NanoVNA Custom Software

[edigi]

Confirmed. Measurement is affected when battery is charging.
When charge LED stops flashing, measurement error disappears.

I tried to measure voltage between GND and VDD on ST-LINK connector.

When powered from the battery: 3.3350 V
When external power connected: 3.3306 V
After some time with ext. power: 3.3325 V (measurement error is still present)
When charge LED stops flashing: 3.3339 V (measurement error disappears)

My NanoVNA creates zig-zags on the display between 295 and 300 MHz when the battery is low or charging. Not a surprise that it's sensitive to supply voltage in that frequency range, as per datasheet the clock chip supposed to operate till 200 MHz...
I heavily suspect that the random lines occasionally appearing is also not just a drawing error but a consequence of this.

As for the torque wrench I don't think that the low frequency range and precision of NanoVNA justifies it (it's more like a waste of money and time for this level), the SMA connector can be tightened without it quite well but NanoVNA is just the perfect tool to check this.

Replacing the MCU is an interesting idea. What would be even more interesting if clock chip could be replaced with something that allows operation up till 2-3 GHz range (means mixer replacement as well...). Probably the resistor bridge would not cut it till 6 GHz but still if clocking (test signal generation) could be improved it would be a significant improvement (sure, not in this price range as significantly more expensive chips would need to be used...).
Note: When there is integer relationship between the clock source of the clock chip and its output signal is quite OK, but when I set 800MHz it's not so clean anymore (it uses 3rd harmonic, so in reality chip is set to 266.6...MHz or so)..

[radiolistener]

My NanoVNA creates zig-zags on the display between 295 and 300 MHz when the battery is low or charging.

Is your NanoVNA has these zig-zag spikes at 300 MHz with low battery only? 
My one always have these 300 MHz spikes... Sometimes it higher, sometimes lower, but it always here...

Here is CH0 and CH1 both terminated with 50 Ohm:


What is your CH1 SWR at 900 MHz?  (just connect CH0 to CH1 with good cable to measure it)

My CH1 SWR is 1.31. Measured with 15 cm RG405:


Cable delay is 1550 ps. I entered it into ELECTRICAL DELAY, and it shows Rs = 41 Ohm at 900 Mhz...

[Flynt]

joeqsmith,i Sent You a PVT Message....

TANKYOU VERY MUCH

[joeqsmith]

As long as you hold the torque wrench somewhere along the actual handle (i.e., the red part shown here, using the model of torque wrench I have here as an example) you will be able to deliver the specified torque or something very close to it.  It (should) go without saying that if you grab the wrench just behind the knuckle, you can apply pretty much whatever force you want in whatever direction you want. 

Makes sense to me - as long as it's designed for that usage, so that it stays it in the torque range its spec'd for - or at least not far enough out of range to break stuff if you aren't doing critical stuff.

Which if you read my response, it pretty much what I had stated. 

Looking at the error of the torque wrench shown, using a knife blade and going from one end of the handle to the other, the variance is about 1" lb. Still well within spec no matter where you hold it.

  Obviously the torque is dependent on the position but my question is how much do different wrenches vary.  I assume many of you have them and would know or at least be able to measure them within some degree of accuracy.   

[joeqsmith]

joeqsmith,i Sent You a PVT Message....

TANKYOU VERY MUCH

I appreciate your enthusiasm for the Nano and your wanting to try other software.  As you were told by _Wim_ and I have stated a few times in this thread, mentioned in my videos and answered numerous time in the comments, the answer is the same. 

Your persisting to SPAM my inbox after clearlying being told no, is part of the reason I won't release it.  I just don't have the time to support it.   

***
As _Wim_ mentioned, you should look at some of the other software being offered here.    It seems that they are off to a very good start and could use some encouragement. 

[Flynt]

 

joeqsmith, I apologize, for what you call "SPAM", but I didn't get any of your Rejection messages, and as I wrote you, I don't read English well and I can't program ... I just needed your software, to work and I was even willing to give you a donation.

I can't use a Translator to convert 10 pages of trhead ...

Okay it does nothing, I will use the original software.

Thanks anyway, even if it didn't cost you anything send it to me ...

