Color Consistency Testing: Why Two Identical Cree XP-G3 LED Modules Can Render Skin Tone Differently
I’ve stood beside dermatologists watching two “identical” exam lights—same fixture model, same spec sheet, same Cree XP-G3 emitter type—render the same patient’s forearm with visibly different warmth and saturation. One light made the skin look neutral and clean; the other introduced a faint but unmistakable yellow-green cast. No dimmer was involved. No aging. Just two modules, same bin code, same lot number, same mounting procedure.
This isn’t anecdote. It’s binning variance in action—and it’s especially dangerous in medical lighting, where subtle erythema or cyanosis detection hinges on faithful color rendering—not just high CRI numbers on paper.
The D2 Bin Isn’t a Single Point. It’s a Zone.
Cree’s standard “D2” chromaticity bin for XP-G3 (4000K nominal) covers a MacAdam ellipse of 7 steps—permissible by ANSI C78.377. That sounds tight until you map it: across real production lots I’ve tested, the actual CCT range within D2 spans 3850K to 4090K. That’s 240K—not trivial when evaluating vascular blush or jaundice.
More insidious is the R9 delta-E shift. R9 measures deep red fidelity—the critical channel for hemoglobin contrast. Within that same D2 grouping, I’ve measured R9 values from 89 to 96, with corresponding delta-E (CIEDE2000) shifts >4.5 against a reference D65 source. That exceeds the clinical threshold where observers reliably detect hue differences in skin tones.
Why the Datasheet Lies Quietly
The Cree XP-G3 datasheet lists “CRI ≥90, R9 ≥80” and “CCT ±200K”—but those are minimum guaranteed specs, not typical performance. They’re derived from statistical worst-case sampling, not lot-level characterization. And crucially: they don’t reflect spectral shape. Two modules can meet R9 ≥80 while having radically different 600–650nm radiant power profiles—one peaked, one flattened. That changes how melanin and oxyhemoglobin reflect.
I think this falls flat because lighting specifiers treat bin codes like ZIP codes: precise delivery points. They’re not. They’re postal regions—with rural outliers and urban density gradients.
A Guided Walk-Through: What to Request Before You Specify
Don’t wait for the fixture to arrive. Demand verification *before* finalizing the order:
- MacAdam ellipse plot per lot number—not just “within D2,” but the actual centroid and ellipse dimensions plotted over CIE 1931 xy. Ask for the 3-step and 5-step boundaries overlaid. If the supplier can’t provide this, walk away.
- Lot-specific spectral power distribution (SPD) reports—with resolution ≤2nm, covering 380–780nm. Pay attention to the 600–650nm hump. Compare R9, R12 (cyan), and R15 (blue-green) alongside CRI and TM-30 Rf/Rg.
- Delta-E (CIEDE2000) vs. D65 at 1000 lux—measured under controlled thermal conditions (Tj = 85°C, not 25°C). This reveals real-world drift.
One project I consulted on required 12 exam lights in a single procedural suite. We requested SPDs for three candidate lots. Lot A had excellent R9 but a suppressed 450nm peak—causing cool, slightly desaturated skin tones. Lot B delivered balanced SPDs but drifted +120K above nominal CCT at operating temperature. Lot C—slightly more expensive—hit all targets: R9 = 94.2, delta-E = 1.8, and near-zero CCT shift. The surgical team confirmed immediate improvement in lesion differentiation.
This works because color consistency isn’t about hitting a number—it’s about controlling variance where it matters most: in the wavelengths that define biological tissue.
Bottom line: “Identical” LEDs aren’t identical. In medical lighting, that difference isn’t academic—it’s diagnostic.