NanoVNA Custom Software

[DiSlord]

Then actually 10 sweeps averaged, not one sweep smoothed.  This makes a lot more sense when looking at your data.

VNA measure by ADC for example 100 samples, then multiplie on 100 sin/cos values table (made FFT for IF) for calculate IF IQ data
But possible get 200 samples or 300 or more.
This look like i try grab more data for FFT bin, but for get more data need more time.

This process not fully average (as i made measure, made another measure and average results), but result look same
This option on H/H4 named bandwidth, H/H4 use fixed IF and possible calculate IF bandwidth.

If you copied the V2Plus4 firmware as a starting point, I expect the calibration accounts for the leakage term.  OWO had brought the need for it to my attention after I started porting my software over to it.

Yes LiteVNA use same calibration alghoritm as V2 (i only add real time interpolation for calibration data, this allow get good measured results if calibration range not same as measured range)

[joeqsmith]

I have a V2P as well from OWO that I do not use.  Now that I know that the hardware is good, if your firmware supports it, I could load what ever version you like into it and provide you with feedback. 

[Schorsch3]

Hi Joe, is it possible to run or scale your software ( NanoVna) in full screen as well?
Great program!

[joeqsmith]

Hi Joe, is it possible to run or scale your software ( NanoVna) in full screen as well?
Great program!

https://www.eevblog.com/forum/rf-microwave/nanovna-custom-software/msg3813893/#msg3813893

[joeqsmith]

Its easy enough to add average to my software.  Rather than having the VNA perform it, I just add a simple FIFO to hold all the sweeps.  As it acquires data, then plots update at the normal rate.  There is an Avgs counter that keeps track of the accumulations.   Once it reaches the target is starts tossing out the oldest sweep and adding the new.  So it's always the average of the last N sweeps.    The PC is basically unlimited compared with the little embedded micro, so this really poses no problem.

Shown with a GHz interdigital filter with no averaging and with it set to 100.   

****
Also shown with 1000 averages including phase. 

****
Looking at S21 for the original NanoVNA with both ports terminated (system dynamic range) after calibration.  Yellow is the raw data.  Violet is the average of 100 sweeps.   Red is using the smooth filter but really stepping on the signal.  Normally, I would rather see the raw data. 

[DiSlord]

For calibration data good if use big averaged measure (in my firmware for H/H4 i use 100Hz bandwidth = ~40x avg on calibration measure as minimum, user can select less bw if need)
This allow get less noise calibration data, and as result less noise measure later after calibration apply

[cdev]

I have the 2.2 and its quite handy for 2.4 GHz antenna evaluation. But not the higher wifi bands. The absolute top for it is 4.4 GHz and above 3 GHz its not flat. But its not super peaky, either. Still useful, IMHO, for relative measurement.

I use the nanovna2 for 2.4 Ghz a fair bit actually


Antennae for 2.4 G are so conveniently small, and easy/fun to build. But without a VNA, not so much.

[joeqsmith]

While looking over OWOs projected specs for the V3,

https://nanorfe.com/nanovna-v3.html

I noticed they are planning to support unknown thru.   I doubt they are planning on producing a four receiver VNA. 

From:
https://www.picotech.com/vector-network-analyzer/picovna/picovna-features

In a VNA a swept sine-wave signal source is used to sequentially stimulate the ports of the interconnect or device under test. The amplitude and phase of the resultant transmitted and reflected signals appearing at both VNA ports are then received and measured. To wholly characterize a 2-port device under test (DUT), six pairs of measurements need to be made: the amplitude and phase of the signal that was emitted from both ports, and the amplitude and phase of the signal that was received at both ports for each source. In practice this can be achieved with a reasonable degree of accuracy with a single source, a transfer switch and two receivers; the latter inputs being switched through a further pair of transfer switches. Alternatively three receivers can be used with an additional input transfer switch or, as in the PicoVNA, four receivers can be used.

Using four receivers eliminates the receiver input transfer switch errors (chiefly leakage and crosstalk) that cannot be corrected. These residual errors are always present in two- and three-receiver architectures and lead to lower accuracy than that of the Quad RX design.

I've never tried it with these low cost VNAs.   

