NanoVNA Custom Software

[joeqsmith]

I looked at the LiteVNA, and remember seeing that it was harmonics based for the upper frequencies and that the nano VNA plus4 series used fundamental frequencies. That turned out to be the closing point for me.

I'm at a loss on your closing point.  So the V2Plus4 is limited to 4GHzish  without harmonics.  It also has no support to use harmonics beyond that limit.   The LiteVNA is limited to 6GHz without harmonics and has support to use them up around 8GHzish.  Note in my previous graph, you can see the LiteVNA using harmonics to 10GHz (not very useful) and the sharp increase at 6GHz where it changes to use them.    Note the low frequency range where the V2Plus4 is very poor.  There is only a small range where my V2Plus4 will outperform my LiteVNA. 

If you followed my waveguide experiments, you know while I did show the LiteVNA directly making measurements in the X-band using harmonics,  to improve the performance I had put together a frequency extender that works to a bit over 12GHz.

I then decide to pay the extra nickel for the pro version because I eventually want to create some crystal filters, and thought somehow that the other VNA versions lacked the capability to do that. Perhaps, being a new uneducated initiate of nanoVNAs, I placed too much emphasis on the marketing copy there.

I've posted many times how the original NanoVNA outperforms all my others when measuring crystal filters.  There's a reason I tell people if you work below 300MHz, that is the best choice.   Above 300MHz, the LiteVNA.  Under $200 total. 

GitHub is really for application development teams and individuals who compile applications from source code. For people who just want to grab a pre-compiled binary and go it is a good bit of a hassle. Have you had a look at Sourceforge?

No.  Only Github.  Easy for me and I suspect for most of the youth who grew up with PCs.  Nearly impossible for the old hams, even with explicit instructions.   They need something that works like their PC file manager.   Google Drive, Microsoft One Drive, Dropbox.   Of course we do have a few special people that try and use their cell phones.

**********
Do us a favor.  Calibrate your V2Plus2 Pro using say 800 or more data points over the entire specified frequency range.   Attach the load to port1 and terminate ports2.  Measure S11 and save it to a Touchstone file.    Without changing any of the setup, measure S21 and save it to a Touchstone file.  Use what ever software they are suggesting to make these measurements (or run it stand alone).   Document the firmware you are using and how you have the VNA powered.   If you don't mind, maybe run it powered off your PC as well as with a battery.  It can make a big difference.     Post the Touchstone files here and I will add them to my graph comparing the VNAs. 

[joeqsmith]

I didn't remember my V2Plus4 being that poor of performance at the lower end so I charged up a battery for it and repeated the test.  Results are now much better.   For fun I ran it with and without the leakage term to show the effects at the higher frequencies.     

Basically, it's right on-top of my LiteVNA.  Much more believable.   The V2Plus4 is now $300. The last LiteVNA cost around $130.  The LiteVNA is good to 6GHz without harmonics and maybe 8 with.  The V2plus4 is limited to 4.    The LiteVNA also has the SD card interface.  I've tried it.   It's also small enough to fit in a shirt pocket for the hams climbing around on their towers...

[joeqsmith]

OWO's site talks about the clones averaging to hit the numbers.  In the previous plot, I had used 10 averages with the LiteVNA.   So for a better comparison, I disabled the averages and used the same set of cables that I used with the V2Plus4.     

Basically, it looks like a wash.  Maybe the LiteVNA is a bit better.

[Fred_B]

Here's a couple of touchstone files. There were made with the battery in the device and the USB cable connected to the PC. The battery was surprisingly difficult to push into the holder when I installed it. When I opened the case up to pull it out, I can see that the holder is bowed out a little bit by the pressure of the springs. I'd need to pry the battery out by sticking something underneath it and prying it up to get it out. I think, I just want to leave it in, until I get some mileage out of it.

I'm still quite unfamiliar with this device, so I hope I got the files done properly. BTW I'm using VNA_Qt. It looks from the display on my PC that my device is doing about 10db worse between 3.5 GHz and 4.0 GHz. That's interesting.

