How to reduce computational resources for the FDTD?

Altair Forum User
Altair Forum User
Altair Employee
edited October 2020 in Community Q&A

I have and FDTD model and I'm trying to reduce the required resources. Are there any tips or options that I can consider to try to reduce the resources required to solve the model?

Tagged:

Answers

  • JIF
    JIF
    Altair Employee
    edited June 2017

    The memory requirement for the FDTD is roughly 150 MByte per 1 million voxels. Without a model to look at, it is rather difficult to make suggestions, but I have listed a few items that you could consider.

    Remove unnecessary free space regions from the FDTD bounding box

    The default behaviour of CADFEKO is to automatically add a free space buffer of size lam0/8 for each of the 6 sides of the FDTD bounding box where lam0 is the free space wavelength at the lowest frequency to be simulated. In some cases this boundary can be set to PEC or to not add this boundary.

    When S-parameters are requested, limit the number of active ports

    Each active port for the FDTD requires the restarting of the time steps. When requesting S-parameters, for example, in a two port solution where only S11 and S21 are required, then deactivate the second port in the S-parameter request in CADFEKO.

    Choose the near field request to be within the existing mesh

    A voxel is stored for each near field point. If the near field point is such that it is outside the bounding box that would have been used without the request, then the number of the voxels will increase and therefore also the computational resource requirements. Also, if the near field points are closely spaced, then the mesh will need to follow this spacing. This could also lead to an increase in the number of voxels in the mesh.

Categories