NanoVNA Custom Software

[joeqsmith]

Joe, you didn't follow the repro steps. If you did, you would have seen that if your assumption were correct, then there is no way I would have completed step 1:

        Flash incorrect FW LiteVNA62-220228.bin to LiteVNA64 device using Solver64 (3a84660); observe successful flash and LiteVNA64 screen displays smaller/mirrored screen output due to incorrect FW

As you presume the unit was correctly prepped to "flash" in step 1 , you should also presume that the necessary steps were completed in step 2 which I stated the same process:
"2. Attempt to flash correct FW LiteVNA64-220228.bin to the device using the same Solver64; "

I can assure you Step 1 completed as I stated and Step 2 did so as well using reasonable expectations (cycling into DFU mode, etc).

If however, you are telling me you attempted to perform the steps as written with reasonable assuptions and were successful, I will retry (for the 6th time - I attempted 5 complete separate times without success even after returning to correct FW using Saver each time, before giving up and resolving that there is a bug.

Since Joe wasn't convinced, I thought I should rule out my system so I copied the (HUGE) NI runtime binaries over to my old Win10 laptop and unfortunately (or fortunately, for Solver) could not repro the issue. This would point to a system issue but it didn't make sense why Solver could flash the FW at all on the first system, so I moved from the front USB3 port I had been using (connected to a bridge chip on the MSI mobo I have) and to one of the native USB2 ports direct on the rear of the mobo, and flashing back and forth was successful. I then moved to the rear mobo native USB3.0 ports and flashing back and forth was successful. So I then went back to the bridge-connected front USB3 port and it was successful 3 times, then failed on the 4th. I then moved back to Saver and flashed successfully back and forth 5 times each without failure. So it seems there may be some intermittent/rarish timing issue with the MSI USB3 bridge chip in the B450 based mobo I have. What I can't explain is why it failed 5 times in succession last night but worked 3 times in a row tonight. Perhaps sunspots (I was also very successful with CB skip last night around the same time

Regardless, good to find out this wasn't a bug in Solver. If you encounter any flash problems, obviously ensure you are following the correct steps as Joe states, but try a mobo-direct USB port just in case.

I have not had a failure yet when programming the Lite's firmware using my software.   This includes the exact versions listed above (shouldn't matter). 

...  you should also presume that the necessary steps were completed in step 2 ..

   Author did not specifically state the VNA was set back to DFU.  Details mater when you are reporting a problem.   

[joeqsmith]

Adding a simple offset to null out the frequency error is trivial.  It's an addition..   

Shown with the Lite's average set to 40 or an IFBW of 100Hz and a span of 500Hz.   Also shown is my PNA with the same settings. 

[jspencerg]

Doesn't it look like the PNA has smaller ifbw that the lite? Learning to read these graphs...

[joeqsmith]

The Lite's skirts are wider and the peak is attenuated.   Put a pulse into a filter and start lowering the cutoff.   What happens?  You could do this with SPICE. 
   

[joeqsmith]

Using the same 2.457MHz crystal, 1kHz span and sweeping the Lite's IFBW from 100Hz (starting at 0) to 4kHz.

[jspencerg]

Nice. Interesting progression and disappearance behavior of side peaks. How many points to that surface?   

Need to get back to practicing with mcp. I have too many study topics on my plate.  It's all interesting.

Really too bad you didn't get that ifbw functionality incorporated for the original nano. I read some discussions you'd had about similar controls for the V2. Locked out of the code :-(

[joeqsmith]

100Hz IFBW = 60k/(15*Avg), Avg = 40.   I tried to collect three sweeps per setting. The Lite was set to 401 data points per sweep.  So the surface is made up of roughly  3*40*401 = 48k points.   

Eventually the side lobes exceed the span and appear to disappear.  Had I used a wider span we would see the images even with an IFBW set to 4kHz.

[joeqsmith]

Same 100Hz to 4kHz IFBW sweep but with a 3kHz span.   5 sweeps per IFBW step. 

[joeqsmith]

Sweeping my original NanoVNA at 30Hz, 100Hz, 300Hz and 1kHz. 

[jspencerg]

Glad to see you think the original nano is still worth collecting data with (since I don't have the lite).  You should keep adding features to version that supports the original  :-)

Just noticed the Solver32.  Why 32?  Does going to 32-bit lose the speed of Solver64?

Looking at these 3-D surfaces brings memories of hand plotting free energy surfaces for phase changes.  3-D plots are trivial now...

[joeqsmith]

Re:...keep adding features to version that supports the original  :-)
I have no plans to do anything more with devices requiring the old protocol or at least anything I will release to the public.   Of course, nothing prevents the firmware for any of these devices from being compatible with the newer protocol. 

Re: Why 32?
It's a hobby and I enjoy experimenting. 

Re:  3-D plots are trivial now...
Many things are now trivial compared with where we were even a few years ago.   Still, I wouldn't want to write a 3D plotting program in assembler. 

I thought about running a longer experiment with the Lite's IFBW where we run through more combos.  For example, if we set the bounds for the Averages to 1-40 but we also include bounds for E0 or 1-15.  This would give us 600 combinations.  Of these, we would have 276 unique IFBW settings.  This assumes Dislord's firmware works as they claim.   In other words, an IFBW of 400Hz could be achieved with Avg/E0 combinations of 10/15, 15/10, 25/6 and 30/5.  We would expect all of these would yield the exact same results based on Dislord's statements.   Of course, the lower the IFBW, the longer it will take to sweep. 

