Beware of cheap underperforming clones

As of 2022 there are many badly performing clones on the market. V2/3GHz NanoVNA uses parts like ADF4350 and AD8342 which are costly and clones have been cutting costs by using salvaged or reject parts.

See official store and look for V2 Plus4/V2 Plus4 Pro versions only to avoid getting a bad clone. We have stopped selling V2.2 versions since October 2020, so all V2 hardware that are not Plus or Plus4 are not made by us and we can not guarantee performance.

NanoVNA V2 Forum

Note: this page is a mirror of https://groups.io/g/NanoVNAV2.
Click here to join and see most recent posts.

A 75 ohm VNA?


Leif M 2021/01/14 10:50

Or a VNA for other resistances. I read that measuring in 75 ohm system loses some accuracy. So how about a 75 ohm VNA?

PETER LAWRENCE 2021/01/14 21:24

Is it valid to simply do the calibrate procedure, but using R75 feeders
and load , instead of R50.



I am building a 75 ohm TVI - DMR filter using multiple stubs out of 75
ohm co-ax.
I need to pass above 470MHz, could live with 1 or 2dB through loss.
tail off below 470, with best rejection 420 -430 MHz.probably only need
20dB rejection.

Pete
------ Original Message ------
From: "Leif M" <leif.michaelsson@gmail.com>
To: NanoVNAV2@groups.io
Sent: Thursday, 14 Jan, 21 At 18:50
Subject: [nanovnav2] A 75 ohm VNA?
Or a VNA for other resistances. I read that measuring in 75 ohm system
loses some accuracy. So how about a 75 ohm VNA?

Leif M 2021/01/14 13:58

Good questions. I hope some guru knows more than I. I have no need a 75 R model, but am just curious.

David Eckhardt 2021/01/14 22:15

I'd suggest broadband transformers. Cal with them in place; end-to-end so
you can use 50-ohm cal standards. I'd suggest Minicircuits as the source
of the transformers. Or, since you have a VNA, design and measure your
own. That way, you know precisely what you are dealing with. You can also
measure the Minicircuits units for the same porpose.

If there were VNAs for every requirement:

About everyone in RF communications: 50-ohms
The TV and CATV industry: 75-ohms
Some video applications: 90-ohms
Audio (TP and TSP): 90 to 120-ohms
Minimum loss group: 300 to 450-ohms
High Power Laser feeds: 2-ohms (slab line)
High Pulsed Power Research applications: 0.5 to 5-ohms


The trade-offs (based on Bell Lab studies in the mid to late 40's for
coaxial cable):
Minimum Loss: 75-ohms
Max power handling capability: 32-ohms
"Happy" medium: 50-ohms
Numerical Average: [75 + 50] / 2 = 53.50-ohms
Geometric Mean: [75 x 50]^1/2 = 48.99-ohms

I'm happy to have a good instrument that is devoted to 50-ohms. I'll deal
with the outliers when they occur during my design and home brewing needs.

Dave - WØLEV

On Thu, Jan 14, 2021 at 9:24 PM PETER LAWRENCE via groups.io <pdlawrence=
btinternet.com@groups.io> wrote:

> Is it valid to simply do the calibrate procedure, but using R75 feeders
> and load , instead of R50.
>
>
> I am building a 75 ohm TVI - DMR filter using multiple stubs out of 75 ohm
> co-ax.
>
> I need to pass above 470MHz, could live with 1 or 2dB through loss.
>
> tail off below 470, with best rejection 420 -430 MHz.probably only need
> 20dB rejection.
>
> Pete
>
> ------ Original Message ------
> From: "Leif M" <leif.michaelsson@gmail.com>
> To: NanoVNAV2@groups.io
> Sent: Thursday, 14 Jan, 21 At 18:50
> Subject: [nanovnav2] A 75 ohm VNA?
>
> Or a VNA for other resistances. I read that measuring in 75 ohm system
> loses some accuracy. So how about a 75 ohm VNA?
>
>
>
>
>
>
>

--
*Dave - WØLEV*
*Just Let Darwin Work*

barnc4br 2021/01/15 06:04

Assuming a 75Ω dummy load is 75Ω, how much error would you get at 75Ω using that to cal instead of the 50Ω std.?
The Nano calibration is normalizing the Smith chart to place 0, ∞, and 50Ω at the appropriate points. Using 75Ω simply creates a 75Ω Smith chart, right? If you connect the 50Ω std., it now appears at 50Ωs on a chart normalized for 75Ω, somewhere between 1 and ∞. And 75Ω appears at 1. Obviously, there is a mismatch between the physical Nano Z and the load Z, but how much of that is normalized?
It seems like it would be more of a problem for S21 tests than for S11. Maybe not.

