cables, one on each port, when I connect the cables with
> the thru, the VNA says the input impedance of port 2 is 53.4 Ohms with
> about 635pf of capacitance. The narrow band hybrid I am making is
> quite sensitive to the impedances on its ports. So, what happens is,
> if I do a one port measurement with the other three ports of the
> hybrid terminated with 50 ohms, I get a S11 curve with a minimum say
> at 150 MHz. When I connect the second port of the VNA to port 2 of my
> hybrid the 90 deg port, the minimum of the S11 curve shifts  up about
> 151 MHz. And when I connect the VNA  to the port 3 of my hybrid the 0
> deg port the S11 minimum shifts down about 149 MHz from the one port
> measurement. Of course port 1 of the hybrid is being driven by the VNA
> and the other two hybrid ports are terminated with 50 ohm loads when
> doing the two port measurements.
>
> So, I searched for some old messages on this and found something that
> said the port two of the VNA has a frequency dependent return loss.
> This is bad for what I am trying to do. So, Does anybody know of a
> workaround for the lack of a stable 50 ohm port 2 input impedance on
> the VNA?
>

--
This email has been checked for viruses by AVG antivirus software.
www.avg.com

This is one of the reasons the nanoVNA costs $200 rather than the $40,000
for a Keysight PNA.

On Sun, Dec 3, 2023 at 10:40 AM Dieter Horst <dieter.horst@arcormail.de>
wrote:

> Hello Frederick,
>
> I‘d simply put an attenuator on port 2 and thereby ‚force’ the return loss
> to a higher level. Something like 3 to 6 dB should do the job.
>
> You may lose some dynamic range though.
>
> Best regards
>
> Dieter
>
> Von meinem iPhone gesendet
>
> Am 03.12.2023 um 09:56 schrieb frederick.brown@gmail.com:
>
> Hi,
>
> I am just starting to use my NanoVNA Pro to measure a four port hybrid
> coupler I am making. I noticed that after calibrating the unit with the
> included cables, one on each port, when I connect the cables with the thru,
> the VNA says the input impedance of port 2 is 53.4 Ohms with about 635pf of
> capacitance. The narrow band hybrid I am making is quite sensitive to the
> impedances on its ports. So, what happens is, if I do a one port
> measurement with the other three ports of the hybrid terminated with 50
> ohms, I get a S11 curve with a minimum say at 150 MHz. When I connect the
> second port of the VNA to port 2 of my hybrid the 90 deg port, the minimum
> of the S11 curve shifts up about 151 MHz. And when I connect the VNA to
> the port 3 of my hybrid the 0 deg port the S11 minimum shifts down about
> 149 MHz from the one port measurement. Of course port 1 of the hybrid is
> being driven by the VNA and the other two hybrid ports are terminated with
> 50 ohm loads when doing the two port measurements.
>
> So, I searched for some old messages on this and found something that said
> the port two of the VNA has a frequency dependent return loss. This is bad
> for what I am trying to do. So, Does anybody know of a workaround for the
> lack of a stable 50 ohm port 2 input impedance on the VNA?
>
>
>
>

How much attenuation to apply to port 2 to stabilize the impedance? A 6 dB
attenuator will assure a maximum SWR of a bit less than 2:1. A 2:1 SWR is
mighty close to 10 dB return loss (RL). A 6 dB attenuator will assure a
worst case RL of 12 dB. But, as previously mentioned you will loose a bit
of dynamic range in any transmission measurement.