[_Wim_]

It seems that CH1 on my NanoVNA has not so good VSWR, is it ok?
(Attachment Link) (Attachment Link)

I think this is caused by the input impedance of channel 2 in your method, not by channel 1. If you connect your calibration load to ch1 (and preferably calibrate the NanoVNA with a smaller bandtwidt, for example 800M to 900M) you can achieve a VSWR or 1.02 (S11 of -40dB) for 800M<->900M, and even better for lower frequencies.

[edigi]

My CH1 SWR is 1.31. Measured with 15 cm RG405:

It's roughly the same (1.33) here.
For the price though there is still nothing to complain.

_Wim_
NanoVNA has CH0 and CH1 and what is meant is what you describe (just naming is different).

[_Wim_]

_Wim_
NanoVNA has CH0 and CH1 and what is meant is what you describe (just naming is different).

 

Mine is indeed also simular than. When calibrated with a 100Mhz span it drops a little to 1.28.

[hendorog]

As long as you hold the torque wrench somewhere along the actual handle (i.e., the red part shown here, using the model of torque wrench I have here as an example) you will be able to deliver the specified torque or something very close to it.  It (should) go without saying that if you grab the wrench just behind the knuckle, you can apply pretty much whatever force you want in whatever direction you want. 

Makes sense to me - as long as it's designed for that usage, so that it stays it in the torque range its spec'd for - or at least not far enough out of range to break stuff if you aren't doing critical stuff.

Which if you read my response, it pretty much what I had stated. 

Looking at the error of the torque wrench shown, using a knife blade and going from one end of the handle to the other, the variance is about 1" lb. Still well within spec no matter where you hold it.

  Obviously the torque is dependent on the position but my question is how much do different wrenches vary.  I assume many of you have them and would know or at least be able to measure them within some degree of accuracy.   

Cool, I was responding to KE5FX, no sorry I didn't see your response or question at the time.

Another option is to work that effect out just from dimensions of the wrenches which might be easier info to get from people.

[radiolistener]

joeqsmith, it seems that you have professional vector analyzer and other equipment.

Could you please measure CH1 input impedance of NanoVNA?
Two sweeps for 1-300 MHz and for 1 MHz - 1.5 GHz will be nice to see how it can be used with direct mode (1-300 MHz) and with extended mode (1-1.5 GHz).

It seems that CH0 has pretty good performance, but CH1 going to SWR=1.32 at 900 MHz which may affect measurements. What is your thoughts about it?

[joeqsmith]

joeqsmith, I apologize, for what you call "SPAM", but I didn't get any of your Rejection messages, and as I wrote you, I don't read English well and I can't program ... I just needed your software, to work and I was even willing to give you a donation.

I can't use a Translator to convert 10 pages of trhead ...

Okay it does nothing, I will use the original software.

Thanks anyway, even if it didn't cost you anything send it to me ...

I just assumed from your acknowledgment to _Wim_ that you understood what they wrote and did not require any further response from me.

Hello Flynt, for the moment this software is not available for download. It was developed by joeqsmith to show what is possible with the Nanovna, to make some educational youtube video's about the NanoVNA and to inspire others to make apps of their own (which has worked already, thanks to 5q5r)

I do not ask for donations, do not enable ads,  have no Pateron's, nor do I beg people to join my channel.   

Suggesting that my making the software available to you wouldn't cost me anything is a naive view.   Considering that you required a personal response from me even though someone else had provided you with an answer, I doubt that by handing you the software that your questions would subside.   Even if I made it free, with no support implied, I suspect you and others would request support.   I would hope you are not suggesting that my time is worth nothing.     

Again, I suggest you support the people who are willing to invest their time to to create an open source platform. 

[joeqsmith]

joeqsmith, it seems that you have professional vector analyzer and other equipment.

Could you please measure CH1 input impedance of NanoVNA?
Two sweeps for 1-300 MHz and for 1 MHz - 1.5 GHz will be nice to see how it can be used with direct mode (1-300 MHz) and with extended mode (1-1.5 GHz).

It seems that CH0 has pretty good performance, but CH1 going to SWR=1.32 at 900 MHz which may affect measurements. What is your thoughts about it?

I suspect that you are correct that the match may not be all the great.  I mentioned that in the last video where I show that T-check.  But it's just $50 and I can't really bitch about it at that price.   

