Grid resolution required to increase too much due to rotation of line edge source

I am trying to do an FDTD EM simulation including a bowtie antenna and an edge source. In the first case the edge source is aligned along x- axis. The number of grid lines required is reasonable.
In a next step i need to rotate the antenna together with edge line source with a certain angle. As a result the line source is not anymore aligned along x-axis. At this point the resolution of the grid is required to be very fine such that the voxeling is correct. Do you know any way how to overcome this problem ?

Unfortunately there is no silver bullet here...
The best way is often to align the edge source with the grid (i.e. instead of rotating the antenna, rotate everything else...).
In case that is not an option (or is not practical for some reasons), you might want to try to use subgridding: place the edge source (and perhaps the neighbouring antenna element) in a manual grid setting folder with the "subgrid level" option set to x3. That will locally refine the grid, which increases the chances that voxeling is correct.
Note that subgrid only works with the aXware solver, though.

You may want to make sure that the endpoints of your source are precisely the corner points of your bowtie. Two neighboring points may appear as one in the GUI, but may be different in the lower decimals. In such a case, you will get a very small mesh step.

An alternative to remedy this is to use a source that is aligned with the grid and modify the tips of the patches of the rotated bowtie such that they touch the tips of the source. Close to the feed, this should not matter a lot. You can check this by comparing the feedpoint impedance and the near-field to those of your original simulation.