A depth of color brought to light by a fresh perspective

A depth of color brought to light by a fresh perspective

What CRI Doesn't Tell You

The Color Rendering Index was standardized in 1965. It averages a luminaire's performance across eight pastel reference samples and reports one number. Nearly every datasheet in the industry still leads with it.

The trouble is what averaging hides. CRI's eight samples are all unsaturated, so a source can render deep reds poorly and barely move the score. Newer standards help — IES TM-30 uses 99 samples across sixteen hue bins — but its headline figures are averages too. A weakness confined to one hue bin disappears into the mean. Specifiers are left reading a constellation of numbers off a page and reconciling them by instinct.

What Cauda Measures

Cauda is our Color Performance Index: eight independent color terms, normalized and combined into a single score from 0 to 100, with every component still readable underneath.

The eight are drawn from CIE 13.3, TM-30–20 and ANSI C78.377 — general rendering, saturated red and blue, fidelity, gamut, hue-bin-1 chroma shift, Duv, and SDCM binning tolerance. The heaviest single weight sits on hue bin 1, saturated reds, because reds govern how skin, food, wood, leather and warm finishes read in the 2200–4500 K range that architectural interiors actually live in. Ra — the number most datasheets lead with — is capped at 8% of the score.

Why the Math Multiplies

Cauda uses a weighted geometric mean rather than an average. The distinction matters. Under an average, a luminaire scoring 100 on seven terms and 40 on one still looks strong. Under a geometric mean it does not, because color failures are multiplicative in practice: the room looks wrong wherever the weak term applies, and the strong terms do not rescue it. Several failure modes — chromaticity well off the locus, loose binning, collapsed saturated red — additionally trigger hard penalty multipliers that compound.

A Luminaire That Passed Everything and Still Fell Short

We measured a high-CRI architectural strip at 3000 K and 2700 K on a laboratory spectroradiometer. Ra above 96. R9 above 85. TM-30 fidelity of 95, gamut of 101. Duv on the locus, inside a 3-step bin. On paper, specification-grade by every conventional reading.

Cauda scored it 75.

The binding constraint was a 2% chroma compression in hue bin 1 — saturated reds rendering flatter than the reference illuminant. Fifteen other bins were neutral or slightly positive, which is precisely why the averages looked fine. The cause was the red phosphor: a broad continuum between 620 and 660 nm rather than the narrow line emission that preserves red reflectance contrast. In a hospitality room or a gallery, cherry wood reads marginally less warm and terracotta loses depth. Nobody files a complaint. But it is exactly the difference a specifier is paid to deliver.

Changing the red phosphor recipe would lift that fixture from 75 into the high 80s without touching CCT, Duv, binning or the fidelity averages. That specificity is the point of the metric.

Where It Goes From Here

Cauda v1.0 is calibrated for 2200–4500 K white light and is explicit about its limits. It says nothing about glare, flicker or circadian content — those are separate problems with separate metrics. The weights are defensible but not yet proven, and the scoring bands are first-draft, pending validation against a wider field set and expert specifier judgment. Profiles for daylight-equivalent and RGBW sources are in development.

We publish the whole formulation, penalties and worked examples included, because a score you cannot audit is not worth specifying against.

Full thesis published on Medium.

by

Matthew John

/