Notes from meeting with Jochen & Jos 23/02/2024

[nvaytet]

Jochen’s reduction workflow

• Our current workflow is based on his script from 2 years ago
• Script has evolved since
• Code in on github in a private repo, but could give us access (will be next week)
• What are the steps when a user arrives?
  • Nexus files created from events taken from kafka
  • Users just want to see reflectivity curves
  • Read in events
  • Do corrections, conversions, normalization
  • I(theta, lambda)
  • Then R(Q), resolution function
  • This covers 90% of applications
  • Also using merging of multiple runs
  • Sometimes, time-resolved (e.g. batteries, discharge), usually slow changes
  • Diffusion studies can be faster (e.g. seconds, or even less)
• Covering Q ranges by tilting the sample:
  • Compute two R(Q) and usually not merging into a single curve because resolution changes when you tilt the sample
• Information of I(theta, lambda) does get lost when collapsing to R(Q) because some parts of the 2D plot have more background (lower theta) and when you collapse you are polluting the areas with less background. This usually means that the peaks stay the same but the valleys between the peaks are less pronounced.

Questions:

• Polarization: how os the data recorded? One file, Multiple files? -> goes into multiple data files. In the future, would like to have in a single file, with a log.
  • Issue with metadata from stream for polarization: probably a bug in filewriter but it was written by Jonas? And has not been updated
• Pixels that can never get any neutrons: do we mask those? Jochen is keeping them.
• Normalization: how is it done? -> We measure one super mirror at low angles. Then, if you know the footprint, you can use it for other angles. Normalization is applied to $I(\theta, \lambda)$.
• What parameters do users change? Mostly just sample inclination and detector inclination. Can read from file metadata, but useful to override for debugging.
  • First time you use the instrument, use samples you know very well.
  • Aligning at high precision can take a long time, use I(lambda, theta) plot to check alignment. Do not trust the values in metadata.
  • Right now, this is done by hand/eye. Would be nice to have procedure to do automatic correction of alignment (by fitting triangles to point to origin), but difficult to make something robust?
  • $\theta$-range and $\lambda$-range was also mentioned as parameters users might want to experiment with