Meshing Guidelines for Geometrical Optics Simulation
•
Flat edges (2D) and planar surfaces (3D) can be coarsely meshed, unless they meet one of the criteria described below.
•
Curved surfaces that can interact with rays should always be finely meshed. The tighter the curvature of such surfaces, the finer the mesh should be.
•
Domains usually don’t need to be as finely meshed as boundaries. A convenient way to refine the mesh in the vicinity of the boundaries is to reduce the Curvature factor in the Size settings window in the mesh sequence. This results in a finer mesh only where the radius of curvature of the surface is small. You might also have to reduce the Minimum element size to avoid warnings in the mesh sequence.
•
Features that compute the density of some quantity on a domain or boundary usually require a finer mesh, because the density term is piecewise discontinuous across elements. This includes the Accumulator (Boundary), Accumulator (Domain), and Deposited Ray Power (Boundary) features. If the mesh is too fine, rays might entirely miss some elements, and then it is necessary to increase the number of rays to avoid “holes” in the deposited power or other density field.
•
Avoid unnecessary edges when possible. Discretization error tends to be larger in mesh elements touching geometry edges.
•
Whenever possible, use the geometry Parts from the Part Library for the Ray Optics Module, rather than creating lenses from geometry primitives like the sphere and cylinder. The built-in lens parts have been constructed in a specific way so as to reduce the discretization error on the reflecting and refracting surfaces.
•
Consider changing the Geometry shape order in the model component settings. Using Cubic or Quartic causes the boundaries to be discretized using higher-order polynomials, which can reduce error by several orders of magnitude.
Similarly, with physics interfaces that solve for a displacement field, such as Solid Mechanics, locate the physics interface Discretization section. A higher shape order such as Cubic Lagrange should be selected if rays are traced in the deformed geometry
•
If the geometry uses parts or primitives, you can reduce discretization error by selecting Use geometry normals for ray-boundary interactions in the physics interface Ray Release and Propagation section. However, this only improves the accuracy if the geometry is undeformed; it has no effect if the geometry is subjected to thermal stress or some other type of deformation.