HFSS problem with cavity filter : version 11 vs later version ?

Help !!!!

Hi,
I simulated a narrow band cavity filter around 20 GHz. Geometry is very simple ; no curve, no imported object, all faces straight, no exotic material....I've optimized my Bandpass filter's response with HFSS version 11. Convergence is very fast with a simple setup (Delta S =0.02 , min passe = 15 --> 10.000 tetraedra last adaptive passe ) and if I tried to make a most stricky convergence (Delta S =0.005,min passes = 30 --> 70.000 t. last adaptive), S matrix still very close to first solution.:smile:

So I tried to open the same project in HFSS v 12. Results are different, my band pass filter hasn't same response, poles are shifted... only 6-7 dB of return losses for the same central setup frequency solution (instead 20 / 22 dB) :
I know meshing method have been updated in v12 (Tau mesh strict or tolerant or Ansoft classic mesh or Auto), so I tried all mesh setting and I always the same reponse... that is very different of my v11's optimization !

I also tried a stricly convergence (and mesh operation) and bad results remain. :-(

Has anyone have the same problem ?
Thanks for your Help !

simple, try it in CST time domain solver, the you know which HFSS version is more reliable.