Cataloguing what light does and what the engine models
The light catalogue is a table of what light does, with a status on every row saying where the site’s light engine stands. On 30 September it held 44 rows: 24 modelled, 15 in preview, and three ruled out, each with a reason that can be checked.
The rows run from a lamp’s inverse-square falloff (light weakening with the square of distance) to the afterimage in the eye, in six groups: sources, at a surface, inside a material, between surfaces, wave effects, camera and eye.
The statuses are Now (modelled today), Preview (built behind a preview switch on the site, waiting for me), Plan (in the design), Camera (belongs to the camera stage) and No (not visible at this scale, with the reason given).
Twenty-four behaviours are modelled and fifteen are in preview
Twenty-four rows are Now, fifteen are Preview, three are No, one is Camera, one is left to the photographs. No row carries Plan any more. Behaviours move from Plan to Preview to Now, and the 30 September update folded in what the latest engine work had built.
Four of the 15 previews landed on 30 September. They are the lamp’s colour from its temperature (Planck blackbody to CIE 1931 to linear sRGB, the chain from a glowing body’s spectrum to a screen colour), the lens’s natural vignetting (darkening toward the corners, by a cos⁴ law), museum glass with a quarter-wave anti-reflection coating, and opal glass that scatters inside itself. The catalogue moved each row as it landed.
The three ruled-out rows are below a pixel or absent in float glass
The three rows are Goos–Hänchen lateral shift, a tiny sideways step in a reflected beam (far below a pixel); diffraction at glass edges (the bands are a few wavelengths wide, far below a pixel); and birefringence, the splitting of light in two by some materials (float glass is isotropic, the same in every direction, and tempered glass shows stress patterns only in polarised light).
Two of those reasons are sizes and the third is a fact about the material, so each can be checked.
One row points the other way. Depth of field, the photographs’ own blur, is marked as being in the photographs themselves: the engine does not add a blur to pictures that already have one.
Some modelled rows are approximations
Fresnel reflection, how much a surface reflects at each angle, is listed as Now (Schlick), the usual approximation. The exact s and p equations are in the polariser preview, which also holds Brewster’s angle at 56.6°, where glass reflects only one polarisation.
Rough transmission through frosted glass is Now (approximate), with the full microfacet model, which treats a rough surface as many tiny facets, in the rough-glass preview. Total internal reflection, where light inside the glass can’t get out, starts at 41.2° in this glass. It is Now, and it is what makes crack faces flash.
Sources and notes (6)
- Walter et al. 2007, microfacet BTDF — the rough-transmission model behind the frosted-glass row
- Schlick's approximation (Wikipedia) — the Fresnel approximation the Now column uses
- Fresnel equations (Wikipedia) — the exact s and p version held in the preview
- Total internal reflection (Wikipedia) — the critical angle row
- Wyman, Sloan and Shirley 2013, CIE fit (JCGT) — the CIE 1931 fit the repo's optics note uses for blackbody colour
- Light catalogue — the notes this was written from
Light & shadowAll 19 parts →
