Re: Opinion need it
You didn't say what kind of topology is used for your filters.
Here are my opinions, for what they are worth:
If you are looking at 3D cavity resonator filters, or waveguide cavity filters with tap resonators, inductive posts, irises, etc., with very high Q (very narrowband), then I would recommend you consider a finite-element frequency domain 3D EM software, like HFSS, Concerto, the CST FEM/freq domain tool, or other finite-element approach.
If you are looking at a wideband 3D (non planar) filter design, then I would recommend a time domain 3D software, like CST time domain solver, REMCOM, IMST or other FDTD or FIT approach software. The time domain codes seem to be more efficient when you have a wideband simulation.
If you are designing planar filters inside of a cavity, then I would suggest that you consider a shielded planar Method of Moments simulator, like Sonnet or AWR EMSight. These tools automatically assume a cavity around your filter. Sonnet even lets you put internal ports very close together inside your filter so you can tune extra coupled line sections or open/short stubs in a fast circuit theory simulator (like AWR or ADS) outside of the slower EM simulation loop.
Unshielded Method of Moments tools (like Agilent Momentum or IE3D) are better for antennas because they assume there are no shielding walls. They can do filter simulations without cavities, but their cavity simulation effects are more of an approximation, I think. For cavities with close spacing to the filter, the simulations must be very accurate because they have a fairly strong effect on the passband edges of the filter.
--Max