Every supplier now claims human-centric lighting. The M/P ratio is the single parameter that separates a calibrated circadian system from a rebrand. A specifier's guide to what it measures and how to require it.
An HCL specification fails in one of two ways: it names a number without knowing who set it, or it names the wrong standard as the authority for the number. This guide fixes both. The sections that follow trace every threshold we publish to its actual source, give six clauses you can copy into a tender, state the basis for the four secondary parameters, and list honestly what our own data cannot yet support.
| Key fact | Source |
|---|---|
| WELL v2 Feature L03 daytime threshold is at least 150 EML [136 melanopic EDI] at eye level for at least 4 hours. | WELL v2 Feature L03 |
| Higher L03 tier: at least 275 EML [250 melanopic EDI]. | WELL v2 Feature L03 |
| CIE S 026:2018 defines the α-opic metric system. It is a measurement standard, not a threshold source — it sets no limits. | CIE S 026:2018 scope |
| EML and melanopic EDI differ by a fixed factor: 1 EML = 1/1.103 m-EDI, so 150 EML = 136 m-EDI and 275 EML = 250 m-EDI (the pairs as published by WELL; the raw conversion gives 249.3). | Unit conversion (D65 reference spectrum) |
| Flicker is limited by regulation: PstLM ≤ 1.0 and SVM ≤ 0.4 under EU 2019/2020. Our own design target is tighter: SVM < 0.15. | Commission Regulation (EU) 2019/2020; SUMLINK internal target |
| Catalogue colour consistency is SDCM 3-step as the working baseline with CRI 90 (2-step and CRI 97+ are project-specific tightenings, not the baseline). | SDCM: `data/dsh_migrated/light_engine.csv` (`sdcm` column — LE013 = 3, LE015 = 3, LE014 = 2/3 at 2700–4000K). CRI: `bom.csv` (`cri` = 90) and `light_engine.csv` (90/93/95/97 options) |
Metrics and units — the six terms that get confused
| Term | Symbol / unit | What it is | Set by |
|---|---|---|---|
| Photopic illuminance | lux | Light weighted to cone (visual) sensitivity | CIE photopic V(λ) |
| Melanopic EDI | m-EDI, lux | Melanopic irradiance expressed as the equivalent D65 illuminance at the eye, per CIE S 026:2018 | CIE S 026:2018 |
| Equivalent melanopic lux | EML | Same quantity on a different reference; differs from m-EDI by the factor 1.103 | WELL / industry convention |
| M/P ratio | dimensionless | Melanopic illuminance ÷ photopic illuminance — a ratio, so it is illuminance-independent | Industry convention; used in our design targets |
| Melanopic DER | dimensionless | Melanopic daylight efficacy ratio — the equivalent ratio referenced to D65 daylight | CIE S 026:2018 |
| Circadian stimulus | CS | A separate model (the Rea/Westland CS index), not an α-opic quantity — not used in this guide | Rea & Figueiro CS model |
The distinction that matters commercially: m-EDI and EML are absolute quantities measured at the eye; M/P is a ratio. A luminaire with an M/P of 0.45 at 250 lx and the same luminaire at 500 lx have identical M/P and a m-EDI differing by a factor of two. Any specification that requires an M/P ratio without also requiring an illuminance has specified nothing about the biological dose.
This is the section the rest of the lighting industry does not publish, because publishing it exposes how loose most specification language is.
