Some users might have this question about using a TFSF with a oblique substrate, like a setup shown in the screenshot below and this is an incorrect use case!!
I guess the driving force of setting up a model like this may be trying to workaround the angle dispersion, for a broadband source. Unfortunately, this workaround introduces other problems. The TFSF source works by running a 1D simulation on one of its corners to obtain the “background” information. The “background” information contains a 1D spatial index profile and this is used to distinguish the scatterer. And it also assumes the same spatial 1D profile at all corners. Clearly, this is not the case for this setup and therefore it is not a correct use case. For more info about the correct usage of TFSF, please visit the correct usage and TFSF on periodic setting.
If the reason is the broadband source with an angle, one way is to run multiple single frequency simulations.