Unsolved problems with electromagnetic software

[surykatki]

I was wondering about this: what are current problems in electromagnetism which are hard or impossible to solve with available software (because of any reason). I would really appreciate if you could share your experience in this area.

Thank you in advance and best wishes!

 

[adel_48]

The most requuired problems are those which require prohibitingly high simulation time, such as large finite antenna arrays and finite FSS.

I think there has been a review paper of Raj Mittra on this subject you may try to search for it on IEEE

 

[antennas321]

Yyes I agree with adel. One of the most difficult problems that I deal with in EM analysis is how to address large finite arrays (finite frequency selective surfaces are in the same class of problem).

Typically, small arrays can be analyzed using a full-wave solution (FEM, MoM, etc.) and then one makes the jump to an infinite array solution using periodic boundary conditions. The problems that fall in between these two problem sizes are still difficult to address (arrays in which edge effects are still not negligible and mutual coupling is high).

 

[plasma]

Also calculating active impedance in scanning angel of phased arrays

pl

 

[Pushhead]

Hi,

Generaly speaking, I think that any "electrically large problem" presents a challange to EM solvers (either time or frequency domain).

To be more specific, think of a transmission-reception problem in an urban scenario, or an accurate analysis of an antenna radiating over a plane/car.

P.

 

[adel_48]

wee, you cannot make it that general, since for urban propagation, you are not seeking extremely high accuracy since your model already lacks it (you cannot make a very accurate model of an urban environment) and for such (extremely large) problems approximate methods such as the GTD are the best solutions.

for cars and planes, they are still solved by either hybrid techniques and/or FDTD. But of course they require extremely long time. and again you havethe problem of inaccurate model

 

[surykatki]

Thanks for the discussion!
Of course large problems are difficult to solve because of time/memory constraints but are there any problems which are not limitted by the computations which are still hard to solve? Or small problems which turn out to require a lot of resources and are hard to solve with modern EM software?

 

[adel_48]

Well, may be for certain numerical techniques there are still unsolved problems, for example in FDTD if you have a strongly resonant structure (e.g. a high Q cavity) it would take very long time for simulation.

 

[doug freeman]

Well, may be for certain numerical techniques there are still unsolved problems, for example in FDTD if you have a strongly resonant structure (e.g. a high Q cavity) it would take very long time for simulation.

But in this case I would never use a time domain code for a high Q structure because of this issue, so maybe there is no sense in trying to develop techniques to develop this where another method is strong.

I have had some problems with multilayer spiral inductors (3-4 layers), volume meshing (FEM and FIT) give good results but long solve time, surface meshing (Sonnet/momentum) have quick result but have some errors when calculating losses. Maybe this is an area which would be worth looking at.

 

[antennas321]

plasma has a good point about scanning phased arrays. Predicting the active element pattern and active impedance can take a prohibitive amount of time with an EM solver. It is the active response you need since it includes mutual coupling between the array elements.

 

[jian]

Hi, All:

Scan angle of phased array is touch. For a finite size phased array structure, there might be hundreds of elements and each element may have a feed or port. Regular simulators will have problems on it. Iterative matrix solvers may help on solving a large finite array with one feed. However, if each element has a feed, it will be very time consuming even it might be possible.

Something we have implemented into the IE3D 11.5 might be of interest to you. We have implemented the IE3D FASTA scheme. It is very efficient for large arrays with multiple ports. I tried it on an 8x8 array with 1-feed. Regular matrix solver may take 2 GB RAM and couple hours. IE3D FASTA takes about 40-70 MB RAM and about 12 minutes to solve it. The good thing about it is that it will not slow down and increase the RAM requirement much even with each patch having a port. The IE3D 11.5 will be released later this month and early June. Interested users can give it a try when it is released (www.zeland.com). I am from Zeland Software, Inc.

 

[surykatki]

I've heard that e.g. dielectric resonator antenna (DRA) problems were very hard to solve with most of the existing software (in reasonable time). Do you know any software which is well suitable for DRAs?

 

[adel_48]

yes, materials with high dielectric constants require very fine mesh in it which enormously increases the simulation time. However this is only the case for volume solvers such as FEM, FDTD and volume equivalence MoM. The surface equivalence MoM can easily handle homogeneous materials with high dielectric constants. I think there has been an example of an antenna with a dielectric constant of 85 on the site of wipl-D may be you would like to see it

 

[surykatki]

yes, materials with high dielectric constants require very fine mesh in it which enormously increases the simulation time. However this is only the case for volume solvers such as FEM, FDTD and volume equivalence MoM. The surface equivalence MoM can easily handle homogeneous materials with high dielectric constants. I think there has been an example of an antenna with a dielectric constant of 85 on the site of wipl-D may be you would like to see it

What about resonating features of such an dielectric structure? I think it also increases simulation time.

 

[adel_48]

here I was speaking about frequency domain simulators, resonance does not require more simulation time. wipl-D is a frequency domain MoM software. you will find more details about it in www.wipl-d.com.

I think that IE3D has also this capability in a much better graphical interface, but I couldn't find a lot of examples about 3D structures in it.