VNA beginner question

[Phil1977]

Hi,

I have a NanoVNA that´s laying since quite a long time in the drawer and that was so far mainly used to measure the length of coax-cables.

Currently I´m playing with micro-coax cables and I´m trying to figure out if e.g. SMA-crimp connections are good or bad. It seems like the reflection measurement is most sensitive to impedance mismatches, so I´m starting to get a calibration that is as good as possible.

Can anyone tell me if the following spectra show something generally wrong?

That´s the 50 load directly after calibration


That´s the 50  load just after opening it up and re-attaching to the SMA-socket:


In best case both images would be identical - but are a) SMA connections just generally that unreproducible, are b) RF-connections just that sensitive that these variations are normal or is c) something wrong with my VNA or with the calibration kit?

I´m using a NanoVNA V2 clone. I have three 50 -calibration loads and this one is the most stable of all three.

[mwbajor]

Im confused, I'm not familiar with the NanoVNA and I could be looking at it wrong but both S11 and S21 plots look almost identical.

Are you asking why after calibration you don't see identical S-parameters? Are you asking why you're not measuring exactly 50ohms?

If so, the answer is you are close to 50Ohms, (0.06%close) which for most cases is definitely good enough. You're not going to get anything perfect.

If you are looking for interruptions in the cable connections, you might be able to use TDR (time domain reflectometer) mode if the VNA has it. That will give you the distance to the fault and a rough idea of how bad it is. Bigger reflection, bigger break or VSWR mismatch.

[Phil1977]

Thanks for the reply, sorry for not asking more precisely...

It´s about the S11 plot. In the measurement directly after calibration, the reflected power is down to -70, partially to -80db at low frequencies.

Just after detaching and reattaching the calibration load it´s about 10db more. Did you not see it because this is just to be expected or should even a cheap NanoVNA be better? What is a reasonable SNR you can expect there?

I used the TDR more quite frequently to check cables - but now I wanted to find out how good or bad some coax crimps are. The impedance mismatch was never that bad that there was a strong reflection, but it seems it really doesn't take much mismatch in a RF conductor so that you get e.g. a -40db reflection. In other words, I´m looking for these small mismatches and I´m asking myself can a pocket sized VNA be good enough to find them?

[Marsupilami]

This looks like a very reasonable change between insertions.
Note though, that the ~70dB what you're seeing is nothing realistic anyway.
During cal what you're telling the analyzer is:
-Here, take this load. It is 50.0000000 Ohms. My word is law.
Then the analyzer says -Sir, yes sir!
So when you're performing a measurement right after you get something very good as the analyzer goes
-Well I don't know, but you told me before this was exactly 50.00000 Ohms, so who am I to argue.

The proper way to evaluate how good your cal is to use an independent, known load.

But thinking about just repeatability this is good. The ~10dB difference should be looked at relative to full scale, not as a 10x difference.
The error between a reading of -70dB and -60dB is something like 0.0001% of full scale.

HTH

[virtualparticles]

@Marsupilami is correct. The residual directivity, which sets the error floor for your S11 measurement is probably -15 dB for +/- 3.3 dB, 3 Standard deviation results. That's because the load standard that comes with those tiny VNAs has an uncertainty of about -25 dB at best. Anything below -25 dB has no real meaning at all. -40 dB, for instance, simply means that the real reflection is no worse than -25.