Dave - WØLEV

On Sun, Dec 3, 2023 at 6:29 PM Tom W8JI <w8ji@w8ji.com> wrote:

> First, be sure some of it is not aggravated by your cables. I have to be
> careful to use real 50-ohm cables on the test bench. It isn't uncommon to
> find 55-ohm cable sold as 50-ohm.
>
> You could use a good 5 or 10dB pad at port two, and calibrate S12 or S21
> with that pad.
>
> You could also correct port 2 a bit with a small surface mount shunt
> resistor. The 635pF series equivalent capacitance is negligible at 150 MHz.
>
> 73 Tom
>
>
> On 12/3/2023 3:56 AM, frederick.brown@gmail.com wrote:
>
> Hi,
>
> I am just starting to use my NanoVNA Pro to measure a four port hybrid
> coupler I am making. I noticed that after calibrating the unit with the
> included cables, one on each port, when I connect the cables with the thru,
> the VNA says the input impedance of port 2 is 53.4 Ohms with about 635pf of
> capacitance. The narrow band hybrid I am making is quite sensitive to the
> impedances on its ports. So, what happens is, if I do a one port
> measurement with the other three ports of the hybrid terminated with 50
> ohms, I get a S11 curve with a minimum say at 150 MHz. When I connect the
> second port of the VNA to port 2 of my hybrid the 90 deg port, the minimum
> of the S11 curve shifts up about 151 MHz. And when I connect the VNA to
> the port 3 of my hybrid the 0 deg port the S11 minimum shifts down about
> 149 MHz from the one port measurement. Of course port 1 of the hybrid is
> being driven by the VNA and the other two hybrid ports are terminated with
> 50 ohm loads when doing the two port measurements.
>
> So, I searched for some old messages on this and found something that said
> the port two of the VNA has a frequency dependent return loss. This is bad
> for what I am trying to do. So, Does anybody know of a workaround for the
> lack of a stable 50 ohm port 2 input impedance on the VNA?
>
>
>
> <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient>
> Virus-free.www.avg.com
> <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient>
> <#m_-3560535667607674059_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>
>
>

--

*Dave - WØLEV*

I used the two cables that came with the VNA.

The post I read stated that this problem is due to the lack of 12 term correction in the software. I guess that means the firmware in the DNA is operating with the eight-term correction.

The NanoVNA has been out long enough that this should be a well-known problem with a well-known solution.

I suppose I could make a small board to go on the end of the VNA before the cable with a microstrip line and a shunt resistor for each frequency of interest.

Yeah, thanks for that.

I used to work for an antenna engineer and he first had bought a mini VNA to tune the antennas we were making. I noticed in the use of thhe mini DNA that the numbers it was spitting out failed to add up. Meaning that the the figures the machine was producing for the reactance and the resistance of the load failed to agree with the figure given by the machine for the impedance of the load according to the equation. X^2 + R^2 = Z^2.

We had a different older VNA that cost about a grand. When I tested the antenna with that one all the numbers agreed according to the equation, has one would expect.

When I posted on the mini VNA users form about the problem, the only response I got was the mini VNA is a very accurate machine.

That's solution might work for me because with this project the output levels from the hybrid are quite high only 3 DB down or so. This was also the solution and the other post that I read. I should get an attenuator set.

Minicircuits has decent pads. VAT-6A are about $20 Qty 1

If you want to go the extra mile, calibrate the attenuator. You’ll need
external software to de-embed the (small) mismatch from the attenuator







> On Dec 3, 2023, at 6:29 PM, frederick.brown@gmail.com wrote:
>
>

> That's solution might work for me because with this project the output
levels from the hybrid are quite high only 3 DB down or so. This was also the
solution and the other post that I read. I should get an attenuator set.

_._,_._,_

* * *

Hi, for improved return loss it is common practice to use an attenuation of 3-6dB on port 2.
For example, the Key N9912A specifies a port VSWR of 1.5:1, or *14dB* return loss.
With a 3dB attenuator this is improved to 14 + 3 x 2 = *20dB* , with only 3dB loss of dynamic range.
The V2 Plus4 Pro has similar port 2 return loss, so the same attenuation values can be used.
Thanks

Good answer. When I worked in test lab, it was our practice to add a 3dB
pad to the DUT end of all test cables. Also, we used very high quality
phase stable cables, microwave frequencies, but they cost $1000 for a three
foot cable.

