Author Topic: [Solved] NanoVNA says impedance of my 3200R resistor is 542R-1250X at 50 MHz  (Read 7979 times)

0 Members and 1 Guest are viewing this topic.

Offline niconiconiTopic starter

  • Frequent Contributor
  • **
  • Posts: 366
  • Country: cn
I recently purchased a NanoVNA for playing with radio electronics, mainly HF radio for now, and I've already measured some filters and the result looks good. Now I'm trying to understand the behavior of toroid impedance transformers and baluns - it seems every amateur radio operator can wind them, but I never get it, so I decided to do some experiments.

I etched a single-layer FR4 test fixture with some holes and two SMA connectors, so I can solder different components and see how it measures. My methodology is basically: (1) solder a transformer on it, (2) solder a resistor dummy load, (3) do a one-port S11 measurement to find the return loss (or SWR), and keep tweaking it and repeats the steps. If the results look reasonable, I assume the transformer is doing impedance transformation correctly, then I can wind another identical transformer, connect them back-to-back, remove the resistor, and do a two-port S21 measurement to find out the insertion loss.

It worked fine at HF. But I started to get strange results when I move into VHF (output transformer needed for an upconverter, and it's only 30-60 MHz). As a sanity check, I removed everything from the test fixture, leaving only the SMA connector, and a 3200R metal film resistor and did a single-port measurement. NanoVNA reports the impedance 3200R-411X at 50 kHz as expected, but at 50 MHz, astonishingly, it becomes 542R-1250X (Z=1362R). I think it's why I was getting strange S11 measurement for my transformer - even a resistor doesn't behave like a resistor for me. See the Smith Chart attached.

I've never used a VNA before and I don't know understand the RF black magic here (I thought 100 MHz would be black magic, but apparently even 30 MHz is black magic). What could be the reason of these measurements? Is it a real result, or it's just a bogus artifact of the VNA (or, should a do a SOLT calibration on my test fixture)? Is it the parasitic elements in my test fixture? Parasitic elements of the resistor (but I heard ordinary metal film resistor works up to 400 MHz without seeing significant change of impedance)? Transmission-line effect? I suppose there can be some transmission-line effects, but I didn't expect to see an effect this great.

Update: Case closed. It doesn't work because I shouldn't attempt to do it. A VNA is not designed to work under significant impedance mismatch. If you measure S11 for a mismatched load, the results are bogus.
« Last Edit: October 15, 2019, 09:52:25 pm by niconiconi »
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2852
  • Country: us
If it is a wire-wound resistor, I can totally believe this reading.  The R is low due to interwinding capacitance, and the X is there due to inductance.

See if you can find a real carbon composition (bulk carbon element) resistor and check it.  Even carbon-film and metal-film resistors often have helical grooves in the element that create inductance.

Jon
 

Offline exe

  • Supporter
  • ****
  • Posts: 2647
  • Country: nl
  • self-educated hobbyist
photo of resistor?
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
Note that best accuracy of 50-ohm matched RF instruments is measuring impedances which are as close to 50Ohms as possible. The rest of the questions could be answered by following video:

https://youtu.be/pXjFS2MhuqI?t=1
 

Offline niconiconiTopic starter

  • Frequent Contributor
  • **
  • Posts: 366
  • Country: cn
Here is the resistor.

I think it is a metal film resistor, is it? I purchased them from a random online vendor, the vendor claims it is "precision metal film resistor", 5% or 1%, don't remember. I hope they don't sell wirewound resistors masqueraded as fake metal film resistors in Shenzhen, but please double-check.  :-DD :-DD :-DD...

Also, the board has been reworked for nearly 30 times with a lot solder and nasty paint residue from the magnetic wires. Perhaps these substances are somewhat conductive or capacitive?
« Last Edit: October 15, 2019, 08:38:03 pm by niconiconi »
 

Offline niconiconiTopic starter

  • Frequent Contributor
  • **
  • Posts: 366
  • Country: cn
And yes, the coax has been calibrated with a set of open, short, and 50R load standard before all measurements. Otherwise the capacitance of the coax distorts the measurement so much and creates bogus results. After calibration, a 50R RF dummy load measures as a clear dot on the Smith Chart.