| Number in circulation | Who actually set it | What it means | Checkable path | Status |
|---|---|---|---|---|
| 150 EML [136 m-EDI], ≥4 h, at eye level | WELL v2 Feature L03 (circadian lighting design) | Daytime threshold for an occupied space | WELL v2 L03 text, captured for this project | 來源核對 |
| 275 EML [250 m-EDI] | WELL v2 Feature L03, higher tier | Enhanced daytime threshold | Same source | 來源核對 |
| ≤10 m-EDI evening, ≥3 h before bed | Brown et al. 2022, PLoS Biology 20(3):e3001571 | Pre-sleep melanopic exposure limit | Peer-reviewed paper (doi:10.1371/journal.pbio.3001571) | 來源核對 |
| M/P ≥ 0.9 morning; 0.5–0.8 daytime; <0.4 evening | SUMLINK programme design targets — no standard sets these | A procurement proxy that is easy to write into a schedule | Not a standards citation; treat as our design intent | 內部設計目標 |
| PstLM ≤ 1.0, SVM ≤ 0.4 | Commission Regulation (EU) 2019/2020 | Regulatory limits for temporal light modulation, at full load | EU regulation; single lighting regulation | 來源核對 |
| SVM < 0.15 | SUMLINK internal design target | A deliberately tighter engineering target, not a legal limit | Internal | 內部設計目標 |
| SDCM 3-step | SUMLINK catalogue baseline | Colour consistency between adjacent fixtures | `data/dsh_migrated/light_engine.csv` — `sdcm` column: LE013 = 3, LE015 = 3, LE014 = 2/3 at 2700–4000K. Note: `bom.csv` has no `sdcm` field, and lists SDCM as an open gap in `hcl_gap` for several rows | 已生產驗證 |
| SDCM ≤ 2-step | Project-specific tightening | Optional, geometry-dependent | Project specification | 選擇性要求 |
| CRI ≥ 90 | SUMLINK catalogue standard | Colour rendering baseline; 97+ available as an option | `bom.csv` (`cri` = 90); `light_engine.csv` lists 90/93/95/97 options | 已生產驗證 |
| Δu'v' ≤ 0.006 | ANSI C78.377 Duv tolerance | Chromaticity tolerance for a given nominal CCT at initial measurement | ANSI C78.377 | 來源核對 |
| Δu'v' ≤ 0.007 over 6,000 h | ENERGY STAR | Chromaticity-maintenance limit — a different quantity from the above | ENERGY STAR lamp programme requirements | 來源核對 |
| Illuminance 500 lx, UGR ≤ 19, Uo ≥ 0.60 (offices) | EN 12464-1:2021 | The actual workplace-lighting requirement | EN 12464-1 | 來源核對 |
| — | CIE S 026:2018 | Sets no threshold. It defines how to calculate and report α-opic quantities | CIE S 026:2018 scope | 來源核對 |
Read the fourth row of that table again, because it is the most consequential. The three-band M/P schedule that appears in SUMLINK content is our design intent, not a citation. It is defensible and useful — it converts a photobiological threshold into something a commissioning engineer can check at the panel — but it is not a standard, and any specifier who writes "per CIE S 026:2018, M/P ≥ 0.9 in the morning" has attributed a requirement to a document that contains no requirements. That sentence appears in the published version of this article, and it is wrong.
The same discipline applies to the metric standard itself. CIE S 026:2018 tells you how to compute melanopic EDI from a spectral power distribution and sets the reference conditions (D65 reference spectrum, α-opic weighting functions, the five α-opic quantities). It does not say how much is enough. The "how much" comes from WELL, from the peer-reviewed literature, or from your own design targets — and each of those must be named separately.
Every number below is calculated, not observed. The calculation chain is short, and knowing it is what lets a specifier interrogate a datasheet.
| Step | What happens | What to declare |
|---|---|---|
| 1. Measure the SPD | A calibrated spectroradiometer records radiant power against wavelength across the visible range at a defined geometry | Instrument, its calibration uncertainty, the measurement plane and the drive condition |
| 2. Apply the weighting functions | The melanopic weighting function is derived from the melanopsin photopigment (peak sensitivity near 480 nm) and normalised in CIE S 026:2018; the five α-opic functions share the same construction | Which α-opic quantity is being reported — melanopic, in an HCL spec |
| 3. Compute the α-opic quantities | The weighted irradiance is converted into an equivalent illuminance; for melanopic, that is m-EDI | The reference spectrum used (D65) and the conversion convention |
| 4. Report the ratio | M/P = m-EDI ÷ photopic lux | Both terms, never the ratio alone |
Four reference conditions decide whether two M/P figures are comparable, and datasheets routinely omit all four:
On uncertainty: because these quantities are computed from a spectral measurement, their uncertainty is inherited from the instrument — calibration, stray light and spectral bandwidth all propagate. Anyone who reports an M/P to three decimal places without an uncertainty statement is over-stating the precision of the measurement. For procurement purposes, treat the figures as indicative bands, and require the last digit to be interpreted with care.
