Validate CL/PF Rays #1342
Description
In #1338 I found some discrepancies between the MATLAB PFT and Python regarding the orientation of their frequency grid for PFT. This generates a different (reflected) PFT representation which may or may not require different usage with shifts/phases/filters and symmetry concerns all down the line. Different usage was required for estimating the abinitio shifts. My suggested patch was to take the MATLAB convention and make calling code match MATLAB, but it was decided to take a more typical polar convention here. Unfortunately it is not clear i) what convention was expected/used when much of the CL code was ported ii) when/where that convention matters.
I ran a sanity check conjugating the pfi and pfj rays in some areas or commonline_base and was still able to achieve a reconstruction. I'm inclined to believe the CL matrix build may work with either set of rays, but I have concerns about the corner cases...
I'd like to see
- the algorithms reviewed for any canonical convention regarding the rays
- areas where the convention matter or are agnostic identified; patched if needed
- all numerical differences between the PFT identified/isolated
- Differences in CL mat identified/isolated
- Currently in code I am only aware of a threading bug in the MATLAB mex code that was fixed in Python.
- Currently in results, the tables are interchangeable up to MATLAB (1based) conversion. Most of the table agrees, but some elements are +-1 mod 360, and some which concern me are 180+-1.