I'll try a SMD resistor later.
 

Offline niconiconiTopic starter

  • Frequent Contributor
  • **
  • Posts: 366
  • Country: cn
If it is a wire-wound resistor, I can totally believe this reading.  The R is low due to interwinding capacitance, and the X is there due to inductance.

Jon

Oops, I wrote the wrong sign. The measurement is actually 542R MINUS 1250X. If it's inductive, I would've accepted it. But you see, on the Smith Chart it's clearly CAPACITIVE. It's why I suspect the results are completely bogus. Perhaps it's out of calibration due to the effects of the fixture, or perhaps this VNA is not even designed to measure anything two order-of-magnitude greater load than 50R... Or there is a hidden capacitor somewhere I cannot find.
« Last Edit: October 15, 2019, 09:01:36 pm by niconiconi »
 

Offline niconiconiTopic starter

  • Frequent Contributor
  • **
  • Posts: 366
  • Country: cn
I did not have 50R SMD resistors in the partbin for now, so I tried more THT resistors in my collections. First I tried a 47R resistor, the VNA shows pretty accurate and flat result throughout the spectrum. I later tried a 910R resistor, at 50 kHz it's accurate, but at 50 MHz, the real part becomes 693R and capacitive. Previous strange results during transformer testing was probably an isolated issue due to forgetting to calibrate the coax, not related to this.

Anyway, I think... Case closed. I learned it the hard way: One should not attempt to measure S11 using a VNA if there's a significant impedance mismatch. It's a stupid thing to do, because the result will be completely unreliable. Trap for young players!  :palm:
« Last Edit: October 15, 2019, 09:53:32 pm by niconiconi »
 
The following users thanked this post: MagicSmoker

Online Bud

  • Super Contributor
  • ***
  • Posts: 7276
  • Country: ca
You got it correct, none of the VNAs would accurately measure such large impedance.
Facebook-free life and Rigol-free shack.
 

Offline virtualparticles

  • Regular Contributor
  • *
  • Posts: 149
  • Country: us
au contraire!! One can make extremely acccurate impedance measurements with a VNA. See "Make Accurate Impedance Measurements Using a VNA", Microwaves and RF, June 21st, 2019 written by yours truly.
 
The following users thanked this post: sab, hendorog

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
au contraire!! One can make extremely acccurate impedance measurements with a VNA. See "Make Accurate Impedance Measurements Using a VNA", Microwaves and RF, June 21st, 2019 written by yours truly.

Thank you sir, that is what the crowd has come to see :)

Link here:
http://www.clarke.com.au/pdf/CMT_Accurate_Measurements_VNA.pdf
 
The following users thanked this post: virtualparticles

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
au contraire!! One can make extremely acccurate impedance measurements with a VNA. See "Make Accurate Impedance Measurements Using a VNA", Microwaves and RF, June 21st, 2019 written by yours truly.
Thank you sir, that is what the crowd has come to see :)
Link here:
http://www.clarke.com.au/pdf/CMT_Accurate_Measurements_VNA.pdf
Note that said accuracy can be achieved with CMT VNA that has whooping 140dB dynamic range and very good reflection measurement accuracy. NanoVNA is far from that. Actually it would be good to have similar impedance error chart for nanoVNA. Keysight impedance measurement handbook is another good paper to read for VNA beginners:

