After concluding yesterday that there are really too many dealiasing failures with German data currently, I increased the size of the window that determines how many neighbouring velocities are used in estimating the correct velocity. This seems an improvement, though dealiasing errors will remain.
