Numerical correlation between non-visual metrics and brightness metrics—implications for the evaluation of indoor white lighting systems in the photopic range

Lighting research and vision science have a long history, accompanied by the dynamic evolution of light source technologies from thermal radiators such as tungsten and halogen incandescent lamps, to discharge light sources such as mercury, sodium, and fluorescent lamps, to the current technology of white and colored LEDs (Light Emitting Diodes)^1–6. At each stage of this historical evolution, different visual tasks and visual metrics were defined to meet specific social, visual, and industrial needs. In the early decades of the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$20\textrm{th}$$\end{document}20th century, visual performance, including contrast vision, reaction time, visual acuity, and glare, was the focus of scientific and technological considerations. Thus, photometric quantities such as illuminance, luminance, uniformity, and glare index were defined using the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V(\lambda )$$\end{document}V(λ) function, the spectral luminous efficiency function for daytime vision^7,8. With the development of fluorescent lamps and metal halide lamps, with more possibilities to modify the lamp spectra, several aspects of color quality such as color rendering index (CRI), correlated color temperature (CCT) and chromaticity of white light have been addressed in scientific literature and also in regulations for practical lighting^9–11. With the continuous improvement of LED technology, new aspects of color quality at higher cognitive levels, such as color preference, color memory or color saturation, have been scientifically studied. These aspects have been introduced in practical applications, e.g. museum lighting^12–15.

With the discovery of intrinsically photosensitive retinal ganglion cells (ipRGCs) containing melanopsin pigments^16–18, two main lines of research on non-visual ipRGC-based effects have emerged. On the one hand, in lighting and sleep research, experiments have been conducted in laboratories or real-world settings (e.g., nursing homes, hospitals, schools, offices) to investigate the relationship between light intensity, spectrum, time, duration of light treatment, and non-visual outcomes (e.g., attention, sleep quality, alertness) using conventional photometric and colorimetric parameters such as vertical illuminance in lx, luminance in cd/\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m^2$$\end{document}m2, color temperature, and spectral irradiance distributions^16–24.

Since 2005, there has been a lot of research and discussion about how the receptor signals (rods, cones, and ipRGCs) combine to produce the non-visual effects of light. Section “The 2018 Circadian Stimulus (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2018}$$\end{document}CS2018) model” of this article describes the circadian stimulus (CS) models in the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2018}$$\end{document}CS2018 and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 versions of Rea et al.^1,2,25. Melanopic Equivalent Daylight Illuminance (mEDI, in lx) has been introduced according to the CIE publication³. These two metrics, CS and mEDI, are currently proposed for scientific discussion worldwide. They are also subject to further investigation.

From a physiological point of view, these two metrics represent different opinions on how to define the quantities for non-visual effects. According to the above mentioned CIE publication, the non-visual effects should be based on the signals of the ipRGCs, which express the effect of the melanopsin pigments, so that the mEDI metric can be used for the calculation and evaluation of lighting systems regarding the aspect of their non-visual effects. The CS model of Rea et al, in the 2005 and 2018 versions^1,2, is based on the idea that the non-visual effects come either from the ipRGC channel (if the opponent channel signal \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(b - y \le 0)$$\end{document}(b-y≤0), i.e. the blue channel has a weaker signal than the (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L+M$$\end{document}L+M) or yellow channel, for example in the case of warm white light sources), or a combination of the ipRGC channel and the signals of the combinations of the (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L+M$$\end{document}L+M), S cone and rod channels if (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b - y > 0$$\end{document}b-y>0). The cases (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b - y \le 0$$\end{document}b-y≤0) and (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b - y > 0$$\end{document}b-y>0) correspond to white light correlated color temperatures (CCTs) of, empirically, about (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CCT \le 3710\,K$$\end{document}CCT≤3710K) and (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CCT> 3710\,K$$\end{document}CCT>3710K), respectively.

