Commit 8a1d604a authored by Vacaliuc, Bogdan's avatar Vacaliuc, Bogdan
Browse files

slides: revise 1, 2, 3 per 2026-04-21 scientist review; add 3a-3e detail slides

Slide 1 (current vs ideal): drop the lr_reduction (sibling) box from
the TODAY panel; the tangled-state story is tight enough with
quicknxsv2 + mr_reduction + Mantid alone.  Label every inter-box
arrow "Calls" so scientists do not read them as inheritance
relations.  Reduce bullet density.  On the IDEAL side, make the
two-instrument story visible: quicknxsv2 UI and an REF_L UI both
Call into ref_core.

Slide 2 (API mapping): the right column is now labelled
"lr_reduction (next branch)" with the subtitle "Mantid-centric ·
REF_L · unpolarized (standard reduction)".  All references to
lr_reduction:new_workflow removed.  Title shortened to "mr_reduction
↔️

 lr_reduction".  This is the Mantid-centric pair; the
new_workflow + quicknxsv1 cross-reference is deferred to after the
hack-a-thon per scientist feedback.

Slide 3 (silent defaults): collapse from 5-column (Parameter / GUI /
auto / Physics / Observable) to 4-column (Parameter / GUI / auto /
Effect icon).  Icons are inline SVG pictograms summarising each
tension.  Rewrote the ErrorWeightedBackground row text per
scientist feedback: the uniform-mean explanation stays, the
inverse-variance-mean text is dropped, and the observable is
reframed as "backgrounds used when high-intensity off-specular
reflections are present unfairly bias low-count reflectivity."
Icon for that row is a detector diagram showing bright off-spec
ovals intruding on the background band — simplified from the
scientist's sketch slide3-ErrorWeightedBackground-Explanation.jpg.

New detail slides 3a through 3e, one per silent default:
  3a  ErrorWeightedBackground — physics (plain mean formula + why
      inverse-variance is not used for REF_M) + observable (detector
      contamination diagram + R(Q) cartoon showing an unphysical dip)
  3b  CropFirstAndLastPoints — physics of edge-bin trimming +
      observable showing both paths' R(Q) with the autoreduce path's
      dropped endpoints marked
  3c  RoundUpPixel — specular-pixel rounding in constant-Q binning,
      number-line diagram of 126.9 (GUI) vs 127 (autoreduce)
  3d  AcceptNullReflectivity — failure-mode difference, side-by-side
      panels showing the GUI's zero curve vs the autoreduce
      RuntimeError stack
  3e  UseSANGLE — current agreement but with a semantic inversion
      trap (UI checkbox "Trust DANGLE" unchecked → Configuration.use_dangle=False →
      MRR UseSANGLE = not use_dangle = True).  A rename of use_dangle
      without auditing the call site would silently flip every θ.

Co-Authored-By: default avatarClaude Opus 4.7 (1M context) <noreply@anthropic.com>
parent 4422e2e0
Loading
Loading
Loading
Loading
−60.9 KiB (179 KiB)
Loading image diff...
+98 −126

File changed.

Preview size limit exceeded, changes collapsed.

−2.67 KiB (288 KiB)
Loading image diff...
+9 −9

File changed.

Preview size limit exceeded, changes collapsed.

−100 KiB (204 KiB)
Loading image diff...
Loading