I've had need to run fairly lengthy meep simulations and found it more convenient to run a simulation, save the Near2Far data, and then at a later time extract whatever farfields I needed. I have also been dumping and loading the chunk layout, due to testing "split_chunks_evenly=False".
The farfield results were not what I expected, even though the near field results looked normal. I created a super simple "toy model" to make sure I wasn't missing anything obvious, a plane wave propagating through an empty 3D box with a single Near2FarRegion at the far end.
I still saw saw strange, asymmetric farfields and so with a bit more digging I discovered that the issues only occurred while running on multiple cores, and only while loading in the chunk_layout. It also might only occur while the simulation area is not a perfect square.
I've had need to run fairly lengthy meep simulations and found it more convenient to run a simulation, save the Near2Far data, and then at a later time extract whatever farfields I needed. I have also been dumping and loading the chunk layout, due to testing "split_chunks_evenly=False".
The farfield results were not what I expected, even though the near field results looked normal. I created a super simple "toy model" to make sure I wasn't missing anything obvious, a plane wave propagating through an empty 3D box with a single Near2FarRegion at the far end.
I still saw saw strange, asymmetric farfields and so with a bit more digging I discovered that the issues only occurred while running on multiple cores, and only while loading in the chunk_layout. It also might only occur while the simulation area is not a perfect square.
Code below: