Light Sources: How LED Changed Colour Work

A lighting revolution that solved an energy problem and created a colour problem

Between roughly 2010 and 2020, most of the developed world replaced its lighting. Incandescent lamps were phased out by regulation, compact fluorescent lamps had a brief unloved reign, and light-emitting diodes took over almost everything: homes, offices, shops, galleries, street lighting, car headlamps, torches and camera-adjacent video panels. The energy case was overwhelming and the transition was, by the standards of infrastructure change, extraordinarily fast.

For anyone whose work depends on how things look, though, the transition was disruptive in a way that rarely got discussed outside trade publications. Incandescent light had one enormous virtue: it was a continuous thermal spectrum, essentially the same physics as the sun, just cooler. It rendered colour almost perfectly because it contained every wavelength in a smooth curve. LEDs do not work that way at all, and the consequences ripple through retail, galleries, textiles, dentistry, printing and anywhere a colour judgement is made under artificial light.

How white LEDs actually make white

Almost no white LED emits white directly. The dominant approach is a blue LED chip — typically emitting around 450 nanometres — coated with a phosphor, a material that absorbs some of that blue light and re-emits it at longer wavelengths as a broad yellowish band. The blue that leaks through plus the yellow that comes back combines to look white to the eye.

The eye is easily satisfied here because human colour vision compresses the whole spectrum into three signals. Two lights can look identical while containing completely different mixes of wavelengths. That is fine until you put an object under the light. Objects reflect selectively, and if the light contains very little energy at the wavelengths an object reflects most strongly, that object will look dull, dark or simply wrong.

The characteristic gaps

  • The cyan dip. Between the blue peak and where the phosphor emission ramps up, there is usually a trough around 480 to 500 nanometres. Turquoises, teals and some greens can look flat or greyish under cheaper LEDs.
  • The deep red shortfall. Standard phosphors fall away above roughly 630 nanometres. Deep reds, burgundies, skin undertones and the red end of wood finishes lose vibrancy. This is the single most noticed failure in retail and in galleries.
  • The blue spike. The narrow pump peak is disproportionately strong. Optical brighteners in paper and textiles, which absorb ultraviolet and near-blue and fluoresce, behave differently under LEDs than under daylight or fluorescent light, so whites can look inconsistent.

The metrics, and why one of them misled a whole industry

The traditional measure of how faithfully a lamp renders colour is the colour rendering index or CRI, published in 1965 and revised in 1974. It works by illuminating a set of test colour samples with the lamp and with a reference source of the same correlated colour temperature, then measuring how much each sample shifts. The general index, Ra, averages the first eight samples.

Those eight samples are all moderately saturated pastels. None of them is a strong red. This mattered little in the era of incandescent and fluorescent lighting, but LED manufacturers, quite rationally, engineered towards the number that was specified. A lamp could score Ra 82 and still render a saturated red poorly, because no saturated red was being tested. Buyers who specified a CRI figure believed they had bought colour quality and had in fact bought a score.

Metrics worth asking about

  • R9. The ninth test sample, a strong red, is not included in Ra but is reported separately by serious manufacturers. A high Ra with an R9 near zero is the classic warning sign. Ask for R9 above 50, ideally above 90 for critical work.
  • TM-30. A newer method using 99 colour evaluation samples spread across hue and saturation. It reports Rf, a fidelity score, and Rg, a gamut score indicating whether the lamp saturates or desaturates colours on average. Two lamps with equal fidelity can differ markedly in whether colours look punchy or washed.
  • Correlated colour temperature. Reported in kelvin, it describes whether the white is warm or cool. It says nothing about rendering quality. A 2700 K lamp and a 4000 K lamp can both be excellent or both be poor.
  • Duv. A measure of how far the light sits from the true black-body locus, usually towards green or towards pink. A lamp with a correct kelvin figure but a green cast is a common and deeply irritating fault, and Duv is the number that catches it.

Consequences in real workplaces

Colour matching under LED is where these abstractions become expensive. A textile buyer approving a dye lot, a printer approving a proof and a paint retailer mixing a match all rely on judging two samples under a defined light. If the light is not what the specification assumes, the judgement is not portable.

Standardised viewing booths solved this decades ago by using filtered fluorescent tubes engineered to approximate the standard daylight illuminant D50 or D65. LED viewing booths now exist and the good ones are genuinely excellent, but the transition period produced a lot of mismatched judgements as facilities upgraded piecemeal. Two sites in the same company can disagree about a colour simply because one replaced its booth lamps and the other did not.

