BLM calculations (python)

[phockett]

Things to do:

• Normalisation
• Vectorization/parallelization
• Verification
• Saving of Xarrays
• Tensor formalism
• Matrix formalism
• AF-BLM calculations
• ADM (and other, esp. external values) normalisation options/routines.

[phockett]

Testing code (local only) D:\code\ePSproc\python_dev\ePSproc_MFBLM_dev_tests_050919.py

Notebook demo (ongoing): ePSproc_BLM_calc_demo_Sept2019.ipynb

[phockett]

Basic code now working & verified. Pretty slow.

[phockett]

Should be able to get a good speed-up on repetitive function calls with caching: https://docs.python.org/dev/library/functools.html#functools.lru_cache.

See also numba development branch for other methods.

[phockett]

Normalisation by SF and B00 > X-section now in MF-BLM calcs (dev version).

[phockett]

Initial version of AF-BLM code added. Running, not verified (almost certainly wrong).

[phockett]

Updated AF-BLM code for Xarray ADMs. Sort-of working. Careful testing still required.

[phockett]

Tensor formalism now basically working. Debugged for N2 (z,x,y) case, and verified against existing loop code. A few issues left to clear up, mainly consistency and tidy-up, plus some new capabilities.

TODO:

• Check for source of small differences (but >threshold) between old and new calcs. Look to be Eke dependent, so likely SF and/or/with thresholding method.
• Test for other pol geoms and sym mixing.
• Plotting.

[phockett]

Tensor formalism also now used for LF case, for direct comparison with ePS GetCro outputs, see https://epsproc.readthedocs.io/en/dev/methods/LF_AF_verification_tests_260620_tidy.html.

Still TODO:

• Verify for AF case.
• Degeneracy factors for general/AF case.
• General tidy & rewrite dev code.

[phockett]

Revisiting AF case, looks like 'it' degeneracy needs some work - implicit frame rotation here? For N2 pi-g test case, get -ve X-sections with current calculation, which implies phase issues and/or renorm issue.

[phockett]

ALSO: seem to have an inconsistency with (x,y) pol geometries here - should check source code & fix. Likely due to mix-up in frame defns., i.e. need to define (x,y) in LF not MF?

[phockett]

See #26 for ongoing issues for geometric/tensor versions.