NanoVNA Custom Software

[joeqsmith]

You connect external generator to ch1 port?

You see 60k IF mirror (at 4MHz - 2*60k = 3.88M).

Scope was on port1 to monitor VNA's frequency and CW generator in series with crystal was on port2. 

Shown with the same 500kHz span and hardware setup using the original NanoVNA.   Note span of two peaks now only at 10k.   

[joeqsmith]

If we set the span to 5kHz and compare the original NanoVNA with the Lite.   

With the same setup, I program the Lite's E0 register to 0x1E and see a similar shape.   

My question is the problem with the Lite's narrow band measurement the firmware's filter settings?

[DiSlord]

It's not clear to me what you want to measure in this way.

You send to CH1 port CW frequency 4MHz. Lite sweep in span 500kHz.
Measure result - this CW frequency go to mixer and mix vs sweep signal. And output IF sweep in range 500kHz.
So on display you see internal digital filter around 60k (additional you see mirror IF then on -60k)

Change AVG / E0 settings you change this filter width

NanoVNA Eddy v0.5 firmware use 5k IF (my fw for H/H4 use 12k IF) and not allow change filter width (my firmware allow select it), so you also see 2 peaks (additional mirror IF at -10kHz) on H.

For measure crystal need connect it between NanoVNA ports. Then main frequency correct sweep in measure range, pass thru crystal, and measured. Main problem on measure low width filters like crystal need use less width internal filter.

This my measure series connection 4MHz XTAL between ports, as can see exist strange result - this from wide RBW on measure, but i can reduce it (by increase AVG), as can see 5x avg better, but not good, select 10xAVG and get good result.

If you want use VNA like SA, need use step depend from internal measure filter width.

Additional Lite allow (in last fw) measure and show  raw data (CONFIG->EXPERT SETTINGS-> DISPLAY SAMPLES)
in this mode channel data contain raw data samples, and you allow see on diaplay it:
I send to 1 port 30MHz (and stay on 30MHz CW in raw samples mode) and see IF samples on port (in linear format).
 
Additional i can enable FFT and see spectre (DISPLAY->TRANSFORM->TRANSFORM ON) in Logmag
Now i can see signal SA in frist 512 points
This mode also used for debug, and not friendly use.

PS gf correct answer how all work

[joeqsmith]

I wanted to remove the effects of the output being chopped.  End goal is to use the Lite for narrow band measurements.    Data from the original NanoVNA is very similar to what I see with my vintage VNAs.   The Lite is not even close.   I want to know if this is because of how the filters are implemented or is it a limitation of the hardware?   

[joeqsmith]

Sure enough, if I increase the averages to 10, the results are greatly improved.    Shown compared with data taken from original NanoVNA from 2020. 

[DiSlord]

Can you show how V2Plus4 measure this filter?

[joeqsmith]

Do you have firmware that will run on it that you want me to try?    I need to test firmware programming function for it anyway so no problem.   It's an early version of the hardware.

Attached is with the original firmware. 

[DiSlord]

V2Plus4 have close source code (DSP and measure part).
I can provide only UI, points count, calibration changes. But can't change any related to measure.

PS H/H4 better work on low frequency < 10MHz (and have little less noise on < 100MHz, in this range i can get up to -100dB noise floor in S21 measures on my H4, but need use low RBW)

In Lite if remove DC block capacitor and ESD protection diode possible get better results on low frequency range < 10M (but this huge increase chance burn rf switch)

[Alextsu]

I wanted to remove the effects of the output being chopped.  End goal is to use the Lite for narrow band measurements.    Data from the original NanoVNA is very similar to what I see with my vintage VNAs.   The Lite is not even close.   I want to know if this is because of how the filters are implemented or is it a limitation of the hardware?

This might be a consequence of RF signal processing chain differences between original NanoVNA and V.2 mods. All V.2 mods need to switch RF signal between output and internal paths to perform measurements.

[gf]

All V.2 mods need to switch RF signal between output and internal paths to perform measurements.

But is this really a problem if you grant enough settling time between switching and taking the measurements?
Sure, a large settling time does not make the sweep faster, of course. But speed vs. narrow-band is always a trade-off.
Does the Lite actually have 3 receivers, or does it switch the receiver as well?

[Alextsu]

All V.2 mods need to switch RF signal between output and internal paths to perform measurements.

Does the Lite actually have 3 receivers, or does it switch the receiver as well?

Yes, the Lite switches the receivers.
There had been 2 main problems with the narrowband filter measurements before.
First came from RF switching,
the second from DFT (FFT) IF signal processing.
But the latest FW for the LiteVNA solves both of them and is well suited for narrowband measurements.

[joeqsmith]

I've never used the H4 due to firmware.   Maybe your later firmware has improved but going forward, I have no plan to support the original protocol.  The released software for the original NanoVNA and the H4 will most likely not change.   

It's a shame they locked down the firmware for the V2+4.  My experience with the firmware for both the original NanoVNA and H4 wasn't pleasant.   Even the V2+ (not the +4) had a major problem where it would hang at random.  I suspect the narrow band limitations with the V2+4 are also with the firmware.   From what I have seen with the LiteVNA, even your unreleased firmware is proving to be very stable and I greatly appreciate your efforts. 

