Mesh Operations in HFSS
First of all, HFSS is quite challenging.
I am new to HFSS and have a question. There is the option "Mesh Operations" in HFSS. Where I can refine or define my own mesh for the problem.
How and when do I have to apply the operation in order to take effect ? Before or after the "Analysis" process ? For example, I have a problem with 30000 tetrahedra (which is shown to me in "Mesh Statistics" (right-click on Analysis)) and want to reduce the FEM part number to 5000 for an airball, so computation time doesn′t take too long.
How can I do that ? Online help doesn′t take me any further.
Thanks, Sebastian, Germany.
Under the "HFSS"->"Mesh Operations"->"Initial Mesh Settings", you can adjust the initial mesh to be more coarse.
Good Luck!
Yes I have read about that. The Problem is, that I have HFSS 11 beta and the feature ain′t available yet in that version. There must have been another way in earlier versions.
When i finish a simulation (here 7 el yagi antenna) it tells me "normal completition of simulation". When I apply mesh operations after that it won′t do any difference in results.
So what do I have to do to apply mesh ops in HFSS 11 beta ?
Sebastian, Germany.
Hi Sebastian,
mesh density determines the accuracy of results. From my FEM solver experience, only 30000 tetrahydra is very few for such a model with several wavelength dimension. Remember that FEM needs to mesh the entire volume (air or any other material). This is different from NEC or other MoM solvers that only mesh the conductors!
Best regards
Volker
My computer is too old anyways. A 1500 MHz Athlon with 1,3 GB RAM. It takes ages to analyze. My goal is to learn HFSS not to work with it here.
7 el Yagi ... it is about 70000 tetrahedra by now. Is an example. No conversion yet. I am not really clean with all the analysis, there are still errors in it, something I don′t do correctly. For a 3 el yagi I get good results. 7 el might be too much. grrrrrrrrrrrrrrrrrrrr. Damn computer.
Anyways, I don′t know to apply mesh operations. In HFSS beta 11.
Sure, but with too coarse mesh you will get nonsense results. FEM requires much higher computer resources than MoM analysis. You need to sample the E&H fields with fine enough resolution, and that cubed in 3 dimensions. For your electrically large model (several wavelength), that is a large model, and adaptive mesh refinement will take mesh count up even more.
Such yagi antenna is simple in MoM, but a large model in FEM. Get a fast multi-core machine with >32GB and HFSS makes sense to use. On your computer, it's the wrong tool because the method requires massive numbercrunching with much RAM.
Ah ok thanks. Now it makes more sense. I used to work with EZNEC Pro (NEC-2) and there a 7 el Yagi is no big deal. It takes some minutes to complete accurately. I see, FEM is a completely different principle. Didn′t realise that. Well, ok, I got quite far with HFSS with my limited resources and can handle basic problems up to accurate solutions. Before I move on, I will take some holiday away from HFSS. Otherwise I get bored to death with my machine. Then comes part II, advanced functions (Optimization etc.). I teach HFSS myself with help from a professor in Braunschweig and the world wide library.
But coming back to my first question: How can I alter the mesh to an applied mesh operation in HFSS 11 beta ? The function of the actual version ain′t available in there.
Thanks so far Volker, Sebastian (NRW).
Then, Volker, what is the criteria for FEM models to be "large" and "small" ? The number of wavelengths ? And the resulting number of Tetrahedra ?
Seb.
Yes, Sebastian, in the end it's the number of tetrahydra/unknowns. Large analysis volume requires a minimum number of tetrahydra, because you need to sample the spatial wavelength fine enough everywhere in the 3D volume, including all the air. Bigger antenna => bigger size in wavelength => more tetrahydra in initial mesh. Using less tetrahydra isn't possible/useful then, because it would undersample the fields and result in nonsense data.
Of course, large and small also depends on your computer resources.
I'm sorry I can't help with the location of HFSS mesh settings because I use another FEM solver.
OK, then a 7 el Yagi is not possible to analyse with my machine. Never mind, my goal is to learn the principles of HFSS, so I can use it professionally if i get accepted for doctor scholarship. Learning on Your own is hard. Yesterday I celebrated "basic knowledge" (step 1) with alcohol Step 2 is advanced knowledge, which will follow next weeks.
What program do You use ?
Hi Sebastian,
I agree that learning EM simulation takes time ... I'm doing this for almost 20 years now - full time
My present work is automated EM workflows with Keysight ADS, so I'm using the Keysight MoM and FEM solvers (Momentum, EMpro and ADS FEM).
Best regards
Volker
Ah ok. I am into antennas, I am a physicist with major in electronics and antennas. Usually I worked with EZNEC, but since I want to apply for doctorate scholarship I have to be familiar with FEM solver (FEKO, CST or HFSS). I read a lot about FEKO and know the program from theory, but can′t handle it, because my computer isn′t modern enough and the program is too expensive. Since antennas are also my hobby, it was a personal interest I learned HFSS with. I divided my knowledge of HFSS into two parts, basics and advanced. As I said, I finished the basics quite successfully and will move on with advanced in a couple of days. But it took me three month to really get into HFSS and now I can use it with relative ease.
Thank You for the answers. Sebastian.
Hi Sebastian,
I agree, EM is an exciting topic. If your area of interest is antennas, then you might want to have a look at the time domain solvers also. Depending on your license, HFSS and CST offer both frequency domain (FEM) and time domain solver types. For antennas and other electrically large problems with small port count, time domain can be the better choice.
Best regards
Volker
Hello Volker,
do You know a non - commercial program for antennas (except 4nec2) ?
Sch?ne Grüsse, Sebastian.
Hi Sebastian,
I think that these NEC MoM codes are a good, efficient choice for wire & yagi antennas, where you want to model 1D current segments.
My own work usually requires meshing of 2D or 3D currents (areas and volumes, not just wires), so I'm not familar with the details of NEC codes.
You might have seen the sticky thread on free EM codes in this forum. But same issue as HFSS: heavy numbercrunching for models that are easy in NEC MoM, so you better use that.
Best regards
Volker
Thanks Volker for the overview. I realized, that NEC is MoM, and HFSS is FEM. And that those two Methods are not comparable. Do You know FEKO ?
Seb.
Does FEKO offer different solution types ?
Hi Sebastian,
today all the large EM solver manufacturer offer different solver methods. FEKO is used a lot in Germanys automotive industry, but I haven't used it myself.
You can see the FEKO methods with 1s google search: https://www.feko.info/product-detail/numerical_methods
Best regards
Volker
Thanks for the link. When I worked at Hirschmann Company, the guys of the automobile section worked with FEKO. Then, I didn′t understand much of it, I just had learned EZNEC and FEKO was too big for me.
Today, I have read the manual of FEKO and know about its working principle. Actually, I have never worked with FEKO, just Empire Excel and HFSS. I can work with HFSS but not with Empire.
I will have a look on the different methods offered by FEKO.
Greetings, Sebastian.
Thanks for the link. When I worked at Hirschmann Company, the guys of the automobile section worked with FEKO. Then, I didn′t understand much of it, I just had learned EZNEC and FEKO was too big for me.
Today, I have read the manual of FEKO and know about its working principle. Actually, I have never worked with FEKO, just Empire Excel and HFSS. I can work with HFSS but not with Empire.
I will have a look on the different methods offered by FEKO.
Greetings, Sebastian.