Simulation of large structures

[olzanin]

cst 30000 mesh cells

Hello,

which software is suitable for simulations of large structures (10x10x10 m) at frequency of about 5 GHz. The most software tools can not handle with such large number of unknowns caused by this high frequency.

Applications:
-human in some room: BAN
-two antennas in closed structure: communication inside of some sensitive body.

I know WIPL-D, HFSS and CST. WIPL-D can not because of limited number of unknowns. HFSS is frequency domain tool but I need time domain like CST. Have somebody done some experience in this topic?

For all proposals I will be very thankful.

 

[ali110]

remcom wirelessinsite

hi ,
I think IE3D is suitable for that. because in this software you can divid the structure in some active regions.

I hope it'll be helpful

 

[asdfaaa]

PO, GTD, UTD method maybe helpful to your application.
Regards,

 

[irfan1]

You may consider using the 64 bit capable CST with the appropriate hardware. For instance, a dual 64 bit xeon - 3.2 GHz configuration with 6 to 12 GB of DDR2 RAM. such a machine would sound so expensive in the past. but nowadays you can get such a nice computer under 5000$

 

[rautio]

Let's think about this one...

5 GHz wavelength is under 1 CM (free space). Let's give it the benefit of the doubt and say it is 1 CM.

The volume of interest is 1000 wavelengths on a side. Any kind of meshing based EM analysis requires minimum of 20 cells per wavelength. So, that is 2x10**4 cells per side. This means a volume mesher requires 8x10**12 mesh cells. I think we can forget about volume mesh for the entire structure.

If we further assume that there is some complexity of surface metal in the volume, surface meshers (like Sonnet for whom I work) top out at about 30,000 subsections. If you go to iterative solvers, you might be able to get up over 100,000, but now solution stability and accuracy become an issue. I don't think surface meshing is going to be useful either.

GTD (Geometric theory of diffraction) will provide an approximate solution, provided the structure is not too complicated.

Let's say we could get an accurate solution of some kind at 5 GHz. Now, if the temperature changes a little bit, and say the struture changes size by 0.001%, now it will have changed in size by 1 full wavelength, you now have a completely different structure with a completely different answer. And we are assuming that we could even build the initial structure accurately.

Am I saying there is no practical solution? No way. The way to attack this problem is from the point of view of random scatterers. There is a lot of research in this area. For example, rather than dealing with a specific back scatter solution, you deal with averages and standard deviations. If you don't know exactly what the structure is to within a very small fraction of a wavelength everywhere, then it must be treated as a stochastic process.

They might exist, but I don't happen to know of any commercial tools for this. It is certainly a problem that could be handled at a number of good research universities, Ph. D. research kind of stuff.

 

[rautio]

Hmmm...no one caught my slip, including me until just now. Wavelength is under 10 cm, not 1 cm. It is 100 wavelengths on a side and 8x10**9 mesh cells. I think that still puts volume mesh out of the question.

 

[ricy]

hi olzanin,

take a look to https://www.remcom.com/WirelessInSite/index.html
It is a optical tool based on GTD/UTD; maybe it is that what you need?

We use it for WLAN and base station antennas in urban environement.


best: ricy

 

[grandizer]

I think you must deal with a part of the structure and not the total structure
FEMKO can deal witha large problem
also FLomerics..
I hope that help

 

[olzanin]

Hi,

Thanks for your replays.
Have you heard something about the mode-matching method? The way is to divide the structure into several blocks calculate the S-matrix and then combine them. Because of this analytical method the computation would be very fast.
I heard CST could do it.

Best regards

Olzanin

 

[langhnaj21]

any one can use available software packages in market like xfdtd by remcom, ansys(multiphysics) or CST.
other than MEEP is GNU based code so one can try it also.

 

[ff539]

Dear all,
recently I have been working on simulations of Through the Wall radar scenarios in CST (simulations of room with objects like pieces of furniture and human body phantoms inside, plus some radar structures outside the wall; these simulations must be performed at UWB frequencies like 1-5GHz).

My license for CST is going to expire soon and we were thinking if it is worth renewing it or switching to xFDTD.

Any good piece of advice is more than welcomed.
Thanks