Accuracy about difference versions of HFSS?
I used HFSS 8.5 and 9.2 to simulate a waveguide polarizer. With the two versions , i found that the VSWR and ampltude are similar, but the phases do not agree. In 9.2, the phase difference is 90 degree and in 8.5 it is 84 degree. I want to know which version is more accurate and more raliable. I hope i can share your experience.
Urgent and look forward to your replies!
Thank you very much!
I feel that u should check the mesh size in both the cases.
Hi,
with experience, I have got excellent acccuracy results with MICROSTRIPES, 3D EM Solution
you may want to give a try (HFSS is great solution too)
regards
AW
with the xperience ver 9.0 is better
i presently use HFSS 9.1 and i've found it definitely more reliable than previous versions..
i would stick to the version 9 results if i was in your case...
regards,
wiztronix
I found version 9.0 9.1 9.2 are not reliable, version 8 and version 10 are better.
Regards
Emund
I wonder this question too.
i have some experience from two antenna team/lab.
one choose 8 the other choose 5, the reason both are the accurate.
i take the version 10. any comment on pl.
Wow. This thread was an eye-opener for me. I work for Sonnet. A critical part of software development is something called "regression testing". We have a few thousand circuits in our regression testing for Sonnet, some have been there for over 20 years, and they still give the same results we got over 20 years ago. We constantly run the regression testing while doing development to make sure we get the same answer for the same circuit all the time. While there are occasional exceptions, we find different versions of Sonnet give almost exactly the same answers, there might be a last digit change that is in the numerical noise, but that is about it. If there is a bug of some kind that we fixed, then there might be a more significant change, but this is rare.
If the results change on most regression test circuits by even a few degrees in phase or a few tenths of dB in magnitude, regression testing becomes unusable. If, say, 90% of all the circuits in your regression test fail (i.e., give different results, thus requiring a human to look at the data and make a judgment as to correctness), then the regression test becomes useless. This is not a tenable situation for a commercial tool. I do not think this could possibly be the case with any major commercial tool. My guess is that your example is a rare exception.
Fixhyxl, my advice is to see if you changed something like meshing density or such, that might explain the difference. If there is no difference in the input data, you should contact support, they should be able to figure it out. And if you see this happen with Sonnet, or with any EM tool, my advice is the same. Good luck.
i think it depends on your structures
reliabilty of HFSS after ver9.0 is very good..only difference with versions is in computation speed
whenever you are doin 3d modeling you hav to b careful abt defining boundary conditions and your structure that decided the reliability of modeling