Color temperature

Color temperature measures the “warmth” or “coolness” of a light source by analogy to the temperature of an idealised black body that would emit similar light. The unit is the Kelvin (K).

Reference points:

• 1700 K: match flame, candlelight. Very warm.
• 2700 K: incandescent “soft white” bulb. Yellow-orange.
• 3000 K: warm-white LED. Slightly less yellow.
• 4000 K: “neutral white” LED, office fluorescent. Reads slightly cool.
• 5500 K: noon daylight. The reference for most photographic white-balance.
• 6500 K: overcast daylight, “D65” — the colour-space white point for sRGB.
• 10,000 K: blue sky on a clear day, distinctly cool blue.

Counter-intuitively, lower Kelvin values look warmer (more red/orange) and higher values look cooler (more blue). This is because a low-temperature black body emits mostly long-wavelength (red) light; a high-temperature body shifts toward shorter wavelengths.

Practical use:

• Light bulbs. Pick 2700-3000 K for warm residential, 3500-4000 K for kitchens and offices, 5000+ K for tasks needing daylight-accurate colour discrimination.
• Photography and video. White balance corrects the colour cast of the ambient light. Setting WB to match the actual colour temperature produces neutral colour reproduction; deliberately mismatching creates a warm or cool tone.
• Display calibration. Most graphic design uses D65 (6500 K) as the white point. Some print-prepress workflows use D50 (5000 K) to match the standard viewing booth.

Worked example

You photograph a white sheet of paper under three light sources and the camera is on auto-white-balance. Under a 2700 K incandescent bulb the raw sensor reads roughly R:200 G:160 B:90 — heavily red/orange-shifted. Under 5500 K daylight: R:200 G:200 B:200 — neutral. Under 10,000 K open-shade north-light: R:140 G:180 B:230 — blue-shifted. White-balance correction multiplies each channel by a gain to push the result back to neutral white. For 2700 K source: gain B by ~2.2× and R by ~1.0×; for 10,000 K source: gain R by ~1.6× and B by ~0.87×. Mismatched WB (camera set to daylight 5500 K but photo taken under tungsten 2700 K) produces the orange-tinted “wedding photo by lamplight” cast that an entire generation of film photographers had to embrace.

The blue-light health discussion of recent years centres on colour temperature: high-CCT light (above ~5000 K) suppresses melatonin production more than warm light, which is why apps like f.lux and Night Shift mode shift display output toward 2700-3000 K in the evening. Whether the effect on sleep is clinically meaningful is debated, but the mechanism — short-wavelength light triggering circadian alertness signals — is well-established in chronobiology.

When and why it matters

Colour temperature matters whenever multiple light sources mix in a single scene — daylight through a window plus 3000 K LED ceiling fixtures plus a 2700 K desk lamp produces a colour cast that no single white-balance setting can correct. Cinematographers solve this by gelling lights to match (CTO/CTB gels shift colour temperature in defined increments). Office designers solve it by specifying a single CCT across all fixtures and avoiding adjustable-CCT LEDs near windows. Photographers solve it by shooting RAW and white-balancing in post. The everyday consumer mistake is buying a mix of 2700 K and 4000 K bulbs for the same room — the human eye adapts to one or the other but the mismatch reads as visually unsettling without anyone being able to name why. For display work, calibrating to D65 (6500 K) is the universal default; D50 (5000 K) is correct only if your output workflow ends at a print proofing booth lit to D50. Reference: US DOE — Lighting choices.

Correlated colour temperature (CCT) vs true black-body: the strict definition assumes the source is a thermal black-body emitter — which incandescent bulbs and sunlight approximately are, but LEDs and fluorescent tubes are not. LEDs emit narrow spectral peaks; the “3000 K” label is a correlated colour temperature (CCT), meaning “looks similar to a 3000 K black body to a typical observer” even though the spectrum is shaped differently. Two 3000 K LEDs from different manufacturers can look subtly different despite the matching label — the supplementary CRI (Colour Rendering Index, 0-100) score quantifies how accurately the source renders test colours; CRI ≥ 90 is excellent, ≥ 80 is acceptable, < 70 is visibly off for photography and design work. Reference: CIE 015:2018 — Colorimetry.