In recent years, roughly between 2018 and 2022, numerous international scientific discussions and analyses have been conducted to specify the correct metric for the non-visual effects, using data sets from experiments by different research groups on nocturnal melatonin suppression as a validation basis. The following important research results were achieved during this period: Improvements of the CS model (versions 2005 and 2018) in the two years 2020 and 2021, taking into account the exposure time t (in hours) and the visual field, and modeling the contributions of the ipRGC channel, the S cones, the rods and the (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L + M$$\end{document}L+M) channel with improved terms, for both cases (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b - y > 0$$\end{document}b-y>0) and (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b - y \le 0$$\end{document}b-y≤0)^2,25. This improved formulation was validated using data sets for melatonin suppression² and is described in the present article.Publication on “Recommendations for daytime, evening, and night-time indoor light exposure to best support physiology, sleep, and wakefulness in healthy adults” by a group of sleep researchers and neurophysiologists²⁶.Based on the analysis of experimental data on nocturnal melatonin suppression and using mEDI as an input metric for non-visual effects, Giménez et al.⁴ defined a formula to predict melatonin suppression with exposure time and pupil dilation as additional parameters. This new metric was found by a machine learning method and is a non-linear transformation of the mEDI metric. This formula, now called the Gimenez formula in this article, is compared to the CS metric in its 2021 version.Parallel to the dynamic and intensive development of metrics for non-visual effects of light on humans, the development of metrics for brightness and visual clarity has experienced a renaissance with the development of quasi-monochromatic and phosphor-converted white LEDs with different correlated color temperatures and chromaticity coordinates. This new discussion started at the end of the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$20\textrm{th}$$\end{document}20th century by Fotios and Levermore⁵, with the development and analysis of new psychophysical methods in 2012²⁷, and has been continued by the authors of the present article^6,28,29. This renaissance could be explained by the fact that the brightness of white LED light sources of the same luminance but different spectra (different correlated color temperatures, chromaticity and color saturation enhancements) are perceived at different brightness levels. The root of these perceptual differences can be argued as the luminance or the signal of the luminance channel (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L + M$$\end{document}L+M) is not solely responsible for the brightness perception, which includes additional contributions from rods, S cones, opponent channels^5,30–32 and the ipRGC channel^33,34. This argument is demonstrated in the brightness experiments of the PhD thesis of Pepler³⁵.

With the above considerations in mind, the non-visual effects of light can be modeled either with a combination of signals from the ipRGC channel, the S cones, the (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L + M$$\end{document}L+M) channel, and the rod channel (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 model), or with the ipRGC signals alone (mEDI, CIE publication S 026:2018³). Similarly, a brightness metric (denoted by M) can be constructed by using an exponential function of the (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L + M$$\end{document}L+M) signal, the ipRGC signal, and the S cone signal⁶. From the point of view of lighting research and engineering, the following research questions arise: Is there a reasonable and usable correlation between the melanopic equivalent daylight illuminance (mEDI) and the circadian light \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CL_A$$\end{document}CLA in the CS models in the 2018 and 2021 versions; as well as the equivalent luminance \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L_{eq}$$\end{document}Leq according to Eq. 1 of Fotios and Levermore⁵ (see section “Brightness perception and modeling” of this article) for brightness perception?Is there also a useful correlation between the values of the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 metric, the Gimenez values for melatonin suppression, and the brightness perception metric according to the brightness metric in⁶?What is the difference between the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 metric values and the Gimenez melatonin suppression values for the same light source spectra? Is this difference acceptable for practical use of these metrics?If there is useable correlation between the different metrics and the difference between them is small enough to be in an acceptable range then a converting formula can be developed to transform one metric to the other with sufficient accuracy so that lighting researchers, sleep researchers or lighting engineers can design and evaluate a lighting system with several metrics recommended today although these metrics were established from different human physiological viewpoints. In the next sections, the brightness metric⁶, the model \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 of Rea et al.² and the formula of Giménez et al.⁴ will be presented before the correlations and relationships between the metrics for the non-visual effects and brightness will be described based on a calculation of 884 measured light source spectra.

