MLFMM Solution is not Converging

Ab28
Ab28 Altair Community Member
edited October 2020 in Community Q&A

Hi,

I am simulating a reflectarray of microstrip patches on a finite-sized substrate using FEKO 2019. Array size is 200mm x 200mm,  and simulation frequency  is 14GHz. Each microstrip patch has two small-sized elliptical holes in it. There are a total of 441 patches.  I am using a Coarser mesh because, otherwise, I run out of memory. Currently, I have 128GB of memory. The solution method I am using is MLFMM with box size of 0.15*Lambda and SPAI (8192) precondition. I ran out of memory with box-size greater than 0.15*Lambda. After running for many hours FEKO gave convergence error. I then turned-on MLFMM stabilization. This time the solver didn't give any error. However, it never gave any result as simulation kept running for many more hours and solver kept trying for convergence without any success and without any end in sight. Finally, I myself aborted simulation. I can't attach CADFEKO Model file here as it is of 83MB!

 

What could be the possible reason for convergence-error? Coarser Mesh? Complex geometry with fine details (due to holes)? Use of EFIE? Please suggest some remedy.

Tagged:

Answers

  • Mel
    Mel Altair Community Member
    edited October 2020

    The coarse mesh and the small box size is likely to blame.

    You may benefit from a FEM/MLFMM solution. Read through this thread.

Categories