Some old plots from 2017 of the residual phase-noise at 10MHz when the signal is fed through an SMA-to-SFP board, SFP to optical, 2m optical fibre, then SFP back to electrical, and another SFP-to-SMA board.

One version of the board is on github:

As a comparison the noise-floor of the 3120A and a frequency distribution amplifier is also shown.