Sadly, I just don't use a VNA enough in my electronics hobby to justify getting a newer system.  Now if Keysight, RS or Copper Mountain wanted to donate a new 2-port 6GHz system, I may have to change my stance on donations.      Imagine the fun one could have with a modern higher end VNA..

[hendorog]

joeqsmith, it seems that you have professional vector analyzer and other equipment.

Could you please measure CH1 input impedance of NanoVNA?
Two sweeps for 1-300 MHz and for 1 MHz - 1.5 GHz will be nice to see how it can be used with direct mode (1-300 MHz) and with extended mode (1-1.5 GHz).

It seems that CH0 has pretty good performance, but CH1 going to SWR=1.32 at 900 MHz which may affect measurements. What is your thoughts about it?

This app note has good info on the effect that Port 2 (Port 1 on the Nano) load match has on measurements. It can usually be mitigated with a good attenuator, at the cost of some dynamic range.
http://anlage.umd.edu/Microwave%20Measurements%20for%20Personal%20Web%20Site/5965-7709E.pdf

[radiolistener]

Just tested, with 10 dB attenuator it is usable up to 200 MHz (SWR=1.01). With 20 dB attenuator up to 700 MHz (SWR=1.03).

But it reduces dynamic range... Is it possible to fix CH1 input to get swr=1.00 up to 300 MHz?

Currently it drops down to 45 Ohm at 300 MHz.

[rhb]

I had mentioned in the last video I made about the limited resolution (distance) when making TDR measurements.    I talked a bit about how you may be able improve it but with it being a $50 unit, I don't see much of a point.   Well except for the pure fun of it.....

So I talked it over with a friend of mine and we had a good laugh about the whole project.  He offered one idea that should help improve the setup but it's all a guess if the Nano can actually be used this way and get any useful information from it.   Of course, cost will need to be ignored which is part of the humor.... 
 
First baby step,  modify my software to support the range extender.   It may not look like anything is going on in the attached plot, but what you see the Nano using Channel 1 for S11.  That's not a misprint.   Granted, it's only 100MHz.       Like I said, baby step.   

***
Should mention, yes, it is indeed a 100 ohm resistor.

A distance resolution of 1 mm requires timing to 5 ps, 10 ps for a reflection.  At 900 MHz 1 degree is 3 ps.  So if the nanoVNA reads phase to even 10 bits of accuracy, 1 mm resolution for a broadband reflection is not a problem.  Just do a linear fit to the phase in the frequency domain.

It gets more complicated if there are multiple reflections of different bandwidths, however, a sparse L1 aka basis pursuit can reliably solve that.  Though it might take a fairly powerful desktop to do that in less than several minutes.

I'm trying to squeeze TDR into the nanoVNA MCU with a swap out for an STM32F303CCT6 part with 48 KB of RAM and 256 KB of flash.  That's a completely different can of worms.

BTW What do you mean by "range extender"  is this frequency or time?

Reg

[hendorog]

Just tested, with 10 dB attenuator it is usable up to 200 MHz (SWR=1.01). With 20 dB attenuator up to 700 MHz (SWR=1.03).

But it reduces dynamic range... Is it possible to fix CH1 input to get swr=1.00 up to 300 MHz?

Currently it drops down to 45 Ohm at 300 MHz.

When you say usable, I am not sure if you are talking about the effect of that SWR on the error corrected measurement, or just the SWR of the port itself.
SWR of 1.03 is 30dB RL, which is outstandingly good and will have a very minor effect. In the doc I posted they are talking about improving 18dB RL for the VNA port - which is about 1.3 SWR. And remember that was an HP document, so a high value instrument, and we are talking about a $50 device here.

The OP posted earlier about his 'range extender' and showed an S11 measurement on port 1 and he has posted before about a transfer switch.
I don't know exactly what he has there, but it is pretty interesting. If it still permits capturing S11 on port 0, then a full 2 port cal is possible which does correct for load match.

There may be other mathematical approaches, such as 1 port 2 path, which might be able to cancel out some or all of that error, I don't know for sure if they do.
Here is an implementation, which is part of the lib I have been using. It does say '...can be fully corrected...'
https://scikit-rf.readthedocs.io/en/latest/api/calibration/generated/skrf.calibration.calibration.TwoPortOnePath.html#skrf.calibration.calibration.TwoPortOnePath

[radiolistener]

Is it possible to calculate chart like this from S1P data (magnitude and phase per frequency)?



