v0.29.1 — The beam turns on the coating

A correctness release: no new features, no API change. Everything here is visible in the viewport
rather than in the numbers — and no traced number changed. The green_doubler and michelson
segment digests are identical to v0.29.0, because the tracer reads ports and matrix_world, never
the mesh.

Fixed

• Reflective optics are seated on the plane they reflect from (#25).
  A mirror's coating stood 3 mm proud of the REFLECT plane, a dichroic's 1.5 mm, a grating's 2.5 mm,
  and a curved mirror's apex a further 2 mm — so the drawn beam folded inside the glass. For a
  curved mirror the seated point is the apex, not the rim: the mesh's spherical cap carries a fixed
  cosmetic sag, so seating the rim would still have turned the beam 2 mm off the face it visibly
  strikes.
• The mount follows the glass. Sliding the mesh alone left the KM retaining ring — the physical
  stop for the mirror edge — floating 3.6 mm clear of the face it retains, with the plate bore
  cutting through the coating.
• The deformable mirror is seated too. It sits in the tracer's REFLECTIVE set and folds at its
  REFLECT plane like any mirror, but the first pass missed it; its face stood 2 mm proud.
• IN ports sit on the coated face (#29).
  Beyond tidiness this corrects auto-alignment residual reporting, which had been aiming at a
  point off the interaction surface: on the bundled benches GD_KTP's angular error drops from
  0.4135° to 0.000007° and GD_Dichroic's position error from 1.2186 to 0.5999 mm, while
  MI_M_fixed's genuine 3.9725° misalignment is reported unchanged. The metric stopped charging for
  a standoff; it did not go blind.

Documented

• The corner cube is a thin-element idealization (#30).
  The internal path is deliberately not modelled — the element carries no depth parameter and the
  mesh apex is a cosmetic proportion. The omitted term is 2d: for a hollow cube exact and
  identical for every ray in the aperture, independent of both entry point and incidence angle
  within the acceptance cone. A solid cube is 2nd at normal incidence only. In an interferometer
  the missing 2d is a constant piston — it shifts fringe position without distorting the
  wavefront.

Install

Drag the zip into Blender, or use the one-click link on the
project page which also subscribes you to
the extension repository so future releases arrive as native updates.

optical_alignment_sim-0.29.1.zip
size    459381 bytes
sha256  8eb1387f46f5f4d4dfe7268abce6d156098f41a9bfe627b778a6e4fa739fa86e

Requires Blender 4.2.0+. Tested on 4.2.3 and 5.1.1.