Measure one workplane under one declared instant, one set of shades and one
set of luminaires.
The governing model is stated and bounded rather than implied:
Beam illuminance is DNI * max(0, dot(n, s)), zeroed when any solid stops
the ray to the sun;
Sky illuminance is the isotropic vault DHI * V, where V is the fraction
of a cosine-weighted hemisphere that reaches the sky unobstructed and above
the reference ground. An unobstructed ground-facing-up plane has V=1; tilted or shaded planes retain sampling error;
Each luminaire contributes I * max(0, dot(n, d)) / r^2, zeroed when a solid
stands between the fitting and the point;
Ground-reflected light is not modelled, and says so as an unsupported
metric rather than being quietly folded into the sky term.
Anything the solver cannot answer is reported rather than approximated. A sky
model it does not implement returns an unsupported run; a study with
neither a sun nor a fitting returns not-run; a contrast ratio over a plane
with a dark point returns a gap, because max / 0 is not a large number, it
is not a number.
Settings text covers the declared study fields and preserves list order; equivalent covered JSON values repeat the digest, while ignored fields, signed zero and possible hash collisions limit identity claims.
The cosine-weighted sky estimate is exact only for an unobstructed plane
whose hemisphere lies wholly above the reference ground; a tilted or shaded
plane's sky term depends on diffuseSamples. This identity is mathematical: binary64 light products and summation can overflow, and nonfinite computed metrics throw even when scalar inputs were admitted.
Measure one workplane under one declared instant, one set of shades and one set of luminaires.
The governing model is stated and bounded rather than implied:
DNI * max(0, dot(n, s)), zeroed when any solid stops the ray to the sun;DHI * V, whereVis the fraction of a cosine-weighted hemisphere that reaches the sky unobstructed and above the reference ground. An unobstructed ground-facing-up plane has V=1; tilted or shaded planes retain sampling error;I * max(0, dot(n, d)) / r^2, zeroed when a solid stands between the fitting and the point;unsupportedmetric rather than being quietly folded into the sky term.Anything the solver cannot answer is reported rather than approximated. A sky model it does not implement returns an
unsupportedrun; a study with neither a sun nor a fitting returnsnot-run; a contrast ratio over a plane with a dark point returns a gap, becausemax / 0is not a large number, it is not a number.Settings text covers the declared study fields and preserves list order; equivalent covered JSON values repeat the digest, while ignored fields, signed zero and possible hash collisions limit identity claims.
The cosine-weighted sky estimate is exact only for an unobstructed plane whose hemisphere lies wholly above the reference ground; a tilted or shaded plane's sky term depends on
diffuseSamples. This identity is mathematical: binary64 light products and summation can overflow, and nonfinite computed metrics throw even when scalar inputs were admitted.