How do I know if my NanoVNAV2 has been damaged? I know this question has probably been asked numerous times, but a keyword search of "damage" didn't yield useful information. Any direct answer or a link to existing answers would be appreciated.

I have not been looked at the uncalibrated raw data for a long time. I know SOLT calibration can easily hide the effect of a broken test fixture, so now I find myself start wondering whether my VNA is still working at all, after years of use. I heard sometimes the ESD-damaged RF switches at the ports can have subtle failure modes that still enable measurements, but with degraded input and output impedances.

An obvious and rigorous answer is to use another scalar or vector network analyzer to measure the impedance of Port 1 and Port 2 of the suspected NanoVNAV2, at reduced power levels. If |Z| or Z still looks reasonable, it's probably fine. You can even hack together a crude setup using a spectrum analyzer and a directional coupler to make a return loss bridge. I have these tools, but it sounds like a tedious and somewhat tricky experiment (both the output frequency and injection frequency must be adjusted at the same time to make them spaced apart). Are there any shortcuts or quick checks that I can do?

The images attached are the uncalibrated measurement results read by NanoVNA-Saver and plotted by a custom script.

For the isolation test, I used a matched load at Port 1 and a 6 dB attentuator at Port 2 (I couldn't find another RF load at the moment, but an attenuator approximates it enough). I found the uncalibrated S11 chart has a resonance dip at 2.4 GHz, at other frequencies, the worst return loss measured was about 15 dB. The measured S21 has isolation better than 80 dB below 3 GHz, above 3 GHz it degrades to 50 dB. For the "direct" test, I used a short coax cable to directly connect both ports. The measured S11 plot shows multiple resonance peaks with 40 dB return less, likely due to quarter-wave transformation with an imperfectly matched Port 2, the worst return loss is better than 10 dB. The measured S21 plot ranges from 1 dB of gain to 4 dB of loss.

Are these results to be expected by the raw readings with calibration removed? What are some other quick checks that I can do?