Hello Alex,
OptiFDTD could be a solution. However, there are a number of considerations for FDTD:
1. Computational resources required will be significant (as is typical when modelling a relatively large structure like a fiber) to model a domain that is 130 um x 130 um x 15 um would take 9Gb and a minimum of 5 hours with the coarsest mesh we would recommend. This would only really be able to support a few periods (assuming it is not a long period grating) at best so we would need to rely on periodic boundaries which will have limited usefulness when it comes to chirped gratings. This is based on the assumption of using standard SMF-28 but if thinned cladding is used, it would go a long way to reducing computational resources.
2. The way we would model roughness would be through scripting and loops that would create random structures where roughness is required.
Regards,
Optiwave Support Team