As I mentioned on the call, in the paper cited below the authors show experimentally that with a fiber length of 7km and mean DGD ~ 2ps that there is significant relative rotation of the polarization states of two pumps separated by 5nm. See Fig. 3(b).
This experiment is done in C-band so the fiber loss is lower than in O-band and the effect will not be as strong over long spans, but because of the wide channel spacing the fiber PMD is still sufficient to depolarize the signals in a relatively short distance. This both reduces the effect of FWM for aligned polarizations and increases it for alternating polarizations.
We can have more discussion at the meeting, at which time I hope we will hear more details about the simulations in liuxiang_3ca_1_0517, and how they differ from this published experiment.
Regards,
John