The issue of how to set the number of cells in EME solver

- October 18, 2024 at 6:43 am
  
  2022160206
  Subscriber
  
  I recently encountered this problem while using the EME solver for simulations. When we change the number of cells, we get different sweep results, as shown in the figure below.
  I understand that the results will be more accurate with a larger number of cells, but what is the minimum value I can rely on? Additionally, if my Group 2 is a taper structure, can I trust the EME sweep results?
- October 21, 2024 at 9:33 pm
  
  anna.wirth-singh
  Ansys Employee
  
  Hello, thanks for your question.
  The minimum number of cells depends on how complex or quickly changing your geometry is, and what is an acceptable level of error in your simulation. The best way to determine the minimum number of cells is to do convergence testing. That is, run the simulation for an increasing number of cells, and you should see the result settles to a particular value as the number of cells is increased. If increasing the number of cells does not cause more stable results, then perhaps there is a different issue that needs to be investigated. Please see this page for a more detailed explanation of convergence testing: Convergence testing process for EME simulations – Ansys Optics
  If you have not already viewed it, I also recommend checking out this example as a starting point for waveguide tapers with the EME solver: Linear waveguide taper – Ansys Optics. In this example, 19 cells are used for the tapered region, so that may be a good starting point. The EME sweep results should be accurate if the number of cells is large enough due to the use of continuously varying cross-sectional subcell (CVCS) method (explained here: MODE - EigenMode Expansion (EME) solver introduction – Ansys Optics).
  I hope these resources are helpful, and please reply back if you have further questions.