Data Reduction Notes

Notes from canSAS 2024 Data Reduction Topical Presentation/Discussion

Information:

BILBY is a useful playground given its 5 detectors and its ability to work both monochromatic with velocity selector or ToF, providing variable resolution.

Interventions by Adrian Rennie (Uppsala University, Adrian.Rennie@kemi.uu.se), Tyler Martin (NIST, tyler.martin@nist.gov), Brian Pauw (BAM, brian.pauw@bam.de), Andrew Jackson (ESS, andrew.jackson@ess.eu), Glen J Smales (BAM, glen.smales@bam.de)

Topic of autoscaling traceability:

• Brian Pauw (BAM) suggests to avoid entirely the issue of stitching by scaling each configuration to absolute units [if absolute scale is excellent in every instance and therefore no adjustment is necessary, which is not always the case].

Topic of transmission values:

• Adrian Rennie (Uppsala University) states that the physics behind transmission should be used to make a proper fit; the modelling of transmission vs wavelength can go therefore beyond a simple polynomial especially if there are edges in the energy range experimentally accessed.

• Glen J Smales (Diamond) asks that how the transmission is determined must be indicated. What solid angle range is used for that? And when you’re measuring the incident flux somewhere else (further upstream) than the “transmitted” flux, is it actually comparable?

Topic of uncertainties:

• Andrew Jackson (ESS) asks how to account for uncertainties that are not a result of scattering counting statistics? E.g. uncertainties on transmission factors.

• Brian Pauw (BAM) ask how multiple uncertainties could be handled. How to deal with correlation between errors? It remains an open question [probably would have to be dealt with by the data analysis software].

• Adrian Rennie (Uppsala University) informs that to his knowledge most data reduction packages do not include correct treatment of correlated random errors such as those arising when the same detector efficiency corrections are used for several components in a measurement series (sample, background, etc.).

Topic of overall data reduction process:

• Tyler Martin (NIST) mentions Reductus, a graph-based description of the sequence of data reduction. See publication https://doi.org/10.1107/S1600576718011974. Andrew Jackson indicates that ESS is doing something similar.

• Brian Pauw (BAM) mentions his universal data correction work, explaining that by experience it becomes easy to see if there are mistakes, and at what stage of the data corrections.