Over the past six decades, many research studies have been conducted using colored and conventional white light luminance experiments and modeling. The most important models are those of Guth et al.³⁸ Ikeda et al.³⁹, Kokoschka et al.⁴⁰, Nakano et al.³², Palmer⁴¹, Ware and Cowan⁴², which led to a summary paper of the CIE (International Commission on Illumination) in⁴³, which tested all models until 2001. All models included in this fundamental paper considered the contributions of the opponent channels (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L - M$$\end{document}L-M) and (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S - (L + M)$$\end{document}S-(L+M)) indirectly by implementing the chromaticity x and y into a joint function with the luminance of the achromatic signal (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L + M$$\end{document}L+M). In a dissertation on photopic brightness in indoor lighting in 2017, Pepler³⁵ varied the spectra of polychromatic white light sources and the luminances in the photopic range on a homogeneous and diffusely reflecting wall in a real room without daylight incidence and found in a comprehensive psychophysical experiment that under the defined test conditions with white light, the most consistent model corresponding to the subjective evaluations of the test subjects is a model by Fotios et al. from 1998⁵, in which the so-called equivalent luminance (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L_{eq}$$\end{document}Leq) can be defined according to Eq. (1). This model divides the signal of the S-cones (S) by the signal of the V(\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda$$\end{document}λ) function and then calculates a metric to the power of 0.24, see Eq. (1).1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} L_{eq} = L_v \cdot (S/V)^{0.24} \end{aligned}$$\end{document}Leq=Lv·(S/V)0.24In Eq. (1), the exponent of photopic luminance (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L_v$$\end{document}Lv) remains 1.0, i.e., luminance remains uncompressed. To calculate the signals S or V, the relative spectral radiant flux of the light source must be multiplied by the spectral sensitivity function of the S-cones or by the V(\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda$$\end{document}λ) function, respectively, and this product must be integrated over the visible wavelength range.

With the discovery of a new type of ganglion cells, the intrinsically photosensitive retinal ganglion cells (ipRGCs), described in several scientific publications, including Hattar et al. in⁴⁴, some research has been conducted to answer the question whether ipRGC signals would also contribute to the perception of lightness in the photopic visual field. According to recent neurophysiological studies, there are some reasons to assume that ipRGCs interact with the visual channels in at least two different ways (see Zele et al.³³): \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M_4$$\end{document}M4-subtype ipRGCs project to the LGN and contribute to human light perception (see Brown et al.⁴⁵).a group of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M_1$$\end{document}M1-subtype ipRGCs establish signaling connections with upstream dopaminergic amacrine cells. Luminance signals can be transmitted to the outermost sublamina of the inner plexiform layer, influencing the state of light adaptation (see Prigge et al.⁴⁶).As a result, studies by Zele et al.³³ in 2018 and Yamakawa et al.³⁴ in 2019 had found ipRGC signals in brightness perception. In 2015, Bullough et al.³⁶ established a model for brightness that considers the contribution of the luminance channel V(\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda$$\end{document}λ), S-cone S(\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda$$\end{document}λ), and ipRGC (melanopsin, Mel(\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda$$\end{document}λ)), which is described in Eq. (2).2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} B_2= V(\lambda )+ 0.6 \cdot g \cdot S(\lambda ) + 0.5 \cdot Mel(\lambda ) \end{aligned}$$\end{document}B2=V(λ)+0.6·g·S(λ)+0.5·Mel(λ)The S-cone contribution multiplier g in Eq. (2) depends on the level of adaptation and increases as a function of light level. In the Bullough et al.³⁶ model in Eq. (2), the contributions of the luminance channel, S-cones, and ipRGC signals are integrated as a linear function into the brightness metric \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$B_2$$\end{document}B2. Brightness perception was analyzed and modeled in⁶ based on psychophysical experiments performed by the authors of the present article. For these experiments, 25 absolute spectra of multiple LED combinations (white LEDs and colored LEDs) with 5 different correlated color temperatures between 2700 and 10,000 K and 5 horizontal illuminances between 45 and 2000 lx with a relatively high color rendering index in the range \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$89 \le \text {IES TM-30-20 } R_f \le 93$$\end{document}89≤IES TM-30-20Rf≤93 were used. The resulting brightness model is shown in Eq. (3) (TUD stands for “TU Darmstadt”).3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} M_{TUD,VT2023}=8.9974 \cdot \left[ E_v^{0.2629} \left( S^{0.074} + 0.5 \cdot G^{0.0424}\right) \right] -1.3307 \end{aligned}$$\end{document}MTUD,VT2023=8.9974·Ev0.2629S0.074+0.5·G0.0424-1.3307This model (Eq. 3) contains the combination of an illuminance term compressed with the power function \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E_v^{0.2629}$$\end{document}Ev0.2629 and two terms with the compressed S-cone andipRGC signal. The optimization based on the experimental data had also shown that, from a mathematical point of view, the contribution of the S-cones and the ipRGC channel is crucial. It was the intention of the model builders to present this brightness model for lighting applications in the photopic range with white light.

