Measure colour. Keep the meaning.

RGB values describe mixtures of particular primaries. A triplet from a camera, a web file and a working space need not identify the same stimulus. Before comparing values, establish the coordinate system, encoding function and scale. Colorimetry provides an agreed description of colour stimuli through a standard observer rather than through whichever device happens to be available.

XYZ is useful as an intermediate representation. Device independence does not mean absence of conditions: the observer, input data and normalization still matter. XYZ computed from a known sRGB pixel describes the specified signal. It is not a measurement of light emitted by your monitor.

In colour-matching experiments, an observer compares a test stimulus with a mixture of three primaries. Some matches require adding a primary to the test field; the original coordinate system records that as a negative coefficient. This is a description of the matching arrangement, not negative physical light.

Colour-matching functions reduce a spectral distribution to tristimulus values through weighted integration. XYZ is constructed as a transformation of a matching system. Its functions are neither three camera filters nor direct cone sensitivities. An XYZ vector does not retain the original spectrum.

CIE 1931 2° and CIE 1964 10° are different standard sets of functions for different visual field sizes. The angles do not describe screen resolution, camera rotation or an accuracy percentage. Changing the observer can change results calculated from spectra.

A standard observer does not guarantee identical perception for every individual. Observer variation matters particularly with narrow-band sources and comparisons between display technologies. This lab uses RGB specifications referred to CIE 1931 2°. It does not offer a fictional 10° switch without the spectral data needed to recalculate it.

X, Y and Z are tristimulus values, not red, green and blue percentages. Y relates to photometric luminance through the normalization being used; X and Z complete the colour specification. Y is not nonlinear video luma Y′, nor is it a direct measure of perceived lightness.

In relative notation, reference white is often normalized to Y = 1 or 100. Converting relative Y into cd/m² requires an absolute reference. Y = 0.5 in this lab does not mean that the screen emits 50 nits. Calculations retain floating-point values; clipped sRGB preview patches do not modify them.

Chromaticity is computed as x = X/(X+Y+Z) and y = Y/(X+Y+Z). Multiplying all three XYZ values by the same positive factor does not change these ratios. A dim stimulus and a brighter one may therefore occupy the same xy position. Add Y to recover the full XYZ description.

At X = Y = Z = 0, normalization divides by zero. Ideal black has undefined chromaticity: displaying D65 or (0,0) would invent data. An inverse xyY conversion must also check for nonzero y and valid coordinates.

The spectral locus connects the chromaticities of monochromatic stimuli. The line joining its ends is the line of purples: these mixtures do not correspond to one wavelength. The coloured background of a familiar diagram is only an orientation aid; an ordinary monitor cannot faithfully reproduce every stimulus shown on it.

This lab deliberately draws only RGB triangles and the calculated point, without an invented spectral horseshoe or a claim to render the whole visible gamut. Switching xy to u′v′ changes the coordinate system, not XYZ. Axis labels and numerical readouts change together.

An additive RGB system places three primary chromaticities at the vertices of a triangle. Nonnegative mixtures fall inside it. This useful two-dimensional projection contains neither peak luminance nor black level, power limits or the luminance available to individual saturated colours.

A point can lie inside the chromaticity triangle while its required XYZ still needs an RGB component above one at the chosen scale. A negative component after a valid conversion can indicate an out-of-gamut colour rather than a faulty matrix. Do not clip intermediate values just to produce a neat preview patch.

A white point defines the reference neutral of a colour representation. sRGB and Display P3 use D65 despite having different primaries. With Y = 1, D65 is approximately XYZ [0.95046, 1, 1.08906]: a neutral need not have equal XYZ components.

Under another reference white, equal R/G/B may represent a different chromaticity. Do not force a white XYZ vector to equal components as a supposed balance correction. When whites differ, decide whether to preserve absolute colorimetry or estimate corresponding colours for a new adaptation state.

Equal distances in xy do not correspond to equally visible differences. Triangle area must not be advertised as a percentage of all perceived colours, and an ordinary XYZ residual is not automatically ΔE. Colour-difference evaluation uses specified spaces, formulas and conditions.

Even CIELAB is an approximation rather than a universal appearance model. Lightness, chromaticity, surround and adaptation do not reduce to one plotted point. Numerical agreement is a useful technical check, not a promise of identical appearance under every viewing condition.

First identify what the signal represents at the selected stage: a decoded source, a working representation or a display output. A space name without encoding and context can be incomplete. A flat-looking image alone cannot identify the correct input.

Check input assignment, transform stages and the quantity shown by each scope. CIE Chromaticity, Waveform and RGB Parade answer different questions. Chromaticity does not replace level monitoring, and a browser patch does not replace a controlled viewing path. This experiment models neither particular OFX settings nor a Resolve project.

Take x = 0.3127, y = 0.3290 and Y = 0.18. The formulas X = xY/y and Z = (1−x−y)Y/y give approximately X = 0.17108 and Z = 0.19603. Doubling XYZ raises Y to 0.36 while x and y stay unchanged. This follows from normalization, not a special application feature.

Now change X alone. The sum changes, so both x and y change. Adding red in RGB and increasing X in XYZ are different operations. Name the basis before interpreting a slider movement.