Definition and service boundary
Circadian lamp spectrum testing is a measurement service for the optical radiation emitted by a lamp, luminaire, or lighting system under agreed operating conditions. Its central task is to measure spectral power distribution and derive specified radiometric, photometric, colorimetric, and retinal-photoreceptor-weighted quantities. The test object may be a fixed-spectrum lamp, a tunable-white or multichannel luminaire, a programmed system, or a defined point in an installed environment. Each object calls for a different measurand, geometry, and scope of conclusion.
This is a testing and evaluation service, not lamp manufacturing, lighting-design consultancy, clinical research, medical diagnosis, or a light-therapy prescription. A laboratory report states what was measured for the submitted item in defined states. It cannot by itself prove better sleep, a shifted body clock, or treatment of a disorder. It also does not automatically cover photobiological safety, electrical safety, electromagnetic compatibility, lifetime, glare, temporal light modulation, or compliance of a building installation. Those questions require separately named methods, criteria, and responsible parties.
Terms such as circadian, daylight-like, or night-friendly are product positioning rather than measurements. Relevant metrics need a named system, unit, and calculation version. A laboratory may measure and explain technical results, but a report must not be presented as regulatory approval, a health certification, or proof of human outcomes unless a separately authorized certification or research process actually supports that claim.
Commissioning and sample control
A useful commission begins with the intended decision, not with a marketing metric. The client should state whether it needs total source output, directional emission, or the light likely to arrive at a bedside, work plane, or eye position. Total spectral radiant flux, spectral radiant intensity, spectral irradiance, and spectral radiance describe different conditions. An integrating-sphere result cannot be substituted for exposure at a specified distance and direction. A spatial assessment also needs a defined plane, position, height, orientation, field of view, obstruction, and ambient-light condition.
Sample records should include model, serial or batch number, optics, driver, controller, firmware or application version, rated settings, and condition on receipt. Tunable products need a mode matrix covering brightness, correlated colour temperature, channel mix, scenes, dimming protocol, and relevant transitions. Testing only the brightest or warmest setting does not characterize the full programme. When sensors, schedules, or cloud logic alter output, the method must define how the state is frozen or triggered and logged so that identically named scenes do not silently produce different spectra.
Before testing, the parties should agree on supply type, voltage and frequency, mounting orientation, warm-up or stabilization criterion, ambient temperature, airflow, and installed accessories. Sample count must match the claim. One unit can describe that unit but cannot establish production consistency. Batch or unit variability needs a sampling plan, sample size, and acceptance rule. Shipping damage, software updates, repairs, ageing, contamination, and substitution of a sample should remain traceable throughout initial and repeat testing.
Spectral measurement and technical method
Laboratories normally use a calibrated spectroradiometer with input optics suited to the measurand. An integrating sphere can provide total spectral radiant flux, a goniometric system can characterize angular distribution, and a defined detector position can provide spectral irradiance. Field work requires ambient-light control and a recorded spatial geometry; laboratory work requires control of stray light, temperature, supply, and thermal stabilization. Spectral range, wavelength increment, bandwidth, integration time, and averaging should support the requested metrics without detector saturation or inadequate signal.
A raw spectrum is not a context-free fingerprint. Array instruments can be affected by dark signal, wavelength error, non-linearity, finite bandwidth, and spectral stray light. Input optics add possible cosine-response and field-of-view errors. The calibration chain should suit the spectrum and signal level under test. The laboratory should verify or correct dark response, wavelength scale, responsivity, linearity, and stray light, while preserving the relationship between raw and corrected data. Narrow-band multichannel LEDs require particular attention when the calibration source and test source have substantially different spectra.
The confirmed spectrum may support illuminance, chromaticity, correlated colour temperature, agreed colour-rendering or colour-quality metrics, and the five alpha-opic quantities, equivalent daylight illuminances, and daylight efficacy ratios defined by CIE S 026. Every derived value should refer to the same confirmed spectrum, geometry, and operating state. Flicker or time-varying spectral modulation needs a separate method with adequate temporal resolution; a static or time-averaged spectrum is not a temporal light modulation test.
Circadian metrics and limits of evidence
CIE S 026 is a metrology system for the potential stimulation of the L-, M-, and S-cones, rods, and melanopsin pathway. Metrics such as melanopic equivalent daylight illuminance allow optical environments to be compared on a consistent basis, but they are not sleep scores or percentages of circadian repair. Correlated colour temperature is also only an aggregate chromaticity descriptor. Two spectra can have similar CCT and different photoreceptor-weighted results, while a blue peak or short-wavelength percentage alone does not represent the complete effective stimulus reaching the eye.
