NanoVNA V2 Forum

Note: this page is a mirror of https://groups.io/g/NanoVNAV2.
Click here to join and see most recent posts.

Calibration problem


Salil Tembe 2020/07/05 08:40

I am using the same cal-kit that came with NanoVNA v1. Any one among you used that and got good results? I am failing to get anything good.

After calibration, I connected a RG405 with 50ohm load at the other end. The s11 was as bad as -25dB. It should have been better than -35 to -40dB (had been tested on high end VNA).

hwalker 2020/07/05 11:12

On Sun, Jul 5, 2020 at 08:40 AM, Salil Tembe wrote:

I am using the same cal-kit that came with NanoVNA v1. Any one among you used that and got good results? I am failing to get anything good.

After calibration, I connected a RG405 with 50ohm load at the other end. The s11 was as bad as -25dB. It should have been better than -35 to -40dB (had been tested on high end VNA).
====================================================
Salil,
My first V2 purchased from Tindie came without a cal kit or cables. I used the cal kit from my hugen NanoVNA-H 3.4 and had return measurements better than 45dB on the 50ohm standard up to 2 GHz. More problematic was the RG-316 cables that came with my hugen NanoVNA-H 3.4. I later ordered a hugen S-A-A-V2 and the ss405 cables that came with it are giving me much more reproducible measurement results. The return loss measurements for both V2's are pretty much the same but less noisy with the ss405 cables.

- Herb

Salil Tembe 2020/07/05 21:22

I tried with RG402 cables again. Everything seems ok below 2GHz. It gets bad above 2GHz, especially in the 2.5GHz to 3GHz range. I am not sure if it's my calibration or the cal-kit itself or the cables.
To be sure, I purchased ultra low loss 18GHz cables which are about to arrive in few days. I will report back when i try with those cables. I am also using good quality mini-circuits adapters.

MarkZ 2020/07/06 06:01

I tried an experiment, to see what was up with the unit I bought via Tindie a while ago.

I used nanovna-saver and a setting of 1023 data points. Then I calibrated with the Rosenberger kit from SDR-Kits (no cable, just connected straight into the SMA adapter).

Then, I did four sweeps:
1) immediately resweep the Rosenberger 50ohm load (which is known to be 49.35 ohms)
2) sweep an amazon 50 ohm load which is known to drop off at high frequencies
3) sweep an Amphenol 50 ohm load (the less expensive cal kit from SDR-Kits or available via Digikey)
4) resweep the Rosenberger load (since I didn't let the system warm up this should drift)

Images included.
-Mark

MarkZ 2020/07/06 06:43

I tried the experiment again with my other v2 from Tindie (yes, I bought two; both from the first production run). This one has no battery in it and I won't put the graphs in because they're virtually identical to the first set of graphs except interestingly there is no drift in the fourth graph. I tried a few times and consistently it was almost exactly the same as the first graph. If I had to guess that fat battery stuck between the two boards in my first unit is causing additional heating.

-Mark

OwO 2020/07/06 22:03

-20dB logmag(S11) is considered good for a cheap 50ohm load. Even the
calibration standard that comes with the V2 from Tindie is around -30dB
at 3GHz. It's physically impossible to do much better than this (the
geometry of the SMA mating transition has inherent reflections). The
only way professional cal kits do better is they come with kit
parameters that tell you the exact magnitude and phase of the reflection
at each frequency, not by having < -30dB reflection coefficient. Not
that it matters, -20dB reflection means 1% of the power is reflected,
you should never try to tune your antenna or amplifier below that.

reuterr 2020/07/06 11:00

Hello Mark,
After calibration with a 50 Ohm load from nanoVNA at the end of a 50 cm RG316 coax cable,
I measured S11 and S21 from 50 KHz to 3 GHz, see the attached screen shot.
S11 was better than 40 dB.

But it was confusing, that the mark at 200 MHz tells -40.8 dB instead of -60 dB from the chart.