The concept of Rea et al., covering the models \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2005}$$\end{document}CS2005, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2018}$$\end{document}CS2018 or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 is based on the design of a phototransduction circuit which regards the following mechanisms: The phototransduction of photoabsorption, signal generation and conversion into a frequency-coded form, and the processing of the signals of the different channels (LMS—cones, rods and ipRGCs) exhibit subadditivity. Additivity is assumed, e.g., in the definition of illuminance or luminance with the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V(\lambda )$$\end{document}V(λ) function or the mEDI metric, when the spectral sensitivity of the receptor system is multiplied by the spectral radiance or spectral irradiance of the incident radiation and all effects at each wavelength between 380 nm and 780 nm can be integrated by summation to the final effect of the total polychromatic radiation. No signal reduction is expected. In contrast, a possible subadditivity occurs when the effect at a wavelength \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda _1$$\end{document}λ1 is reduced when interacting with radiation of a certain wavelength \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda _2$$\end{document}λ2. In neurobiology, subadditivity can be explained if the neural circuit for phototransduction contains a spectral opponent channel. In vision, two spectral opponent channels are known, a (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$L - M$$\end{document}L-M) channel and a (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S - (L + M)$$\end{document}S-(L+M)) channel. In the context of non-visual effects, the spectral opponent channel (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S - (L + M)$$\end{document}S-(L+M)) (also referred to as [d=TUD]\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b - y$$\end{document}b-y\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$B - Y$$\end{document}B-Y) is taken into account. This is an important difference between the CS conception and the conception of the mEDI metric, which is defined by the CIE³ and assumes additivity of the nonvisual pathway. However, subadditivity was found to be essential in the experiments of Figuiero et al.^47,48.The CS concept followed the idea that a non-visual effect consists of two components, a spectral component and a quantity component. The spectral [d=TUD]sensitivity functioncomponent, denoted by the circadian light \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CL_A$$\end{document}CLA which will be described later, expresses the spectral generation of a stimulus at different receptors and channel systems (LMS-cones, rods, ipRGC) at a certain state of the spectral opponent channel (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(B - Y > 0)$$\end{document}(B-Y>0) or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(B - Y \le 0)$$\end{document}(B-Y≤0)). The definition of the mEDI metric does not distinguish between cases.The quantity component in the models from \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2005}$$\end{document}CS2005, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2018}$$\end{document}CS2018 up to the model \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 takes into account the exposure time, the characteristic of the visual field due to the spatial distribution of the ipRGC receptors on the retina, and the absolute magnitude of the circadian light value \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CL_A$$\end{document}CLA.A conversion from \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CL_A$$\end{document}CLA to the circadian stimulus CS in the model versions \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2005}$$\end{document}CS2005 and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2018}$$\end{document}CS2018 was based on the data sets of Thapan¹⁷ and Brainnard¹⁶ with quasi-monochromatic stimuli for nocturnal melatonin suppression. It has been improved and validated in 2021 by data sets from a variety of research groups. Therefore, the CS metric is also valid for lighting design processes for both evening and nighttime lighting.

The numerical study of Giménez et al.⁴ pursued similar intentions as the CS models with the following research conception: The authors of this study aimed to build a metric with mEDI (Melanopic Equivalent Daylight (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$D_{65}$$\end{document}D65) illuminance) as the starting parameter and extended the analysis to include the contributions of LMS cones based on a correlation analysis of 29 different data sets of nocturnal melatonin suppression published in scientific papers.The co-parameters were the exposure time and the pupil state of the subjects during the experiments (with or without pupil dilation). The metric to be found should be a metric for predicting nocturnal melatonin suppression similar to the above mentioned \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 model of Rea et al.The data analysis was based on the Random Forest (RF) method, a machine learning approach to solving classification and regression problems. The model was constructed in two steps. In the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1\textrm{st}$$\end{document}1st step, mEDI illuminance, photopic illuminance, rhodopic EDI (for rods), L-opic EDI, M-opic EDI, and S-opic EDI were subjected to separate correlation analyses at different exposure times for narrowband and polychromatic light spectra. From 21 to 10,000 lx, the mEDI metric showed the best correlation coefficients. The S cone EDI outperformed the mEDI metric only in the range below 21 lx.