I have not considered modifying the hardware.   If you have benchmark data for the front end mods you mention, please post them.

I've been cleaning up my software and working on making it a bit more child proof.   Normally, I wouldn't waste any time on it but with your new firmware for the Lite it was very easy to get into trouble and have to kill the application with task manager.

Currently I use LabView 2011.   Flipper and I have been testing a 64-bit version of the software.  NI offers the community edition free of charge now.   This includes the application builder (supports EXEs).   However they are 32-bit only and I would need to upgrade my license.   $$$ 

***
It appears that Community 2021 is 64-bit.   I will need to read their license contract but it may be an option.   I doubt they would give me a price brake to upgrade.   

[joeqsmith]

Last night I installed the 64-bit runtime engine and built the software with the latest 64-bit version of LabView.   It may provide some advantages when using the Lite.   It seems to require less processing and screen update rates are a blur.  Looking over the license for the community version,

As a condition to the License, and unless and only to the extent that this Agreement expressly permits otherwise, you must not

(iv) publicly perform or display the Software;

LabVIEW Community Edition and LabVIEW NXG Community Edition

You may distribute or transfer applications you create with Software, but only if you comply with the terms of the Agreement regarding distribution of Authorized Applications and such distribution of the Deliverables is for non-commercial, non-industrial, and if you are a degree-granting educational institution non-teaching and non-research, purposes only.

Except for the limited distribution rights set forth in this license, in no event may you distribute any software or code created with the Software.

When I finally get around to making a full demo of the software, I may compare the 32 and 64 bit versions.  If users feel it's something they want, I may start a Patreon account to cover the cost of the upgrade.   

[joeqsmith]

Getting back to the narrow band measurement using Dislord's pre-release firmware,  I thought would take the time to run one last filter across multiple VNAs.  This was from back in Aug 2020, so a long time coming (sorry for the delay Grandchuck):
https://www.eevblog.com/forum/rf-microwave/fun-with-crystal-filters/msg3180002/#msg3180002

Shown with the latest 64-bit software without any calibration using 10 averages (register 0x44 0x40= 0x0a). 

[joeqsmith]

Using METAS so show how the V2Plus4 compares with my original NanoVNA.  Notice how the gain is also off.  This data was taken with both VNAs calibrated.   I can install a wideband device like an attenuator on the V2Plus4 and do not see this level of error in the gain.

[joeqsmith]

Comparing the original NanoVNA with the Lite. 

[joeqsmith]

Shown comparing my original NanoVNA and LiteVNA with my two old systems.   To make this measurement with the HP8754A, requires an external source so I didn't bother.   

[joeqsmith]

METAS allows normalizing.  Here I am using my old Agilent PNA as a reference.   Shown is both the LiteVNA's linear and log mag relative to the PNA.   Once the firmware is released and I will most likely add a simple button to set these registers automatically. 

[jspencerg]

Hello,
Regarding the NI clauses: You could demonstrate with the bought 32-bit and distribute the 64-bit home version. Seriously,  getting users to pony up some support is very reasonable.
Regarding your firmware support: Which platforms/authors will you continue to support operability with your software?
Primary reason for posting:  I thought I understood from the first comparison graph that v2+4 was doing much better than the original at that ~3Mhz frequency. The further comparisons showed the original doing very well in comparison to the other machines.
Why did the original do poorly in comparison to v2+4 and then do well against others?
I'll rewatch your original posting.  Would be interesting to see what frequency range the original's performance drops below the v2+4 (i recall the issue of harmonics applying).
Thanks.

[joeqsmith]

LabView is old and very expensive.   As the user base dies off and people search for lower cost and license friendly solutions,  I assume they came out with the community version in an attempt to try and recapture future sales.  IMO,  NI would have been better off summing that lawyer speak into 5 sentences or less in layman's terms.  I'm not sure I can ethically distribute my software using the community version without hiring a lawyer to interpret their license agreement.   I'm sure some sales or marketing director at NI has it all figured out.

Cost I would imagine by the time I cover the income taxes, sales takes and cost of the license, we are upwards of three.   

Regarding your firmware support: Which platforms/authors will you continue to support operability with your software?

 
Interesting that you would ask as LabView now appears to support the Mac and LINUX under a single license.  This was never the case before.  If you wanted to target a different OS, you bought a separate license.   Still I have no use for LINUX or the Mac and doubt I would invest anything to support them.  I did try using LabView on LINUX when NI first released it.  It was a total time waste.  Pretty much my experience with LINUX throughout the years which is why I will not support it.     

I have no plans to develop software for any devices using the original protocol.  I have no idea which of these low cost VNAs are compatible.   If your VNA is backward compatible with the V2+, V2+4 or Lite, then it should work.  Any testing I do will be with the V2Plus4 and the LiteVNA64.

