An optimum grid setting should be one which provides good representation of the physical geometry with minimum requirement on grid resolution (does not require extremely small grid step), so it is always a trade off. I would suggest that you use different grid step and simulation period, until you get a set of results which do not vary anymore with changing grid step and period, then you will be more confident of the simulation result since they appear to be invariant with respect to simulation set up.
It is important to check voxels (the discretized model) before running the simulation.
You may improve grid resolution further on significant parts of your structure. I suggest you to use “connectivity check" tool under simulation once you generate voxel. It will give warning when there are floating sources and/or sensors. You can try this tool with your original setup and my setup.
For an FDTD simulation, the most important thing is that you get a correct voxelized model to simulate, if the voxels are wrong (do not represent the actual device) then all the simulation results will be wrong. I strongly suggest you to go through the user manual for voxeling and gridline tips, it will be of great help for your simulation tasks.
The gridder works in the following way:
To obtain a "uniform" grid, set the Priority to 0, such that no baselines are created.
I'm trying to obtain a cubic uniform grid. I set the grid priorities to 0, and all the Maximum Steps and the Geometry Resolutions to be the same. But yet, in Options\Main Grid\Min Step and Max Step are not equals for all the axes.
How can we obtain a cubic uniform grid?
to get a uniform grid, you need Geometry Resolution = 0 (meaning no constraint on the grid coming from the CAD objects), and Priority = 100 meaning this setting always has precedence over the other ones.
You might also need to delete the other settings (or, equivalently, move all the objects inside this "Uniform" grid setting).
Geometry Resolution = 0
Priority = 100