With mEDI illuminance as the initial parameter, other components such as LMS cone signals, exposure time (duration), and pupil dilation were added to the set of input parameters, and a four-parameter logistic function was constructed and compared to the available data sets. The accuracy of the regression analysis was expressed as the root mean square error (RMSE). The optimal random forest model was the model with the lowest RMSE and the least number of predictors. In addition, the coefficient \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r^2$$\end{document}r2 was also used. The results of this analysis can be summarized as follows: Adding L and M cones did not improve the model quality compared to the combination of S cones and ipRGC alone.The combination of ipRGC and S cones resulted in a higher correlation coefficient and a lower RMSE value compared to ipRGC alone (mEDI). The difference was rather small, so the Giménez research group decided to ignore the S-signal portion in their model.The logistic function model therefore includes mEDI, exposure time and pupil dilation, see Eq. (11).11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} Suppression_{melatonin}=\frac{0-100}{1+\frac{log_{10}(mEDI_{melanopic} \cdot 10^6)}{9.002-0.008 \cdot \Delta t_{expose} - 0.462 \cdot dil_{pupil}}} \end{aligned}$$\end{document}Suppressionmelatonin=0-1001+log10(mEDImelanopic·106)9.002-0.008·Δtexpose-0.462·dilpupilIn Eq. (11), the symbols have the following meaning:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$suppression_{melatonin}$$\end{document}suppressionmelatonin= melatonin suppression (in %)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$mEDI_{melanopic}$$\end{document}mEDImelanopic = melanopic EDI (lx)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta t_{exposure}$$\end{document}Δtexposure = exposure duration (in minutes)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$dil_{pupil}$$\end{document}dilpupil = pupil dilation applied: 0 = no, 1 = yesCompared to the CS model of Rea et al., the value of melatonin suppression is up to 100% at infinite illuminance. In the opinion of the authors of the present article, the most important difference between the CS model in the version \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 (or \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2018}$$\end{document}CS2018) and the model of Giménez et al., is the aspect of value scaling. (Note that both models were built using regression methods for nocturnal melatonin suppression data. Both models have been validated using partially similar data sets from well-known research groups).

In this article, the concept of non-visual parameters such as the circadian stimulus modes of the Circadian Stimulus by Rea et. al. in the USA (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2018}$$\end{document}CS2018¹, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021²), melanopic equivalent daylight (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$D_{65}$$\end{document}D65) illuminance, mEDI of the CIE S 026/E: 2018³ and the latest formula of Giménez et al.⁴ for nocturnal melatonin suppression⁴ are briefly described to understand their structure and characteristics. Also, the equivalent luminance of Fotios et al.⁵ or the brightness of the TU Darmstadt⁶ are briefly introduced.

Then the calculations and analyses based on the databases of 884 light sources (26 conventional incandescent lamps and filtered incandescent lamps, 252 fluorescent tubes and compact fluorescent lamps, 419 LED lamps and LED luminaires, and 185 daylight from measurements) were implemented.

Summarizing the results of this article, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 is a significant improvement over \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2018}$$\end{document}CS2018, which did not work in the warm white region (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CCT \le 3710\,K$$\end{document}CCT≤3710K). The correlation between the brightness metric \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M_{2023,TUD}$$\end{document}M2023,TUD, the Giménez metric (based on the mEDI metric) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CS_{2021}$$\end{document}CS2021 based on \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$CL_{A,2021}(CLA\,2.0)$$\end{document}CLA,2021(CLA2) is good or even very high. Consequently, these three metrics can be converted by linear formulas (see the equations Eqs. (12)–(14) with acceptable accuracy from an engineering point of view.