Looking at the code in dsp.c, nanovna.h, and nanovna.c,
It looks like the two synthesizers are set for an offset frequency of 5kHz - that is, the LO to the mixers is 5kHz higher than the frequency coming out the CH0 port.
The stock edy555 version samples for 48 samples at 48 ksps, so it grabs 1 millisecond. That's multiplied by sin and cos at 5kHz (in a table in nanovna.h) integrated in dsp_process() to produce the values that get processed into the measurement.
So, to a first order, the detection bandwidth is 1kHz (= 1/T). The effective band pass filter is a sin(x)/x kind of shape (I didn't see any windowing or weighting).
so interfering signals (i.e. broadcast band interference), if within, say, 20 kHz, of the test frequency might have an effect.
The "selectivity" (in radio receiver terms) is set by the audio low pass filter formed by and the input Z of the ADCS (10k?, set in tlv320aic3204.c). I didn't look long enough to figure out if edy555 programmed the chip to have a low pass filter (it has some ability to do FIR filtering)
