Improve accuracy at backscattering direction

[GoogleCodeExporter]

For some cases the accuracy of the DDA results is especially bad at (or near) 
the backscattering direction. This has been mentioned before, but was recently 
exemplified by Anne Virkki et al. at ELS14 - 
http://www-loa.univ-lille1.fr/ELS-XIV/documents/abstracts/ELSXIV_abs81.pdf

First task is to understand when and why this is happening. Probably, this 
correlates with importance of shape errors. So the issue is relevant when the 
shape errors dominate at backscattering (compared to discretization errors).

Second task is to somehow improve the situation. Currently, the best option 
seems to be the usage of FCD formulation + extrapolation. Further improvement 
should probably be by addressing shape errors, e.g. weighted discretization. 
The latter is touched upon in issue 12.

Original issue reported on code.google.com by yurkin on 9 Jul 2013 at 9:10

• Blocked on: #12