# Optics Photons travel from the source to the camera entrance. The optical model holds the mirrors, obscurers, the focal surface and the optical responses. ```{toctree} :maxdepth: 1 optical-elements telescopes shadowing limitations ``` ## Tracing a photon 1. The source supplies position, direction, wavelength, time, weight and id. Star light (`star_photons`) is sampled on a disk (radius 1.2 times the dish radius, perpendicular to the optical axis); a source at finite distance emits from its position. 2. Incoming leg: nearest of primary mirror, secondary back side (if it blocks incoming light) and obscurers. Lost photons record where and on what. 3. Primary: the nearest valid mirror hit gives a specular reflection, scattered by `mirror_reflection_random_angle` (two Gaussian components). For facets the reflected direction is rotated; for polynomial mirrors the normal is tilted by half the angle. 4. Secondary (dual mirror): same, with the after-primary obscurers. 5. Focal surface: plane or polynomial. Position, direction, time and incidence angles are recorded. Tracing stops at the camera entrance; pixels, light guides and the camera window are not part of the optical model. ## Mirror scatter conventions `mirror_reflection_random_angle` follows the convention of the simulation it comes from. It is not a measured RMS of the reflected-ray angle. Per reflection, one Gaussian width is drawn (the second component with the given fraction), then two independent deviations a1 and a2 with that width: - **Segmented mirrors** (`method` `outgoing_angles`): the reflected ray is rotated by a1 about one tangent axis of the facet and by a2 about the other. A rotation about a tangent axis moves the ray by sigma times the sine of the angle between ray and axis, so the mean squared deviation is sigma^2 (1 + cos^2 i) for incidence angle i: sigma per axis at normal incidence, and the in-plane component shrinks with the cosine of i. The axes are orthonormalised against the normal at the hit point (on a curved facet the vertex tangent is not perpendicular to it). - **Polynomial mirrors** (`method` `surface_slopes`): the slopes dz/dx and dz/dy at the hit point each change by half of a1 and a2, so the reflected ray deviates by about the full angle where the surface is flat. A slope is not an angle: a slope change ds tilts the normal by ds cos^2(theta) radially and ds cos(theta) tangentially, with theta the tilt of the surface. On strongly tilted parts of a secondary the effective width is smaller than the nominal one, and anisotropic. Compare PSFs, not parameter values, between this convention and one defined as a ray angle. Both are uncorrelated between photons, equal over the whole mirror and independent of wavelength. They describe the PSF core; a wide scattering halo is not modelled. Weights multiply the responses met (unless geometry only). Random numbers follow the common rules of [](../stream/random-numbers.md), so results do not depend on threads, blocks or photon order. Times use the group index stored in the model.