I took a look at your simulation and I have a few suggestions.
- The cc-fiber Structure Group, was using a dialectric material in the core which was close to 1.4. Since the ring structure is much bigger than the wavelength the effective index of all the modes will essentially be that of the core material. You can double check your indices and meshing in FDE solver (See Below).
- The material model that you were using was fitting the $ Al_2O_3 $ near telecom wavelength (see below). The material properties at 1550nm are a transparent dielectric, but aluminum oxide has strong absorption at 10 $\mu m$ which significantly effect the solutions.
Try to improve the fit by adjusting fit parameters like bandwidth, fit tolerance and imaginary weight.
- Finally the FDE region is a bit too close to the waveguide, and so you have some unphysical modes due to coupling with the boundary conditions.
Single mode wave-guides are not good at coupling to free space, but in FDTD you could terminate the waveguide and place monitors after to see the divergence. Since your waveguide is multi-mode you may find the modes couple to a Gaussian beam efficiently. You could look at how efficiently using FDE in MODE. Maybe consider a horn antenna to improve gain and analyze it with the EME solver. If you would like simulate a complete optical set-up we recommend the Zemax interoperabillity.
Check out these links for more information
Far field projections