1 P. Boulenguez, S. Carré, M. Perraudeau, C. MartinsonsRisque Photobiologique à la lumière bleue Comparaison de deux approches de mesure sur 15 lampes à LED et fluocompactes Hello, My name is Pierre Boulenguez and I will present a work on behalf of the Lighting Engineering team at CSTB, the French public Building Science Institution. The talk is entitled BLUE LIGHT HAZARD – Comparison of the Photobiological Risk Groups Of Fifteen Lamps Assessed using the Uniform Spectrum Assumption and a New Imaging Spectrophotometry Approach P. Boulenguez, S. Carré, M. Perraudeau, C. Martinsons RetinaLED

2 Résumé L’évaluation du Risque photobiologique à la lumière bleue (BLH) faisant l’hypothèse d’un spectre uniforme sous-estime le niveau de risque (principalement pour les LED). Une nouvelle approche basée sur l’imagerie spectrale permet de résoudre ce problème. As an overview, I will argue that a common procedure to assess the blue light hazard of a light source which relies on the assumption that the spectrum of the source is spatially and directionnaly uniform, leads to an underestimation of the blue light hazard. This underestimation is particularly significant for light emitting diodes. To address this problem, I will present a new approach based on an innovative imaging spectrophotometer.

3 Plan Le risque à la lumière bleue L’imagerie spectrale (méthode 2)Contexte Le problème des LED Méthode 1 : le spectre uniforme L’imagerie spectrale (méthode 2) Acquisition Traitements Étude comparative (Méthode 1 / Méthode 2) Lampes testées Résultats & Discussion In more depth, the outline of the talk is as follows. I will start by reminding ourselves some fundamentals aspects of the blue light-hazard, and notably its assessment using the « uniform spectrum » assumption. I will next present the principle of our innovative imaging spectrophotometer and its application to blue light hazard assessment. I will ultimately compare the two methods on the basis of the assessment of a range of sources representative of the French consumer market.

4 Contexte IEC 62471:2006 (Standard) Risque à la lumière bleueSécurité photobiologique Risque à la lumière bleue Ultraviolet Risque actinique Proche ultraviolet Infrarouge Effets thermiques sur la rétine The standard sixty two four seven one « Photobiological safety of lamps and lamp systems>> classifies five potential photobiological hazards of general lighting systems. Namely: the actinic ultraviolet, the near ultraviolet, the infrared radiation, the retinal thermal, and the blue light hazard: the subject of this talk. A “group of risk” can be determined for each of these hazards, labeled from High Risk to Exempt. Informally, the group of risk for a given source and a given hazard is dependent on the dose of radiation received, which in turn is dependent on the observation distance and on the exposure duration. IEC 62471:2006 (Standard)

5 Risque à la lumière bleue𝐵 𝜆 𝑉 𝜆 Focusing on blue light hazard, let us remind ourselves that the concern about exposure to blue and dark-blue radiations arose in the mid twentieth century from retinal burns on pilots performing frequent oversea flights. Experiments on monkeys in the seventies showed that this risk is particularly pronounced for radiations in the range four fourty to four sixty nanometers. An action spectrum was subsequently determined more precisely and is called B of lambda. It is plotted here alongside the luminous efficiency V of lambda. Note that the y-axis on the graph is logarithmic. The complete understanding of the mechanism behind blue light retinal injuries is still an active field of research in ophthalmology. What is known is that it is a relatively slow photochemical process, which involves the accumulation of toxic lipofuscin granules. Children and teens are particularly at risk, because the transmittance of the crystalline in the blue end of the spectrum is much higher at younger age. It evolves from approximately ninety percent for a newborn to only thirty percent at eighty years. Effets sur la rétine Dégénérescence maculaire liée à l’âge Risque particulier pour les enfants & adolescents