https://literature.cdn.keysight.com/litweb/pdf/5950-3000.pdf
 

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
au contraire!! One can make extremely acccurate impedance measurements with a VNA. See "Make Accurate Impedance Measurements Using a VNA", Microwaves and RF, June 21st, 2019 written by yours truly.
Thank you sir, that is what the crowd has come to see :)
Link here:
http://www.clarke.com.au/pdf/CMT_Accurate_Measurements_VNA.pdf
Note that said accuracy can be achieved with CMT VNA that has whooping 140dB dynamic range and very good reflection measurement accuracy. NanoVNA is far from that. Actually it would be good to have similar impedance error chart for nanoVNA. Keysight impedance measurement handbook is another good paper to read for VNA beginners:

https://literature.cdn.keysight.com/litweb/pdf/5950-3000.pdf


I think the point of the document is that using the correct method makes an enormous difference to the accuracy which can be obtained - whatever VNA used.

I agree it would be good to get a similar chart for the nano, but I expect it is going to be very frequency dependent in the nano. So picking a limited frequency range will likely be needed.

No doubt a CMT VNA is way better than the Nano, but the paper is based on the CMT datasheet spec, which I assume is worst case over the entire span - all the way to 6.5GHz.
The Nano even in its fundamental range below 300MHz will still be worse, but maybe at or below 10 MHz it might not be terrible in comparison - as we are comparing it at 10MHz to the worst cast CMT VNA spec at 6.5GHz. No maths was harmed while writing in this sentence, so feel free...

There is still fixture removal maths and a common mode transformer required for series measurement. Also I assume the lack of 2 port calibration will be an issue. Maybe 2 port 1 path?
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
I agree it would be good to get a similar chart for the nano, but I expect it is going to be very frequency dependent in the nano. So picking a limited frequency range will likely be needed.
Indeed. NanoVNA have many ranges. The higher the range the worse performance: 1~100MHz, 100~300MHz, 300~450MHz, 450~900MHz.
Quote
There is still fixture removal maths and a common mode transformer required for series measurement. Also I assume the lack of 2 port calibration will be an issue. Maybe 2 port 1 path?
Everything is there already. NanoVNA have both, fixture removal (as delay setting) and full SOLT calibration for S11 and S21. Note that NanoVNA do not have hardware for S22 and S12, accordingly no calibration for those.
 

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
I agree it would be good to get a similar chart for the nano, but I expect it is going to be very frequency dependent in the nano. So picking a limited frequency range will likely be needed.
Indeed. NanoVNA have many ranges. The higher the range the worse performance: 1~100MHz, 100~300MHz, 300~450MHz, 450~900MHz.
Quote
There is still fixture removal maths and a common mode transformer required for series measurement. Also I assume the lack of 2 port calibration will be an issue. Maybe 2 port 1 path?
Everything is there already. NanoVNA have both, fixture removal (as delay setting) and full SOLT calibration for S11 and S21. Note that NanoVNA do not have hardware for S22 and S12, accordingly no calibration for those.

I think there is a bit more to it:
Fixture removal is more than just fixing the delay - as described in the paper. Maybe that is good enough, but it will add to the error.

The need for the common mode transformer is also mentioned in the paper. I don't know how much error that will add either.

The SOLT calibration on the nano, and in the software I've seen for it, is just a response cal on the through. It doesn't correct for load match or source match. That will increase the error for series measurements. There is an improved cal type called enhanced response which improves this, but still doesn't correct for the load match of the CH1 port. You can use an attenuator to improve that match, and that will degrade the dynamic range.
http://literature.cdn.keysight.com/litweb/pdf/5965-7709E.pdf

There is another type of calibration which was implemented on the 8753 + T/R test set called 1 path 2 port, which requires the DUT be manually reversed. I think that cal type can be almost as good as a full 2 port cal if the switching hardware existed in the nano.

I'm not sure if I am up to working out the maths on these errors...

By pulling the data from the nano and using the calibrations available in scikit-rf or equivalent, the maths issues here could be resolved. Still need to build or buy a balun though.


 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
The SOLT calibration on the nano, and in the software I've seen for it, is just a response cal on the through. It doesn't correct for load match or source match.

