MARK REA and JEAN PAUL FREYSSINIER explain why lighting regulatory policies that combine efficacy and color rendering elements are fundamentally misguided, and how the industry could simply ensure that customers get both energy-efficient and quality light.
We believe that it is both important and appropriate for governments to be concerned with limiting wasted electric energy. Light sources that can provide the many benefits expected from lighting at a lower electric power should be formally encouraged or regulated by governments. But some policies that modify or patch efficacy guidelines with a color rendering index (CRI) requirement intended to ensure quality of light are fundamentally flawed. This article will explain how a revised look at the definition of the lumen could yield simple regulatory metrics that ensure efficient and quality light.
One of the important benefits provided by lighting is color quality. Several national governments (Australia, Canada, and the United States)1 and at least one state government (California through the California Energy Commission, or CEC 2016 policy)2 formally recognize the importance of color quality by including minimum requirements of the color rendering properties of light sources.
The idea behind such policies is that tighter restrictions on light source luminous efficacy (photopic lumens per watt) should not penalize color quality. The metric of choice by government regulators for characterizing the benefit of color quality has been the general CRI developed in the early 1960s. Briefly, CRI is a calculation procedure comparing the chromaticities of eight standard color chips illuminated by a reference source with their chromaticities when illuminated by a practical light source. The greater the chromaticity shift, the lower the CRI. The lower score applies even if people prefer the color quality rendered by the practical light source more than that provided by the reference source!
FIG. 1. Results from research performed by Narendran and Deng show a negative correlation between CRI and subjects' general preference and skin tone preference under different light sources.
Fig. 1 shows results from a study of color rendering by Narendran and Deng3 that clearly demonstrate the inability of CRI to characterize the color preferences rendered by different light sources. These results strongly suggest that a minimum requirement for CRI will not guarantee peoples' satisfaction of the color quality of illumination provided by a light source.
In more recent work, Fig. 2 shows results from an experiment where subjects were asked to a) rate different LED light sources in terms of how appealing they render the colors of plastic blocks (columns 1 and 3) and b) their willingness to purchase those plastic blocks based on how they looked under those light sources (columns 2 and 4). Their responses were plotted as a function of CRI (columns 1 and 2) and of gamut area index, GAI (columns 3 and 4), a measure of color saturation or color vividness provided by a light source. GAI is a metric that the Lighting Research Center (LRC) at Rensselaer Polytechnic Institute developed in 2010 as a complement to CRI.
Why CRI fails as a patch
Both Fig. 1 and 2 clearly show that higher values of CRI do not necessarily mean better color quality. Logically then, CRI should never be used by government regulators as a patch to luminous efficacy requirements to ensure that light sources will provide the expected benefit of color quality.
The fundamental problem is not, however, replacing one color quality patch with another one (e.g., GAI). Rather, the solution is to revise the current definition of luminous efficacy that is based upon photopic lumens per watt. The photopic luminous efficiency function, V(λ), sets the wavelength range and the weighting of those wavelengths in the numerator of the luminous efficacy calculation (i.e., lumens). V(λ) was developed in 1924 to represent "the spectral sensitivity of human vision".4 It is now known that V(λ) is based on just two of the five photoreceptors in the human retina, the long-wavelength (L) sensitive and the middle-wavelength (M) sensitive cones. The other three retinal photoreceptors provide humans with sensitivity to shorter wavelengths, so V(λ) is an inadequate and incorrect representation of human spectral sensitivity.
The spectral bias of V(λ) is particularly important for color quality, because it heavily discounts the significance of the short-wavelength (S) sensitive cone required for trichromatic color vision. (For background on color science, see the four-part series published by LEDs Magazine on the topic.) Rather than continue adding compensatory color quality patches to photopic luminous efficacy requirements, the definition of the luminous efficiency function underlying the lumen should change to be inclusive of all photoreceptors in the human retina.
FIG. 2. Research revealed subjective ratings of overall appeal of color of plastic blocks and willingness to purchase those plastic blocks based on how they look under eight different LED light sources, as a function of CRI (columns 1 and 2, negative correlation) and GAI (columns 3 and 4, positive correlation) for the LED sources evaluated. Observers rated each of the eight LED light sources sequentially, in random order at an illuminance of 420 lx inside a white diffuse viewing booth. One of the three tile sets was predominantly blue (55%, top row), one was predominantly red (55%, middle row), and the third one had equal proportions of red and blue blocks (35% each, bottom row). The relatively small number of green and white blocks was constant in all three tiles.
Fig. 3 shows the action spectra of the five photoreceptors in the human retina, taking into account pre-retinal filtering, along with the photopic luminous efficiency function, V(λ), currently used to define the lumen, and the universal luminous efficiency function, U(λ), proposed as a replacement. As Fig. 3 clearly shows, U(λ) better represents the spectral sensitivity of human vision than V(λ). Since luminous efficacy defined in terms of U(λ) would no longer be inherently biased against the S-cone, government regulators would no longer need compensatory color quality patches to modify or complement their luminous efficacy requirements.
FIG. 3. The graph shows the action spectra of the five photoreceptors in the human retina, taking into account pre-retinal filtering, along with the photopic luminous efficiency function, V(λ), currently used to define the lumen, and the universal luminous efficiency function, U(λ), proposed as a replacement.