Source: CIE S 026:2018 for the metric definitions and reference conditions; the instrument and uncertainty discussion follows standard spectroradiometric practice. We hold no in-house SPD measurement record for BY001.
Most factory datasheets do not state an M/P ratio, and not always through evasion: plenty of manufacturers have simply never commissioned a spectral measurement of their own product. The following request resolves the question in one round trip.
Please provide, for each luminaire model and CCT step: (1) the full spectral power distribution at the specified drive condition; (2) the calculated melanopic EDI and M/P ratio per CIE S 026:2018, with the measurement plane, the CCT and the dimming setpoint declared; (3) whether each value is factory self-reported, measured on production samples, or independently measured by a third-party laboratory — and if independent, the laboratory and report reference; (4) for tunable-white models, Δu'v' at each dimming setpoint across the tuning range, not endpoint values only; (5) PstLM and SVM at full load and at representative dimmed setpoints.
Note what that request does and does not ask for. It does not ask whether the product is "verified" by the seller's own brand — a label a buyer cannot audit. It asks for the provenance of the measurement, which is checkable: a laboratory name, a report reference, and a stated setpoint. That is the difference between a specification and a slogan.
Two things a specifier should expect to find. First, suppliers whose products genuinely carry a distinct spectral design can usually produce an SPD curve, because they had to commission one to design the product. Second, a supplier who cannot produce an SPD at all has no basis for any melanopic claim, and the honest response to the request above is "we do not have that data" — which is a legitimate answer, and much more useful to you than a plausible-looking number.
Basis: SUMLINK procurement practice (internal). The request template asserts no external figure; the source requirements it names are CIE S 026:2018 (calculation basis) and the provenance rule stated in clause 6 of the request template.
Each clause below states what to require, and — in the last column — the basis that lets you defend it if a supplier pushes back.
| # | Clause | Text to use | Basis |
|---|---|---|---|
| 1 | Spectral and melanopic reporting | "The supplier shall provide the full spectral power distribution and the melanopic EDI and M/P ratio calculated per CIE S 026:2018, at each available CCT step and at each declared dimming setpoint, with the measurement plane and drive condition stated." | CIE S 026:2018 (calculation); WELL v2 L03 (if a WELL project) |
| 2 | Illuminance, uniformity and glare | "General office lighting shall achieve 500 lx maintained illuminance on the work plane, uniformity Uo ≥ 0.60, and UGR ≤ 19." | EN 12464-1:2021 |
| 3 | Colour consistency and CCT tolerance | "Colour consistency between adjacent luminaires of the same type shall be within SDCM ≤ 3-step; where the layout makes colour variation conspicuous, ≤ 2-step shall apply. Measured CCT shall be within the ANSI C78.377 tolerance for the nominal CCT." | SUMLINK catalogue standard (3-step); ANSI C78.377 |
| 4 | Chromaticity stability across dimming | "For tunable-white products, Δu'v' shall be declared at each dimming setpoint across the full tuning range. Chromaticity drift from the nominal white point shall not exceed 0.006 in u'v' coordinates at initial measurement." | ANSI C78.377 Duv tolerance |
| 5 | Temporal light modulation | "At full load and at the 25% and 10% dimming setpoints: PstLM ≤ 1.0 and SVM ≤ 0.4. The design target is SVM < 0.15." | Commission Regulation (EU) 2019/2020 (limits); SUMLINK target (stricter) |
| 6 | Evidence and provenance | "The supplier shall state, for each declared value, whether it is factory self-reported, measured on production samples, or independently measured; independently measured values shall cite the laboratory and report reference." | This guide; procurement auditability |
Clause 5 is worth a note. The regulatory limits are the weakly flickering test that a product must pass to be placed on the EU market — a floor, not a quality standard. A visible flicker problem in a dimmed office usually appears at low dimming setpoints, which is exactly where most datasheets stop reporting. That is why the clause names 25% and 10% explicitly.