Glenn Elmore 2021/01/15 06:18

" If you connect the 50Ω std., it now appears at 50Ωs on a chart normalized for 75Ω,"

No. In an ideal (TEM) environment and connector type/fixture, the center of the chart will be the reference impedance you used, whatever that may be. The center of a Smith Chart is not 50 ohms - it's the reference load.
Thus apart from imperfect fixture/connector environment (non-TEM, imperfect equivalence between calibration devices and measurement environment) the Smith Chart continues to give the correct answer.

A perfect 75 ohm match on a Smith Chart referenced to 75 ohms will reveal 1:1 SWR in the same way that a perfect 50 ohm match on a 50 ohm referenced chart will.  Of course, measuring 50 ohms in a 75 ohm environment or vice versa will both show a mismatch.  On a 50 ohm chart, a 75 ohm resistance will show up as a point to the right of center, on a 75 ohm chart a 50 ohm load will be to the left of center. But the measured values in each case will still be correct.

SWR, return loss results depend upon the measurement reference - they are relative measurements.

At this time, nanoVNA seems to have 'baked in' the idea that 50 ohms is some kind of chart center. Unfortunately even the Touchstone files in nanoVNA-saver have 50 ohms hard coded into their headers. This is not the general case.

Glenn n6gn

barnc4br 2021/01/15 06:32

Yes, that is what I think I said. "1" on a Smith chart is the center. That is normalized. *MY error* of course, is that the 50Ω load is going to appear between "1" and *0* , or to the LEFT of "1" or center. and NOT ∞

John Cunliffe W7ZQ 2021/01/15 06:40

Did you ever locked into the offerings of the cable company filter providers. They make high pass, low pass bandpass and notch filters for every frequency range even specific made ones do not cost the world. Probably will safe a lot of time, money and space.

The HPF 500 will probably do the trick for you, why re-invent the wheel?

https://www.soontai.com/HPF.html

John

Kadir Mariño Abreu 2021/01/15 08:24

Wideband impedance adapters exist, but some are expensive.

https://www.fairviewmicrowave.com/matching-pad-50-ohm-sma-male-75-ohm-f-female-3-ghz-fmmp1006-p.aspx?gclid=EAIaIQobChMItLCpzKqe7gIVBYTICh0wrgFhEAQYBCABEgJMovD_BwE

https://www.amazon.com/-/es/859955-50-A-75-Ohm-convertidor/dp/B003NPWGAQ

https://www.cseonline.com.au/Z7550-FMSF-SMA-female-50-ohm-to-F-male-75-ohm-Matching-Transformer.html

https://www.signalbooster.com/products/50-ohm-to-75-ohm-converter-adapter-connector

Glenn Elmore 2021/01/15 08:28

Very weill, I misunderstood your post then. You asked

"Assuming a 75Ω dummy load is 75Ω, how much error would you get at 75Ω using that to cal instead of the 50Ω std.?"

The answer is "it depends". If you have a perfect fixture, and perfect noise-free calibration you will have no error whatsoever.  OTOH, if your calibration devices or the calibration/fixture and calibration/measurement environment itself is not good you can have error, though for small gamma, for DUTs which are relatively close to a match (your good load reference) that error will be small.  For measurements relatively near the center of the chart, which 75 ohms on a 50 ohm chart might be considered to be, you are probably in pretty good territory. But as you go in the direction where your measurement environment isn't ideal, in this case if your high impedance (Open standard) measurement were poor, then as the DUT's characteristics get closer to that area the errors will increase.

Glenn n6gn

barnc4br 2021/01/15 09:29

Yea, I was just referring to error introduced because the Nano is still a 50Ω device even though a 75Ω load was used for cal. The chart is a 75Ω chart now when caled with a 75Ω load. And an open or short is still an open or short regardless of the load used. I'm just wondering what the effect on accuracy of the Nano is when using a 75Ω load instead of a 50Ω given everything else being equal.
I guess any calculations using 50Ω or assuming a 50Ω match would be in error, I just don't know which ones they are.

Dragan Milivojevic 2021/01/15 18:30

https://www.youtube.com/watch?v=OP-mVp8kqNs

On Thu, 14 Jan 2021 at 19:50, Leif M <leif.michaelsson@gmail.com> wrote:

David Eckhardt 2021/01/15 17:55

Designing, building, and measuring is the best method of learning the ropes
of RF. Sure, everything we could possibly want (or think we need) is
available somewhere for a price. But home brewing is a wonderful learning
tool, and once learned, one takes the knowledge throughout life and builds
on that acquired knowledge.

