Speaker
Description
Automated data analysis at synchrotron beamlines is often limited by the
complexity of managing many processing parameters across large sets of related
measurements. Although the conversion of raw detector data into physically
meaningful representations is largely deterministic, existing approaches
typically embed parameter handling directly in analysis scripts, making
automation, reusability, and reproducibility difficult. We introduce the Spice
concept, a declarative approach for managing analysis parameters in a
scan-aware manner.
Spice separates parameter definitions from the analysis
code and organises their validity along the chronological sequence of
explicitly typized scans. Update boundaries can be automatically assigned
by the machine and refined by the user. The versioned update
mechanism has is built around a hierarchical built-in defaulting mechanism.
This allows parameters to be applied consistently across multiple scans
with drastically reduced user intervention, while still supporting local
adjustments where necessary.
Using X-ray diffraction as an example, we demonstrate how
Spice enables a clear separation between routine, instrument-specific data
preparation (“cooking”) and the subsequent scientific interpretation. We
describe an implementation within the Nx5d framework and discuss how the
approach supports reproducible analysis, automated reprocessing, and
provenance tracking.
The Spice concept provides a general architectural pattern for scalable
and maintainable data processing at synchrotron and other large-scale
experimental facilities, in particular where enabling second-pass
data analysis after FAIRdata/OpenData principles is a priority.
