Simulation software for aerospace antenna design

Hello,

I'm an engineer, designing antennas (generally on airplanes). I would prefer not to disclose the company, for confidentiality reasons.

I have been browsing through this forum many a time, but have been unable to formulate an answer to my question.

What software (CST, HFSS, Feko, ..) is best suited for my work and why? Of course we have heard what the vendors themselves claim, but I'm also interested in what actual users think.

Does anyone inside the same (or a related) industry have any suggestions? What do you guys find important about a package when using it?

Cheers.

Hello colleague

My interest (EMC and antenna design) is little bit broader and slightly different from yours but I think that my experience should help you.

I suppose that you want whole aircraft simulations. Precisely, large scale simulations at least (for example, because you want to know the change of antenna characteristics situated on real aircraft)

I have experience with these softwares: CST, PAM-CEM, FEKO (little bit)
1. CST:
Good importability of CAD models (I recommend NASTRAN)
Good precision and meshing techniques (sometimes, when you are dealing with large and complex models, you can hit on problems with optimal mesh - but they are solvable)
User friendly.
Simulation times are sometimes a little bit longer than FDTD solvers.
Now, there are several modules that you can use together with benefits - MWS, DESIGN STUDIO, Cable Studio.
Quick response/help for users and good SW developement in time.
(From v.2009, there is a Cable Studio for harness simulations. However I have no practical experience with this module because I am not using it now. SW licenses are sometimes a little bit expensive as you probably know ) ).
(CST use FIT instead of ordinary FDTDs - Because of it, it can have better mesh representation and this means better precision)

2. PAM-CEM:
Good importability (if you know what you are doing; I recommend NASTRAN too)
Relatively good precision (on FDTD; If you compare elmag. field results of large and complex model, you get practically same results as from CST. Of course it depends on many things! So you must be careful what exactly you are comparing).
My experience with antennas in this software is limitted, but I suppose that some simple wire antennas can be made there (I am using it for EMC only).
NOT so user friendly as CST! (horrible manuals; I recommend some training or you will spend many weeks burried in manuals - there is possibility that you end grey-haired too ; Sometimes I think that they made it difficult and user un-friendly by purpose - because of paid training courses ) )
Postprocessing .... I usually use Matlab for simple graphs (currents, E/H points etc.), istead of their PAM VIEW. But it is my opinion. They are saying that Visual CEM has very good postprocessing (I think it can be true .... I can tell you next mont ).
(PAM-CEM have CRIPTE module for cable harness simulations - VERY GOOD, VERY PRECISE and very usefull for EMC! Probably there is no use in it for you but it is worth to mention it.)
(Now, they shifted PAM-CEM to "new coat" and they are trying to sell it as Visual CEM - there are still some problems with interfaces and it is buggy. )

3. FEKO:
Very good and precise MoM.
Good importability (I recommend IGES and STEP for geometry and STL or NASTRAN for mesh representation - I haven't any problems there. But you must know how to prepare the CAD geometry for import ).
User-friendly? I supose yes - I made my first a "litte bit complex" model in few hours after I started with it. Excellent manuals.
I had only few opportunities to work with feko, so that is all.
(One thing, MoM methods are good for relativelly simple geometry - for example if you want to simulate model that has only external surfaces of the aircraft (we can say for example "shell model"), it is good choice. However, more complex aircraft models should be problem - divergence; extremely hardware expensive for such tasks).
Now, they are trying to implement CRIPTE in FEKO (good news for me ).

Summary:
for antenna simulations try CST (if you have money). Other option is FEKO.
for EMC: PAM-CEM or CST (pamcem is less expensive). The second option is FEKO.

There are many other SWs you can use. (XFDTD, SemCAD, CONCEPT 2, etc...)
for example see: http://emclab.mst.edu/csoft.html
P.s. look at CONCEPT 2 - it is very interesting MoM (and maybe not too expensive)

I would already like to thank the people that responded, both here on the thread and privately.

The previous author did interpret my intentions correctly. The objective is not just simulation of an antenna in free space, but indeed with the inclusion of (part of) the airplane itself, to study the total radiation pattern. In the near future, I might be doing similar things with satellites as well. Their total size, compared to the wavelengths, makes it a bit challenging.

Initially I thought MoM solvers (with Fast Multipole Methods) would be ideal, for their capability to handle large objects more efficiently than other approaches.
But, as you mentioned, they seem to run into stability problems when the model becomes more complex. I've only recently started looking at what I can accomplish with approaches like FDTD or CST's FIT.

Have you tried CST's new MoM? I know Feko seems to be the leading MoM solver, but I saw on CST's webinar that the latest version also offers MoM/FMM, which could be interesting. A shell model is usually enough for I want to accomplish.

I should probably check out PAM-CEM. I have to admit I don't even know it. Although grey hair is beyond my ambitions, it looks like it's worth to at least look into it. :)

Fro aerospace antennas you most effectively would use Method of Moments tools. Classic tools are CONCEPT-II and FEKO.

CONCEPT-II is kind of University based, so don't expect from there nice userfriendly GUI. Also I have not heard anything about their development progress - for me it looks like project is frozen.

FEKO provides good techniques for high frequencies and also MLFMM. And if you would review lates Quarterly (http://www.feko.info/knowledge-base/...rly-march-2010) you'll find that they have introduced anisotropic composite materials now, which might be important for your industry. But working with these materials woud require from user a lot of inputs and very clear knowlage of applied materials characteristsics, which I am no sure are easily available.

