I am a recent graduate of West Virginia University using a trial license. An important part of my senior deign project was accurate simulation of group delay of a co planar wave guide, but I have noticed that increasing the number of frequency points in the simulation significantly changes the observed group delay curves and the simulations seem to have some other inconsistencies.
I have attached two projects, one using the built in co planar wave guide model and one using an Analyst model which I believe is set up correctly. Can anyone tell me which one is guaranteed to be accurate if either or what mistakes exist in the models? Also, what is the reason for the inconsistency in the group delay measurements relative to the number of frequency points?
I have observed that the group delay magnitudes at some points on the curve, especially the zero frequency point, increase as the number of frequency points increases.
I believe you are experiencing some EM structure set up issues. First this is a co planar structure and there is no need for the increase in simulation time from using the full 3D Analyst solver. This is a structure that should be simulated in AXIEM. The mesh of the structure is way over done, make sure that you are edge meshing, and set the density to normal. Then increase the grid size to 0.1 mm and look at the mesh, that's what it should look like. Also please use AFS, there is no need for this structure to simulate all 1000 points. Lastly really think about the physical dimensions of your structure and setup. You have set the boundary conditions to basically be some sort of grounded CPW because of the distance between your top and bottom BCs in comparison to your freq range. If you sweep the distance in the AIR to the top boundary or change it to approx open you will see the effect it is having. With a few of these changes the group delay is much closer to the closed form model prediction.
I believe I have made all the corrections you suggested but I am still seeing a very significant difference in observed group delay for the EM model and the built in model. Could you perhaps show me your results compared against the built in model and help me diagnose what I am still doing wrong? Additionally, for group delay simulation in general, should I be preferring the electromagnetic model or the built in model?