Nano does full phase/magnitude SOLT calibration for both reflected S11 and thru S21 and not as you say "just a response cal on the through". Please explain what you mean by saying so and why you think it is so.
 

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
The SOLT calibration on the nano, and in the software I've seen for it, is just a response cal on the through. It doesn't correct for load match or source match.

Nano does full phase/magnitude SOLT calibration for both reflected S11 and thru S21 and not as you say "just a response cal on the through". Please explain what you mean by saying so and why you think it is so.

Ok you have prompted me to have a closer look at the code and the nano is a bit better than I thought. It isn't what I would call 'Full calibration' though:

Where I'm coming from is, in order of accuracy:
1. Response calibration is the lowest form of through calibration. This is just a normalisation and corrects for Transmission tracking.
2. * Enhanced Response calibration. Uses the OSL information to correct for source match as well. It doesn't correct for load match. You can use an attenuator to improve load match.
3. 1 path 2 port calibration. I _think this should_ correct for everything and be almost as good as a Full 2 port cal, but requires a manual DUT reversal.
4. Full 2 port calibration, the ultimate, but it is not possible on the nano.

* I now think the Nano implements Enhanced Response cal. Whereas previously I thought it was just a Response cal. So one step better than I thought. Which is good.

I'm not sure if this is widely known. I haven't seen this laid out anywhere.

Here is a page on ER and Response cal:
Response:
http://ena.support.keysight.com/e5071c/manuals/webhelp/eng/measurement/calibration/basic_calibrations/thru_response_calibration_transmission_test.htm
Enhanced Response:
http://ena.support.keysight.com/e5071c/manuals/webhelp/eng/measurement/calibration/basic_calibrations/enhanced_response_calibration.htm

Edit: all of the above is related only to the _through_ measurement - i.e. S21. Perhaps that wasn't clear either.
S11 measurements are based on the OSL cal. That is fine as long as you disconnect two port devices and terminate them with a good load. Otherwise the measurement will be influenced by the nano not correcting for Load match.

« Last Edit: October 19, 2019, 03:26:57 am by hendorog »
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
3. 1 path 2 port calibration. I _think this should_ correct for everything and be almost as good as a Full 2 port cal, but requires a manual DUT reversal.
No. You are drifting away from topic. Accurate impedance measurements mentioned in the paper you linked uses only S11 and S21. As you already find out, Nano does both.

Quote
Edit: all of the above is related only to the _through_ measurement - i.e. S21. Perhaps that wasn't clear either.
S11 measurements are based on the OSL cal.
Obviously. "T" letter in the SOLT abbreviation means "through". And S11 do not  have "through" because it is single port :D

Quote
That is fine as long as you disconnect two port devices and terminate them with a good load. Otherwise the measurement will be influenced by the nano not correcting for Load match.
Terminating open ports of DUT is impedance measurement basics, but thank you for reminding - perhaps someone will benefit from this info.
 

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
3. 1 path 2 port calibration. I _think this should_ correct for everything and be almost as good as a Full 2 port cal, but requires a manual DUT reversal.
No. You are drifting away from topic. Accurate impedance measurements mentioned in the paper you linked uses only S11 and S21. As you already find out, Nano does both.

Nope. The point is valid. The link I referenced doesn't mention that cal type, which is why I pointed that out as a separate item.
It is a cal type implemented on the 8753. The 8753 supported an 85044 T/R test set.

Actually being able to measure S11 and S21 isn't the point. The point is how accurately you can measure them. That accuracy is dependant on the type of cal you do. This entire discussion is about how to measure various impedances accurately is it not?
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
3. 1 path 2 port calibration. I _think this should_ correct for everything and be almost as good as a Full 2 port cal, but requires a manual DUT reversal.
No. You are drifting away from topic. Accurate impedance measurements mentioned in the paper you linked uses only S11 and S21. As you already find out, Nano does both.

