NanoVNA Custom Software

[purpose]

Morning Mr. Smith,
Well I did read the manual... well the crystal pertinent bits, just so I could give it a try before I went to bed.
Fired straight in and am currently characterising away. I also forgot to go to bed.
I'll certainly delve deeper into your writings and give any feedback I note.
Very interesting to see the very different responses of these Chinese made jobbies. No wonder they're dirt cheap.
When they're all done, I can't wait to press sort.



Once again
Very many thanks
Peter

[joeqsmith]

If the goal is only to press sort and see what comes up, just select save 10-50 times for the same crystal and try it.   If the goal is to actually measure crystals, the plot you show doesn't look right.  You can waste a lot of time taking bad data.   I suggest your time is better spent making sure things are correct before going further.   

First thing are those spikes in the frequency.  There is no way the crystal should shift frequency like that.  It wouldn't recover that fast.  Something else is going on, maybe noise.  Assuming the cables are tight and you're not touching anything, maybe it's your Nano.  Some versions of firmware would show bad readings at random.  That's part of the reason for the regression testing.  I haven't seen this problem for some time.   It should be obvious that running the same firmware removes a lot of questions and may save you a lot of time.   

Start by selecting the Xmsn Rectangular from the Main menu.   With your short installed in place of the crystal,  it should read well within 0.1dB after proper calibration (Capture1.PNG).  Make sure you understand the difference between the Nano's internal calibration and the software's.  Post a screen shot.

Assuming that the gain and noise look good, I would select Record and let it run for 10 minutes or more.   Press Record again to turn off the recording.  Select Advanced, Surface, Add Waveform and then load the file you saved the data to.   Next select Narrow Band Zoom and select S21 Mag.   You should have << 0.1dB noise P-P.   Notice mine after touching the setup takes a dive but is still within 0.045dB P-P (Capture2.PNG).  Post a screen shot.

Assuming you are not seeing a problem up to this point, select the Main menu (you should still be seeing << 0.1dB) and swap out the short for one of your 10MHz crystals.  You should see a VERY smooth curve (Capture3.PNG).    The peak is the Attenuation.  You show 20dB which seems very low.   Post a screen shot.

Select Record and save the data to a new file.  Let it run for another 10 minutes or more.   Don't touch anything and stay away from the fixture.  If you have a lot of air flow in the room, my standard practice it the beach towel.   It's too bad that the Nano doesn't have an ambient temperature sensor.  People would rather have a worthless battery indicator.   I would make a note of the temperature if you plan to try and reproduce the test or are collecting the data over long periods of time.   100ppm/degC with even a 2 degree shift and a 10MHz is easily detected. 

Select Record again to stop the recording.  Select Surface and select Flush.  Now select Add Waveform.  Looking at both S21 Mag/Phase, it better be very clean.  (Capture4 & 5)

If all of that seems good, I have no idea what those spikes are.  If a second crystals shows the same problem and you are running identical firmware, it seems there is a problem with the Nano, cables, not tightening things, external noise source....  You need to figure it out before wasting time measuring parts.

If you find the problem and get similar data, I would next validate that the C0 measurement is working.  If you don't have a good RLC meter,  I suggest installing a known capacitor.  Maybe a few values from 10pF to 100pF.   Depending what you have for parts, their tolerances can be pretty wide.  Just make sure that the numbers are in the ballpark and you should be fine.  If you insert a 100pF capacitor and it reads 50pF, something is obviously wrong.  Try another capacitor.  This shouldn't be a problem based on the data you posted for your jig.  Again, that assumes you have properly calibrated the setup. 

***
Very poor grammar for even me.     

[purpose]

Evening Mr. Smith,
If the goal is only to press sort and see what comes up, just select save 10-50 times for the same crystal and try it.
OK. Sometimes I act like a schoolboy.
You can waste a lot of time taking bad data.
A good job I actually went to bed.
Start by selecting the Xmsn Rectangular from the Main menu.   With your short installed in place of the crystal,  it should read well within 0.1dB after proper calibration (Capture1.PNG).  Make sure you understand the difference between the Nano's internal calibration and the software's.  Post a screen shot.

Next select Narrow Band Zoom and select S21 Mag.   You should have << 0.1dB noise P-P.   Notice mine after touching the setup takes a dive but is still within 0.045dB P-P (Capture2.PNG).  Post a screen shot.

Quite possibly not narrow band zoom.
Assuming you are not seeing a problem up to this point, select the Main menu (you should still be seeing << 0.1dB) and swap out the short for one of your 10MHz crystals.  You should see a VERY smooth curve (Capture3.PNG).    The peak is the Attenuation.  You show 20dB which seems very low.   Post a screen shot.