Dave - WØLEV

On Fri, Jan 15, 2021 at 2:40 PM John Cunliffe W7ZQ <n2nep@rochester.rr.com>
wrote:

> Did you ever locked into the offerings of the cable company filter
> providers. They make high pass, low pass bandpass and notch filters for
> every frequency range even specific made ones do not cost the world.
> Probably will safe a lot of time, money and space.
>
> The HPF 500 will probably do the trick for you, why re-invent the wheel?
>
>
> https://www.soontai.com/HPF.html
>
>
> John
>
>
>

--
*Dave - WØLEV*
*Just Let Darwin Work*

Glenn Elmore 2021/01/16 07:18

Essentially a "VNA" has no prescribed impedance.  Internally there is a 'perfect' ADC or other detector. Between that idealized location and the physical location of calibration devices there is real hardware which is not ideal. The model for this imperfection is called an "error adapter".  Calibration is the process of placing definitional devices, called calibration standards,  beyond it and taking measurements. It is these measurements that allows the VNA to characterize the error adapter and subsequently mathematically remove the effects of that imperfect hardware using a model.

Typically the physical environment for that hardware, which may be both internal to the VNA and external in the form of cables and fixtures is in a common impedance and often a common connector type, but it need not be.  For the nanoVNA the hardware which includes the directional device (resistive bridge), switches, transmission line, SMA connector and possibly external  connector and cabling may be nominally 50 ohms but it isn't perfect and neither is this required.  On my nanoVNA, the second port does not look precisely like 50 ohms after calibration with known good standards. This is normal.

When a measurement is performed after calibration the results are essentially the internally measured data modified by the error adapter model and measured values. The results isn't perfect but it can allow an uncorrected VNA, one with perhaps -20 dB or worse deviations from an ideal to become significantly better. High end VNAs which have hardware accuracy of, say, -20 dB can be corrected to -50 or better accuracy, depending upon a host of factors.

The definition of the measurement characteristic impedance within that arbitrary hardware environment and using a TEM model for its contents is also imperfect and there are always limitations to the measurement. Noise, systemic errors, repeatability all enter in to place a limit on the precision and accuracy of even the most careful measurements.

When one performs an open-short-load (OSL) calibration on a VNA it is the load, possibly modified by plane extensions , line loss or other imperfection models that prescribes what the resulting measurement reference mpedance will be. It is not the physical hardware that does this. There is not only nothing truly 50 ohms, or any other impedance, about the overall system. There is  only a user-supplied definition of what the reference impedance is to be considered.   If one uses imperfect 50-ish ohm internal/external hardware  with any reference impedance which doesn't exactly match this, that is to say any reference at all, the correction for this and the resulting data is dependent upon that mapping.  If one uses a 70 ohm or any other load device for calibration the machine is being told "this is the definition of chart center, return all results in relation to this (and the definition of the 'short' and 'open' ) .

If the proffered load has greatly different characteristics, different connector type and 'connectability' some of the assumptions will cease to be very close to reality and the returned results may be considered to be "in error" and not so useful.  Using a 70 ohm load as calibration reference in a generally 50 ohm hardware environment for a standard really isn't any different from measuring a 70 ohm DUT in more purely 50 ohm environment with 50 ohm standards. The same deviations from the model apply in each case, the reference load and the DUT are still being compared with measured and modeled data and errors in this process still apply, whether those differences affect the calibration process or the measurement process they are similarly present in the results.

The practical bottom line is that a "well connected", quality 70 ohm standard can make a perfectly fine calibration device for all practical purposes. Even if the VNA internals have a view that "50 ohms is all there is" if measured data is interpreted in light of the chart center actually being that of a 70 ohm reference the measurements can be useful.

As the situation changes such that the calibration standards' environment no longer is representative of the measurement environment the usefulness of the results degrades. Connecting a wirewound 50 ohm resistor to the calibration port using clip leads and calling it "50 ohms" may give results that aren't very useful, depending upon how much effect those deviations and non-repeatability have at the frequency and connector type being used.

To give yourself confidence, I suggest calibrating the normal way that you trust and measure the candidate '70' ohm calibration reference and see if it looks like a dot at 70 ohms on the chart. If it does, you can proceed to recalibrate using it instead of the 50 ohm standard and have reasonable confidence that data measurements in the vicinity of 70 ohms are also of useful quality.

Glenn n6gn

barnc4br 2021/01/16 08:06

Thank you Glenn!!!
Great explanation! That was my thought, I just did not understand the "error adapter" and if that was effectively calibrated out. My 75Ω load is a simple dummy load for a TV splitter so now I need to see how "dummy" it is.

David Eckhardt 2021/01/16 17:04

Glenn, you can explain far better than I can write. Something that might
help the group is to explain the concept of normalization as applied to the
chart center (or any other function). I could try, but you are far better
at putting it in words. I prefer a white board to explanations of this
sort.