Firmware version: 20220301

[joeqsmith]

Fred,

I used a flat-top battery in mine which seems to fit.  Leaving the battery in with the USB connection is how I collected that last data, so perfect.

Looking at your S21 data, notice the header states S11 rather than S21.

# MHz S MA R 50
!   freq        S11       
0.050000  0.00132 -161.92

If I plot your s21 data, the first thing I notice is the magnitude is VERY high.   The next thing is the phase is stable.  Think about it, the two ports are not connected to one another and both are terminated.  Wouldn't you expect it to be random?   

I suspect this is actually is S11 and not S21.   I've attached my S21 data for my V2Plus4.  I expect your Pro to be similar. 

[Fred_B]

I was going to say the S21 data is wrong. I'm still learning out how to use the thing. I have a different capture file with both the s11 s21 in it. I have to wait till I'm back in my lab to upload it to. That will be late tonight.

When I was looking at the graph on the screen, I noticed that everything looked about the same as what's in your graph except at the very end between 3.5 GHz and 4 GHz where my device seemed to jump up about 10 dB to about -60 dB where yours looks like about -70 dB.

[joeqsmith]

Sounds good Fred.  Look forward to it.

When I first received the V2Plus4 I had posted in this thread how poor the performance was in this region.   OWO at that time was still involved and had explained that there was only one program available that handled the leakage correction.   This isn't something that is normally used because it can lead to other problems.   It is possible that the software you are using doesn't enable it by default.   I posted for you the effects of turning it on and off with my software.   Once I have your data, we can see how it compares with my V2Plus4 with the leakage correction disabled. 

[Fred_B]

It seems the only way I can figure out how to get VNA_QT to export a file with S21 in it is like this.

[joeqsmith]

Note how your new file's header is:
# MHz S MA R 50

MA selects the magnitude angle.   Ideally, this would have been in dB or decibel angle.  The software I am using to plot the data does not appear to handle two different formats.   It also doesn't appear to allow a conversion.  No problem, so I converted it with my software.  Then I noticed you have a bunch of 0's in your data.   It's like your software doesn't have enough resolution.  While my software will handle that condition, the viewer will not.    So, to work around the problem, I have to fudge your data to remove the 0s.   Maybe QT has a workaround.  I've never used it.

Attached is showing your Pro compared with my V2Plus4.  Your Pro has slightly lower noise. 

[joeqsmith]

Showing your Pro compared with my LiteVNA, which again, doesn't use harmonics until we move beyond 6.3GHz.   I would say they are a wash as far as which has the lowest noise.  Again, there is no averaging going on as OWO was suggesting. 

[joeqsmith]

From OWO's site:

improved USB protocol allows full speed streaming of measured data to the PC.

I hooked up an antenna to the LiteVNA and decoded a local AM radio station.  Is that full speed streaming?   

Note: beware of clones of our older designs claiming to go to 6GHz but when tested are not able to measure accurately above 1GHz! To make measurements at these higher frequencies requires the entire signal path to be redesigned using RF rated parts and materials, impedances need to be tightly controlled, and all parts must be rated to handle the higher frequency, which adds significant expense. Even open source designs are not able to do it at below $300-500 BOM cost, and still have noticeable drift and repeatability problems.

   I have shown a lot of data off that LiteVNA. 

Many users have compared the NanoVNA V2 Plus4 with professional instruments. When using the same calibration kits, the measured data and smith chart matches well.

Some what true depending on the device we are measuring.  All of the low cost VNAs I have looked at use a squarewave drive.  Video below shows two of my vintage VNAs compared with the low cost ones.  There is no comparison, the low cost ones can't measure anything meaningful.   These problems I am sure are obvious to the professional but maybe not to the amateur radio hobbyist.   