Another tool you might review is EMCoS EMC Studio (http://emcos.com/index.php?id=72). This is suite equipped with one of the fastest Method of Moments cores and providing effective interfaces for Harness Databases (like KBL-data and so) for hybrid MTL-MoM solution for both - radiation and susceptibility problems. So, with Hybrid methods one can include into simulation very complex harness bundles with complex shielded cables, stochastic properties and attached devices models (Spice, VHDL-AMS) at the terminations. That is what is not available in classic MoM codes. Here are some examples for Aerospace simulations - http://emcos.com/index.php?id=126 - kind of simple, but might be useful.

FEKO and EMCoS EMC Studio provide also Parallel versions for Clusters - so virtually any large-scale structure can be simulated well.

With latest versions CST and HFSS have also introudiced now Integral Equations Solvers (MoM), but those were not yet tested by time and society and I believe they could not yet cover everything what is available in products, which are working with MoM over a decade already.

Oh, that ability to include layers of anisitropic material was as yet unknown to me. Very cool.

I do notice they mention that -thin- layers are supported. I'm supossing they are using some thin layer approximation? I would be surprised otherwise, though. I was unaware that MoM/MLFMA was already extended to anisotropic materials.

In fact, how do these solvers cope with isotropic homogeneous bulk materials? Most examples I can see seem to focus on perfectly electric conducting surfaces(and thus don't model the inside). Do the MoM/FMM solutions mentioned support the dielectric model as well, with general permittivity and permeability? This I know to be theoretically possible, from literature. Often not all components can be described with PEC alone.

I have been giving 'Points' to people that helped me. I'm not entirely sure what they do, but I'm hoping you're happy with it. :)

Thanks for appreciation. I have now idea what to do with those Points, but let them be :)

On materials - in MoM you can generate closed boundary surface and specify so-called 'medium' inside providing complex Permittivity/Permeability parameters. In EMC Studio those are easily specified as frequency dependent, while in my last try with FEKO I didn't find such an option - for different frequencies I had to simulate separate project (not sure about now).

This works very fine. In practically unlimited frequencies range (from very low to 20-30 GHz tested). Simulation time is longer when you require dielectric, but accuracy is good. So ferrites, lossy absorbing materials and so on could be modeled well. The only problem is again to get corresponding materials properties right. Even manufacturers sometimes could not guarantee characteristics of their products - so there might appear some specific issues, but in general this is manageable.

BTW, here is also example on ferrite-antenna MoM simulation - not frequency dependent, but you can get the idea - http://emcos.com/docs/SmartEntrySystemsSimulation.pdf.

Hello Aneirin and e_m_c

CONCEPT-II lives. Now, they are developing it for the HIRF-SE project (http://www.hirf-se.eu/hirf/).

Thanks, good to know this. It can be useful when you must deal with composite materials.
Materials can be measured easilly if you know where By the way, if you have problems with material parameters try this
"Dielectric Constant and Loss Data" (parts 1-4) and "TABLES OF DIELECTRIC MATERIALS"(1-5) from MASSACHUSETTS INST OF TECH CAMBRIDGE LAB FOR INSULATION RESEARCH.
(if you hit on Dielectric Constant and Loss Data" - part 3 or TABLES OF DIELECTRIC MATERIALS parts 2 and 4 please tell me I am searching for these parts for years! )

This is why FEKO try to implement CRIPTE code It can be used for this job with many benefits.

I think CONCEPT-II provide parallel processing too (at least my next job is to test it for my employer).

Yes you are absolutely right. They cannot catch the train so quickly (but they are trying it )

To Aneirin: thanks for points (I don't know how to use them ... but I'll figure it out )

CONCEPT-II surely looks interesting, especially for a free code. From what I can tell by its website, it does seem to support parallel solution on a cluster, which could be quite interesting.

A question related to the choice of software package - what are your general rules of thumb in deciding what method to use? There's MoM, which seems favourable to handle large scattering problems (when accelerated by FMM-type approaches), there's FEM/FIT and then the time-domain approaches like FDTD.

Are there packages that support hybrid solution? I know, for instance, it's theoretically possible to combine MoM (for instance on the outside medium, including the plane) with FEM (which could be more appropriate to do the inside of the antenna itself).

rezna, thanks for the news on CONCEPT-II! Seems, like project they are involved in is quite strong and has I think not bad budjet for these times :) I did not yet clearly imagine whole project and the CONCEPT-II piece value there, but anyway this is something which is better to follow for updates.

Aneirin, FEM-MoM is supported by FEKO (http://www.feko.info/applications/Bi...ectromagnetics), but I am not sure about performance.

Hi Aneirin,

For your case, maybe you can try CST MICROSTRIPES. It's include in CST Studio Suite 2010. It has a pretty method solver that is Transmission Line Matrix (TLM) and very suitable for EMC/EMI analysis. You can read the brochure from www.cst.com.

Depending on the wavelength, FDTD might still be an option.
Here are some airplane examples done with the Empire FDTD solver:

http://www.empire.de/page33.html

Those FDTD simulations look quite cool. From my university times, though, I have a sort of suspicion about FDTD's accuracy. Simulations I used to do myself, seemed to show that the inherent problems like dispersion errors and uniform cubic mesh imposed some rather troublesome restrictions.

The frequency in those examples is probably a little too low for what I want to do, but I assume that perhaps with some graphical card acceleration (CUDA), FDTD would get within reach.

An interesting option to look further into - thank you. I'll also give you some Helper points. :') I so feel like Santa Claus.