ShadowFinder compared each pixel with the median of a 1-pixel circle about the
geometry's beam centre. On a tilted detector that centre is the point of normal
incidence, not where the beam lands, and the rings of constant 2theta are not
circles about it: on a 2theta arm swung by 5-10 deg at 50 mm the circles cut across
the background's steep fall-off and the far side of the detector read as shadow.
Every tilted home-source SMV set in the open arm masked 14-41 % of its detector.
A pixel's ring is now distance * tan(2theta) from the detector matrix, its azimuth
for the arm search the lab azimuth, and the mean is divided by the obliquity
(distance / l_z)^3 - the part of the solid angle that varies along a 2theta ring -
alongside the polarization factor. Host and device share one helper. Untilted, all
three are exact identities, so untilted sets keep their mask bit for bit.
Named-set before/after on the same commit (16 sets): 5reo, 5uth, 6h2p_1p89A, 4bwl,
lyso_x06da_ref unchanged. Shadow 26.5->5.9 % (3mc4), 40.6->10.4 (3r6o),
39.4->6.4 (5vml), 37.1->9.2 (3meb), 33.1->6.8 (5cc8), 13.8->5.8 (3p85);
completeness 77->91, 43->98, 80->84, 43->78, 45->80, 86->87 %. The 0.2 deg-tilted
in-house PILATUS4 sets move by a few hundred pixels; the 20/30 deg tilted
small-molecule sets mask 0.4->0.8 and 0.1->0.6 %, SHELXL R1 unchanged or better.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PXc1zD1yxHqaR1DGdCTHL4
leonarski_fandClaude Opus 5.5
c207012a13