73, Rudi DL5FA

Salil Tembe 2020/07/06 11:33

Thanks for your responses. I have decided to get some good quality connectors and cables and a huber and suhner termination.
https://in.element14.com/huber-suhner/65-sma-50-0-1-111-ne/rf-coaxial-sma-terminator-50-ohm/dp/2473838?st=50%20ohm%20termination Looks good?

Siegfried Jackstien 2020/07/06 18:42

you have to push the slider below the scan to the 200mhz mark ... for me
it look that your slider is at the far left and so is the mark in the scan

then when you push slider right spot the reading should be ok

how many dots did you use to measure that scan??

greetz sigi dg9bfc


Am 06.07.2020 um 18:00 schrieb reuterr@web.de:

Siegfried Jackstien 2020/07/06 18:46

spec says up to 18 ghz yes ... but it also says at 4ghz swr is 1.1 ...
hmmm is that good enough for calibration??

greetz sigi dg9bfc

Am 06.07.2020 um 18:33 schrieb Salil Tembe:

reuterr 2020/07/06 11:59

Hello Mark,
In order to be sure, I measured again, and made a photo of the nanoVNA-V2 screen.
See the attached photo.
It shows the same values like with the program VNA View.
In this case a 50cm RG402 coax cable was used.
73, Rudi DL5FA

Salil Tembe 2020/07/06 12:19

Siegfried, 1.1 SWR It should be fine I believe.

Another problem I have is the adapter delay compensation. So, the NanoVNA pc software lets you compensate it but didn't find any option to do it on the device itself. Got any tip for that?

Kurt Poulsen 2020/07/06 22:03

Hi Salil
There will always be impedance transformation in a cable reducing the return loss measurement when you are not calibration at the end of the cable. By definition when calibrating at the end of the cable with a 50 ohm load which in the calibration setting are supposed to be pure 50 ohm and then measure the load you use for calibration the return loss it then by definition infinitum, as that is what the calibration setting is defining for a pure 50 ohm load, so what you measure is the noise floor of the S11 port0 of the NanoVNA.
However there are small contribution for the internal crosstalk so in theory a correct statement
Kind regards
Kurt

Kurt Poulsen 2020/07/06 22:32

Hi Mark
I have done the Rosenberger kit characterization for many years and have measurements up to 6GHz. To use the male kit together with the NanoVNA S-A-A-2 you need for the NanoVNA saver the L/C coefficients for short and open and for the load actually a individual set of data where you only have the DC resistance very accurate measured by SDR-Kits, but not the frequency dependence and also the small shunt C is has embedded. For the QT software you need a set of s1p for S11 SOL and a s2p file for the S21 Thru part of calibration.
I will when I get the time produce these data and publish here, as well a document how to top tune the load using a UT401A or RG402/G rigid line which also is used for testing any calibration of any NanoVNA.
But as said when I get time I have several project ongoing right now
Kind regards
Kurt

Dragan Milivojevic 2020/07/06 23:01

Datasheet states 1.05 up to 4GHz which is equivalent to ~30 dB of RL.


On Mon, 6 Jul 2020 at 20:55, Siegfried Jackstien <
siegfried.jackstien@freenet.de> wrote:

Siegfried Jackstien 2020/07/06 21:09

dc to 4 g = 1.05

4 g to 10 g = 1.1

10 g -14 g  = 1.15

14 g -18 g = 1.25

i took the second line ... yes i have read the specs :-) ... i thought
for calibration it should be better ... not?!?

dg9bfc sigi

Am 06.07.2020 um 21:01 schrieb Dragan Milivojevic:

Dragan Milivojevic 2020/07/06 23:55

More is always better. The question here is do you need it.