Clause 3 is the other common trap. The published version of this article required SDCM ≤ 2-step as a baseline. Two-step consistency is a real and achievable specification, but it is a tightening that a supplier may reasonably price above the catalogue baseline, and applying it indiscriminately to every fixture on a floor plate buys colour consistency nobody can see. Our catalogue standard is 3-step; ask for 2-step where the geometry makes variation conspicuous (a continuous run of fixtures, a wall-wash, a cove) and accept 3-step elsewhere.
A specification that is only checked at the design stage is a specification that gets substituted. Spread the checks across the project.
| Stage | What to verify | Method |
|---|---|---|
| Design | SPD, m-EDI and M/P per CIE S 026:2018, Δu'v' across the tuning range, PstLM and SVM at 100%, 25% and 10% | Paper review — with the provenance stated for each value |
| Delivery | Model numbers match the specification, and certification claims are model-specific rather than family-wide | Compare against the quoted driver model, not the brand |
| Commissioning | Illuminance at the work plane, uniformity, glare, and the CCT at each schedule node | Lux meter plus the panel readout; spot-check at a real desk |
| Post-occupancy | Whether the evening node still matches the daylight window after a season changes | Re-check the evening node against local sunset at each changeover |
Basis: SUMLINK QA procedure (internal).
The primary parameter, and the one most often mis-reported. Require it per CCT step and per dimming setpoint, calculated per CIE S 026:2018. A single headline M/P is not a specification; our own design-target curve spans 0.45 at 2700K to 1.06 at 6500K, and a buyer comparing two products needs the same CCT steps from both.
These are two different measures of temporal light modulation and they are not interchangeable. PstLM quantifies perceptible flicker; SVM (stroboscopic visibility measure, per CIE TN 006) quantifies the stroboscopic effect — the banding or phantom-array artefact that appears when something moves in modulated light. SVM = 1 corresponds to the threshold at which an average observer begins to see the stroboscopic effect, so our design target of SVM < 0.15 sits deliberately well inside imperceptibility, and the regulatory limit of ≤ 0.4 sits at the edge of the legal floor rather than at a comfort standard.
The important operational point is the measurement setpoint. Regulation (EU) 2019/2020 tests at full load. Real offices rarely run at full load, and dimming is where poor driver design shows up. Require the figures at 100%, 25% and 10%.
SDCM (standard deviation of colour matching) expresses how far a luminaire's chromaticity sits from the nominal point in MacAdam-step terms. Industry practice treats 3-step as the level at which adjacent luminaires of the same type are not visibly different, and 5-step as noticeable. Our catalogue standard is 3-step; where a layout makes variation conspicuous — continuous runs, wall-washes, coves — a 2-step requirement is worth writing, and worth paying for. Elsewhere it is money spent on a difference nobody can see.
This is the parameter our published article conflated, and the confusion is widespread. Two distinct requirements travel under the same symbol:
| Requirement | What it constrains | Typical basis |
|---|---|---|
| Initial chromaticity tolerance | How far the measured white point may sit from the nominal CCT at first measurement | ANSI C78.377 Duv tolerance (±0.006 in u'v' terms) |
| Chromaticity maintenance | How far the white point may drift over rated life | ENERGY STAR-style limits (≤ 0.007 over 6,000 h) |
| Chromaticity stability across dimming | How far the white point moves as the product is dimmed | No single standard; must be declared per setpoint |
A tunable-white product can pass the first and fail the third: it holds its nominal white point at full output and shifts visibly at 20%. That is why the clause below asks for Δu'v' at each dimming setpoint, and why "0.006 u'v' units" alone, quoted without saying which of the three requirements it refers to, is not a specification.
Our catalogue engine-and-driver combinations span 1800–6500K at their widest (`bom.csv` row `H032-LE016-D007-PENDING-25W`; other rows are 2200–6500K and 2700–6500K), wider than the 2700–6500K band our schedules use. A wider range is not automatically better: the useful question is where the melanopic curve flattens. Below about 2700K the M/P ratio approaches our design-target floor of 0.45 and stops falling appreciably — which the final section discusses.