R&L electronics - still selling clones despite being asked not to by the original developers

Achievable performance (V2 Plus4 Pro, calibrated, IFBW=0.8kHz, AVG=20)

Funny as they have talked about the clones using averaging to make their hardware appear better than it is.  You want averages, attached plot compares your Pro with the LiteVNA, 100 averages.  Personally, I normally will take the noise over the speed.   

[Fred_B]

VNA_Qt leaves a lot to be desired. When I captured that the display was using units of dB. I had the two traces set for s11 magnitude and s21 magnitude. In the export menu there were only basic capture and export items without allowing any selection of just what was to be captured or exported beyond Port 1 report 2.

I got a copy of Solver64 and it seems to lacks the ability to control my device. It appears to link up, but it fails to trigger a regular or calibration sweep.

VNAsaver fails to run on my computer because they've updated too python 3.9 which fails to run on Windows 7. So I'm stuck with VNA_Qt, so I can get an external hard drive and load up Windows 10 or Linux distro.

I'll post a screenshot of the VNA_Qt sweep. From that you can see that it's pretty much the same as yours except in the 3.5 to 4.0 GHz range where it looks to be about 10 decibels worse.

[joeqsmith]

I got a copy of Solver64 and it seems to lacks the ability to control my device. It appears to link up, but it fails to trigger a regular or calibration sweep.

Odd as I would have thought it would use the same protocol as the V2Plus4.   I did have a guy write me once who also described a case where he could not get the VNA to sweep with my software.   He sent me some screen shots and it was easy to see that he never turned on the sweeps.       True story.     I have ran into people who would try to run my software without NIVISA installed.  Of course that doesn't work out very well.    Then there was the guy telling me that it wouldn't sweep who was entering m in place of M. 

If you wanted, you could make a video showing everything from turning on your VNA to trying to run it.  Maybe there is something obvious going on.   It's also possible OWO purposely changed the protocol to prevent the use of third party software.   Seems like a dumb idea but you never know.

From that you can see that it's pretty much the same as yours except in the 3.5 to 4.0 GHz range where it looks to be about 10 decibels worse.

From my post showing the leakage correction disabled, looks like my V2Plus4 is about 20dB worse in this upper region.  It's possible that they had improved the hardware so that without the correction, they gained 10dB.   What's odd is they don't allow you to select it and that the Touchstone file doesn't reflect what is shown on your screen. 

[joeqsmith]

I downloaded the latest QT and Saver software.   Both connect to both my LiteVNA and V2Plus4 and I am able to run sweeps and display the data.   I suspect your Pro version uses the same protocol and should just work. 

One thing I noticed is the last time I had tried to use that QT software (Pretty sure this is what it was) it would hang on every press.  It's actually the reason I decided to port my software over to the NanoVNA in the first place because it was so poor.   This new version at least seems to have solved those basic problems.   I tried the export and it does appear there is a bug with the S21 export.   Still your S2P is fine outside of the zero problem.  Again, I suspect another bug in QT and I was not able to find a work around.   Of course, you take the 20 Log 0....   

I have a some error checking in my import function to handle bad Touchstone files and just added another case for the zero and made a note it was to work around a bug in QT.  I tossing out the bad data so you can at least use the good parts.     

[Fred_B]

So, I got solver working somewhat. I and setup and read sweeps, although when I try to do a cal, it prompts me for a full 2 port cal what ever I do. The manual said if there is an absence of any attached transfer relay, solver would just do a 2 port one path cal. All I have connected is the nano. Also on the advanced tab all the plots are void of data. Is that normal?

When I try to export a touchstone file from solver all I seem to be able to do is s11 in a s1p and s21 in a s2p. Is there any way to just export s21 data in to a touchstone file?

Here are some screen shots:

[joeqsmith]

What was the problem?  I'm thinking I have heard them all by now but curious if you have a new method...

You have selected a valid port number for transfer relay,  the software was able to open the port you requested and did not detect the the feedback and set the fault.  The fault has no other effect than to let the operator know there is a problem with their transfer relay.  When you run a calibration, the software runs through all combinations because you have told it you have a transfer relay attached.    The advanced tab will display all four S parameters with the transfer relay installed.