On Mon, 6 Jul 2020 at 23:12, Siegfried Jackstien <
siegfried.jackstien@freenet.de> wrote:

Siegfried Jackstien 2020/07/06 23:28

my day job is measure and control things in the industry

and if you wanna measure a thing more or less exact ... you need a
calibration that is BETTER as what you wanna measure

so if you want to calibrate a nwa ... i guess the standard should be 40
db or better ... not 30 db (or less on top end)

i know that especially on the high end its difficult to get to such low
number ... but it is possible

so i would say a good standard starts at -60 ... going slowly up to
maybe 40 on 3ghz and only go up to 30 at 6ghz and above

so it should be 40 or better on the higher end of the vna2/saa2

dg9bfc sigi

ps REAL good calibration standards are expensive ... guess why

Am 06.07.2020 um 21:55 schrieb Dragan Milivojevic:

Salil Tembe 2020/07/06 21:43

Kurt,
It would be fantastic if you can publish the s1p files. Take your time.

MarkZ 2020/07/08 05:20

Hi Kurt,

I'd love it if you could post values for nanovna-saver for the Rosenberger calibration kit (or any others). I'm including my absolute guess at values (other than R which is crisp) to point out the various non-linear settings.

To Others: there were some posts to me that didn't seem to be questions. I hope I didn't miss anything. The one note about noise in calibration - when I finished calibrating with the load the S11 was virtually a straight line at -44dB - kind of the definition of calibration. Doing the sweep right after that caused a 'lot' of noise to appear because at these levels any noise at all shows up on a chart as noisy ripples.

-Mark

tom Kawala 2020/07/08 14:30

Whatever VNA, cheap or high end, I would always get the best coax possible.
Does not have to be expensive. I got from eBay semiflex " Huber+Suhner
Sucoform 141 Hand Formable SMA Plug to Plug Lead 400mm OM0898 " 2 pcs for
30 quid cut them in half. One soldered with short, one with load, one open
and one for measurements. Calibrating for each use, I used this for
matching antennas in sub 1GHz bands and the S11 results with NanonVna's
and Smith chart match produced way better tuning than previously with
expensive VNA, default wires and cal kit in a professional lab where they
do measurements for a living. I followed their advice to make my own cal
kit and wires with the most predictable connection to my boards.

On Wed, 8 Jul 2020 at 13:20, MarkZ <mzachma@gmail.com> wrote:

ok1vaw 2020/07/08 06:47

Could you place a picture of the soldered load, open and match? I am quite curious as far to achieve a good standard is not easy. For example leaving open a perpendiculary to its axis cutted semirigid cable has a significant stray capacitance and it is able to radiate nicely.

Kurt Poulsen 2020/07/08 17:07

Hi OK1VAW
Not really in the lower frequency range as the NanoVNA operate. An UT141A rigid line has an capacitance of 0.538pf "to the world" and that is not much of an antenna See attached proof
It is however a good sniffer looking from where a radiation exist on an PCB

Kind regards

Kurt

tom Kawala 2020/07/08 16:41

blob of solder for short. 2x 100R for load. cannot be bothered with
covering the open ends. so the other two used for open and measurement have
nothing attached. cut is done with dremel blade.

On Wed, 8 Jul 2020, 16:08 Kurt Poulsen, <kurt@hamcom.dk> wrote:

Siegfried Jackstien 2020/07/08 16:25

lay one 100 ohm resistor flat and put the second upright (standing on
the side) ... i have seen that in an article where somebody measured
especially the higher end ... and best results (on a pcb with a 50 ohm
stripline) was with one resistor flat and the other on its side

maybe better one to left and one to the right (like a "T") ???

greetz sigi dg9bfc


Am 08.07.2020 um 15:41 schrieb tom Kawala:

Kurt Poulsen 2020/07/08 22:07

Hi All

Sorry, I had a bug in the model I made for UT141A. It is practically the same for RG402/G
The fringe C for the open line is 41.056fF (41.056e-15F) this value can be used for adding in a calibration kit file both for the open and the load where a further 50Ff shall be added for the solder used for the two 100ohm resistors. If you need the one way delay then divide the capacitance by 50 and you get the delay in ps and remember one way as used for the NanoVNA saver, but for other VNA’s twice the value.

