In BoundaryMode/WavePort cases with lossy BCs, the solve can take a large number (e.g. >~100) of GMRES iterations, while it usually converges in <5 iterations with PEC boundary conditions, as mentioned in #903 (comment). This is because the sparse-direct preconditioner used in the BoundaryMode solve only uses the real part of the system matrix, so convergence is degraded when there are significant imaginary contributions. This issue is to investigate this and improve the BoundaryMode preconditioner so it handles those cases better.
In BoundaryMode/WavePort cases with lossy BCs, the solve can take a large number (e.g. >~100) of GMRES iterations, while it usually converges in <5 iterations with PEC boundary conditions, as mentioned in #903 (comment). This is because the sparse-direct preconditioner used in the BoundaryMode solve only uses the real part of the system matrix, so convergence is degraded when there are significant imaginary contributions. This issue is to investigate this and improve the BoundaryMode preconditioner so it handles those cases better.