Now, I could have, and may have had an enable feature for the transfer relay at one time.  I chose to set it up this way because when I don't have the transfer relay attached, the port is null and the software knows it is not attached.   Saves me having to change the settings, because I am lazy. 

For exporting, I am not sure if the software will even save both S11&S22 into a file today.  There may be other problems with the export as well. .... Let me restate that.  There may be other problems with the software as well!!!  Again, it's not an end product and changes based on what ever experiments I happen to be running.   You may run into some things you feel are outright bugs that me be intentional for some test I was running.   

The manual is very old and I am sure there are a lot of sections that are no longer valid for the current release. 

[Fred_B]

The problem with the sweep was the port settings. If I set only the first one to the port for the VNA, then it would link but that was all. When I set the first two for the com port for the VNA, and it would link and sweep and do what would be expected, although I stopped for the day at that point before trying to do a cal.  Before I made that last post I went and set all three port settings to the port for the VNA, and everything worked as normal except except for the calibration. So I guess I have to set the one for the transfer relay to a null port and the calibration issue will be solved.

I exported an s2p touchstone file and that appears to have the s21 contained in it.

Hopefully, everything for the crystal measurements will work. I'll build a jig after I finish my current project. I was beginning to feeling mighty disheartened that I paid twice as much as need be for a device that was going to be a problem using solver for that.

BTW The leakage terms setting seems to fail to do anything noticeable in the measurements at the high end of the s21. Is that the proper place to look for it? Does that need to be saved into the configuration file before it becomes effective?

[joeqsmith]

The problem with the sweep was the port settings. If I set only the first one to the port for the VNA, then it would link but that was all.  When I set the first two for the com port for the VNA, and it would link and sweep and do what would be expected, although I stopped for the day at that point before trying to do a cal.

Ok, that is a new one. I would never have tried to tell the software that the VNA and transfer relay are the same device.  If it were actually a product, I think we would need to add a lot more protection for the users.   

It makes no sense that the VNA wouldn't sweep if it links as the two ports are not related.   I can add an enable for the transfer relay (pretty sure I had it at one time) if you feel that would make things less confusing.  You would need to create a new defaults file to support it.   I actually use the transfer relay from time-to-time and like not having to select it but am willing to change it. 

Hopefully, everything for the crystal measurements will work.

"Works" has different meanings to different people.  The last thing I bought from eBay "worked".  Want to know how many hours and $$ it took to get it to "work" by the my definition (which is that it meets spec)?   

Again the original $50 NanoVNA does a much better job of "working".  For the LiteVNA, V2Plus/4 and suspect your Pro, narrow band measurements was an afterthought.  The hardware was already set in stone with many poor choices and Dislord made an attempt to work around it in firmware.   Maybe it will "work" well enough for you.

I exported an s2p touchstone file and that appears to have the s21 contained in it.

Guessing it does something but I doubt very much that it is correct.   Maybe once you get a little further long, you can let me know.

BTW The leakage terms setting seems to fail to do anything noticeable in the measurements at the high end of the s21. Is that the proper place to look for it? Does that need to be saved into the configuration file before it becomes effective?

Well, I am not sure.  You have told the software you have a transfer relay (which you don't), you then go through a full 2-port cal (which of course makes no sense as you don't have the hardware in place to do it).  Then you attempt to collect some meaningful data.  There are other things you could be doing as well that would effect the leakage corrections.    It's also possible, but doubtful, that your VNA is so good, enabling it has no effect.   

[joeqsmith]

OWO's site shows a noise plot for your VNA using the following settings:   Achievable performance (V2 Plus4 Pro, calibrated, IFBW=0.8kHz, AVG=20).   Below 100dB.   

Using my software with the LiteVNA, I set it to 20 averages, 800 Hz IFBW  and disabled the leakage term.  I think the real question is how long did it take them to acquire that data compared with the LiteVNA using my software.  For me, that was several minutes. 

