Re: Feko Suite 4.2

Although I have not used the CG card, but obviously CG=conjugate gradient. So the linear system is solved iteratively with CG. Even if the matrix is filled in full, you could have convergence problem. If many elements are set to zero, then the corresponding coupling is not considered, the iteration could converge to the wrong answer or not converge at all.

Oops! I had taken CG card for computer graphic card. This is because recently there has been growing interest in using computer accelerator boards, such as those from NVIDIA and ATI, for engineering/scientific simulations.

Really! Do you need special compiler and/or hardware to take advantage of those special "processing" power?

As to software, yes. There are many teams around the world working on the topic, e.g. in Stanford University. They have a C compiler for that. As far as I can remember, it's free to d/l. It seems at least two or three fold increase in computational power, in comparison to the latest versions of P4 could be obtained. Just google it, i.e. graphic card scientific engineering ...

BTW, recently a company has developed an FDTD tool with an FPGA accelerator. WOW, it's very fast. I'll mention its link later.

Here is the address of the company which has developed a 3D FDTD tool with an FPGA accelerator:
http://www.emphotonics.com

Hello All,

I'm new to Feko and it seems like this is the right group of ppl to ask:

How do I setup a simple coplanar waveguide structure in Feko? Also, is there any way to shift the reference planes in a CPW structure?

Hi dr_em,

for no gnd plane co-planar waveguide, add extra edge extension at feeding /port location, using ae card for center conductor and add ae source for side conductor, feed between labels adjust phase to the center conductor to be 180 difference to the two side coductors (Gnd).

Hope this helps.

Cheers,
Element7k