Re: which software to simulate and optimise Ku microstrip fi
Hi mahdithdn -- Thanks for taking the time to give your opinion rather than remaining silent. It is very much appreciated.
Basically, I don't know where this "volume meshing is slower but more accurate than surface meshing" idea came from, but it is very wide spread and it is totally wrong.
First, if you have infinite computer resources, all correct EM tools will give essentially the same answer provided you use a sufficiently fine mesh. For time domain tools, like CST, you have to also get enough of the impulse response. For FEM tools, like HFSS, you also have to get the current to converge.
Thus, we have to check how much error there is as a function of analysis time. This is easiest to do for a simple circuit for which we know the exact answer. One such circuit is the stripline standard discussed above. Do the test on all tools that you can get hold of. Look at the calculated Zo and compare it to the known exact answer. Any difference is error. Refine the mesh and plot the error as a function of analysis time.
If you can't get Zo from the tool, make the line width so that the exact Zo is exactly 50 Ohms and the length is exactly 1/4 wavelength, then 100*mag(S11) = percent error in Zo. Mag(S11) = 0.01 means 1% error in Zo. (You probably should do this for volume meshing tools, they usually have an independent 2-D solver that calculates Zo. You want to evaluate the 3-D solver, not the 2-D solver.)
We did this for CST and we were given results for HFSS. What we found is that CST is clearly better than HFSS. That is why we decided to represent CST in North America. Of the volume meshing tools it is simply the best. If CST had not been clearly better (or even just a little bit better), we would have stayed out of volume meshing.
We also found that for this circuit at 0.1% error, Sonnet is about 1000X faster than HFSS. If we back off to 3% error, it is now only about 100X faster, for this circuit and for our specific set up. (The volume meshing FEM HFSS, not the surface meshing MoM Ensemble HFSS was used.)
But...Don't believe me or anyone else!!!! After all, I work for a (kind of) competitor of all these tools. Do the test for yourself, on your computer, using stripline dimensions and substrates similar to whatever you use in your work.
If nothing else, use the Ensemble part of HFSS for your planar circuits (I have not seen test data on Ensemble, but if it is a correctly done, being an MoM surface meshing code, it should achieve a given level of accuracy much faster than any volume meshing code).
You will find that volume meshers are great for 3-D arbitrary structures and surface meshers work great on planar structures, and in practice there is actually very little that either tool can do well in the other's territory.
In the US, the national motto is, "In G0d we trust." To that, as scientists and engineers we add, "All others must have proof." So, don't trust me. Get proof.