For your reference here are the FFT's of the audio signal from CH1 while
connected to CH0 at various frequencies. 50 MHz, 299MHz, 1200MHz and
2100MHz, all with offset of 4kHz
Next the isolation measurements for 50MHz and 2100MHz.
This is done in my own HW so there are no changes in ADC settings, apart
from the SI5351 everything else remains the same.
With strong signals the noise floor increases max 10dB but the SNR
differences are as expected due to decrease in the 4kHz harmonic mixing
signal.
As you can see FFT'2 are rather clean. Only the 4kHz and its harmonics.
Close observation of the 4kHz shows its starts to wobble more at higher
frequencies explaining why it does not fit perfectly in the 480 samples you
use.
Hope you will be able to see the same once you have done the windowing of
the FFT input signal.
Erik,
I've separately emailed my results having added windowing to the FFT
signal. While my results are much improved from previously, the noise
spectrum is nowhere near as clean as yours. I'm assuming that although the
front-end of your setup is like the nanovna, the audio backend is different
and that you don't use the TLV3220 for ADC (your sample rate is > 200KHz?).
The nanovna has a very strong 1KHz component - probably a result of the
1KHz codec frame-rate. The CPU does a lot of processing each frame which
might be injecting noise.
Your observation of slight 4KHz wobble is a worry in that it might force
the need for a windowing function on larger samples.
Rgds,
Dave