On Tue, Dec 5, 2023 at 10:44 AM nanov2support <edelswartz123@gmail.com>
wrote:

> Hi, for improved return loss it is common practice to use an attenuation
> of 3-6dB on port 2.
> For example, the Key N9912A specifies a port VSWR of 1.5:1, or *14dB*
> return loss.
> With a 3dB attenuator this is improved to 14 + 3 x 2 = *20dB*, with only
> 3dB loss of dynamic range.
> The V2 Plus4 Pro has similar port 2 return loss, so the same attenuation
> values can be used.
> Thanks
>
>
>

It might be worth saying why this kind of RL on the port is ok.
For the vast majority of measurements, the calibration will measure the Port 2 return loss, and factor it in, so S21 measurements (or S22) measurements will be correct.
The challenge comes when you're measuring something that is sensitive to the load impedance (some amplifiers, some filters, mixers, etc.) then, it's worth improving the RL of Port 2 (and possibly, increasing the drive from Port 1 to get the dynamic range back).

If you want a real challenge, measure all this stuff at high power, or characterize the changes in performance as the load impedance changes (Load pull testing, generically). For instance, the frequency and output power of a magnetron is highly dependent on the load impedance.

This is sort of the rabbit hole down into which you can descend when it comes to making measurements with a device like a VNA.

Ah, Load Pull Testing! Did a lot of that qualifying cell phone PAs. The
companies Focus and Maury Microwave make high-end automated
electro-mechanical tuners for load pull. The VNA is used to calibrate the
tuner for a band of interest. Then the tuner could provide you with any
impedance (R +jX) at a given frequency while you then measured the
amplifier's output gain with the VNA and checked for stability. We would
often check all phases of a 10:1 SWR. This was of course automated as it
took a long time and we measured a batch of parts at high and low
temperature.

On Tue, Dec 5, 2023 at 9:59 PM Jim Lux <jimlux@earthlink.net> wrote:

> It might be worth saying why this kind of RL on the port is ok.
>
> For the vast majority of measurements, the calibration will measure the
> Port 2 return loss, and factor it in, so S21 measurements (or S22)
> measurements will be correct.
>
> The challenge comes when you're measuring something that is sensitive to
> the load impedance (some amplifiers, some filters, mixers, etc.) then, it's
> worth improving the RL of Port 2 (and possibly, increasing the drive from
> Port 1 to get the dynamic range back).
>
>
>
> If you want a real challenge, measure all this stuff at high power, or
> characterize the changes in performance as the load impedance changes (Load
> pull testing, generically). For instance, the frequency and output power of
> a magnetron is highly dependent on the load impedance.
>
>
>
> This is sort of the rabbit hole down into which you can descend when it
> comes to making measurements with a device like a VNA.
>
>
>
>
>
>
>
>
>
>

"for improved return loss it is common practice to use an attenuation of 3-6dB on port 2"

This is interesting. A 6 dB attenuator seems to be the popular suggestion. You say that this 'improves' the return loss. It improves the loss figure by hiding the power of the reflection in in the dissipation of the attenuator. The problem is still there. The method just kind of sweeps it under the rug so to say.

I have seen this effect in simulating antennas with a length of coax feed line. The loss of the coax attenuates the SWR mismatch of the antenna. For me I guess this means that the impedance mismatch at port 2  of the VNA will affect the performance of my hybrid less by however much the reflection is reduced in the attenuator. I give it a try.

Cheers,

Fred B

And the price paid, such precision and firmware updates are not in the manufacturers financial interests.



These are not professional folks.

I have never had a problem with calibrations. Please let me know what to
look at for the problem(s).

