Actually there's a substantial difference between radiation boundaries and PML. The numerical problem of meshing requires a finite volume to be used for Finite Element method. PML acts as an absorber, in order to avoid reflections at the domain edges. The reflections, in such a case, are due to discontinuities between the discretized domain/volume containing the structure and the non-discretized domain/volume. Using PML, this numerical reflections are minimized. Radiation is a boundary, used to indicate that wave can propagate through those directions towards infinity (free space). Here's a definition I found on a presentation,
for HFSS:
Radiation Boundary vs. PML
-Recommended spacing for radiation boundary is
lambda0/4 while for PML, it is lambda0/6.
-In general, PML boundaries can be placed closer to the
structure
-Depending on the angle of incidence, some reflections
will exist for radiation boundaries
-PML boundaries have zero reflections
-
Radiation boundaries is a boundary condition
--
Definition is the 2nd order boundary condition that
approximates free-space
-PML is part of the solution space
-Definition is a set of “fictitious” biaxial anisotropic
material
An example of use of PML is at
www.emtalk.com/tut_3.htm.
For characterizing an EBG unit cell, have a look at the attached file. It may be helpful. Another solution, with respect to the one you find in this file, is to enforce an incident plane wave and use E-H boundaries and a radiation boundary in place of the waveport (in HFSSv11 is possible to choose a radiation boundary with incident fields). Another way for characterizing a unit cell is to consider periodic boundaries (master-slave) and enforce the incident plane wave by correctly setting a
floquet's port.
Hope this helps.
I.