Perform background subtraction in S(Q)-1 to improve low-r region of pdf in POALRIS reduction #40646
Description
Is your feature request related to a problem? Please describe.
The low-r region of pdf < first interatomic distance is largely spurious (from backgrounds, inadequate subtraction/normalisation) etc.
The data can be cleaned up by subtracting a smoothly varying backgrund in Q (similar to l;ogic of top-hat convolution in gudrun).
For example this is a spline fit to the background with the one-sided robust cost function I have previously talked about.
The pdf is much improved, particularly at low-r as expected.
After #40639 a constant has been subtracted - but the background subtraction still proves more effective
Describe alternatives you've considered
Top-hat convolution detailed in Gudrun manual and Section VI of this paper
https://hal.science/hal-00513296/file/PEER_stage2_10.1080%252F00268970903025667.pdf
But Gudrun only does this for data evenly spaced bins (see p88 of Gudrun manual)
https://www.isis.stfc.ac.uk/OtherFiles/Disordered%20Materials/Gudrun-Manual-2017-10.pdf
Which is not the case for POLARIS data - particularly not after banks are merged using StitchByBackground - see #40182
Additional context