6 Limite d’exposition cône de vision

7 Sensibilité à la lumière bleueSensibilité spectrale au risque « bleu » Moyenne spatiale 𝐿 𝐵 𝑥 , 𝜔 = 𝜆 𝐿 𝜆 𝑥 , 𝜔 ,𝜆 𝐵 𝜆 d𝜆 10s: Luminance BLHW Position, Direction Luminance spectrique Sensibilité spectrale 1h: 20cm In the aforementioned sixty two four seven one standard, the classification of a source with respect to the blue light hazard is based on exposure limit values expressed in terms of spatially averaged blue light hazard weighted radiance. The blue light hazard weighted radiance, noted Lb, at a position x on a surface, in a direction omega of emission, is simply the spectral radiance weighted by the B of lambda action spectrum. This spectral weighting is formalized by the presented equation. Yet, this blue light hazard weighted radiance is not intended to be used as is, but should be averaged over finite solid angles, dependent on the exposure time. For instance, consider the following compact fluorescent lamp. Its height is twenty centimeter and the lamp is also observed at twenty centimeter. Then, the extent of the averaging surfaces for the blue light hazard weighted radiance for ten seconds, one hour, and eight hours exposures are schematized on the left. The rationale for this spatial averaging arise from the spreading of the image of a point source on the surface of the retina : by the finite pupil diameter, and by involuntary eye movements for prolonged exposures. 8h:

8 Les problèmes avec les LED« Pic » dans le bleu Variations spectrales 1h 8h 10s 5cm 𝐵 𝜆 LED (3000K) Blue light hazard recently has recently been under intense scrutiny because of its link with light emitting diodes and the emergence of these sources for general lighting system. Let’s consider the following “average” spectrum of a typical white-light LED. The peak in the blue end of the spectrum arises from the fact that most white-light LEDs are composed of a blue light emitting die, which excites a phosphorescent material embedded in a surrounding plastic dome. As consequences, the average emission spectrum display a peak in the blue end, but also, the emitted spectra depend on the optical thickness of phosphorescent material traversed, that is to say, the exact emission position on the surface of the source and the direction of emission. The spatial averaging of the blue light weighted radiance, again illustrated here on a three LEDs luminaire, can be much smalller than the total surface of a single LED. Therefore, the uniform spectrum assumption leads to severe inaccuracies in the blue light hazard assessment.

9 L’hypothèse d’un spectre uniformeHypothèse du spectre uniforme spectrophotomètre + Cartographie de luminance Traitement d’image Repérage du maximum sur une cartographie de luminance pondérée 𝐿 𝜆 𝑥 , 𝜔 ,𝜆 = 𝑈 𝜆 𝜆 𝑈 𝜆 𝑉 𝜆 d𝜆 ∙𝐿 𝑉 ( 𝑥 , 𝜔 ) Position, Direction Spectre uniforme (spectrophotomètre) Luminance (vidéo luminancemètre) To quantify the blue light hazard associated with a given source therefore implies to detect the maxima of the spatially averaged blue light hazard weighted radiance for various exposure times, that is to say, for various averaging surfaces. A practical approach to perform this assessment relies on the assumption that the spectrum of the source does not vary with the emission position and the emission direction. A spectrophotometer can be used to measure this spectrum assumed uniform. Let’s note this spectrum U of lambda. The scaling of this spectrum dependent on the emission position and direction can also be measured using an imaging luminancemeter. The spectral radiance can then be recovered using the following equation. In turn, a blue light hazard weighted radiance image of the source is obtained from this quantity. Maxima associated with various averaging surfaces can then be computed using simple image processing technique. In this study, we have combined a moving windows approach with mipmaps.

10 Comment éviter l’hypothèse d’un spectre uniforme ?

11 Acquisition imageante et spectraleCamera ILT CCD, réponse linéaire, 5°C 12 bits + HDR Filtre pilotable (LCTF) Balaye tout le spectre visible Bande passante étroite Objectif macro To answer to this question, I present an innovative imaging spectrophotomer. It hardware part consists of a composite detector made first, of a radiometry-grade camera. The camera is equipped with a cooled interline charge-couple device. Its initial dynamic of 12 bits is extended using a classical geometrical high dynamic range acquisition technique. This camera was fitted with a Liquid Crystal Tunable Filter: a filter of which the transmittance can be dynamically tuned, sweeping the spectrum with Gaussian like narrow bands. Finally, a narrow field of view optic was added for the Blue Light Hazard assessment application

