The software is able to ignore the odd harmonics of that square wave.
Works something like this:
Two Si5351 outputs are used.
CLK1 is what goes out to the DUT.
CLK0 is always 5khz higher than CLK1, they are mixed
in the three SA612A mixers, the resulting 3 channels of audio
are digitized by a 3 channel 16 bit ADC and the resultant data
fed to the ARM processor. All three channels of audio are
at 5khz, but have different amplitudes and phase.
The 5khz audio is filtered out using DSP techniques (an FFT)
inside the processor, any products resulting from harmonics will
not be 5khz.
For example, assume the signal to the DUT is 100 mhz.
So CLK1 is a 100mhz square wave out to the DUT,
it is mixed with CLK0 at 100.005 mhz resulting in that 5khz audio.
The 3'rd harmonics will be at 300.000 and 300.015,
resulting in 15khz audio which is easily rejected.
When operating at 600mhz, we're beyond the maximum frequency
that the Si5351 can produce. So CLK1 sends out a 200mhz
square wave to the DUT, and we mix that with CLK0 at 200.00166 mhz.
The third harmonic of CLK0 is 600.005 mhz, and it is now
only the third harmonics of CLK1 and CLK0 that produce
a 5khz mixing product.
Jerry, KE7ER
toggle quoted message
Show quoted text
On Mon, Sep 2, 2019 at 08:42 PM, Reginald Beardsley wrote:
FWIW Here is a display from my 8560A of the spectrum of the nanoVNA source
producing a 10 MHz CW signal and another showing how it is connected.
The strong odd order harmonic content is probably the major limitation on the
accuracy of the unit. I'm not aware of a way to correct the issue in software.
If someone is, please explain. I'd really like to know about it.
Have Fun!
Reg
Warren Allgyer