It depends on the level of accuracy you need. We usually use large mesh size for first time simulations to check if the simulation is okay. Later, we use small mesh size for more accurate simulation results. As a rule of thumb, I choose the mesh size to be 10% of the object/gap dimension.
In FDTD, you have the option of using conformal meshing. This automatically generates a graded mesh which has more accurate steps at the interfaces, curves, slots and small features. However, relatively larger mesh sizes will be used for the rest of the simulation domain. The thing that saves time.
In cases where you check the conformal mesh and still your critical dimension is not well resolved, you will need to use override meshing.
In a nutshell, it is recommended to use conformal meshing as much as possible, and sometimes mesh override regions to force a smaller mesh size in critical regions.
It is important to mention, that there are actually to limitations to mesh size:
- critical dimensions and interfaces
- the wavelength is smaller in higher index materials so the conforming mesh takes this in consideration. The mesh size in a material of n=2 is half that of n=1. So, in general, you don’t have to worry about this.
I encourage you to have a look on this page:
One more tip to go, there is a monitor called refractive index monitor. You can use it to check the exact structure that will be simulated. Let me put it this way, if you draw a circle, this doesn’t necessarily mean a circle will be simulated. It depends on the meshing.
Here is how it looks without overgrid mesh:
Now, I added an overgrid mesh for x and y only of 0.0001 um and checked again. Looks more like a circle with no staircase shape.