Dave - WØLEV

On Sun, Dec 28, 2025 at 5:28 PM W4JDY1953_G via groups.io <W4JDY1953=
gmail.com@groups.io> wrote:

> And the price paid, such precision and firmware updates are not in the
> manufacturers financial interests.
>
>
>
> These are not professional folks.
>
>
>
> *From:* NanoVNAV2@groups.io <NanoVNAV2@groups.io> *On Behalf Of *
> frederick.brown@gmail.com
> *Sent:* December 3, 2023 19:05
> *To:* NanoVNAV2@groups.io
> *Subject:* Re: [nanovnav2] Input Impedance of Port 2
>
>
>
> I used the two cables that came with the VNA.
>
> The post I read stated that this problem is due to the lack of 12 term
> correction in the software. I guess that means the firmware in the DNA is
> operating with the eight-term correction.
>
> The NanoVNA has been out long enough that this should be a well-known
> problem with a well-known solution.
>
> I suppose I could make a small board to go on the end of the VNA before
> the cable with a microstrip line and a shunt resistor for each frequency of
> interest.
>
>
>
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient>
> Virus-free.www.avast.com
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=emailclient>
> <#m_-5920165853195463421_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>
>
>

--

*Dave - WØLEV*

As alluded to in a previous post, the nanoVNA is a 1-path 2-port instrument using a 6-term error model that is calculated from calibration.  Even after calibration It measures only S11 and S21. It doesn't automatically turn the device around and stimulate it from the 2nd port.  For a full 2-path measurement one needs 12 error terms and a test environment that reverses stimulus/response ports for you. But you can do that manually with less expensive VNAs like this one.

* using VNA View (nanonovna-qt)
* Perform the best calibration you can  with the best standards you have. Being certain to account for the measurement planes for the standards.
* Perform a  first  measurement of your DUT. Save the results using S1* capture.
* Physically and carefully turn your device around, repeat the DUT measurement and save those second results with S2* capture.
* Save both those result to a Touchstone file  using Export s2p.

You now have complete S-Parameters for a two-port device, measured from the perspective of a 'perfect' VNA located at the calibration planes. VNA port imperfections have been "calibrated out".
Use your preferred tool, perhaps something like QUCS-S, to import the Touchstone file for analysis.

BUT, remember that although you have an accurate measurement unless your application provides just as perfect an environment there will be mismatch errors in practice.    If you are really being precise you can proceed to measure your application environment and then use tools such as Python's scikit-rf  to predict exactly what the in-use results will be.

This points out our limits.   The original HP 8510 used very carefully fabricated air dielectric coaxial lines as impedance standards. These went into the sliding-load calibrations as well as stand-alone "air lines".  Gun rifle bear machining technology was used to do this.  Even so, reflection measurements much below -50 dB became very difficult.  You can see why if you calculate line impedance vs dimensional error.  Sag of the center conductor, gold plating roughness, skin effect and the presence of a (very small at 50 ohms) TM mode of propagation in coax all served to set practical limits.

I'm mentioning this to point out the difference between theory and practice.  You probably care about using your coupler in some real environment that isn't as precise as your standards.  Unless you take this into account some of your extra effort may be wasted.   Try measuring 'good' coaxial cable impedance some time.  I think you'll find that commercial products aren't 50.0 ohms, but sometimes something pretty far from it !

Best,
Glenn n6gn

Glen, not just "50-ohm" coax, but I've measured window line claiming to be
450-ohms. Values from 370 to 480 ohms are common. In my case, it all
comes out in the matching network(s).

The "old" HP always pushed the limits on what they could manufacture and
sell. It was expensive for a reason. Some of us got used to that
extremely low level of inaccuracy and precision.

However, the average amateur should keep in mind the cost of the NANOVNAs
and TinySAs is many orders of magnitude less than the traditional HP
equipment and the R&S equipment of today. These inexpensive pieces of RF
test equipment are more than "good enough" for our amateur uses! Early on,
I complared the NANOVNAs to my HP 8753C. Not bad at all and beyond the
performance the amatuer demands. But I'm sure glad to have instruments I
can easily carry to the feedpoint of my antennas!

