That depends on the exact setup you're studying. How complex is the geometry? Are you interested in single frequency or broadband results? FDTD/FIT solvers (like CST MWS) are probably suited quite well, since curved structures can be modeled using conformal solvers. Finite element solvers can probably handle the curved structures better, however they might require significantly more memory depending on complexity and will usually yield single frequency results as opposed to time domain solvers, that can extract multiple frequency results given a proper input pulse.

I agree with Mr Kuehne. Further more for large structures (dim >> wavelength) CST Studio has also an IE solver that improves accuracy and speed. Looking at the market it seems that CST and a skilled EMC competence are a good starting point

This is a very open ended question as Kuehne and Orlandi have quite rightly said. For large problems you might also consider FEKO. For high frequency EMP problems you can have a huge frequency range and might really be looking for a time domain solution (FDTD or TLM, both available in CST. FDTD is also available in EMPIRE) although you would clearly need to control the size of your model. You might also look at COMSOL, which allows you to link different modules together. Be aware that you have to provide the linkage, I believe.

I would recommend using numerical codes based on method of moments. Since you are dealing with high frequencies, it is better to reduce the computational domain, and method of moments uses surface integral equations. Regarding the arbitrary curves, you don't need to worry too much about that since there are very good meshers available and they would handle most of the geometries. About the specific commerical software you can use:

[1] HFSS recent versions have integral equation formulation which is basically Method of moments (MOM)

[2] Wipl-D, this is also MOM based, less pricey than HFSS, I think that it includes quad mesher though I am not 100% sure about this

[3] If you want to develop your own code excellent start would be book by S. Makarov "Antenna and EM Modeling with Matlab", which comes with a number of interesting examples. There is also a free version of Antenna Toolbox which contains improved codes from the book.

For a gentle introduction on numerical electromagnetics I recommend David Davidson "Computational Electromagnetics for RF and Microwave Engineering".

You can find a brief result comparison of various TD as well as FD solvers applied on small aircraft model in the paper ""HIRF interaction with metallic aircrafts.

A comparison between TD and FD methods" from Salvador Garcia of the University of Granada at. al.

Priscilla,

to the question "Which numerical modelling technique is best suited for an EMC/EMI problem and why?" no direct answer can be given. I dare to say that you should use analytic expressions for at least in the case that there is a large distance between the source and the victim (EMP ??). In particular when the victim structure is fine compared to the distance in-between you'll render into numerical problems

On the opposite, if the distance in-between is large and the 'source' can be represented by a plane wave then it will depend on the structure whether a simple approach like NEC would work or that the effort with FEKO or CST is worthwhile.

Do consider a first or second order estimate before spending the effort by going into details and check whether the results are still in line.

Priscilla,

As other people have indicated the choice of computational technique for a particular problem is usually quite dependent on the details of the geometry, materials, frequencies etc involved. I can easily think of EMC/EMI problems in which different tools have specific advantages over others. There is a lot of good advice above.

You do mention high-frequency circuits and EMP suggesting time-domain tools may be more appropriate as they can deal more efficiently with the large bandwidths involved in looking at pulsed source. Also some can deal directly with the non-linear effects arising from the presence of circuit elements. If you actually want to directly model coupling into a circuit board then dealing with frequency dependent inhomogeneous materials may also be important which would suggest a technique like FDTD/FIT. The more advanced commercial tools, like CST, but there are others, have sub-cell and conformal models to deal with curved surfaces.

If the problem is electrical very large then FIT/FDTD may not be feasible and another approach, or even a hybrid approach using more than one tools may be more suitable. You need to start with a very clear definition of your problem and what you want to achieve in terms of results.

Depending of geometry, materials, field sources and frequency range, results of interest, you have three fundamental options: FEM, BEM/MoM/IEM (including their combination) and FDTD/FIT, but sometimes PEEC or TLM may give better results. Electromagnetic modeling is an art. There are no definitive recipes. The most popular isn't always the best approach. You should compare the advantages and drawbacks of these method, in order to choose the best approach, appropriate for your problem. But first of all, understand deeply your problem and (re)formulate it in correct manner. Do not expect valuable answers to vaguely formulated questions.

For EMI / EMC, I highly recommend Ansoft Maxwell as a Commercial, Off-the-Shelf (COTS) solution. The scripting language makes importation of various geometries and geometry modification / buildup very easy, and the technique uses hybridized FEM / MoM. Basically, adaptive elements are used with FEM in the places where the ambient field is to be solved, and MoM is used on the boundary of the problem domain to enforce the Sommerfield Radiation condition. With Maxwell, the user has the power to turn a few knobs, choose a few options, and automate all of this! BUt on the downside, it's an expensive piece of software, and it requires some pretty good compute power. Still, the adaptive FEM mesh makes for small sized tetrahedrons in reqions where the field changes rapidly, and large tetrahedrons where the field changes slowly, optimizing efficiency. Overall, nice piece of software.

All the earlier post are very usefol.

There is no one single 'best' technique. It depends on what you wanted to get out from the simulation. It will determine how the problem is formulated for the simulation and that is ultimately depends on how much computational resources you can gardner. e.g. presence of thin membranes/layers will bias the simulation towards TLM. However FDTD will not use up as much computing resources.

For a EMP type of problem, suggest you use a time domain approach like FDTD or TLM.

Please refer to http://en.wikipedia.org/wiki/Computational_electromagnetics for an overview of different computational techniques.

A big thanks to all for the insights.

Ian you mentioned that FDTD/FIT may not be feasible for a electrically larger problem. Also a hybrid approach should work well.

I'll try and work out on a hybrid algorithm for my problem which consists of a small and a large scale structure!! It might get the job done hopefully.

Dear Lady,

use the numerical System Mephisto. This is a 2 DTLM program and I have many simulations results in EMC with my students.

Yours Sincerely

F. Gräbner