A few years ago, I have tried to use my VNA-H as a spectrum analyzer. For sure there are many design ands hardware caveats dealing to limited relevant results. However, below are some interesting results that I have got. Sorry parts of comments on pictures are in french. Port 2 is used as input with 0 dBm max level

Large span measurements on a SINGLE signal source (CW) are possible, but results about relative levels are false. The reason is the amplitude-frequency response provided by the VNA is driven by the calibration corrections values (THRU). Here is the same signal (50 MHz) displayed on an Ultra SA and the NanoVNA-H. There is a difference of 23 dB on the harmonic 2 part.

01 and 02 pictures

Small span measurements provide confusing results for several reasons : low IF values (here 12,5 kHz) dealing to a false product (mirroring), and also spectrum aliasing according to span range value. Here is the same 10 MHz CW signal displayed on the NanoVNA-H with changing only the span range ! Totally unusable ...

03 and 04 pictures

The only relevant measurements that I got is in using the NanoVNA-H as a "modulation analyzer" for AM and FM (narrow bandwidth). Span range value is 10 kHz in order to avoid the previous artifacts.

- AM : relative sideband level measurement : AM rate 30% (picture 05), AM rate 100% (picture 06)

- FM : Bessel Null method (first zero) : AF=1 kHz, deviation = +/- 2,5 kHz, no null (picture 07) AF=1 kHz, deviation = +/- 2,4 kHz, perfect null (picture 08)

Comparing the previous results to a real modulation analyzer (Racal Dana) proved to be relevant.

Jean-Roger