Another point to ask:
Using HSPICE models for connectors for the diff signals, does Hyperlynx assume an ideal common-mode current throught the connector? Is it possible to see the effects of adding more ground connections nearer to the signals to the model and seeing the reduction in crosstalk, due to smaller loop area? I have attempted this by connector the spice nodes to either NC or DGND.
Are they HSPICE models or just SPICE models? I'm guessing they are probably just a passive SPICE model, but just wanted to check.
On the waveform issue, the behavior does seem odd. Have you checked to make sure that the outputs on those signals are set to the right settings (i.e. out and output inverted). It looks like they could both set to output and the differential sum is 0V which would give you that non-switching behavior with a little bit of "noise" due to any asymetry in rising and falling switching edge.
If you can post the design example, maybe I can take a look. The alternative is to submit an SR and the support guys can most likely get you going with this fairly quickly.
On the adjacent grounds in the connector, HyperLynx does assume an ideal reference (as do most circuit simulators that aren't fullwave 3D). Whether you see an impact on the singal by tying multiple GND nodes together will really depend on how the model is constructed and if there are inductive and capacitive coupling matrices between the signal and those other nodes. If there are, then you should see the impact of that in your HyperLynx simulation.
The models are HSPICE and I am using the HyperLynx-SI simulator.
I have checked the signals and they are set to the right settings. I will submit a SR, since this issue is relatively import to my work and I require a solution as soon as possible.
As for the connector model, I am not very familar with spice models in this context, however I have uploaded the model file for you to have a look at.
HM_erm8-02_erf8-05.mlm.zip 9.4 KB