Changes for 2024.1.1
To do:
-
Remove warning message for version 2024.1 ; make it the default version -
Update documentation (especially activation command) -
Old configuration files should work with new versions (for now at least 2 issues) -
sigma-clip: one should be able to enter only one threshold instead of 2 required by config file -
"integrator-config" should always generate the full configuration file (i.e "--advanced" parameter is confusing) -
sigma-clip: investigate discrepancy issue wrt 2022.X -
Investigate median filter issue -
Use integrate1d_ng
Notes
The discrepancy for sigma_clip_ng
results is due to the pyFAI versions.
- sigma-clipping is "dangerous" to use when doing pixel splitting (concurrent threads rebinning)
- When there are large Bragg peaks, the Poisson error model is not correct (very different distributions are mixed together). Better use
hybrid
. - thresh=0 sets the Chauvenet mode
For median filter, one has to use the trim_n_pts
parameter. If set to None (default !), it makes medfilt1d
crash because npt_azim
is not provided. Note that it should be a power of two.
Edited by Pierre Paleo