The metamerism problem — where two samples match under one light and diverge under another — became more pronounced too, because LED spectra are so spiky and so varied between manufacturers. A pair that matched in the studio can visibly split on a shop floor lit by a different LED product, and both parties can be measuring correctly.

Practical mitigations

  • Specify the illuminant in writing. A colour approval that does not name the viewing light is not an approval, it is a preference expressed on a particular afternoon.
  • Buy lamps by spectrum, not by marketing. Request the spectral power distribution, R9 and TM-30 data. Reputable manufacturers publish them; the absence of data is itself information.
  • Keep batches together. LED output varies between production runs, a phenomenon managed by binning — sorting chips into tolerance groups. Mixing bins in one room produces visible differences between fittings, especially on a white wall.
  • Watch the dimming path. Incandescent lamps warm as they dim, which people find flattering and expect. Many LEDs hold colour temperature constant while dimming, which reads as cold and clinical. Warm-dim products exist and cost more.

What LEDs made possible that was not possible before

It would be unbalanced to treat this as a story of loss. LEDs enabled things that thermal sources never could. They switch instantly and can be modulated at high frequency, which underpins everything from camera-synchronised studio lighting to sensing systems. They emit little infrared, which is why museums can now light a fragile watercolour closely without cooking it, a real conservation gain.

They can also be tuned. Multi-channel fixtures with four, five or six emitters — adding amber, lime, deep red or violet to the usual mix — can fill the spectral gaps deliberately and can be adjusted for the specific object being lit. Galleries increasingly use tunable fixtures to bring out a particular pigment. Retailers use them to make produce and meat look fresh, which is either good merchandising or mild deception depending on where you draw the line.

Where tunability is genuinely useful

  • Conservation lighting. Low ultraviolet and low infrared output means less photochemical and thermal damage per unit of visible illumination.
  • Circadian schemes. Shifting colour temperature across the day in hospitals, care homes and offices, based on the effect of short-wavelength light on melatonin. The evidence is promising rather than settled, and installations often overclaim.
  • Studio and horticultural work. Precise spectral control lets photographers match ambient light exactly and lets growers target the absorption peaks of chlorophyll.

Flicker, drive electronics and the things a spectrum does not show

Spectral quality is not the only way an LED installation can go wrong. An LED is a semiconductor driven by electronics, and the electronics introduce failure modes that thermal sources simply did not have. The most common is flicker: because an LED responds almost instantaneously to changes in current, any ripple in the driver output appears directly in the light output. An incandescent filament has thermal inertia and smooths this out; an LED does not.

Cheap drivers, and many dimmers retrofitted from the incandescent era, produce substantial modulation. Most people do not consciously see it, but it is implicated in headaches and eye strain, and it is unmistakable on camera as banding or rolling bars in video. Anyone photographing product, artwork or textiles under LED lighting will meet this problem before they meet any spectral one.

What to check besides the spectrum

  • Percent flicker and flicker index. Traditional measures of modulation depth. Lower is better; reputable manufacturers publish figures, and a driver that does not disclose them usually has a reason.
  • Dimmer compatibility. Trailing-edge dimmers generally behave better with LED loads than leading-edge types designed for resistive filaments. Mismatches cause buzzing, flicker at low levels and premature driver failure.
  • Thermal management. LED output and colour both drift as the junction heats. A fixture with poor heat-sinking will shift colour over the first hour and lose output over its life faster than the rated figures suggest.
  • Lumen and colour maintenance. Published as L70 or L80 figures for output and as a maximum colour shift over rated life. Two lamps with the same initial spectrum can diverge visibly after a year of use.

How to think about lighting choices now

The honest summary is that LED has raised both the ceiling and the variance. The best LED lighting available today renders colour better than incandescent ever did, with continuous spectra achieved through multi-phosphor blends or violet-pump architectures, and it does so at a fraction of the energy cost. The worst LED lighting, which is abundant and cheap, is markedly worse than the incandescent lamp it replaced, and the packaging will not tell you which one you have bought.

For anyone whose work involves judging colour, the discipline that matters is documentation: know what the light in your space actually is, keep the spectral data for it, do not mix products in a room where colour decisions are made, and treat any approval made under unspecified light as provisional. The physics changed underneath the whole profession in about a decade, and the practices that assumed a smooth thermal spectrum need to be rebuilt on the assumption that they no longer have one.

Read next: What Color Is Apricot? Shades, Codes and How to Use It