Primary reason for posting:  I thought I understood from the first comparison graph that v2+4 was doing much better than the original at that ~3Mhz frequency. The further comparisons showed the original doing very well in comparison to the other machines.
Why did the original do poorly in comparison to v2+4 and then do well against others?

Sorry but I am not sure I follow.  When you ask why something did poorly by comparison, it tells me very little without disclosing what metrics you are using.   My V2Plus4 has never been able to make narrow band measurements.  I've stated that several times and talked about it during my review of the V2+4.   I've said if you work below 300MHz and plan to make narrow band measurements, stay with the original NanoVNA.  If you want to experiment with PDN measurements, the original NanoVNA still out performs the others.   According to Dislord, the H4 does a better job at the lower frequencies.  That may be true today but the last time we had that discussion, I posted a fair bit of data showing otherwise.  I suspect the problems were caused by firmware.   The whole VNA has to work, firmware needs to be stable .... for me to spend any time with it.   

Since buying the Lite, I have not done much with the other products.  I am interested in seeing what Dislord comes up with for benchmark data at lower frequencies with the DC block and transient protection removed.  IMO, that's it's only weak spot.   Still, for the cost, who can complain?  All of these products have been well worth their price.   

[jspencerg]

Good evening,  The graph comparing the original to the v2+4 had a line for the v2+4 which showed the multiple crystal effects while the original's line was a weak hump. Perhaps the labels got switched?
We hobbiests are benefitting from the NI home license and your efforts.  It's right give back some and update your version.  Labview and everything else is expensive.  I read these poor opinions about Labview, but I can't think of a better prototyping or research platform. Creating tools with graphical programming environment of Labview has to be faster/easier than python.
Thanks.

[joeqsmith]

Good evening,  The graph comparing the original to the v2+4 had a line for the v2+4 which showed the multiple crystal effects while the original's line was a weak hump. Perhaps the labels got switched?
We hobbiests are benefitting from the NI home license and your efforts.  It's right give back some and update your version.  Labview and everything else is expensive.  I read these poor opinions about Labview, but I can't think of a better prototyping or research platform. Creating tools with graphical programming environment of Labview has to be faster/easier than python.
Thanks.

I am guessing you're referring to: METAS_NanoVNA_V2P4_magphase.PNG
https://www.eevblog.com/forum/rf-microwave/nanovna-custom-software/msg4018888/#msg4018888

I don't know what is meant by a weak hump.  The labels are not switched.   If anything, I would call the V2+4 the weak hump as it shows more attenuation than it should.   I also show how the original compares with my other VNAs.   I don't show the V2+4 in the other comparison plots because it is so poor.  Maybe copy the graph you are referring to into paint and add some circles or something to indicate what you are calling a weak hump.  To me, it couldn't be more clear but when your the one doing the work, everything seems clear.

Yes, I do see a lot of negative comments and misinformation on LabView.  It's not a big deal.  I just assume they are ignorant about it is all.  A lot of my time working on this software has not been coding.  That's the easy part.  A fair amount is spent doing research.   

Git doesn't provide many metrics to estimate the number of users but I doubt my software is very popular. Also, based on feedback, I estimate the majority of the people who have tried the software to be radio hobbyists.  There may not be any advantage for this group to move to a 64-bit platform.    It may take some time to accrue enough to cover the cost. 

Because it seems very few people took advantage of the manual, I doubt I will put any effort into a new one.   Its a lot of work just to have people asking basic questions that are covered in it.   Most likely I will make a demo video for the LiteVNA similar to what I did for the original NanoVNA and V2+4  and call it good.   

[jspencerg]

I'll review your original video about the crystals and study on it some more.  Thanks for extra details.
Thanks for writing the manual.  Real manuals seem to have gone out of style.  It's all online forums, posting/searching questions and reading faqs. I prefer a good reference. Let the community figure it out is the new support.

[DiSlord]

I prefer not do hardware mods (use for firmware tests only original hardware)

Last H/H4 hardware (v3.6) show more good perfomance in noise floor. But i compare old H and H4 and V2Plus4 (one year ago) in impedance measures vs Agilent 4294A in 2-30MHz measures.
My H4 can see max 7-8k and give bigger error on measure < 1Om, V2Plus4 allow see near 12k (but also exist errors on < 1Om).

Last beta Lite firmware contain AGC mode for S11 measures, it allow get more dynamic range (but exist small 0.04dB shifts due not linear opamp? on measure gain level change)

V2/Lite/V2Plus4 use one measure channel and switch for Reference/Reflect/Thru. On crystal measures this of course affects, but setting a larger delay does not solve the problem.
Reduce measures RBW (set device AVG or E0/E1 registers) allow get correct results (but use external sotware AVG not reduce RBW and not allow get it) i think need show/select measure RBW value (not AVG as now)

PS as i see main problem for Lite noise level on S21 measures near 85-95dB (but for cheap device all good, i hope Hugen can improve this in future), S11 measures show very good results.

[DiSlord]

Shown with the latest 64-bit software without any calibration using 10 averages (register 0x44 = 0x0a).

You mean 0x40 register (set average) ?
0x44 = 1 get only S11 measure
0x44 = 2 get only S21.