Take one luminaire family with our design-target M/P curve, and compare a tuned programme against the conventional static setting it replaces.
| Time | Programme CCT / lux | Programme m-EDI | Static 4000K at 420 lx | Difference |
|---|---|---|---|---|
| 08:00 | 5700K / 500 lx | 460 | 277 | +66% melanopic dose at the start of the day |
| 14:00 | 4000K / 420 lx | 277 | 277 | Identical — the conventional setting matches by accident at this hour |
| 18:30 | 2700K / 250 lx | 112 | 277 | −60% melanopic dose in the evening |
Programme values are SUMLINK design-target arithmetic: m-EDI = M/P ratio × illuminance, using the design-target curve in the table below. Static column: M/P 0.66 × 420 lx.
The table makes the commercial argument in one line. A static installation delivers exactly the same melanopic dose at 18:30 as at 08:00 — it has no dynamic range at all. The tuned programme delivers a 4:1 ratio between the start and the end of the day. What a client is buying when they buy HCL is that ratio, and it is measurable at the eye with a spectroradiometer.
It also shows the ceiling of what a warm setting can do. At 18:30 the programme is delivering 112 m-EDI, which is far above the ≤10 m-EDI evening guidance in the peer-reviewed literature. A 2700K source at 250 lx is a sensible end-of-day office setting; it is not a melatonin-neutral environment, and no schedule confined to a 2700K platform minimum can be. Our own design-target floor is M/P ≈ 0.45 at 2700K, so an evening target of "M/P < 0.4" cannot be met by spectrum alone — it requires reducing illuminance as well, or accepting a higher melanopic dose than the sleep literature would ideally want.
This section is the reason the article can be trusted with the rest of it.
What we can assert:
What we cannot assert:
The honest version of the argument in this article is unchanged, and it is stronger than the version with the badge: the M/P ratio is a computable, checkable quantity, most of the market does not report it, and the way to get it reported is to ask for the SPD and the provenance rather than for a label.
Basis: this section is an inventory of our own gaps, read from `bom.csv` (`hcl_status` = `HCL_PENDING`; `hcl_gap` lists M/P Ratio, m-EDI, SVM, PstLM, Flicker % and Δu'v'), `light_engine.csv` (the same fields blank) and `field_source.csv`.
They measure the same quantity on different references, and they differ by a fixed factor of 1.103. One EML equals 1/1.103 melanopic EDI, so 150 EML is 136 m-EDI and 275 EML is 250 m-EDI. WELL writes both because its requirements were originally published in EML while CIE S 026:2018 reports in melanopic EDI.
No. CIE S 026:2018 defines how to calculate and report alpha-opic quantities, including melanopic EDI, and sets the reference conditions for doing so. It sets no limits. Thresholds come from elsewhere: WELL v2 Feature L03 for workplace circadian lighting, peer-reviewed literature for pre-sleep exposure, or your own project targets.
Write the threshold you can defend and name its source. If the project is pursuing WELL certification, use the L03 daytime threshold of at least 150 EML (136 m-EDI) at eye level for at least four hours. Do not write 'per CIE S 026:2018, M/P 0.9 or above', because that standard contains no such requirement.
Not necessarily. Ask for the full SPD, the calculated melanopic EDI and M/P per CIE S 026:2018, and the provenance of the measurement: factory self-reported, production-sample measured, or independently measured. A supplier who answers that they do not have that data has given you a straight answer.
Require PstLM of 1.0 or below and SVM of 0.4 or below, the limits under Commission Regulation (EU) 2019/2020, and ask for them at full load and at the 25% and 10% dimming setpoints rather than at full load only. If flicker matters for the application, set an engineering target tighter than the regulation; our design target is SVM below 0.15.
No. It has been withdrawn from this guide. Until an independent measurement record exists for a given product, there is nothing behind the badge. What we report instead is the measurement status of each value: factory self-reported, production-sample measured, or independently measured.
Every luminaire in the catalogue carries its measurement status and its driver protocol in the product data. → Browse the catalog