[joeqsmith]

Using my homemade standards to calibrate the V2Plus4.  Next I measured an SMA to SMA adapter, short section of semi-rigid and a longer section of RG400/U to simulate an unknown thru.  Then wrote a simple program to determine the Unknown and feed the data back through with corrections. Plots show the raw data in green and violet after correction.   

[ON7CH]

Which values for SMA Calibration Standards did you use?

Hereby the data that I use for my SMA calibration standards.

[joeqsmith]

Which values for SMA Calibration Standards did you use?

When making the videos for these low cost VNAs, I've used the ideal model for both my home made FR4 standard as well as the ones supplied.   I've also presented data without calibrating the V2Plus4 at all or just normalizing the data as with the following example: 

https://www.eevblog.com/forum/rf-microwave/shunt-capacitance-of-1206-smd-resistors-jeroen-belleman-december-2010/

In that case, just a quick visual was good enough to see the trend.  If you read that entire 2-pages,  a gentleman here had helped me to characterize a set of home made N and SMA standards that I use with my vintage Agilent PNA.  I posted a little more detail about that here:     

https://www.eevblog.com/forum/projects/20db-rf-attenuator-seeking-feedback-to-improve/msg2987982/#msg2987982

Below is a review I made for OWO's  V2Plus4.  If you take the time to watch it, you will see just how loose I am when it comes to calibration. 

[ON7CH]

Joe, with your help I found the instructive home page of Mario Hellmich relating to the characterizing of a SMA calibration kit.
Very interesting, recommended for a VNA user.

[joeqsmith]

If you're looking for a low cost, ready made solution, you may want to consider:

https://appliedeminnovations.com/index.php/product/vna-calibration-kit/

They appear to be a decent quality but I don't see where they spec the number of mate cycles.   I would hope the plating is better than what you will find with the parts supplied with these low cost VNAs. 

[ON7CH]

Can someone help me with the definitions/coefficients for this SMA Wiltron calibration sets.
See attachment.


It concerns the standards type 22S50 and 22SF50.
I searched the internet with no results. I can't find a detailed data sheet.

Thanks in advance.
Guido

[joeqsmith]

Can someone help me with the definitions/coefficients for this SMA Wiltron calibration sets.
See attachment.
(Attachment Link)

It concerns the standards type 22S50 and 22SF50.
I searched the internet with no results. I can't find a detailed data sheet.

Thanks in advance.
Guido

Guessing you have a similar story as this person:
https://www.eevblog.com/forum/testgear/a-general-question-about-vna-calibration-kit-definitionsconstants/

Maybe you will find some useful advice there.

[joeqsmith]

State of the software.....

I corrected the autoscale for the antenna scanning feature, added averaging and fixed a few other minor problems.  I decided against releasing this version as I started to give some thought to more complex calibrations.   

While looking over OWOs projected specs for the V3,

https://nanorfe.com/nanovna-v3.html

I noticed they are planning to support unknown thru.   I doubt they are planning on producing a four receiver VNA. 

From:
https://www.picotech.com/vector-network-analyzer/picovna/picovna-features

In a VNA a swept sine-wave signal source is used to sequentially stimulate the ports of the interconnect or device under test. The amplitude and phase of the resultant transmitted and reflected signals appearing at both VNA ports are then received and measured. To wholly characterize a 2-port device under test (DUT), six pairs of measurements need to be made: the amplitude and phase of the signal that was emitted from both ports, and the amplitude and phase of the signal that was received at both ports for each source. In practice this can be achieved with a reasonable degree of accuracy with a single source, a transfer switch and two receivers; the latter inputs being switched through a further pair of transfer switches. Alternatively three receivers can be used with an additional input transfer switch or, as in the PicoVNA, four receivers can be used.

Using four receivers eliminates the receiver input transfer switch errors (chiefly leakage and crosstalk) that cannot be corrected. These residual errors are always present in two- and three-receiver architectures and lead to lower accuracy than that of the Quad RX design.

I've never tried it with these low cost VNAs.

Using my homemade standards to calibrate the V2Plus4.  Next I measured an SMA to SMA adapter, short section of semi-rigid and a longer section of RG400/U to simulate an unknown thru.  Then wrote a simple program to determine the Unknown and feed the data back through with corrections. Plots show the raw data in green and violet after correction.