I would make a note of the temperature
23 new degrees.



If all of that seems good, I have no idea what those spikes are.
Probably operator error. I'm really not that bright.

Many thanks
Peter

[purpose]

A few spikelets and a hump when I went to get some water.

[joeqsmith]

Start by selecting the Xmsn Rectangular from the Main menu.   With your short installed in place of the crystal,  it should read well within 0.1dB after proper calibration (Capture1.PNG).  Make sure you understand the difference between the Nano's internal calibration and the software's.  Post a screen shot.

Notice how you have ran a complete calibration where in every graph I show the data is normalized.   Do yourself a favor, stop randomly pressing buttons,  go to page 71 and carefully read section 14.4.  It looks like this may have been the cause of that -19dB you first showed as you are now at -3.6dB.   I am guessing that is still not right.   I am not even sure how you would use that full calibration with the fixture.   Maybe you felt you should run a full SOLT then installed the fixture.  No matter.  It you want to get it working, you must follow the directions in the manual.  If something is not clear, let me know so I can address it.       
 
Notice in the data I have presented, the span is set to 500Hz.   You have clearly not selected zoom and are looking at a very wide frequency range.   In the manual where it says Select Zoom, you have to actually select Zoom.   You can't just skip steps.        

Once you have it correctly calibrated and the span set, then run your 10 minute collection.  There is no reason to collect 10 minutes of bad data.   Your setup should be the same as what I show or something is wrong.   Once you have that, don't go any further.  Just post a screen shot of the 3D plot.   I suspect with such a wide span, you may not see the noise problem.

[purpose]

Do yourself a favor, stop randomly pressing buttons
I identify as a toddler... it's my job.
Still, I'm glad my first joeqsmith admonishment was a wrist slap. instead of a punch in the mouth.
I'm going to have to have the night off and read the manual a few times, because following instructions and replicating your charts
are beyond me right now.
This is the latest...

[joeqsmith]

Good Job.  This is what I was expecting.   Now what is causing those spikes?   Are you using the exact same firmware downloaded from the image mentioned in the earlier posts?  If not, I would start there.   Again,  I had seen something like this with the earlier versions of firmware.   

If you are running the same firmware, you could try to move the cables and your jig to see if it has an effect but assuming you are not anywhere near it while it it running, I doubt this is the problem.   If it's not the firmware, I would guess the Nano itself.   You need to get this resolved before moving to the next step.

[purpose]

Just put in the edy555 firm 0.8.0.
Waveform update seems about half as quick.

[purpose]

Just grabbing some data.
Here's the attenuation..

[joeqsmith]

Hit that record and let's see if the firmware had any effect.

If it does, what version was previously loaded?

[purpose]

You want me to do the exact same as my previous? With the 0.8.0 firmware, or with the old?

[joeqsmith]

Oh wait, that was the 3D with the span set correctly!!   Let's see the C0 with 100pF installed in that fixture of yours.  A few caps wouldn't hurt if you have them.   

If you have an RLC meter, you could measure one of your crystals with it and plug it in and try it.   I would expect something below 10pF.   

****

Looking at some of the data I collected for various brands and frequencies, the HC49 package seems to give me between 4 and 6pF.    It looked like that was what you were using so if you don't have any caps, just plug one in and lets see what it comes up with.     

Then clear the graph, and let it collect until the part is stable. 

[joeqsmith]

When you tried the 100 and 33 in the de-5000, what does it show?   

I wonder are we pushing your RLC meter.   Another thing you could do is remove the capacitor from your jig and just run it with nothing and see what you get for C0.  I would expect it to read around 600fF to 900fF.       If you keep selecting C0, I would expect roughly 100fF of variance. 

My old BK meters compare very close to the Nano and the old HP3589A.  I have some tighter tolerance parts from ATC that are very stable which also come out.   But you're getting close now to where I ended up.  I could get all my instruments to basically agree but I was wanting to try and procure a reference crystal that had baby papers to get a handle on the absolute accuracy.   

[purpose]

100pF in the de is 105.7pF, in fixture was 106.181pF
5pf in the de is 5.14pF, in the fixture was 5.59pF
33pf in the de is 32.54pF, in the fixture was 33.356

remove the capacitor from your jig and just run it with nothing and see what you get for C0.
Varies between 500 and 850f.

procure a reference crystal
Best of luck with that, Mr. Smith. Should I ever spot one in the wild, I'll let you know.