For fun I have included my H4 as they allow harmonics to 2GHz.  May be useful for some experiments but you can see why I suggest the LiteVNA for anything over 300MHz. 

[Fred_B]

I've figured out how to operate solver better now. I have XferPort and StagePort both set to unused ports. It's prompting for a two port one path cal. I guess, I was just confused about the interface as I got to know it.

With the two port one path cal, the export s2p seems to export S11, S21, S21, S11. That's all I need.

I did a 20x average of S21 after calibrating with 800Hz IF, leakage terms off.

So, I could have gotten better for half the price... Oh well, at least this will get me though my current project. Then I'll buy a liteVNA and just keep the more expensive, less able one as a backup.

Come to think of it, do you think solver actually changed the IF on the nano to 800 Hz? I see in your settings where 800 Hz is used for the upper frequencies that solver tries to set the IF to 160 Hz for <= 400 KHz.  I wonder if the nano is just ignoring all that. Since there's an absence of any way to monitor from the touchscreen, it's anybody's guess as to what's happening there.

[joeqsmith]

That's correct, I fill the last two columns with duplicate data when there is no transfer relay.  With the transfer relay, you will get all four.

Come to think of it, do you think solver actually changed the IF on the nano to 800 Hz? I see in your settings where 800 Hz is used for the upper frequencies that solver tries to set the IF to 160 Hz for <= 400 KHz.  I wonder if the nano is just ignoring all that.

Looks like again the EEVBLOG forum is having problems with attachments.  Anyway, from OWO's site showing the data for your VNA, we expect 110 dBish average in the lower frequency range where your's is closer to 95.   My guess is this is not a problem with your VNA.  It is very possible that the commands used to set the filters is not the same.  OWO's firmware would have had to follow the same commands as what Dislord implemented or it will not be compatible with Solver.     My V2Plus4 doesn't have support for it, or several other commands for that matter.  The firmware is very dated and not supported. 

I would just join OWO's groups.io and ask there.  Keep in mind they may censor your posts, so if they don't show up you know why.   

***
BTW, I did go ahead and add an enable like I mentioned.  There are a few other problems that I have addressed.  I'm not aware of anyone using the software outside of yourself.  If you don't come across any problems in the next few weeks, I will go ahead and release these changes.   

[Fred_B]

I made some quick simulation models of the calibration standards that came with the nanovna and used them to create touchstone files for calibration with vna_Qt. I used 10x averaging for the calibration and 20x averaging for this screen shot. I was also using the 300mm cables on each port. For the one port measurements I used the thru with the male standards. It helped. Although, I thought calibration would get rid of the cable ripples more in the S11 results.

How does the touchstone correction of the calibration work?

[joeqsmith]

85dB for S21 in the low frequency range is not expected.   

For the most part, everything I have shown with the low cost VNAs has used the ideal model.    I think the only time I have shown using non-ideal standards was for testing the waveguides.   A viewer did call me out for having mistakes in my math but I walked through comparing results with the LiteVNA against my Agilent PNA.  That ended the discussion, so I assume the viewer was a bit lost or there really is a problem and they were unable to articulate their concern.   I'll leave that to other users to sort out as it appears correct to me.    For the waveguides, I have a few standards that I can use for a sanity check.  Actually, the new version of Solver has included the Agilent WR90 standards in the library.  I attempted to make a poor man's set based on them. 

https://www.eevblog.com/forum/rf-microwave/experimenting-with-waveguides-using-the-litevna/msg4796609/#msg4796609

[Fred_B]

I'm wondering if that's some kind of measurement artifact. I see these little discontinuities in the sweep here and there. I made another calibration with touchstone files for the VHF band for my project. And this is what I got, 20x averaging, 601 points, 143 Mhz - 149 MHz.

[joeqsmith]

That's a big improvement from your previous post and more what OWO claims.   I wonder why you can't sweep a wider range like they show and get roughly the same results.   Did they provide you with coefficients for their included standards?