NanoVNA Custom Software

[joeqsmith]

Same procedure, 401 data points, 1.5G to 3GHz range.   I was going to try with the average set to 80 but was starting to fall asleep waiting for the thru.     So I set the average to 10.  Because we were saving the data to the SD card, I ran the Lite off battery to give it the best chance possible.

Shown comparing the above setup (red) with the my software running the Lite the way I would normally use it (dark brown).   We gain about 10dB but it does slow things down.   

Also shown is the V2Plus4(tan) connected to the PC, no average and running off the PCs power.  Sadly, I know using the external power degrades the performance. 

Also shown is data another member had taken for the Jankae LibreVNA (black).   

Feel free to repeat this little experiment and see if your results show any major improvements.

[joeqsmith]

I changed to 80X average and repeated the calibration and measurement.  No other changes were made.   Data was normalized to the original, non averaged data set.  There appears to be some gains at 80 but it takes a very long time to run through it.

[joeqsmith]

I ran the V2Plus4 (brown) from it's internal battery and used the software to average 10X for calibration and collection.  I then compared this with the Lite (black) with the firmware 80X average for calibration and collection (also running from battery).   

[realfran]

I wait for the release of rev.3 to test all these new software features with the LiteVNA, my use is with PC this possibility hare very usefully for it.

[DiSlord]

I try compare LiteVNA / V2Plus4 / LibreVNA in S11 measures

I use 6dB attenuator on port and 30cm cable, calibrate all devices by one calibraion kit.

30cm cable:
Red - LibreVNA
Green Lite
Purple V2Plus4

Lite and V2Plus4 show ~equal results (on V2Plus4 strange spike at begin on > 0dB)

Next measure 6dB attenuator:
Red - LibreVNA
Brown - Lite
Green - V2Plus4
LibreVNA now have strange spike at begin. Lite and V2Plus4 ~equal, V2Plus4 little more noise

All measures made on 1xAVG

[gf]

What was the DUT in the first test? (Seems to be something with almost 100% reflexion?)
And if the DUT in the 2nd test was a 6dB attenuator, how was its 2nd port terminated?
An you wrote "S11 measures", but in the diagram the red trace is labeled "S21"?

[DiSlord]

What was the DUT in the first test? (Seems to be something with almost 100% reflexion?)

Hmm let me see how i write...

I use 6dB attenuator on port and 30cm cable, calibrate all devices by one calibraion kit.
30cm cable:
Red - LibreVNA
Green Lite
Purple V2Plus4

At first i measure 30cm cable cable not terminated. I must see 2xLOGMAG value of cable lost (as long as the wavelength fits on the way back and forth, then beats should already be visible due to attenuation in the cable and not pure 50Om or calibration kit or cable)

For attenuator i must see ~ 2 x LOGMAG of attenuator value (12dB line and small errors)

Colors for stored traces all colors as i write, all measures S11 reflect LOGMAG, at top show colors for live traces (not visible).

[gf]

OK, so the test set was port1 -> 3dB attenuator -> calibration plane, and the DUT was the open-ended cable (or alternatively an open-ended 6dB attenuator for the 2nd test)?

[DiSlord]

First test open ended cable
Second test 6dB attenuator (i also use 3dB attenuator, but not store results, so made mistake)

Most strange results i see from LibreVNA (for cable as i write must be line ~600M, for attenuator strange spike at begin and on 3.2-3.4G)

For V2Plus4 also on cable i see strange up to > 0dB (and also small spikes on measure)

For Lite also exist small error on < 100MHz

[gf]

Agreement of V2plus4 and Lite is not that bad
Another question is of course, what cal kit was used, and whether the standards were simply assumed to be ideal (although they are not), or whether their known/measured impedances were honored for the calibration.

[realfran]

Hello Dislord, my 4 mounts hold V2plus4 not give me the possibility to upgrade with your FW. there is a possibility to erase everything? And put the new firmware?

[joeqsmith]

Another question is of course, what cal kit was used, and whether the standards were simply assumed to be ideal (although they are not), or whether their known/measured impedances were honored for the calibration.

Since purchasing the V2Plus4, I have used the standards supplied with it except for the load, which was sorted from a batch of Mini-Circuits terminators I purchased.   I am not aware if the firmware for any of the low cost VNAs I have support anything beyond the ideal model.   The data I collected for the dynamic range and noise was taken with the ideal model. 

[joeqsmith]

I wait for the release of rev.3 to test all these new software features with the LiteVNA, my use is with PC this possibility hare very usefully for it.

Sorry but no news on a release.    You thought you were confused with 2.x, 3.x isn't going to make things easier.     