Dave - WØLEV



On Wed, Dec 31, 2025 at 8:03 AM Glenn n6gn via groups.io <n6gn=
sonic.net@groups.io> wrote:

> As alluded to in a previous post, the nanoVNA is a 1-path 2-port
> instrument using a 6-term error model that is calculated from calibration.
> Even after calibration It measures only S11 and S21. It doesn't
> automatically turn the device around and stimulate it from the 2nd port.
> For a full 2-path measurement one needs 12 error terms and a test
> environment that reverses stimulus/response ports for you. But you can do
> that manually with less expensive VNAs like this one.
>
>
> - using VNA View (nanonovna-qt)
> - Perform the best calibration you can with the best standards you
> have. Being certain to account for the measurement planes for the standards.
> - Perform a first measurement of your DUT. Save the results using
> S1* capture.
> - Physically and carefully turn your device around, repeat the DUT
> measurement and save those second results with S2* capture.
> - Save both those result to a Touchstone file using Export s2p.
>
>
> You now have complete S-Parameters for a two-port device, measured from
> the perspective of a 'perfect' VNA located at the calibration planes. VNA
> port imperfections have been "calibrated out".
> Use your preferred tool, perhaps something like QUCS-S, to import the
> Touchstone file for analysis.
>
> BUT, remember that although you have an accurate measurement unless your
> application provides just as perfect an environment there will be mismatch
> errors in practice. If you are really being precise you can proceed to
> measure your application environment and then use tools such as Python's
> scikit-rf to predict exactly what the in-use results will be.
>
> This points out our limits. The original HP 8510 used very carefully
> fabricated air dielectric coaxial lines as impedance standards. These went
> into the sliding-load calibrations as well as stand-alone "air lines". Gun
> rifle bear machining technology was used to do this. Even so, reflection
> measurements much below -50 dB became very difficult. You can see why if
> you calculate line impedance vs dimensional error. Sag of the center
> conductor, gold plating roughness, skin effect and the presence of a (very
> small at 50 ohms) TM mode of propagation in coax all served to set
> practical limits.
>
> I'm mentioning this to point out the difference between theory and
> practice. You probably care about using your coupler in some real
> environment that isn't as precise as your standards. Unless you take this
> into account some of your extra effort may be wasted. Try measuring
> 'good' coaxial cable impedance some time. I think you'll find that
> commercial products aren't 50.0 ohms, but sometimes something pretty far
> from it !
>
> Best,
> Glenn n6gn
>
>
>

--

*Dave - WØLEV*

climb a tower with an hp  or agilent on a shoulder strap and find out
that you forgot to bring the calibration kit up to the top of the tower :-)

dg9bfc sigi

Am 01.01.2026 um 19:17 schrieb W0LEV via groups.io:

