ISISNeutronMuon
[ENHANCEMENT] Update MDANSE so that the weights are not rescaled
Is your feature request related to a problem? Please describe. MDANSEs weighting scheme can lead to confusion. For example, when a DCSF calculation of pure Ar36 and pure Ar38 is done they will appear to have similar results. A new user would expect to see that Ar36 would have a much larger S(q,w) than Ar38 since it has a larger b_coh scattering length.
This is because in MDANSE the scattering lengths are rescaled e.g.
Describe the solution you'd like Scattering length and other weights are applied without rescaling them. This change will also fix #530 and #438.
This needs to be discussed. As usual, there is not a single good answer. Historically, both the DISF and DCSF analyses returned a "normalized" S(Q,w) using the weights as you describe above. This makes easier to compare different systems or to check the theoretical predictions (e.g. that F_inc(Q,t=0)=1 or F_coh(large Q,0) = 1). And as most of the experimental data are in arbitrary units, finally it does not really matter if the data are normalized or not. The situation changes for the NDTSF analysis, because in this case the goal was to have a direct view of the relative incoherent and coherent contributions (e.g. in H2O one would have 95% inc + 5% coh and in D2O 21% in + 79% coh). To do this, one needs to work with the sums without applying the weights. So if you want, you can remove the weight rescaling and there is nothing fundamentally wrong, but I am not sure that this is what all the users would expect or prefer. Probably some of them yes, but many others will regret this change. My personal preference is to keep the DCSF and DISF analysis as they have always been (with the normalization applied in the same way as it is applied for the other analysis, and even with the choice of using any arbitrary property as a weight), and the NDTSF analysis as a "neutron specific" analysis where the weighting cannot be chosen and there is no scaling.
Probably unlikely to be implemented, given that most users expect the weighting.