What is the math behind it?

[radiolistener]

When you say usable, I am not sure if you are talking about the effect of that SWR on the error corrected measurement, or just the SWR of the port itself.
SWR of 1.03 is 30dB RL, which is outstandingly good and will have a very minor effect. In the doc I posted they are talking about improving 18dB RL for the VNA port - which is about 1.3 SWR. And remember that was an HP document, so a high value instrument, and we are talking about a $50 device here.

I'm not sure what SWR for VNA input port is acceptable. But as I know it may affect filter response measurement, because filter is intended for 50 Ohm termination.

My primary interest in the frequency range 1...500 MHz. My NanoVNA CH1 (second port for S21) has RL = 23 dB at 500 MHz (SWR = 1.15). Is it acceptable for filters measurement? What error level it may cause? (for filter measurement S11, S21)

[hendorog]

Is it possible to calculate chart like this from S1P data (magnitude and phase per frequency)?

(Attachment Link)

What is the math behind it?

The maths is an inverse FFT.

rhb is a guru in that subject. I'm not, I just use a library. That makes it a one line operation. Example here:
https://github.com/hendorog/nanovna_test/blob/master/NanoVNA%20TDR.ipynb

One line at a time:

Code: [Select]

#Load a directory full of s1p files
duts = rf.read_all('data/measured/')

#Select one of them
s11 = duts['nano calibrated attenuator attempt 2']

#Convert from freq domain to time domain
s11_gated = s11.time_gate(center=0, span=100)

# Plot frequency and time-domain
figure(figsize=(8,4))
subplot(121)
s11.plot_s_db()
s11_gated.plot_s_db()
title('Frequency Domain')

subplot(122)
s11.plot_s_db_time()
s11_gated.plot_s_db_time()
title('Time Domain')
tight_layout()


When you say usable, I am not sure if you are talking about the effect of that SWR on the error corrected measurement, or just the SWR of the port itself.
SWR of 1.03 is 30dB RL, which is outstandingly good and will have a very minor effect. In the doc I posted they are talking about improving 18dB RL for the VNA port - which is about 1.3 SWR. And remember that was an HP document, so a high value instrument, and we are talking about a $50 device here.

I'm not sure what SWR for VNA input port is acceptable. But as I know it may affect filter response measurement, because filter is intended for 50 Ohm termination.

My primary interest in the frequency range 1...500 MHz. My NanoVNA CH1 (second port for S21) has RL = 23 dB at 500 MHz (SWR = 1.15). Is it acceptable for filters measurement? What error level it may cause? (for filter measurement S11, S21)

See the app note I attached earlier, it shows examples of measuring a filter, and shows roughly what your uncertainties will be. Your CH1 RL is a bit better than what they use in their example, but you are in the ball park.

I am not 100% sure on this, but I think that the built in NanoVNA transmission calibration is just a normalisation - i.e 'Response Calibration'.

Using a PC and the library I used in the examples above, you could improve that by doing an Enhanced Response Cal. That corrects for source match error (which is mismatch on Port 0). Mismatch on Port 1 remains uncorrected.

Then you could improve on that further using an attenuator to improve the Port 1 load match, but you sacrifice a bit of dynamic range as you know.

To improve further on that you need a two port calibration.

[joeqsmith]

I had mentioned in the last video I made about the limited resolution (distance) when making TDR measurements.    I talked a bit about how you may be able improve it but with it being a $50 unit, I don't see much of a point.   Well except for the pure fun of it.....

So I talked it over with a friend of mine and we had a good laugh about the whole project.  He offered one idea that should help improve the setup but it's all a guess if the Nano can actually be used this way and get any useful information from it.   Of course, cost will need to be ignored which is part of the humor.... 
 
First baby step,  modify my software to support the range extender.   It may not look like anything is going on in the attached plot, but what you see the Nano using Channel 1 for S11.  That's not a misprint.   Granted, it's only 100MHz.       Like I said, baby step.   