Nope. The point is valid. The link I referenced doesn't mention that cal type, which is why I pointed that out as a separate item.
It is a cal type implemented on the 8753. The 8753 supported an 85044 T/R test set.

Actually being able to measure S11 and S21 isn't the point. The point is how accurately you can measure them. That accuracy is dependant on the type of cal you do. This entire discussion is about how to measure various impedances accurately is it not?

I did mean that NanoVNA S21 and S11 measurement and calibration capabilities are enough to perform measurements described in the article. I suspect that you think that S21 thru calibration do not measure/calibrate Port2 impedance. Well, it does. BTW components such as resistor & inductor (some other components as well) are "single port devices" that do not need to be rotated during impedance measurements.
« Last Edit: October 19, 2019, 12:04:02 pm by ogden »
 

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
3. 1 path 2 port calibration. I _think this should_ correct for everything and be almost as good as a Full 2 port cal, but requires a manual DUT reversal.
No. You are drifting away from topic. Accurate impedance measurements mentioned in the paper you linked uses only S11 and S21. As you already find out, Nano does both.

Nope. The point is valid. The link I referenced doesn't mention that cal type, which is why I pointed that out as a separate item.
It is a cal type implemented on the 8753. The 8753 supported an 85044 T/R test set.

Actually being able to measure S11 and S21 isn't the point. The point is how accurately you can measure them. That accuracy is dependant on the type of cal you do. This entire discussion is about how to measure various impedances accurately is it not?

I did mean that NanoVNA S21 and S11 measurement and calibration capabilities are enough to perform measurements described in the article. I suspect that you think that S21 thru calibration do not measure/calibrate Port2 impedance. Well, it does. BTW components such as resistor & inductor (some other components as well) are "single port devices" that do not need to be rotated during impedance measurements.

I give up. We are miles apart in understanding here.
 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
I give up. We are miles apart in understanding here.

Don't give up. Explain. In more details than  :blah: :blah:   like:

Actually being able to measure S11 and S21 isn't the point. The point is how accurately you can measure them. That accuracy is dependant on the type of cal you do. This entire discussion is about how to measure various impedances accurately is it not?
 

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
I give up. We are miles apart in understanding here.

Don't give up. Explain. In more details than  :blah: :blah:   like:

Actually being able to measure S11 and S21 isn't the point. The point is how accurately you can measure them. That accuracy is dependant on the type of cal you do. This entire discussion is about how to measure various impedances accurately is it not?

It is entirely explained in the previous post.

If you don't understand, then ask specific questions and I'll answer them - within my own capabilities to do so.




 

Offline ogden

  • Super Contributor
  • ***
  • Posts: 3731
  • Country: lv
It is entirely explained in the previous post.

Your previous post was something like "I give up because I am way smarter".  :-// Besides your arrogant attitude, that does not explain much. So please be so kind and either provide link to post or even better: full quote. Then I will continue with questions.
 

Offline hendorog

  • Super Contributor
  • ***
  • Posts: 1635
  • Country: nz
It is entirely explained in the previous post.

Your previous post was something like "I give up because I am way smarter".  :-// Besides your arrogant attitude, that does not explain much. So please be so kind and either provide link to post or even better: full quote. Then I will continue with questions.

Look, you are making comments like: "I suspect that you think that S21 thru calibration do not measure/calibrate Port2 impedance. Well, it does."

I have repeatedly explained the calibration options available for the port 2 calibration, but you say something like that.

Then:
"BTW components such as resistor & inductor (some other components as well) are "single port devices" that do not need to be rotated during impedance measurements."

Sarcastic much? They are only single port if measured in S11 shunt. The pdf which Brian (virtualparticles) wrote explains in detail that there are different methods. There are even pictures on the first page. Then there is a chart which helps us decide which method to use. It's gold.

You just don't appear to be interested in learning, all I am seeing from you is a desire to argue.
 
The following users thanked this post: Jacon


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf