Phase detection accuracy with USRP N210 and CBX-40

Question

I have designed a simple TX/RX system in GRC with a single USRP N210 and CBX-40.

I am transmitting sinusoidal signals at various frequencies within the 2-6GHz range and detecting the amplitude and phase of the received signals at each frequency when connecting a coaxial wire from the Tx port to the RX port of the USRP.

I am doing the FFT at reception and "looking" at the correct sample of my IF (samp_rate/bin_size*N). I therefore detect the amplitude correctly.

However, I observe that the phase is only detected correctly (hence, it has always the same value because the cable connecting both ports has a fixed length), for certain frequencies. In my case, for 2GHz, 2.1GHz, 2.2GHz etc. (every 100MHz).

I would like to have a finer phase resolution and I don't know what of the HW components (could be the PLL) or SW settings could solve my issue. Could it be related to the "tuning policies" of the USRP?

Answer 1

This is a bit of speculation, but the only way I can see this happening is if the LOs are generated from a common (or phase locked) 100MHz reference. In this case, at the 100MHz multiples the PLLs are probably operating in integer-N mode, and so they lock when an edge of the LO coincides with the reference. In this case the two systems are phase locked. When another frequency is chosen that is not a multiple of the reference, the PLLs are operating in Fractional-N mode. In this case, if both SDMs start at an identical time from an identical state (extremely unlikely), the outputs will be phase coincident. Otherwise, they will still be phase locked, but with some random phase offset.

It is possibly to get fractional-N PLLs in phase, but generally this requires some proprietary technique from the PLL manufacturer. Look up using PLLs to run phase arrays, and features like ADIs 'phase resync'. https://ez.analog.com/rf/w/documents/9662/phase-resync-phase-programmability-and-phase-coherence-between-multiple-fractional-n-plls