Photopic and universal luminous efficacy
The table compares the photopic and the universal luminous efficacies of a selection of commercial light sources based on research in Reference 5. Notice that the two white LED light sources with the highest GAI, which is well correlated with color preference (Fig. 2), also have the highest universal luminous efficacy. This example helps demonstrate the fallacy of the myth that there is an inherent tradeoff between color quality and luminous efficacy. Notice too that these two sources do not have the highest CRI.
Consider high-pressure sodium (HPS), a source with high photopic luminous efficacy but poor color quality, as a hypothetical benchmark source for regulating minimum luminous efficacy. Except for HPS, all of the sources in the table would fail luminous efficacy requirements if the minimum photopic luminous efficacy was set at 107 lm/W.
In contrast, all of the sources in the table would pass if a minimum universal luminous efficacy was set at 138 lm/W. The table shows that if HPS were taken as the benchmark source for luminous efficacy regulations, white light sources designed, fabricated, and sold to provide many lighting benefits, including good color rendering, would no longer need a regulatory patch to ensure they remain on the market.
Suggested regulatory actions
In sum, governments want to make sure that increasingly restrictive luminous efficacy requirements do not have collateral negative effects on the expected benefits provided by LED lighting. Among the many quality criteria added to luminous efficacy regulations (e.g., start time, lamp life, flicker), minimum color quality standards are commonly employed.1 CRI has traditionally been the color quality metric of choice among regulators.
Many research results, including those shown here, demonstrate that CRI should never be used as a color quality patch for luminous efficacy regulations. CRI is simply not predictive of user preference or willingness to purchase a light source. More fundamentally, V(λ) is a biased representation of the spectral sensitivity of human vision, specifically discounting the role of S-cones required for good color vision.
Because U(λ) is a more accurate representation of the spectral sensitivity of human vision, utilizing it for defining the lumen in luminous efficacy requirements elegantly and simply obviates regulatory patches for color quality. The myth that there is a necessary tradeoff between luminous efficacy and color rendering disappears once the spectral sensitivity of human vision is accurately represented by the universal lumen. This does not mean that minimum color quality criteria couldn't be included in government regulations, only that they need not be introduced to offset the spectral bias of V(λ).
A comparison of photopic and universal luminous efficacies for a selection of commercial light sources.
A more complete discussion of the universal luminous efficiency function, U(λ), and more detailed, quantitative analyses of the electric energy wasted by using V(λ) in luminous efficacy regulations have recently been published.5-8
1. International Energy Agency (IEA), "Light's Labour's Lost: Policies for Energy-efficient Lighting in Support of the G8 Plan of Action," Paris: IEA, 2006.
2. California Energy Commission (CEC), "Proposed Revised Express Terms, 15-Day Language for Small Diameter Directional Lamp, Portable Luminaires, and General Service Light Emitting Diode Lamps," 2015 Appliance Efficiency Rulemaking Docket Number 15-AAER-6 (CEC-400-2015-044-15DAY-REV), January 2016.
3. N. Narendran and L. Deng, "Color rendering properties of LED light sources," Solid State Lighting II: Proc. SPIE, 4776, 61-67, 2002.
4. Commission Internationale de l'Éclairage, Commission Internationale de l'Éclairage Proc., Cambridge: Cambridge University Press, 1924.
5. M.S. Rea and A. Bierman, "A new rationale for setting light source luminous efficacy requirements," Lighting Res. and Technol., first published on Sep. 10, 2016, doi:10.1177/1477153516668230.
6. M.S. Rea, Value Metrics for Better Lighting, SPIE Press: Washington, USA, 2013.
7. M.S. Rea, "The lumen seen in a new light: Making distinctions between light, lighting and neuroscience," Lighting Res. and Technol., May 2015, 47, 259-280, first published on Mar. 31, 2014, doi:10.1177/1477153514527599.
8. M.S. Rea, "Shedding light on light and lighting," 28th Session of the CIE, Manchester, UK, June 28-July 4, 2015, Vienna: Commission Internationale de l'Éclairage.
Commentary from the editors of LEDs Magazine
This article mentions several issues for which readers may want to peruse some of our recent articles to gain a full understanding of some regulatory policy and industry standards. The authors mention regulatory policy developed by the California Energy Commission (CEC). The CEC action has been especially controversial of late as covered in several LEDs Magazine articles, including one championing the CEC action and others questioning the theory behind the regulations.
The authors also mention the GAI (gamut area index) color metric developed at the LRC and intended as a companion color metric to CRI. Indeed, the LRC's Mark Rea stated in an interview with LEDs Magazine leading up to Strategies in Light 2013 that only a two-metric characterization could accurately describe color quality. And the relatively-new Illuminating Engineering Society (IES) TM-30 standard defines such a two-metric system with the Rg gamut element paralleling the GAI approach.
MARK REA, PhD,is the director of the Lighting Research Center (LRC) and professor of architecture and cognitive sciences at Rensselaer Polytechnic Institute.JEAN PAUL FREYSSINIER, MS,is a senior research scientist at the LRC.