I just tried the HypLynx v9.2 IE3D capability. To start with, I just exported a very simple small 3D area from a board with a embedded differential strip lines with the top plane as GND and bottom plane as POWER. During the port definition stage, I could only choose either lower reference or upper reference for the differential ports. There is no choice to reference both lower and upper planes. So I did the 3D simulation with lower reference for differential ports first and then again with upper reference for differential ports. To my big surprise, when I check the result S-parameter models and corresponding TDR plots, the 3D simulation results is very different from the HyperLynx 2D solver. The 2D solver gives differential impedance very close 100ohm, which has been verified with my TDR scope on real board. However, 3D simulation tells me the differential imepance is ~150ohm and ~135ohm with lower and upper reference respectively. How could this be? I suspect this has to do with way the port reference is defined.