Topics related to Lumerical and more

Manually setting FDTD mesh grid

    • nir.levanon

      In FDTD, I'm interested in comparing two different designs of an opto-electronic device, where one design uses more coarse models of the structures within the device (more rectangles, boxes - elements with less vertices in their polygon) and the other uses more refined models (more details, more vertices in the polygons). In both simulations, the FDTD region perimeters (min/max x/y/z) remain the same, the simulation settings, source settings and so on...

      The mesh type I often use is auto non-uniform with conformal variant 2 meshing refinement.

      The senisble thing to do is to compare both simulations using the FDTD mesh grid from the simulation of the more refined structures models.

      I could use the uniform mesh type, but I believe that these are a few concerns one must be aware of when using it:

      1. By using a larger uniform mesh grid size, the simulation will more accuretly model the device, but at the expense of a longer simulation run time.
      2. With a uniform mesh, the grid points might not all align exactly at the interface positions between materials, which is what mesh overrides over structures usually try to achieve. This could cause the structures to be simulated differently than expected.

      I'm thinking I could possibly set the non-uniform mesh grid of both simulations to be the same by way of adding many positional mesh overrides, however since the device I'm simulating contains many structures I think that it would be way too streneous to be adding the many mesh overrides.

      What are your recommendations for performing such a comparison, without causing the simulation run time to be too extreme, while attempting to keep a high level of accuracy of modelling the refined structure?

    • Guilin Sun
      Ansys Employee

      I personally do not recommend using many override regions. You can simply use one override for the rectangles, and use auto-mesh. This not only greatly simplifies the settigs it is also easy to maintain and modify. The two designs are not the same since they do not have the same geometries, at least in discrete mesh space. Please refer to 

      1.       Understanding Mesh Refinement and Conformal Mesh in FDTD

      2.       Understanding the non-uniform mesh in FDTD 


    • nir.levanon


      Thank you for your reply Guilin Sun.

      I think my main point may have not come through clear enough – How do I compare the results of simulating two different devices with slightly different geometries, when calculating things such as Pabs, refelction and transmission?

      Using auto-mesh on each simulation could cause the vertices of the FDTD mesh grid to be different, which could alter the FDTD calculation and provide slightly different results due to effects from inaccurcy of geometry. 


Viewing 2 reply threads
  • You must be logged in to reply to this topic.