Dave - WØLEV

On Sat, Jan 16, 2021 at 3:18 PM Glenn Elmore <n6gn@sonic.net> wrote:

> Essentially a "VNA" has no prescribed impedance. Internally there is a
> 'perfect' ADC or other detector. Between that idealized location and the
> physical location of calibration devices there is real hardware which is
> not ideal. The model for this imperfection is called an "error adapter".
> Calibration is the process of placing definitional devices, called
> calibration standards, beyond it and taking measurements. It is these
> measurements that allows the VNA to characterize the error adapter and
> subsequently mathematically remove the effects of that imperfect hardware
> using a model.
>
> Typically the physical environment for that hardware, which may be both
> internal to the VNA and external in the form of cables and fixtures is in a
> common impedance and often a common connector type, but it need not be.
> For the nanoVNA the hardware which includes the directional device
> (resistive bridge), switches, transmission line, SMA connector and possibly
> external connector and cabling may be nominally 50 ohms but it isn't
> perfect and neither is this required. On my nanoVNA, the second port does
> not look precisely like 50 ohms after calibration with known good
> standards. This is normal.
>
> When a measurement is performed after calibration the results are
> essentially the internally measured data modified by the error adapter
> model and measured values. The results isn't perfect but it can allow an
> uncorrected VNA, one with perhaps -20 dB or worse deviations from an ideal
> to become significantly better. High end VNAs which have hardware accuracy
> of, say, -20 dB can be corrected to -50 or better accuracy, depending upon
> a host of factors.
>
> The definition of the measurement characteristic impedance within that
> arbitrary hardware environment and using a TEM model for its contents is
> also imperfect and there are always limitations to the measurement. Noise,
> systemic errors, repeatability all enter in to place a limit on the
> precision and accuracy of even the most careful measurements.
>
> When one performs an open-short-load (OSL) calibration on a VNA it is the
> load, possibly modified by plane extensions , line loss or other
> imperfection models that prescribes what the resulting measurement
> reference mpedance will be. It is not the physical hardware that does this.
> There is not only nothing truly 50 ohms, or any other impedance, about the
> overall system. There is only a user-supplied definition of what the
> reference impedance is to be considered. If one uses imperfect 50-ish ohm
> internal/external hardware with any reference impedance which doesn't
> exactly match this, that is to say any reference at all, the correction for
> this and the resulting data is dependent upon that mapping. If one uses a
> 70 ohm or any other load device for calibration the machine is being told
> "this is the definition of chart center, return all results in relation to
> this (and the definition of the 'short' and 'open' ) .
>
> If the proffered load has greatly different characteristics, different
> connector type and 'connectability' some of the assumptions will cease to
> be very close to reality and the returned results may be considered to be
> "in error" and not so useful. Using a 70 ohm load as calibration reference
> in a generally 50 ohm hardware environment for a standard really isn't any
> different from measuring a 70 ohm DUT in more purely 50 ohm environment
> with 50 ohm standards. The same deviations from the model apply in each
> case, the reference load and the DUT are still being compared with measured
> and modeled data and errors in this process still apply, whether those
> differences affect the calibration process or the measurement process they
> are similarly present in the results.
>
> The practical bottom line is that a "well connected", quality 70 ohm
> standard can make a perfectly fine calibration device for all practical
> purposes. Even if the VNA internals have a view that "50 ohms is all there
> is" if measured data is interpreted in light of the chart center actually
> being that of a 70 ohm reference the measurements can be useful.
>
> As the situation changes such that the calibration standards' environment
> no longer is representative of the measurement environment the usefulness
> of the results degrades. Connecting a wirewound 50 ohm resistor to the
> calibration port using clip leads and calling it "50 ohms" may give results
> that aren't very useful, depending upon how much effect those deviations
> and non-repeatability have at the frequency and connector type being used.
>
> To give yourself confidence, I suggest calibrating the normal way that you
> trust and measure the candidate '70' ohm calibration reference and see if
> it looks like a dot at 70 ohms on the chart. If it does, you can proceed to
> recalibrate using it instead of the 50 ohm standard and have reasonable
> confidence that data measurements in the vicinity of 70 ohms are also of
> useful quality.
>
> Glenn n6gn
>
>
>

--
*Dave - WØLEV*
*Just Let Darwin Work*

barnc4br 2021/01/16 13:05

AH, but, calibrating with any load other than 50 places that reference at center (==1) and the Nano says it is 50Ω because there is no way of telling the Nano what the load is and it is programmed to assume it is 50.
Therefore, when I place the 75 on a 50 caled Nano it tells me it is 75. If I cal with 75, it tells me it is 50.
SO, it appears that whatever Z you want to run you need to cal with 50 and the only adjustment you can make is with edelay to accommodate the adapter used.
It seems like if we are calibrating out the error adapter, we should be able to tell the Nano that the load is 75 and then have a 75 Smith chart. But then that seems like too simple of a solution to get a universal VNA.
Is that possible???