***
Should mention, yes, it is indeed a 100 ohm resistor.

A distance resolution of 1 mm requires timing to 5 ps, 10 ps for a reflection.  At 900 MHz 1 degree is 3 ps.  So if the nanoVNA reads phase to even 10 bits of accuracy, 1 mm resolution for a broadband reflection is not a problem.  Just do a linear fit to the phase in the frequency domain.

It gets more complicated if there are multiple reflections of different bandwidths, however, a sparse L1 aka basis pursuit can reliably solve that.  Though it might take a fairly powerful desktop to do that in less than several minutes.

I'm trying to squeeze TDR into the nanoVNA MCU with a swap out for an STM32F303CCT6 part with 48 KB of RAM and 256 KB of flash.  That's a completely different can of worms.

BTW What do you mean by "range extender"  is this frequency or time?

Reg

Frequency (with a lot of issues).  It would be for a very narrow band experiment if I do anything with it. 

Starting on page 29:
http://literature.cdn.keysight.com/litweb/pdf/5989-5723EN.pdf

[rhb]

That's the Raleigh criterion for separating two events.  I was referring to identifying the location of a discontinuity.  Very different problems.

I've viewed the TDR aspect as "Where is the bad spot in my cable?"  and "Where is the impedance discontinuity in my PCB trace?" 

Once you want to look at the reflection from the the SMA side of an SMA to N connector separately from the N side, you have to have 20 GHz or more BW.  And even at 20 GHz in the time domain, it's not easy.  I've been doing a lot of it testing RF connectors using a Tek 11801 & SD-24.

However, I'm fairly certain that, if you have a well defined problem such as the mismatch at either end of an SMA to N adapter, you can solve the problem of the timing to better than the Rayleigh criterion using basis pursuit. However, doing this would require mastering the ugliest and most complex math I ever came across in my life.  It's simple to do in practice, but the logic of how and why it works is painful.  David Donoho has a  proof of a single theorem which takes 15 pages!

I'd like to note that I have serious technical issues with that application note.  It poses a lot of "problems" which are trivial to solve, but presented as proprietary to HPAK despite being sufficiently well known to be basic DSP 101 examples and homework.  I've not read all of it, but much of it is in Dunsmore's book which I *have* studied very closely.  At present I don't know if the issue is terminology and notational conventions, substantive errors or marketing FUD.    While the reflection seismology community has been doing DSP for 20 years longer than the EE community, the EE community chose not use the mathematics community lexicon adopted in geophysics.  Same words, but *very* different meanings.  Having worked in several seismic processing shops, my experience is that it takes about 6 months to map the words people use at lunch to the correct equations.  If it's that difficult in geophysics, going between that and electronics engineering is *much* worse.

In sum, this quickly turns into a major communications problem.  I've seen a lot of arguments which amounted to nothing more than different meanings for the same words.

Have Fun!
Reg

[joeqsmith]

Have Fun!

With no fear of doing something so bad that I would need to mortgage the house to recover, it opens the doors for a major fun fest. 

Playing with the Nano now, up converting to 5GHz  and back down.  Oddly enough, it seems to sort of do something.    Using the home made standards to run the SOL, then attaching the 100ohm load.   

I can't see it being useful in anyway outside of the fun factor. 

Have fun with your soldering and coding.   It sounds like a fun project as well.   

[Bud]

I'm not sure what SWR for VNA input port is acceptable. But as I know it may affect filter response measurement, because filter is intended for 50 Ohm termination.

My primary interest in the frequency range 1...500 MHz. My NanoVNA CH1 (second port for S21) has RL = 23 dB at 500 MHz (SWR = 1.15). Is it acceptable for filters measurement? What error level it may cause? (for filter measurement S11, S21)

Google for "mismatch uncertainty".

[radiolistener]

The maths is an inverse FFT.

rhb is a guru in that subject. I'm not, I just use a library. That makes it a one line operation. Example here:
https://github.com/hendorog/nanovna_test/blob/master/NanoVNA%20TDR.ipynb

One line at a time:

Code: [Select]

#Convert from freq domain to time domain
s11_gated = s11.time_gate(center=0, span=100)


Is this python? Unfortunately I don't familiar with this language. 
Could you please give me the link to the source of that s11.time_gate() function?