Human responses depend on spectrum, intensity, duration, timing, prior light exposure, field of view, gaze, pupil state, and individual characteristics. Total flux on a product bench or a single on-axis reading is not the dose received over time at a user's eyes. A procurement target should therefore specify a reproducible scene and measurand rather than rely on a chip type, CCT, or mode name as a proxy for effect. Evidence from healthy young adults also cannot automatically become an individual prescription for shift workers, children, older people, people with eye conditions, or those taking medication.
Testing can show that an optical specification was met under defined conditions; it cannot alone establish improved sleep, alertness, melatonin change, or clinical benefit. Human-effect claims require research matched to the finished product, exposure schedule, target population, and outcome. A human light-exposure study should fully report the stimulus, but it also needs its own ethics review, informed consent, risk controls, and statistical plan. A routine product test report does not replace those activities.
Quality control, uncertainty, and reproducibility
Credible results depend on traceable calibration, controlled methods, and uncertainty fit for the decision, not on the number of digits shown by an instrument. A laboratory should periodically check wavelength scale, spectral responsivity, dark signal, linearity, repeatability, and stray light, monitor drift with reference sources or control items, and retain calibration status. Integrating-sphere work must address self-absorption, coating condition, and sample size. Goniometric or point measurements must control distance, alignment, optical axis, field of view, and cosine response. Departures from the routine method belong in the report.
Replicates should be collected after stabilization and should distinguish instrument repeatability, remounting variation, repeatability of a mode, and variation between samples. A tunable lamp may drift with thermal equilibrium, driver cycles, or network control, so a brief acquisition may not describe the operating interval. Stabilization curves, time series, and multiple positions may be needed. Retest tolerances should be agreed in advance rather than selecting favourable modes, positions, or averaging rules after the results are known.
Reported uncertainty should identify its coverage, important contributors, and applicable measurands. If the contract calls for a pass or fail statement, the parties should first define the specification limit, how measurement uncertainty enters the decision rule, and how borderline results are handled. Laboratory competence or accreditation should be checked for the actual method, quantity, object, and range. A management-system certificate, accreditation mark, or proficiency-test result does not automatically place every circadian spectrum test inside an accredited scope.
Data, safety, and compliance
A reviewable record preserves the original spectrum, wavelength grid, units, measurement geometry, instrument and calibration identifiers, environmental and supply conditions, sample state, mode settings, correction rules, formula or software version, and anomalies. Derived tables should not overwrite raw data. Manual removal, smoothing, interpolation, or normalization should be traceable. Multi-batch and firmware-upgrade projects need stable naming and version control so that a later reviewer can identify the exact unit and measurement run behind each report.
Routine lamp testing usually requires no personal health data, but customer drawings, firmware, scene files, and unreleased performance may be confidential. The contract should address access, transfer, retention, backup, subcontractor disclosure, and disposal. If work expands into an occupied site or a human study, images, locations, schedules, and participant records may become personal data and need separate handling under applicable law and the study protocol. An ordinary sample identifier does not by itself resolve privacy obligations.
Laboratory controls should also cover electrical, thermal, mechanical, and optical-radiation hazards, especially at short distances, high output, with diffusers removed, or with exposed drivers. Circadian calculations and photobiological safety assessments may both start with spectral data, but they use different purposes, weighting functions, geometries, and exposure conditions. A low melanopic value, warm appearance, or visible-only description cannot demonstrate absence of photobiological risk without the relevant safety method. Conversely, a safety result does not prove suitability for every time, setting, or user.
Deliverables, procurement, and use of results
A complete report identifies the client, sample, date, method and edition, deviations, equipment and calibration status, environment and supply, mounting geometry, stabilization, exportable spectra or raw data for each mode, derived quantities, uncertainty, and authorized signatory. Plots should retain axes, units, and spectral-resolution information. Results should say whether they apply to one unit, a sample statistic, or a field location. Any statement of conformity needs a named specification and decision rule; an undefined claim that a lamp is circadian-compliant is not sufficient.
When procuring the service, compare the laboratory's relevant accredited scope, method experience, instrument suitability, reportable uncertainty, sample capacity, mode matrix, field capability, raw-data format, retest policy, subcontracting, confidentiality, and sample disposition. Price should be tied to the number of units, modes, locations, and add-on tests. A low-cost peak screenshot is not equivalent to a conditioned, traceable result with uncertainty, while a high price or authoritative-sounding name does not raise the evidence level by itself.
Use of the result remains limited to the tested sample, version, and conditions. Research and development teams can compare channel mixes and control algorithms, quality teams can monitor batches and drift, designers can use data as input to scene calculations, and researchers can document a stimulus. Each downstream use should preserve the original boundaries. A change in specification, unit, firmware, optics, or installation should trigger a review of whether retesting is needed. One test must not be extended into certification of every batch or a promise of sleep or health benefit.