FEATHER is an experimental parameter. It can be used to duplicate the de-tuned trap integration of HSPICE. It’s use is not recommended.
As far as trap ringing, I realize it’s disconcerting, but in a sense it’s giving the right answer: The area under each trapezoid is correct, so one knows that the differential equations are correctly integrated. It can be reduced by either (i) using “.options method=Gear” (ii) stipulating a lower trtol, or (iii) stipulating a smaller maximum timestep. Gear is not recommended because it adds a substantial artificial damping to the circuit. I know of two cases where the use of Gear integration let an IC designer believe his circuit was stable until silicon said different and a turning Gear off confirmed.
Unlike some prior art, I don’t smooth trap ringing out it out at all because one needs to know what the gates and flop truly see.