Where {filters} land on the CIE diagram under any light source, how LEDs render colour, and which {filter} gives a wanted colour under another light. {built} are built in; pick a tool below.
Pick two lights, A and B. Every {filter} becomes a colour under A and, from the real spectra, a true colour under B. A chromatic adaptation transform (CAT) only gets the colour under A plus the white points of both lights, and has to predict the result under B. The gap between its prediction and the spectral truth is the error measured here.
What a CAT does. It takes a colour as XYZ, the XYZ of the white under the old light and the XYZ of the white under the new light. It converts to a cone-like space, scales each of the three channels by new white divided by old white, and converts back. That is the idea you described.
Why there is an error. The CAT sees three numbers per colour. Two {filters} with different spectra can give the same XYZ under light A and different XYZ under light B. The CAT gives both the same answer, so it cannot be right for both.
What "true" means. The spectral result is {curveL} times the spectrum of light B, divided by the Y of white B. It is the colour relative to the white of light B, which is what a fully adapted viewer would see. That is the case a CAT with D = 1 tries to predict.
The two views. "One {filter}, all CATs" shows where each CAT lands compared with the true point. "All {filters}, one CAT" draws a whisker from every true point towards where the chosen CAT puts it. Whiskers are magnified so small errors become visible.
Use D = 1 to judge how well a CAT predicts the spectral result. A smaller D under-adapts on purpose, so the error then mostly shows the incomplete adaptation.
The {filter} and both lights are real spectra. The CAT sees them only through three broad, overlapping sensitivities (coloured, from the selected CAT's matrix). Spectra that give the same three numbers under A are indistinguishable to it, even when they split apart under B.
Every light source in one place: the built-in CIE standard sources and your own LED spectra. Each point is one light. Colour the points by CRI or TM-30 to see which parts of the diagram hold good and poor colour rendering, then pick one to see its spectrum and its TM-30 colour vector graphic.
Pick light A and light B, ideally with the same CCT and Duv but different colour rendering. Every {filter} is worked out from its real spectrum under both lights. The chart shows how far each one moves when you swap A for B. The two colour vector graphics show, hue by hue, what each light does to saturation and hue. Two more tabs use the lights chosen here: Lights for a filter and Filter match.
Vertical axis: lightness (how bright, relative to a white of the same light). Horizontal plane: colour, with grey in the middle and more saturated colours further out. Small dot = under A, larger dot = under B, joined by a line; with “Actual colour and brightness” each is drawn in the colour and brightness it really has (the bare light's white is full brightness, so filters that let little light through look dark), otherwise in the colour scale of the “Show” setting. The straight-line distance between the two dots is the apparent change ΔE: it combines a change in colour and a change in brightness in one number. A line that only goes up or down is a pure brightness change; one that only moves sideways is a pure colour change.
Dots show where each light puts the 16 hue bins relative to its own reference light (the circle). A light that matches the reference sits on the circle. Light A and light B are drawn on top of each other so the differences stand out.
The other way round: pick one filter or pigment and see which lights in the library change it the least, or the most. Uses the lights and settings chosen in LED compare (see the strip above).
“Each light’s own ideal reference” compares the filter under every light with the same filter under a perfectly rendering light of the same colour temperature, so the ranking shows which lights show this filter most faithfully. With “light A” the ranking shows which lights are closest to (or furthest from) A. Click a column heading to sort by it, and click a light’s name to open it in the LED library. Colour dots show the colour the filter has under each light (left: reference, right: that light); in the brightness view they are darkened by how much light gets through.
Pick a filter that gives the colour you want under light A (for example your 2700 K LED), then find the filters that give the closest colour under light B (for example a 6500 K LED). Lights A and B are the ones chosen in LED compare (library lights, or synthetic ones with sliders).
The colour is the colour of the light that leaves the filter, in u′v′, so a match means the two lights look the same side by side. It is not corrected for the eye adapting to a warmer or cooler white. “Light out” is the share of the bare light that gets through, at equal lumens of the two bare lights; a filter that matches in colour can still need more or less light. In the chart the cross is the target, the dashed ring is where the same filter lands under B, and the dots are the filters under B; the ringed dot is the best match.
Measure a light (emission) or a pigment sample (reflective) with the i1 Pro. Connect it directly in Chrome or Edge. Each measurement is saved with the calibration it was made with. Lights show their colour numbers at once and can go to the LED library.