> Glen, not just "50-ohm" coax, but I've measured window line claiming
> to be 450-ohms.  Values from 370 to 480 ohms are common.  In my case,
> it all comes out in the matching network(s).
>
> The "old" HP always pushed the limits on what they could manufacture
> and sell.  It was expensive for a reason.  Some of us got used to that
> extremely low level of inaccuracy and precision.
>
> However, the average amateur should keep in mind the cost of the
> NANOVNAs and TinySAs is many orders of magnitude less than the
> traditional HP equipment and the R&S equipment of today. These
> inexpensive pieces of RF test equipment are more than "good enough"
> for our amateur uses!  Early on, I complared the NANOVNAs to my HP
> 8753C.  Not bad at all and beyond the performance the amatuer
> demands.  But I'm sure glad to have instruments I can easily carry to
> the feedpoint of my antennas!
>
> Dave - WØLEV
>
>
>
> On Wed, Dec 31, 2025 at 8:03 AM Glenn n6gn via groups.io
> <http://groups.io> <n6gn=sonic.net@groups.io> wrote:
>
> As alluded to in a previous post, the nanoVNA is a 1-path 2-port
> instrument using a 6-term error model that is calculated from
> calibration.  Even after calibration It measures only S11 and S21.
> It doesn't automatically turn the device around and stimulate it
> from the 2nd port.  For a full 2-path measurement one needs 12
> error terms and a test environment that reverses stimulus/response
> ports for you. But you can do that manually with less expensive
> VNAs like this one.
>
> * using VNA View (nanonovna-qt)
> * Perform the best calibration you can  with the best standards
> you have. Being certain to account for the measurement planes
> for the standards.
> * Perform a  first  measurement of your DUT. Save the results
> using S1* capture.
> * Physically and carefully turn your device around, repeat the
> DUT measurement and save those second results with S2* capture.
> * Save both those result to a Touchstone file  using Export s2p.
>
> You now have complete S-Parameters for a two-port device, measured
> from the perspective of a 'perfect' VNA located at the calibration
> planes. VNA port imperfections have been "calibrated out".
> Use your preferred tool, perhaps something like QUCS-S, to import
> the Touchstone file for analysis.
> BUT, remember that although you have an accurate measurement
> unless your application provides just as perfect an environment
> there will be mismatch errors in practice. If you are really being
> precise you can proceed to measure your application environment
> and then use tools such as Python's scikit-rf  to predict exactly
> what the in-use results will be.
> This points out our limits.   The original HP 8510 used very
> carefully fabricated air dielectric coaxial lines as impedance
> standards. These went into the sliding-load calibrations as well
> as stand-alone "air lines".  Gun rifle bear machining technology
> was used to do this.  Even so, reflection measurements much below
> -50 dB became very difficult.  You can see why if you calculate
> line impedance vs dimensional error.  Sag of the center conductor,
> gold plating roughness, skin effect and the presence of a (very
> small at 50 ohms) TM mode of propagation in coax all served to set
> practical limits.
> I'm mentioning this to point out the difference between theory and
> practice.  You probably care about using your coupler in some real
> environment that isn't as precise as your standards.  Unless you
> take this into account some of your extra effort may be wasted. 
>  Try measuring 'good' coaxial cable impedance some time.  I think
> you'll find that commercial products aren't 50.0 ohms, but
> sometimes something pretty far from it !
> Best,
> Glenn n6gn
>
>
>
> --
> *Dave - WØLEV
> *
>
>
>

Dave,
It's not just instrument precision, which even nanoVNA has a lot of if good standards and techniques are used. I agree that it is plenty for most  ham use can and that portability is a very valuable feature of all the tiny products.

Fundamentally, the theory around high impedance lines, whether coaxial or balanced is WRONG.  I've written about this in Another Look At Transmission Lines in RSGB's Radcomm+ *,* a draft of that article available *Text of the Article ( http://www.sonic.net/%7En6gn/Another%20Look/Another%20Draft%202019.pdf ) and Accompanying Figures ( http://www.sonic.net/%7En6gn/Another%20Look/Figures_draft%202019.pdf )* for those who do not have RSGB membership *.* I'll likely publish an even more complete article and descriptions of measurements expanding on this  soon.

While one RSGB  text from the last 125 years was cautious about describing the theory for  balanced lines outside of ~200-600 range, the standing opinion that "276*log(b/a)* is precise is clearly wrong by inspection.  For very large spacing/diameter ratios this exceeds twice the impedance of free space and is clearly not possible.  This is discussed in my previous writing about SWTL.

The problem is even deeper than this though. Our very use of S-Parameters is limited to the idea of a measurement plane and calibration plane where a 'perfect' VNA would report what it sees as a stimulus/response ratio.  But this is a plane.  It does not account for high impedance where there can be a  longitudinal component of fields that violate the planar presumption so violate the presumption of both the VNA and S-Parameters  and therefore invalidate the entire context.  More about this in publications to come, perhaps.

Glenn n6gn
Fort Collins, CO

One presumably calibrates on the ground before climbing the tower. (at least, that's what I've done) And to be fair, there's that cool Anritsu "cal kit on a keychain" thing they used to have, with Open, Short, and Thru all one assembly.

https://www.anritsu.com/en-GB/components-accessories/products/osl-and-tosl-series