OWO was recently posting about the V3 and they were requesting feedback on features.   

https://www.eevblog.com/forum/testgear/nanovna-v3-(6ghz)/msg3847832/#msg3847832

With the squarewave drive, I don't see a lot of value in a full 2-port system.   We may be stuck with using a external transfer relay.  Which gets back the the topic of calibration.  For the couple of you actually trying to use my software, I am sure you realized that the calibration is pretty limited.    The main reason for this is because of pure laziness and the lack of performance on these low cost VNAs.   Because the original NanoVNA worked much like my old
HP8754A, I stayed with the same software structure.   The calibration for the NanoVNA is actually a bit more advanced than what I wrote for the HP8754A which isn't saying much.  Consider the 8754A at best could normalize the data where my software allowed me to run a SOL on it. 

My software normally applies the error correction as soon as it sees a valid sweep.  This works fine for the 1-port VNAs but as we add the transfer relay,   the software has to make multiple requests to read all 4 four S-parameters.   This means a change to how the software was structured.   All doable but I doubt the gains are worth it.

As I continue to ponder my software for the low cost VNAs, the first problem I see is we need to ditch the current calibration file format.  This means the new software will not be compatible with the old files.    If the transfer relay was selected when a calibration was performed, the software prompts the user to run an SOL on both ports.   The new software would work the same way, however when selecting the 2PSwp (2 port sweep) it will test the state of the transfer relay and that the unit was calibrated.   It will then enable the full SOLT or SOLR (unknown thru) model.   This would only be useful when using the Advanced, 2-Port Plot tab.    When only S11/21 are required, the software will continue to normalize the thru but will use the 12-term model if SOLT is selected with the transfer relay.   

If I get something that I think may be useful to users, I will release it.   Going forward, I don't intend to put any more effort into supporting the original NanoVNA outside of correcting problems.  We have the LiteVNAs on order and are expecting them soon.  Looking forward to seeing how their performance compares with the V2Plus4.

Attached screen shots of version 3.0.


 

[joeqsmith]

I'm looking for a reference book that covers the unknown thru in detail (along with other techniques).   

[joeqsmith]

We received the LiteVNAs and ran a few tests on them.  Within a few minutes of testing we ran into a small snag with the firmware but the hardware appears like it was worth the wait.  I should have mine in the next few days.

[KE5FX]

I'm looking for a reference book that covers the unknown thru in detail (along with other techniques).

You have Dunsmore, right?  See 3.4.3.1 in the newer edition.

[joeqsmith]

I'm looking for a reference book that covers the unknown thru in detail (along with other techniques).

You have Dunsmore, right?  See 3.4.3.1 in the newer edition.

I  was just discussing that book with another member.   I am not of fan of his mixing terms, for example 3.35 the use of E11 in place of ESF.  Why not make it as clear as possible?   3.4.3.1, it is summed up nicely with "... of which is beyond the scope of this book..." 

Don't get me wrong.  I think it is a very good book and recommended for anyone interested in the topic but I am really looking for a book dedicated to calibration that keeps it on a practical level for my aging brain.   It's a fun topic but as I mentioned to the other member, with the software I have written and as much as I have used VNAs, where I stand today, I know very little about them. 

[joeqsmith]

The LiteVNA arrived.   

[joeqsmith]

The claim 50kHz to 6.3GHz.   My software doesn't care so I set it to 6.6GHz.  Looks like 6.47GHz is the upper end.    Setting it to 6.4GHz, it seems fairly accurate.  The counter is GPS referenced. 

[joeqsmith]

 I was really interested in seeing how low it would run.    My friend Flipper and I had ran some simple remote tests with this exact unit a few days ago and I suspected there may have been a firmware problem when running at the lower end.  Now that I have the unit, I can see it was a timeout in my software that needed to be increased to handle the slower rates.   

***
9kHz is the lower limit of the SA.  I think when we ran it with the scope, we were down below 10kHz before it fell off.    We did not try to collect data down this low. 

[joeqsmith]

Similar results when looking at a 3.68MHz crystal as the V2Plus4,  both poor.  No surprise.   OWO had mentioned a version of firmware that would allow the V2Plus4 to make narrow band measurements.  I never looked into it. 

[joeqsmith]

OWO had asked about using a short length of coax and looking at linearity.    Here I am using a M-M attached to a F-F  to create a small extension, calibrating at the end, then removing the extension.   Using the same settings,  I repeated the test with the V2Plus4.