David Eckhardt 2021/01/16 21:27

I'll try to explain. There are two approaches to using the Smith Chart:
1) Center is whatever you calibrate to and, 2) A normalized chart
center. The first is obvious.

Normalization is the process of setting the chart center to unity, 1. That
is accomplished by dividing the chart center by the value of the cal
standard. The true, non-normalized, values are recovered by multiplying
the normalized chart values by the value of the cal standard used. If you
cal with 50-ohms, the chart center is 50-ohms. If you cal with 150-ohms,
the chart center is 150-ohms. With a normalized chart, you multiply the
values - both real and reactive - by the cal standard used. Normalization
is a very useful mathematical operation. Take the federal budget -
strictly an example. Say 50 years ago it was $1E9. Say, at present it is
$2.4E15. That's a lot of increase, but by how much?! To better get a
handle on the increase, divide or normalize the present budget by dividing
it by the budget of 50 years ago: 2.4E15 / 1E9 = 2.4E6. So the increase
based on or normalized to the 50-year old budget is 2.4E6 times that of the
50-year old budget. Use of a normalized chart can be very useful when
dealing with something outside the standard set for most RF communication
applications, that being 50-ohms. A couple of days ago, I emailed out how
Bell Labs came to standardize on 50-ohms. Here it is again:

Glenn Elmore 2021/01/16 14:09

Not only is it possible but it's the way most commercial VNAs operate, there is a setting for changing the Zref. Somehow that feature hasn't yet been properly adopted within nanoVNAs. Even though nanoVNA-saver calibration window has a field for different Zref's of non-ideal calibration devices, that value is not properly used afterwards in displayed results. Neither are the Touchstone files saved properly indicating that reference impedance, even though the Touchstone header has provision for precisely that entry. It's presently hardcoded at "50" by the Python code.  This situation seems to be leading some people to think that there is something truly special about 50 ohms and that multiple VNAs are necessary. This is fallacious.

Until and unless this situation changes it's probably best to remember that the center of the chart is always the reference impedance, that of the calibration device for OSL calibrations, and that if a device other than 50 ohms is used for calibration all values should be scaled accordingly.  This is easy for simple impedance and admittance but it does require  more complexity when formats such as SWR or return loss are considered since these depend upon the reference impedance in more complex ways. VSWR= (1 + | S 11 |)/(1 - | S 11 |) but since S11 changes depending upon the reference impedance, VSWR must be understood in that context. A 100 ohm DUT is a mismatch on a 50 ohm chart but it's a perfect match with 1:1 SWR on a 100 ohm chart.

Glenn n6gn

barnc4br 2021/01/16 15:00

David - I understand the chart center. The issue is as far as Nano is concerned the chart center is 50 regardless of what you use to cal. So when I use 75, it is placed in the center and Nano says it is 50. If I use 50 to cal and connect 75, it places it at 75 and says it is 75.
Am I missing something?

David Eckhardt 2021/01/16 23:43

Thanks, Glenn! I'll confess I use the HP 8753C more than the NANOs. But
for the average ham, the NANOs are the next best thing to sliced bread.....

Dave - WØLEV

PS: Someday we need to meet in person...... after Covid?

On Sat, Jan 16, 2021 at 10:09 PM Glenn Elmore <n6gn@sonic.net> wrote:

> Not only is it possible but it's the way most commercial VNAs operate,
> there is a setting for changing the Zref. Somehow that feature hasn't yet
> been properly adopted within nanoVNAs. Even though nanoVNA-saver
> calibration window has a field for different Zref's of non-ideal
> calibration devices, that value is not properly used afterwards in
> displayed results. Neither are the Touchstone files saved properly
> indicating that reference impedance, even though the Touchstone header has
> provision for precisely that entry. It's presently hardcoded at "50" by the
> Python code. This situation seems to be leading some people to think that
> there is something truly special about 50 ohms and that multiple VNAs are
> necessary. This is fallacious.
>
> Until and unless this situation changes it's probably best to remember
> that the center of the chart is always the reference impedance, that of the
> calibration device for OSL calibrations, and that if a device other than 50
> ohms is used for calibration all values should be scaled accordingly. This
> is easy for simple impedance and admittance but it does require more
> complexity when formats such as SWR or return loss are considered since
> these depend upon the reference impedance in more complex ways. VSWR= (1 +
> | S11 |)/(1 - | S11 |) but since S11 changes depending upon the reference
> impedance, VSWR must be understood in that context. A 100 ohm DUT is a
> mismatch on a 50 ohm chart but it's a perfect match with 1:1 SWR on a 100
> ohm chart.
>
> Glenn n6gn
>
>
>

--
*Dave - WØLEV*
*Just Let Darwin Work*

barnc4br 2021/01/16 16:10

So Glenn,
Using a 75 load yields inaccuracies because the firmware uses 50 in calculations instead of 75.
Using 50 load yields inaccuracies because 50 does not properly normalize the error adapter for 75 and any results are skewed.
So the answer is, use your gut feel.
The Nano is a 50Ω device, because the firmware can't handle anything else.
Apparently, it is a non-trivial matter, and any thoughts about testing something other than 50 without a Z transformer are bogus.

