Rob
Forum Moderator

Try "neighbouring cell" on the backflow direction. The radial velocity looks very odd on the last image. 

To clarify, you're running 100k particle tracks and displaying those? Have you got stochastic tries turned on?