12 Traitement des acquisitions spectralesNiveau de gris Sensibilité spectrale d’un pixel Longueur d’onde centrale du filtre Spectre à mesurer g 𝜇 = 𝜆 𝑟 𝜇,𝜆 𝑠 𝜆 d𝜆 Reconstruction spectrale So, a range of images, typically eighty, are acquired for different transmittance of the liquid crystal tunable filter, as illustrated here. To recover a spectrum for each pixel is an operation called spectral reconstruction. Consider the following equation. G of mu denotes is the of pixel’s values for different responsivities r of u lambda of the composite detector. S of lambda denotes the spectrum associated with the pixel, and is the unknown in the equation. It is worth noting that such form is known as a Fredholm’s integral equation of the first kind, a classical mathemetical ill posed problem. I will not discuss the regularization procedure in detail. Simply, an approach baptised narrow-band regularization, which relies on the special transmittance of the liquid crystal tunable filter, was used. For blue light hazard assessment, the spectral image obtained is weighted against B of lambda. Maxima associated with averaging surfaces, again using image processing techniques such mipmap & moving window, allows to qualify a source according to the sixty-two four seven one standard.

13 Étude comparative

14 Lampes testées Sources Hypothèse d’un spectre uniforme (Méthode 1)LEDs & fluocompactes Usage domestique Distributeurs grand public Hypothèse d’un spectre uniforme (Méthode 1) Imagerie spectrale (Méthode 2) To assess the two methods, we have selected a range of light emitting diodes and compact fluorescent lamps, more or less representative of the French consumer market. Each of these sources were assessed for blue light hazard using the spectrophotometer and imaging luminancemeter approach, that is to say, under the uniform spectrum assumption, and using the new imaging spectrophotometer approach. The results of this study can not be presented here to their full extent due to the lack of time, but can be found in the paper that accompanies this talk. As an excerpt, we have extracted two interesting figures.

15 Distance de risque (Source à 3 LEDs) Discontinuité FoV Method 2The first one concerns the hazard distance, or the distance from the source at which no adverse effect is expected to occur for a given exposure. This hazard distance is plotted for the three LEDs source here. The x-axis spans three hours. The discontinuity at the end of the graph, exactly at x equal ten thousand second, is a fov discontinuity in the standard sixty two four seven one. Let’s now zoom on the leftmost part of the graph. Two considerations are to be made here. First the two sharp increases illustrated by the red straight lines are due to the fact that, initially, an averaging surface is centered on a single LED. As this surface increases, it will meet the second LED, then the third. A second consideration is that Method 1, under the uniform spectrum assumption, undervalues the hazard distance. That is to say, a distance is considered safe when it actually is not.

16 Luminance BLHW Méthode 1 Méthode 2 (Source à 5 LEDs) Points chaudsThe As an excerpt, let’s consider the two following blue light weighted radiance maps of the five LEDs source illustrated in the center. The image on the left was obtained using the spectrophotometer and imaging luminancemeter approach. The image on the right was obtained using the new imaging spectrophotometer approach. It clearly appears on this figures that the strong blue emission directly above the LED dies are lost using the former method, as can be seen by the hotspots lost from the figure on the right. ECHELLE ! 𝐿 𝐵 𝑥 , 𝜔 = (W∙ m 2 ∙ sr −1 )

17 Étude comparative

18 Étude comparative

19 Conclusion Synthèse PerspectivesMise en application des deux méthodes sur un ensemble de lampes L’hypothèse d’un spectre uniforme Sous évalue le risque Spécialement pour des temps d’exposition très courts Besoin d’une analyse spectrale ou selon BLHW Perspectives Mesure goniométrique Projet Rétinaled / INSERM Suivi des travaux IEC/CIE & ICNIRP Exposition chronique This comparative study confirmed that a risk exists for some sources, although most lamps were classified as exempt. We have also presented a new imaging spectrophotometry approach that overcomes the uniform spectrum assumption, problematic for an accurate blue light hazard assessment. In perspectives, due to the rising concern over the blue light hazard, camera weighted against B of lambda appear on the market, although they are still expensive. Besides, the device could be installed in a goniometer.

20 Merci pour votre attention