David Eckhardt 2021/01/17 00:27

You can always test other 'creations' which are not of the 50-ohm hard
coded impedance using the NANOs. It will read correctly so long as you
don't stray too far from the hard coded value, especially at the high-Z
side of the chart (right side). For example, I have no problem dealing
with 4:1 transformers using a 50-ohm cal. A pure resistive 75-ohm resistor
will read halfway between center and the 100-ohm point to the right. A
pure resistive 100-ohm resistor will read on the 100-ohm point along the
horizontal axis. A pure 25-ohm resistor will read at the 25-ohm point
along the horizontal axis with a 50-ohm cal. However, the source/load
impedance presented to the DUT wil always be 50-ohms with the NANOs
connected directly to the DUT.

The suggestion of minimum loss pads has been made. In that respect I'd
suggest you become familiar with RFCafe.com for designs and tutorials on
all things RF. Also, owning the NANO, you certainly can make you own
transformers with a reasonable stash of appropriate small ferrite
material. You have the tool.

Dave - WØLEV

On Sun, Jan 17, 2021 at 12:10 AM <barnc4br@gmail.com> wrote:

> So Glenn,
> Using a 75 load yields inaccuracies because the firmware uses 50 in
> calculations instead of 75.
> Using 50 load yields inaccuracies because 50 does not properly normalize
> the error adapter for 75 and any results are skewed.
> So the answer is, use your gut feel.
> The Nano is a 50Ω device, because the firmware can't handle anything else.
> Apparently, it is a non-trivial matter, and any thoughts about testing
> something other than 50 without a Z transformer are bogus.
>
>
>

--
*Dave - WØLEV*
*Just Let Darwin Work*

Glenn Elmore 2021/01/16 17:18

Not really. There aren't extra inaccuracies because the firmware is 50ohm-centeric. The hardware and error correction does what it does no matter what the standards were.  If one only pays attention to gamma (mag/phase of reflection coefficient) the right answer results. It's accuracy doesn't change because the FW mangles it. It is what it is, a comparison between the reflection coefficient of a defined-to-be-ideal reference load and the DUT.  The number reported may not be correct but the accuracy of the measurement, ignoring any incorrectly inferred reference impedance is the same.

Calibrate in a load standard (and fixture) that is near that of the DUT and interpret the reported results in light of that impedance. Ignore the calculated R and X since they incorrectly infer that your load standard was 50 ohms. Instead, use the S parameters along with the known load characteristics to determine the result - not what the display might say - or for R+jX scale per the ratio of RefLoadZ/50 ratio.   The nano is not inherently a 50 ohm device, it's an imperfect VNA that doesn't properly accommodate other reference impedances. S11 is calculated against the reference calibration devices.  The S Parameters reported are correct for those standards. It's only when those standards deviate from 50 ohms that the displayed results need to be re-interpreted to account for a different chart center/reference.

Jim Lux 2021/01/16 17:29

On 1/16/21 4:10 PM, barnc4br@gmail.com wrote:
> So Glenn,
> Using a 75 load yields inaccuracies because the firmware uses 50 in
> calculations instead of 75.
> Using 50 load yields inaccuracies because 50 does not properly
> normalize the error adapter for 75 and any results are skewed.
> So the answer is, use your gut feel.
> The Nano is a 50Ω device, because the firmware can't handle anything else.
> Apparently, it is a non-trivial matter, and any thoughts about testing
> something other than 50 without a Z transformer are bogus.
> _._,_._,_

If you measure a device, having used a 75 ohm load for calibration, the
reflection coefficient data (or VSWR) will be correct for a 75 ohm system.

Where it will get wonky is if you then ask for R + X kind of data,
because the firmware assumes always 50, so the answers will be
numerically incorrect - but could be converted algebraically. There are
spreadsheets and code snippets around that can deal with this.

The *accuracy* issue is sort of different. Assuming the correct
calibration, the measurements for a 75 ohm system will be about the same
accuracy (in a percentage sense) as for a 50 ohm system. Where you might
see more accuracy issues is if you were *way* far from 50 ohms -
Measuring a 450 ohm system (after having calibrated with a 450 ohm load)
might not be as accurate because of SNR issues.

Leif M 2021/01/17 13:32

So do I, but many work with TV antennas.