[joeqsmith]

Not sure what you will learn from this but Lite with 1024 points (max firmware allows), 10Xaverge for cal and collect and measuring a cheap 6dB attenuator.  Again, running off battery and saving the data to the SD card.  PC was not used as I assume this is what you were doing.  I used 10X average as you stated this was the minimum you use now.   I have attached the Touchstone file if you want it.   

Thinking of your glitches, I assume you inserted the female to female adapter into the open end of the cable you tested to keep it stable.   

***
It's winter here and the house is dry.  ESD may be another possible cause if what you are seeing is random.

[realfran]

Thank you for your great software

[joeqsmith]

Any calibration You make is only valid for the calibration plane (and compensates for the errors only there) .

You can change the reference plane after calibration. 

Looks like there have been a few changes to the Lite's firmware.  They still don't appear to provide any details about what they changed or why.   Any details about the new release would helpful. 

[DiSlord]

As i know
Only toch screen problems fixed (not work top left corner for 4 inch devices), no any additional fixes.

[Mechatrommer]

You can change the reference plane after calibration. 

how? afaik the term is de-embedding, and its a math intensive procedure and requires full 2-port VNA, which i believe you havent demonstrated it. neither did i.

[gf]

afaik the term is de-embedding, and its a math intensive procedure and requires full 2-port VNA, which i believe you havent demonstrated it. neither did i.

A noteworthy trivial special case of de-embedding is "port extension" which assumes that the network to be de-embedded is a piece of ideal 50 Ohm transmission line. This can serve as a good enough approximation for things like (say) a short female-female adapter between calibration plane and DUT, or similar.

[joeqsmith]

You can change the reference plane after calibration. 

how? afaik the term is de-embedding, and its a math intensive procedure and requires full 2-port VNA, which i believe you havent demonstrated it. neither did i.

The newer firmware for these low cost VNAs supports adding a delay (port extension).  I demo'ed it with my software.  You can certainly use it with a 1-port VNA looking at S11.   

(and compensates for the errors only there)

I was going to say that we are really correcting errors before the reference plane but felt I was being pedantic.  We are correcting for connectors, cables, delay, loss, reflections.....  The VNA measures from the reference plane.     The way it was worded, it seemed like it was suggesting that you had to measure the device right at that magic spot or all bets were off.   Of course, the device we are measuring may be a 1000' of coax, extending far beyond the reference plane. 

[gf]

One problem with Electrical Delay in the v2 firmware is that it subtracts the SAME delay from S11 and S21. But S11 is round trip, while S21 is one way. So if you set an electrical delay in order to extend port1, then it can be either correct for S11, or for S21, but not for both at the same time.

[gf]

Imo it were better if I could set separate port extension delays for ports 1 and 2, where the former is subtracted 2x from S11 and 1x from S21, and the latter is subtracted 1x from S21.

[Mechatrommer]

afaik the term is de-embedding, and its a math intensive procedure and requires full 2-port VNA, which i believe you havent demonstrated it. neither did i.

A noteworthy trivial special case of de-embedding is "port extension" which assumes that the network to be de-embedded is a piece of ideal 50 Ohm transmission line. This can serve as a good enough approximation for things like (say) a short female-female adapter between calibration plane and DUT, or similar.

only that (bolded line), but can we get ideal 50ohm TL at cheap? are we sure the hunglow pigtail M or F connectors/through are really ideal? its ok if we can only "assume" thats fine with me , but in academic world i dont find it that simple. and characterized/ideal port extender can cost 10x of LiteVNA i guess, unless china manufacturing and QC has figured out something. if only want to extend port/plane for insertable dut.. i can only think of, why we dont just recalibrate at new extended plane? if we have both female and male cal kit? (i have ), so we can let the internal math of VNA FW do the job. the difficult part of extending plane is because the DUT is not insertable such as on PCB. i'm not sure though what the reason we want to test an open coaxial cable with reflections everywhere along the line as dislord did and then get into hassle try to extend (or even backward) the ref plane as we like. ymmv.

[joeqsmith]

One problem with Electrical Delay in the v2 firmware is that it subtracts the SAME delay from S11 and S21. But S11 is round trip, while S21 is one way. So if you set an electrical delay in order to extend port1, then it can be either correct for S11, or for S21, but not for both at the same time.
 
Imo it were better if I could set separate port extension delays for ports 1 and 2, where the former is subtracted 2x from S11 and 1x from S21, and the latter is subtracted 1x from S21.

My released software only trims the delay for S11/S22.  The newer version supports S11/S21 (/2 for S21) but currently does not support the transfer relay.   That section of code is the main reason for the change and is still work in progress.   

[realfran]

The software average work fantastic fast, this for the tuning filter is important.
I get a little confused with the TDR I tray to test my LINE stretcher, I need to learn to use it (not a lot of information for coax line stretcher)
When I did 2 port calibrations with the Transfer relay on, I get the same image for both ports works very good.