There is no general answer possible to what method is better or worse.

Besides MEEP as an free and open source FDTD tool, you may consider openEMS (http://openEMS.de) as another free and open FDTD tool supporting 3D Cartesian or Cylindircal inhomogeneous meshes. It uses Matlab or Octave as scripting interface and has a couple of antenna tutorial examples to get started quickly.

Both the algorithms are based on different concept.. I think it is not the aquracy but the domain size that can discriminate the two methods. For 3D objects FEM will be suitable. but for the structures like boards and planar antennas with relatively smaller size MOM will be suitable. remember for MOM the unknown effect is increased with the order of cube while for FEM it is square. ADS Momentum solver increases the complexity with order of nlogn so it is faster then other Momentum solvers present in the market. FTDT is know for its problem solving domain, as it solves very bigger problems which cannot be solved by methods like MoM and FEM. also FTDT is 3D Solver.

Hybrid is the ultimate solution since it will cater all the domain problems. As far as aquracy is concerned I think all are nearly equal.

It also depends on which type of antenna design is of interest, electrically small or large? If you like to simulate a mobile device antenna next to a human body, FDTD will be a good choice, if you like to simulate the antenna on a airplane or on a vehicle, FEM will be more suitable. There are many choices of commercial FDTD, FEM solvers out there, e.g., HFSS, CST, Feko. Try them out and you will know which one suits your application.

To my experience, if you know the structure is going to be fixed and the result you need is just at one frequency, MOM and FEM are fantastic. Though they need more work than FDTD, they will be faster and more accurate.

FEM needs to deal with Boundary Conditions whereas MOM does not need.

However, If you are going to simulate different types of antennas and you want to write the algorithm codes, I really suggest FDTD. The algorithm is very simple and very versatile. Even if you are interested in a specific frequency, you can use the frequency domain version of it called FDFD.

Hi, "In MoM Green's function is required which is a tedious task"...yes this is true, but you have to remember that you cannot derive the Green's function for an arbitrary structure. So, the arbitrary structure has to be discretized and you will be using the Green's function for the discrete elements (which is already known) and finally you have to apply the continuity of the currents at the junction nodes of the elements (see, for example, "WIPL: A Program for Electromagnetic Modeling of Composite-Wire and Plate Structures", B. M. Kolundzija, J. S, Ognjanovic, T. K. Sarkar, R.F. Harrington). So, the complexity of the structure is not a very big issue, because you don't have to derive the Green's function for the whole structure. The issue with the techniques is the size of the structure. For electrically large structures (a very commonly encountered factor in RCS calculation), like an aircraft illuminated by a microwave RADAR, storage and run-time becomes very important and there FDTD, MoM and FEM can trouble you equally. If you are using parallel computing, you can do better with the time complexity, but the space requirement still remains a problem with these techniques. So, in electrically large case, the option is to use fast algorithms like MLFMA. This is very powerful in terms of both time and space complexity improvement. FEKO is a tool which is utilizing the advantage of hybridization of other methods with MLFMA. The second solution to tackle the electrically large problems is to use GPU (Graphics Processing Unit). In essence, MLFMA is far more efficient than FEM, FDTD and MoM to calculate the RCS of electrically large structures.

In the field of antenna a large part of the research interest is being devoted towards the impulse radiating antenna and time domain measurements now a days. In such cases (and also in other cases like high speed interconnects where you are interested to study and improve the transient phenomenon), a time domain algorithm will be better (like FDTD or Time-domain MoM). MEEP (developed at MIT) is a free software on linux platform based on FDTD. In this, you can model dispersive media also. MEEP webpage itself is a great resource for learning. For a quick overview, you may also visit: web.iitd.ac.in/~eez128122.

By the way, if you want to learn more abot GPU, you can google NVIDIA Tesla or NVIDIA Kepler. If you want to learn more about MLFMA, Matrix pre-conditioning etc. you may google Prof. Levent Gurrel or BiLCEM.

@Sayak Bhattacharya

the link web.iitd.ac.in/~eez128122 is not accessible.

Ok, in that case please drop me an e-mail in [email protected], I 'll mail you the pdf.

the mail that I had sent you bounced back. Please mail the pdf to the mail id: [email protected]

For antennae without dielectric materials, MoM is more efficient, whereas, for antennae including dielectric materials FDTD is more efficient. You can use EM solvers using hybrid FDTD/MoM technique for more efficiency and high speed.

I mostly have experience with FDTD and MoM, so I wil compare them. For wire-like antennas: use MoM. For 3-D antennas (example: horn antennas) FDTD is better. For 2-D (example: microstrip antennas) is difficult to favor one over the other. Both have free and open source options (NEC2 for MoM, MEEP for FDTD) and commercial tools also. Regarding interconects, they are 3D problems, so FDTD is a good option as long as the frequency of interest is not very high (the interconnect is not too large compared to the wavelength). FDTD has some troubles with curves.

There is no general answer possible to what method is better or worse.

Besides MEEP as an free and open source FDTD tool, you may consider openEMS (http://openEMS.de) as another free and open FDTD tool supporting 3D Cartesian or Cylindircal inhomogeneous meshes. It uses Matlab or Octave as scripting interface and has a couple of antenna tutorial examples to get started quickly.

Agreed with Thorsten. You have to look into your main objective really either to compromise accuracy (up to certain tolerance) or time consumption to solve the problem. There would be different views from application (sometimes they just want the final answer that gives impact to their systems with some acceptable tolerance) and those who are very much concern on the numerical solution (that can give impact on the time consumption, speed, etc). Scope of your work can also influence the decision in choosing the right technique.

:: just a thought::