Re: adi fdtd
There are some papers. I think they are useful to you maybe.
These are Namiki's original papers. The method described there is stable but has large numerical errors
for larger time steps.
When I tested it (I used the method described in this papers) the main results were
- stability OK
- speed *slower* than FDTD for acceptble errors (< 5%)
when I increased dt to get the same computation speed then there were 20-50% errors.
Part of the problem may have been the implementation, but I find it difficult to believe
that you can get too much out of the older ADI methods.
I don't know about newer mthods. There are papers (I cannot recall them right now) which
claim that they have solved the large-error-for-larger-dt problem but I have not tried them.
My advice is to first make sure that you really want/need ADI-FDTD before you start
implementing it. I may sound very negative but I really would like to learn about a ADI
success story but so far have not heard of any such.
I think that it is most difficult to implement PML in ADI-FDTD.
Painful but doable (there are a few online papers on PML for ADI
from me too