What counts as white. And relative to what.

A colour representation assigns a particular chromaticity to its reference white stimulus. In normalized RGB this usually corresponds to equal maximum components, but those numbers alone do not define the emitted light. Primaries, white chromaticity and encoding complete the definition.

A reference white does not describe every white object in a scene or require removing warm sunset light. Technical neutrality and a creative balance are different decisions. One reference white supports intentionally warm or cool images; changing the mood does not itself require redefining the colour space.

A spectral distribution describes energy by wavelength. Chromaticity xy reduces a spectrum to coordinates for a specified observer. Correlated colour temperature, CCT, is a further summary: the temperature of a nearest Planckian radiator under a defined distance rule. These levels of information are not interchangeable.

CCT alone cannot reconstruct xy, the offset from the Planckian locus or a spectrum. Sources with the same CCT may differ in cross-locus tint and material rendering. A software Tint slider is not a universal Duv scale, and CCT is not an appropriate description for every possible chromaticity.

The D family describes daylight; a Planckian radiator is a different physical model. Their loci are close but not identical. D65 defines a standard spectrum, while an RGB specification normally uses the associated white chromaticity. A D65-calibrated display does not have to reproduce daylight’s spectrum.

The names D50 and D65 do not imply exact CCTs of 5000 and 6500 K. Rounded values near 5003 and 6504 K are commonly given for the canonical daylight illuminants. Matrix work should use documented chromaticities and precision instead of reconstructing a white from its rounded name.

The pair x,y specifies white chromaticity, not luminance. For relative calculations, choose Yn = 1 and derive Xn = x/y and Zn = (1−x−y)/y. Changing the normalization scales the whole XYZ vector without changing its chromaticity.

When comparing tables, check scale and rounding. Slightly different matrices can result from rounded white XYZ values rather than from different adaptation algorithms. The demo uses one explicit xy convention throughout and displays derived XYZ and u′v′ so its assumptions are visible.

The ACES white is specified at x = 0.32168 and y = 0.33767. It is often shortened to D60, but the Academy explains why it is D60-like rather than an exact match to CIE daylight chromaticity at 6000 K. A shared informal name does not turn two slightly different coordinate pairs into the same specification.

An ACES encoding white neither prohibits other scene illuminants nor dictates the final display white. It supports interpretation and conversion. Do not choose set lighting or monitor calibration solely because a project uses ACES.

A display white point is a physical property of its operating mode. File metadata and the working-space selection describe the signal; they cannot by themselves reshape the backlight spectrum or remove surround effects. A setting label and a measurement may differ.

Repeatable evaluation calls for a fixed display mode, white level, surround and output path. Automatic comfort adjustments undermine stable comparisons. This browser page does not measure or calibrate the monitor: its patches encode a numerical example in sRGB, not light at a commanded physical temperature.

Changing primaries and changing a reference white are distinct tasks. Preserving absolute colorimetry keeps the same XYZ stimulus; predicting corresponding colours for a new white generally changes XYZ. Saying different whites always require a CAT is therefore too categorical without a conversion objective.

A relative Bradford conversion can map the source white to the destination white and estimate corresponding colours for other stimuli. It is a model, not spectral relighting. If a profile or managed system has already performed the adaptation, a separate identical stage may apply it twice.

Separate input representation, working space and output in the project. Seeing D65 in one menu does not prove that the whole path is consistent. Identify which operation changes coordinates, which merely assigns interpretation and where the white is already handled.

Before adding explicit technical effects, write down source and destination neutrals. Check a known grey alongside coloured samples: a correct white does not guarantee correct colours. A standalone CAT is neither a replacement for camera decoding nor an automatic solution to mixed lighting.

Three coordinates do not preserve spectral structure. Sources matched at white can interact differently with coloured surfaces, and cameras have their own spectral sensitivities. Matching a grey card therefore does not guarantee that fabric or skin will also match exactly.

Investigate illumination, capture and processing rather than changing the reference white indefinitely. A local correction can be a valid creative choice, but it is not the recovery of lost physical information. A single RGB frame does not provide a universal inverse spectral solution.