A further guide can be found in

http://www.hamcom.dk/VNWA/Some basic knowledge about a transmission line.pdf <http://www.hamcom.dk/VNWA/Some%20basic%20knowledge%20about%20a%20transmission%20line.pdf>

Kind regards

Kurt

Siegfried Jackstien 2020/07/08 22:35

hmmm i see centre conductor but not shield

should not both be there ... and a field between both??

dg9bfc sigi

Am 08.07.2020 um 20:07 schrieb Kurt Poulsen:

Kurt Poulsen 2020/07/09 11:27

Hi Sigi

That is because the model is axesymmetrical that mean for centerline and outward, sp you se only half of the line 😊

Kind regards

Kurt

Siegfried Jackstien 2020/07/09 10:55

ok that i see only half of it is understood ... but i miss OUTER
conductor and the field between inner and outer

see my "rough" drawing

(for clarity also added left side ... the "missing half")

maybe i am complete wrong

i never used that software yet ... but downloaded it and will try to
learn :-)

same as i did with sonnet to develop the poty patch together with mike
willis ... he was using cst studio and i was using sonnet ... both use
different math but end results came very very close

i think it was the same thing like when comparing atlc2 and femm ... not
exact but close to 1% or less ... good enough for ham use :-)

when you then build it in real you have to finetune such an antenna with
an analyzer anyway ... and here we are ... analyzer?? yes ... good
cables???yes.... good calibration standards??? hmmmmmmm questionable
.... grin

greets sigi dg9bfc


Am 09.07.2020 um 09:27 schrieb Kurt Poulsen:

Kurt Poulsen 2020/07/09 20:25

Hi Sigi

This is how the field lines run like inside the PTFE
Kind regards

Kurt

Hvis jeg vælger at Teflon enden ikke har boundary til Air

iz1fks 2020/07/11 06:38

Hello,

finally I have oiught a Nanovna_V2 from hugen with kit and metal case. Is there available any calibration data for the calibration kit to set parameters in nanovna qt or vnasaver instead of using ideal ones?

If I would measure the sma kit with a high end VNA I would only need the s11 of each component? what about the through?

regards
Phil

Kurt Poulsen 2020/07/11 23:38

Hi Phil
I will soon create the s1p files both for the QT software and the L and C coefficients for the NanoVNA saver
Until then see the document I made in November 2019 as a start
Kind regards
Kurt

Siegfried Jackstien 2020/07/11 23:40

hmmm should it not be like this?? (there is currect on outer part of
inner conductor and on inner layer of the shield (not to mention common
mode current travelling on outer side of shield)

greetz sigi dg9bfc


Am 09.07.2020 um 18:25 schrieb Kurt Poulsen:

Kurt Poulsen 2020/07/12 03:44

Hi Sigi
It is electrostatic static simulation end the line are representing as filed strength. The ground had potential 0 V and the center conductor 1V by 0 V there is by definition no potential so that is why the filed line are like they are- It is like a magnet you are approaching the closer you get the higher magnetic force. Remember the Air has and impact and at distance it has potential 0V
Kind regards
Kurt

Barrett Poe 2020/07/12 00:14

Salil,
Are you only doing a one port measurement? If not, what do you do with the other port? I don't think you can account for the termination into port 2 unless you get creative in the current software. Put a termination on your second port and remeasure.

iz1fks 2020/07/12 02:57

Hi Kurt,
thanks for info and for the pdf. I read it some time ago but I was not able to find the file. i suggest to put in the file section of the forum.
regards
Phil

Kurt Poulsen 2020/07/12 13:41

Hi Phil
Will do when I got it updated for both the NanoVNA's and the NanoVNA V2's
Kind regards
Kurt

To reply to this topic, join https://groups.io/g/NanoVNAV2