[joeqsmith]

In the future, I may change the test frequency to 500KHz (seems like it was a bit more stable). 

Try it with a crystal and let's see what numbers it comes up with.   

Also, I am still curious what firmware you had loaded that caused the glitches.

[purpose]

Glitchy firmware was hugen79 1.0.45 NanoVNA-H_20210130.dfu.

Try it with a crystal and let's see what numbers it comes up with. 
The DE?  2.93pF @100kHz

[joeqsmith]

   Yeah, I meant run it for real with the Nano and lets see what numbers you get.  Don't run a bunch of them just yet.  I want to have a look as see if I can improve this measurement first. 

Just an FYI.   The following mentions some of the problems with the firmware in the past.  It's simple enough to test the firmware for headless operation.  Odd after a year it's still a problem.   Hopefully it helps you start to understand why I chose not to release the code to the public.   On one side we have people not smart enough to not plug them into their transmitters and the other unstable firmware.  Icing on the cake was people demanding it.  What a shit show.  You should feel privileged  .     
https://www.eevblog.com/forum/rf-microwave/nanovna-custom-software/msg2710524/#msg2710524

Saunder's standard crystal.  They never responded to any of my attempts to contact them. 
https://www.eevblog.com/forum/rf-microwave/fun-with-crystal-filters/msg3060008/#msg306000

[purpose]

I meant run it for real with the Nano and lets see what numbers you get.
Slight wobble to it, but no glitches.



What a shit show.
I'm a delicate soul. Do I have to?

Saunder's standard crystal.
I shall have a word in their shell like.

peter

[joeqsmith]

Shit show refers to the whole nanoism of it all, nothing to do with you specifically.  Again, had you not been trying to use it to measure filters, I doubt I would have uploaded it.  Good to see someone doing something besides trying to measure SWR.
 
TrC units "(pF)" were removed from the UI.   The TrC test frequency was changed from 1MHz to 300KHz to allow for lower values crystals to be characterized.  TrC measurement now uses an average of 4X sweeps (404 data points).   This improves the calculations variance.   To speed this measurement up, channel 0 is no longer collected during this time.  Overall, it may actually be a little faster.  You may also notice the open circuit capacitance being slightly lower (approx 100-200fF)  as well due to the lower test frequency.   

I tested it with a few different crystals as well as running through my test capacitors.  It seems fine.   No other changes were made. 

[purpose]

Mr. Smith,
I read all of that stuff a few days ago. The gimme generation are a constant source of disappointment.
Just downloaded the latest and about to give her a go.
Anything specific you'd like me to do?

Peter

[purpose]

Mr. Smith,

open circuit capacitance
Much the same @ 500fF-1pF.
The fixture is on a really thick (3.5mm) double sided pcb (with no vias). I'm thinking I need to address the pcb's capacitance.

5pF measures 4.5pF in new software, 5.59pF in old.
33pF measures 32.26 in new, 33.356pF in old.
100pF measures 106.25pF in new, 106.181pF in old.

I'll get back to it.

Many thanks
Peter

[joeqsmith]

If you remeasure the same part several times in a row it should be more stable now.   Attached shows my open circuited jig running at 400KHz.  I ran 130 C0 measurements and the peak to peak variance was 5fF.    I compared it with my two BK RLC meters and also tried a few 5% tolerance caps and it seems to be in the ballpark.   I'm pretty much at the limit of what I have available to to validate it.   

Assuming you are getting stable readings, based on the numbers you're seeing I would just go ahead and try it.   

You may consider building a pull test oscillator and compare the results with it vs your VNA.  I tried a few different methods and the errors were fairly small. 

Let me know how it works out. 

[purpose]

Let me know how it works out.

Will do.

Many thanks
Peter

[joeqsmith]

In order to speed up Zoom, it now works similar to the C0 measurement where Port1 data is not collected.   Also to handle more error in the peak location when Zooming, I now center the peak before changing the span the first time.    No other changes were made.     

[purpose]

Mr. Smith,
You're going to tell me I'm doing it wrong, but I've been shorting the fixture, punching 10kHz into the span and ref/chn ref, loading the crystal, then centering the peak. Span to 500 Hz and centering again. Then off to advanced for the fun stuff. Pressing zoom a dozen times wasn't working for me.

BTW, I was playing with the bandwidth in the glitchy firmware and discovered that it defaults to 4000Hz. Tried it at the more standard 1000Hz and far less glitchy in the xtal data gathering. going down to 330Hz pretty much eradicated it.
Either way, the glitches didn't seem to show up in the 3D graphs.

Nevertheless, I'm back to eddy 0.8.0.

Peter