Roger Need 2021/01/19 10:12

This link discusses the pros and cons of various methods of doing 75 ohm measurements on a 50 ohm system.  One post describes 4 different methods.

https://electronics.stackexchange.com/questions/388710/calibrating-vector-network-analyzer-with-75-omega-coax

barnc4br 2021/01/19 12:04

The other problem not mentioned with calibrating at 75Ω is that the VNA still thinks it is 50Ω. When you place your 75Ω calibrator on after calibration, it is shown as 50Ω. If it were 300Ω it would also normalize the center to 300Ω, which the VNA interprets as 50Ω. So your Smith chart values are bogus. Anything else that is calculated using 50Ω is also not correct. I am becoming convinced that for my purposes with wideband TV devices, just using it as a 50Ω device will probably get me well within the ball park. At least when I put 75Ω on it the Smith chart shows 75Ω. A resonant point will still be at the right freq but may not be at the correct SWR. Since it isn't transmitting an accurate value isn't necessary.

Glenn Elmore 2021/01/20 06:50

No!. The "Smith Chart values", the S-Parameters are not bogus!   They are correct for the calibration load device being used, whatever it may be.  It's the derived values, like R, X and SWR that are incorrect since the software not the measurement per se incorrectly evaluate those complex results as having been measured with a 50 ohm standard as reference.

Scattering parameters, S-parameters, are a relative not an absolute measurement.  S parameters are the ratio of a response compared to a stimulus. They are a stimulus-response measurement, "how much it wiggles when you poke it". They assume a reference impedance, that the system is linear and that the environment is TEM.  Presently, in the nanoVNA v2 the raw data measurement and error correction applied by the software does result in valid S-Parameters. Many of the other values do not get calculated correctly for the non-50 ohm case but the S-parameters do get calculated correctly for the supplied calibration device. The Touchstone file that one exports has the correct S-parameters but the header for that file incorrectly always indicates a reference impedance of 50 ohms, no matter what was actually used. If that data file is then passed to another analysis tool that imports the file, maybe QUCS or something similar, unless the file is corrected for the actual load device errors may result.

Because VNAs make relative rather than absolute measurements they also should not be though of as being similar to signal generators or spectrum analyzers like the tinySA. In contrast, these other instruments assume a (most often) 50 ohm context.  Common signal generators specify their power into a matched 50 ohm load. SAs display power interpreted as in a 50 ohm environment. Impedance is presumed. VNAs make measurements that must include the impedance of their environment to be properly interpreted. For VNAs impedance is not presumed.

Although this thread has displayed concern about DUTs and measurement environments  different from 50 ohms, for many if not nearly all amateur uses simply making all measurements with a 50 ohm calibration load and using the data and derived values in that context will probably not be noticeably different from measuring with a 75 ohm standard.  The subtleties of a step from 50 to 75 ohms at the DUT will usually be insignificant compared to measurement inaccuracy due to noise and non-repeatability.  This situation can change for environments that are radically different from 50 ohms, e.g. measuring sub-1 ohm input/output impedance of power bipolar transistors or measuring very high impedance balanced transmission line, but these applications are less common.

If in doubt or worried, I suggest calibrating and staying in 50 ohms, no matter what is being measured, and if it is desired to have results presented  as if they were measured in a different impedance then calculate those afterwards.

Glenn n6gn

Roger Need 2021/01/20 10:49

Those wishing to make 75 ohm measurements using a 50 ohm NanoVNA can use method 1 in the link I posted above.  Measure everything in the 50 ohm system, and convert mathematically to a 75 ohm system impedance.  This will not be as good an estimate as a VNA equipped with 75 ohm measurement capability but will be reasonable for many users.

The method I use it to calculate the complex impedance from the S11 parameters just like the firmware does for a 50 ohm measurement.  I then convert it to a 75 ohm S11 using these complex impedance values and the system impedance of Z0 = 75.
I do this by exporting a S11 Touchstone text file and importing it into a spreadsheet for the calculations and plotting.  Some may prefer Matlab because complex arithmetic has better support and Smith charts can be plotted.

But there is a simpler method for those not wishing to do the calculations.  You can export a S11 Touchstone file and import it into the freely available RigExpert AntScope2 software.  Once there you go to the settings page and change the system impedance to 75 ohms and it will do all the conversions required.  All the plots including the Smith Chart will be now be based on a 75 ohm system impedance.

Roger

barnc4br 2021/01/20 14:29

So, when I calibrate with a 75Ω load, then go back and read the value of the load on the Smith chart it is really what it says it is, 50Ω? It is magic, right?
The S parameters are measured, therefore they are correct, maybe. If they were calculated, they would definitely not be correct. However, if I normalize it with 50, then the Nano would be effecting the S parameter values for 75.
So, without the ability to program the load value, it is a 50Ω device and calibrating with anything else still introduces errors.
If I could tell the Nano the load is 75Ω then the mismatch in the Nano front end would be normalized out and a 75 load would display as 75 on the Smith data and the S parameters would also be more accurate.

