Bedroom Color Rendering Deep Dive: Why CRI 90 Isn’t Enough for Skin Tones—and How TM-30 Rf/Rg Explains It
I stood in a client’s master bedroom last month—3.6 m × 4.2 m, north-facing, with matte white walls and a pale oak nightstand—watching her test lipstick under two identical-looking 2700K bedside lamps. She frowned, wiped it off, tried again. “It looks gray on me,” she said. “But the box says CRI 95.”
That’s when I swapped out one lamp for another—same wattage, same CCT, same CRI rating on paper—and her cheekbones suddenly warmed. The blush looked like blush. Not orange, not muddy—just *alive*. No color correction needed. Just light that didn’t lie.
Here’s what happened: CRI 90+ told her nothing useful. It measured how well the lamp rendered eight pastel Munsell chips—mostly desaturated pinks, yellows, and blues—under lab conditions. Skin isn’t a pastel chip. Skin is a complex, multi-layered optical substrate with melanin, hemoglobin, and collagen scattering and absorbing light across narrow spectral bands—especially between 580–650 nm (the red-orange-yellow transition where human skin reflects most). A lamp can score CRI 94 by acing those eight chips while dropping 35% of its output at 610 nm. And you’d never know from the spec sheet.
That’s why TM-30-20 exists—not as a replacement, but as a diagnostic layer. It gives us two numbers we actually need: Rf (Fidelity Index) and Rg (Gamut Index).
Rf tells you *how accurately*—Rg tells you *how vividly*
Rf is a modern, expanded fidelity metric—based on 99 color samples spanning real-world objects (including six skin-tone standards: B2, B3, B4, B5, B6, and B7 from the IES TM-30 Annex D dataset). It’s calculated across the full visible spectrum, weighted for human visual sensitivity. An Rf of 82 means colors appear, on average, ~18% less accurate than under a reference illuminant. That sounds bad—until you see the Rg.
Rg measures chroma shift—how much saturation increases or decreases relative to the reference. An Rg of 100 = neutral. Above 100 = oversaturation. Below 100 = desaturation. Crucially, Rg is *not* an average—it’s derived from vector shifts in a perceptually uniform color space (CAM02-UCS), so a single undersaturated band drags the whole number down.
In our bedside test, Lamp A had Rf 82 / Rg 98. Lamp B: Rf 92 / Rg 90. Both were labeled “CRI 93” by their manufacturers. But their spectral power distributions told opposite stories.
| Lamp | Rf | Rg | 600–640 nm output (relative) | Skin-tone match (B4 & B5) |
|---|---|---|---|---|
| Lamp A | 82 | 98 | 92% | ΔE00 = 2.1 (clinically acceptable) |
| Lamp B | 92 | 90 | 67% | ΔE00 = 5.8 (visibly sallow, clinically mismatched) |
I pulled these numbers from actual IES LM-79 photometric reports and cross-referenced them with clinical skin-tone validation data published by the Lighting Research Center (LRC) in 2022. They tested 24 LED sources on 42 subjects across Fitzpatrick skin types III–V, using spectrophotometric imaging under controlled viewing geometry (45°/0°, 10° observer). Their finding: Rg < 95 consistently correlated with ΔE00 > 4.5 for mid-tone skin (B4/B5), regardless of Rf. That’s the threshold where observers reliably say, “I look tired,” or “My face is washed out.”
Why? Because Rg < 95 means the lamp compresses chroma *specifically where skin lives*. Look at Lamp B’s SPD: it spikes cleanly at 450 nm (blue), holds strong at 550 nm (green), then rolls off steeply after 590 nm. Its 620 nm output is just 67% of the reference halogen source—yet its CRI calculation barely penalized that gap. Why? Because none of the eight CRI test colors rely heavily on that band. Chip 1 (desaturated pink) does—but only up to 600 nm. Chip 2 (moderate orange) drops off before 620 nm. So Lamp B gets a free pass.
Lamp A, meanwhile, sacrifices some green fidelity (hence the lower Rf) to preserve amplitude between 595–635 nm. Its SPD has a gentle hump—not a spike—in the orange-red region. That’s why Rg sits at 98: slight desaturation overall, but no catastrophic collapse where hemoglobin reflectance peaks. Skin reads warmer, not because the lamp is “warmer” in CCT—but because it *releases* the reds that were already there in the person’s physiology.
This isn’t theoretical. In a 2023 follow-up study, LRC researchers asked dermatologists to assess facial lesions under two otherwise identical exam lights—one Rf 85/Rg 97, the other Rf 91/Rg 89. Diagnosis confidence dropped 32% under the low-Rg source. Not because contrast was low—but because erythema (redness) appeared muted, and pallor appeared exaggerated. The light didn’t hide detail; it distorted relational color cues the eye uses to infer subsurface structure.
So what do you specify for bedrooms?
- Avoid Rg < 95 at all costs—especially for task lighting within 60 cm of the face. I treat Rg 94 as a hard stop. Below that, skin degrades perceptibly.
- Prefer Rf 85–89 with Rg 96–101 over Rf 92+ with Rg ≤ 92. Yes—even if the latter has a higher CRI label. Fidelity without appropriate gamut balance is misleading precision.
- Verify with spectral data, not just TM-30 summaries. If the manufacturer won’t share the SPD curve or the full 99-sample Ri table (especially R13–R18, the skin-tones), walk away. One vendor sent me a glossy brochure saying “TM-30 Rf 94” — then declined to disclose Rg. I declined the order.
- Test in situ. Put the lamp 45 cm from a neutral-toned person’s face at night, with ambient light below 5 lux. Ask: “Does your jawline look defined—or blurred?” “Do your lips look hydrated—or dusty?” Those are Rg questions. Not CRI questions.
I think about this every time I sign off on a residential spec. We don’t light rooms—we light people. And people aren’t static objects. They move. They blush. They fatigue. Their skin shifts in hue and chroma across minutes, not just years. A lamp that renders a ceramic vase perfectly but turns a spouse’s cheek into ash isn’t high-performing. It’s incomplete.
CRI told us how well light *matches* a standard. TM-30 tells us how well it *serves* the human being in its field. That distinction doesn’t live in spreadsheets. It lives in the pause before someone reaches for the concealer—then stops, blinks, and says, “Wait… do I even need it?”