Circadian design without the hand-waving

CIE S 026, melanopic EDI and what a defensible human-centric specification looks like in 2026.

CIE S 026, melanopic EDI, và một bản spec circadian đủ chặt để bảo vệ bạn trước khách hàng và nhà thầu.

“Human-centric lighting” was a marketing term for most of the last decade. Since CIE S 026 gave the field a measurable quantity and WELL turned that quantity into a credit, it has become something you can specify, calculate and be held to.

The shift is simple to state: lux is no longer the only illuminance that matters.

Why the eye needed a second metric

The photopic curve V(λ), which defines the lumen, describes the cone response that governs vision. It peaks at 555 nm, in the green.

Intrinsically photosensitive retinal ganglion cells — ipRGCs, discovered in the early 2000s — do not contribute to image formation. They contain melanopsin, peak around 490 nm in the blue, and they drive circadian entrainment, melatonin suppression, alertness and pupil size.

So the same 500 lux can be strongly alerting or nearly circadian-neutral, depending entirely on spectrum. The lumen cannot see that difference. That is the gap CIE S 026 fills.

The lumen is defined by V(λ), which peaks at 555 nm. Melanopsin — the pigment driving circadian response — peaks near 490 nm. The gap between the two curves is the whole reason a second metric was needed: the same lux can be strongly alerting or nearly neutral. Curve shapes are schematic.

Melanopic EDI, and the arithmetic behind it

Melanopic Equivalent Daylight Illuminance (mEDI) is illuminance weighted by the melanopsin response instead of V(λ), expressed in the equivalent lux of D65 daylight.

The practical shortcut:

mEDI ≈ vertical photopic lux at the eye × M-DER

M-DER (melanopic daylight efficacy ratio) is a property of the spectrum alone. Typical values:

Source	M-DER (approx.)
2200 K LED	0.35
2700 K LED	0.45
3000 K LED	0.52
3500 K LED	0.65
4000 K LED	0.75
5000 K LED	0.90
D65 daylight	1.00

The same ramp runs through the whole site as the gradient bar. M-DER is the melanopic ratio: multiply vertical lux at the eye by it to get melanopic EDI. It roughly triples across the range — spectrum, not level, is the biggest lever in circadian design. Plotted on a log axis so the warm end stays readable.

Two things follow immediately, and both matter more than any product selection:

It is vertical, at the eye. Roughly 1.2 m above floor for a seated occupant, facing the normal direction of view. Your 500 lx horizontal desk calculation tells you very little — vertical illuminance at the eye in a typical office is often only 30–50% of the horizontal figure.

Spectrum can double or halve your result. 300 lx vertical at 2700 K gives about 135 mEDI. The same 300 lx at 5000 K gives about 270 mEDI. Identical lux, identical energy, entirely different biological outcome.

Time 08:00

CCT 3500 K

Vertical lux 255

Melanopic EDI 160 below daytime target

Scrub through the day, or press play. Level and spectrum move together: the melanopic curve clears 250 lx only while daylight is contributing, and the evening drop is achieved by warming the source, not just dimming it. Schedule is an illustrative design intent, not measured data.

The targets people actually cite

From the 2022 international consensus recommendations, now widely referenced in WELL and in practice:

Period	Target mEDI (vertical, at eye)
Daytime	≥ 250 lx
Evening, 3 h before sleep	≤ 10 lx
Sleep environment	≤ 1 lx

The daytime figure is the demanding one. Hitting 250 mEDI from electric light alone at 3000 K requires roughly 480 lx vertical at the eye — which in a typical open plan means somewhere between 700 and 1000 lx horizontal. That is an energy and glare problem, and often a code problem.

Which is why the answer is usually not the luminaire

Three levers, in order of effectiveness:

1. Daylight. Overcast daylight through a window delivers thousands of mEDI. No electric scheme competes on cost or quality. Seat plan, glazing, blind control strategy and internal partition heights determine circadian performance in a daylit building far more than the fittings do. Push these decisions into concept stage, where you can still influence them.

2. Vertical surfaces. Since mEDI is vertical at the eye, lighting the walls is disproportionately effective. A wall-washed room delivers markedly higher mEDI than a downlit one at the same horizontal lux, and it lowers UGR at the same time. This is the highest-leverage move available to a lighting designer working inside a fixed architecture.

3. Then tunable white. Genuinely useful — but check two things before you specify it:

Constant output across the tuning range. Many tunable fittings lose 20–35% of lumens at the extremes. If your lux calculation used the 4000 K figure and the space runs at 2700 K in the evening, you are below target and nobody noticed.
Who commissions it, and who maintains it. A tunable installation with no owner is a fixed installation at whatever CCT it was left on. Write the schedule into the O&M manual and name the responsible party.

Evening is easier, and more neglected

The ≤10 lx evening target is far easier to hit than the daytime one, and often more valuable — especially in residential, hospitality and healthcare, where late-evening light exposure is directly implicated in sleep quality.

Practical moves: warm CCT after dark (2200–2700 K), low mounting heights, indirect and surface-mounted sources, and a genuine curfew scene rather than a global dim. A 5% dim of a 4000 K scheme still has high melanopic content per lux; a switch to 2200 K at 20% does not.

What a defensible specification contains

Design daytime target: ≥ 250 lx melanopic EDI (CIE S 026) measured vertically at 1.2 m in the seated direction of view, at all regularly occupied workstations, between 09:00 and 15:00, with blinds in their typical operating position. Evening scene (after 19:00): ≤ 10 lx melanopic EDI, CCT ≤ 2700 K. Contractor to submit α-opic toolbox calculations per CIE S 026 for each luminaire type.

Note what that pins down: the metric, the plane, the height, the direction, the location, the time, and the blind position. Every one of those is a place a circadian claim usually goes vague, and every one of them changes the answer.

Two honest caveats

CS and mEDI disagree. Circadian Stimulus, from the Lighting Research Center, is built on a melatonin-suppression model and will give you a different answer from melanopic EDI for the same installation. Neither is wrong; they model different things. State which one you designed to, and do not mix them within a project.

The evidence base is still moving. The direction of travel — bright and cool by day, dim and warm by night — is well supported. Precise dose-response thresholds for specific populations are not settled. Design robustly, avoid promising health outcomes in writing, and be careful about what appears in the client-facing report.

The one-minute check

Stand at the workstation. Hold the meter at eye height, facing the direction the occupant faces. Read the lux. Multiply by the M-DER from the table above. If you are below 250, more downlight will not fix it — you need daylight, brighter vertical surfaces, or a cooler spectrum.