Glenn Elmore 2021/01/20 17:23

When you calibrate with a 75 ohm load and measure that same load  the measurement is the center of the chart. Gamma and |S11|=0.  That's a dot at the center of the chart,  75 ohms. Presently the nanoVNA software proceeds to screw up that correct S-Parameter.   It can't handle a 75 ohm reference impedance and calculates R, X etc based on an erroneous chart center - as though that center were 50 ohms  but the S-Parameter measurement is good. The answer is correct for the reference used. Rendering into other representations by the nanoVNA SW is not.  If you use the S-Parameters, even those exported as Touchstone but treat them as a 75 ohm reference, everything is fine. Just don't trust the nano's internal SWR, R, X .... to get it right except for a 50 ohm calibration device situation. It won't (yet).

The fix for all this is for the nano software to be improved such that it uses the calibration reference as the Z0 in it's internal calculations  of other parameters rather than the hard-coded 50 ohms it presently uses.
Thus, using the nanoVNA stand-alone and all of its representations are only correct if the calibration standard was in fact 50 ohms.  BUT the S-Parameters it returns are fine and can be used separately to obtain correct results. Just ignore all but the S-Parameters for non-50 ohm calls and do it yourself recognizing the reference impedance is what you provided.

Hopefully the software will some day be fixed so that the actual, used calibration reference load impedance will eventually be used for internal calculations rather than the hard-coded "50".   It is this flaw that is causing part of the problem that has people thinking that there are different kinds of VNAs.

Until then, simply measuring with a 50 ohm calibration load, even if one is measuring 75 ohm or other impedance DUTs, will almost always be satisfactory.  The only error introduced will be small and subtle for impedances close to 50 ohms, and even for those not too close. The exceptions are corner cases and if you're measuring DUTs in these areas you should already understand the problems associated with good calibration and fixturing of "strange" DUTs. In general, for quality measurements  a fixture, test-set or connector type, should be as close to that of the intended DUT.

So, yes, the errors occur on the nanoVNA when/because it doesn't handle non-50 ohm references properly.

If all this is confusing, calibrate in 50 ohms and measure everything from the understanding that chart center and all the derived parameters are 50 ohm-centric.

Glenn n6gn

ok1vaw 2021/01/21 00:53

There is one not mentioned aspect. There are circuits (filters) which need to be properly terminated, so in case you not insert a correct impedance transformer (resistive pad, etc.), there is an error presented or you can get wrong results..

Reinier Gerritsen 2021/01/21 18:04

Op 21-1-2021 om 09:53 schreef ok1vaw:
> There is one not mentioned aspect. There are circuits (filters) which
> need to be properly terminated, so in case you not insert a correct
> impedance transformer (resistive pad, etc.), there is an error
> presented or you can get wrong results..
> _._,_._,_
I don't know if it is implemented in one of the PC programs, but in at
least in theory, S-parameters fully characterize a filter. So it should
be possible to calculate the effect of termination with a different
impedance. Some day a clever person will implement it.
Reinier

John Gord 2021/01/21 13:43

The ojisan firmware for the NanoVNA-V2 does mapping to S parameters for different Z0 values.  It has to make some assumptions to do so, since the NanoVNAs only measure in one direction.  It assumes S12=S21 (usually true) and S22=S11 (true for symmetrical devices).
It allows you to calibrate in a 50-ohm environment, then display what the DUT would look like in a different impedance environment.  This is handy for looking at things like 330-ohm impedance IF filters.
--John Gord

On Thu, Jan 21, 2021 at 09:05 AM, Reinier Gerritsen wrote:

Stephen Laurence 2021/01/22 01:24

The Ojisan software also allows the investigation of crystals and crystal filters which cannot be done very well on the standard software. It is due to rapid switching on the inputs which is too fast for crystals.

Unfortunately the software is no longer updated as the author has turned his attention to other things. Nevertheless, I am going to keep this software on one of my V2 devices for crystals and filters with non-50 ohm impedances.  For those with only one device, reflashing the software back and forth is not a major task.

Steve L

n2msqrp 2021/01/22 13:48

Where can I find the Ojisan software. I will also intall it on one of my nanaVNA's.

Mike N2MS

John Gord 2021/01/22 14:24

Here is the site for the Ojisan software:
https://ojisankoubou.web.fc2.com/nanovna/nanovnav2.html
Note that the fancy crystal parameter and selectable Z0 features only work on the V2, not the original NanoVNA.
--John Gord

On Fri, Jan 22, 2021 at 10:48 AM, n2msqrp wrote:

To reply to this topic, join https://groups.io/g/NanoVNAV2

View this thread on groups.io