To speed things up we can limit the Average to 1-10.  This allows an IFBW between 400Hz and 60kHz, with 74 unique settings.  Shown with 2 sweeps per setting. 

[joeqsmith]

Before anyone asks, the drift in the peak is real.  I don't have a calibrated beach towel over the jig and the air temperature will cause the crystal to to drift.

Wrong...

Changing the IFBW settings will change the peak's centroid.  Shown at 4kHz and 60kHz.

[jspencerg]

Re: winding torroids
What was the little white jacketed coax you used when making common mode chokes,  etc?  I've seen twisted pair used in other references. I assume the coaxial has better rf behavior, yes?

[joeqsmith]

Re: winding torroids
What was the little white jacketed coax you used when making common mode chokes,  etc?  I've seen twisted pair used in other references. I assume the coaxial has better rf behavior, yes?

Same coax from the last time you had asked.  Those transformers were characterized to 6GHz.  You can read about them in the manual or search this thread. 

https://www.eevblog.com/forum/rf-microwave/nanovna-custom-software/msg4063153/#msg4063153

[jspencerg]

Last time you referred to blue jacketed coax while the smaller stuff I saw was white. I'm thinking it was the rg188(from comparing spec's), but that does not seem right.  Because:  I have some rg316 with similar specified bend radius limit as the rg188. It seemed,  from watching video and comparing myself,  that the white coax you were wrapping was more flexible than rmin=.5".
So,  the rg188 is smallest you use?
Trivial stuff for you, but all new to me.
I am aquiring supplies to make transfer relay connections using rg400u
Thanks for the details.

[joeqsmith]

Common sizes I use. 

[joeqsmith]

900MHz center, 100kHz span, 2 sweeps per setting,  sweeping IFBF from 400Hz to 60kHz. 

[joeqsmith]

Same setup but at 5MHz. 

[jspencerg]

Could you replace sweeps with center frequency?  Run from 5M to 900M. Still show delta f.

[joeqsmith]

Could you replace sweeps with center frequency?  Run from 5M to 900M. Still show delta f.

I think I understand.  You want to sweep the frequency with some fixed span and fixed frequency step size.  If so, yes I can do this.   However,  think about what you are suggesting.  You have not fully considered what you are asking or I am not understanding.   There's some missing details, or at least one major one...   

[jspencerg]

Just brainstorming: How would peak and lobe profile change as frequency changed?

[joeqsmith]

Re: brainstorming: How would peak and lobe profile change as frequency changed?

Well, our ideas of brainstorming differ.  Think about it.  What causes the peaks and lobes?

[jspencerg]

Haha. Our capacities to constructively brainstorm differ by several orders of magnitude because of difference in referential knowledge.  Fancy way of referring to my ignorance.
I realized, after some thought, that even if one had a dut with adjustable frequency response, the resonance (or whatever) would likely not be the same at each frequency.  So,  that probably would not be a good experiment to investigate the measuring behavior of the VNA.
I discovered why your rg188 seemed so flexible.  The ptfe jacket makes it so.  That is some fancy wire.  I found some mil spec new old stock from 1981 for a good price.
Figuring out wire routing for the transfer relay.  Your images help.

[joeqsmith]

Re: brainstorming

Bingo.  I already sweep the frequency with the span and I also sweep the IFBW.  Nothing there changes except we widen the span.  But to look at narrow band signals... 

We could look at the effects of IFBW on a wideband signal over some very wide freq sweep (basically look at the noise floor) but I don't think we would see anything of interest.   If I were for example to run the previous IFBW sweep using that interdigital GHz filter for example, the lobes are going to stay the same but the noise floor will change.   Even looking at the crystal filters I made is fine.  It's really just when looking at the crystals themselves or a very narrow source.

Your talking to the king of ignorance.   Mine knows no bounds and expands each day.   lol. 

If you are interested in working at these higher frequencies, cable choice is important.  RG174 while very flexible by comparison would make a poor choice.   Pasternak has most of these cables in stock and will sell it in shorter sections.  Digi-key may also carry something you could use. 

The jacket on my blue cables that I showed in many videos with the NanoVNA cracked after a few years of use.   Just too much flexing when running so many demos.

[jspencerg]

Re: cable choices.
I chose the rg400 for transfer relay connections as effort to make best setup knowing I will have minimal flex and  want great quality connection(best I can).
I'm wanting to investigate inductors, chose this high-end rg188 to minimize Impedence changes in wire with frequency.  The rg178 I have is stiff. I'll respect bend radius.
Re: Transfer relay
I read proof for using t-check to examine quality of two-port measurement(using transfer relay in my case).  More reading seems to say that most circuits are expected to be reciprocal for s12 and s21, which seems to relegate use of full 2-port measurements to checking when dut is well- behaved.  Two questions:
1. Are there any non-reciprocal dut's I could study?
2. Do you use full two-port measurements to judge quality of a dut that is supposed to be reciprocal w.re.to s12 and s21?