Color Rendition

You are here

The emergence of SSL technology quickly challenged long-standing performance metrics used for assessing color quality, particularly the International Commission on Illumination (CIE) Test Color Method, commonly known as CRI. Over several years of collaborative technical leadership, DOE has been instrumental in moving the lighting industry to replace CRI in order to better characterize the color performance of LEDs.

In 2015, the Illuminating Engineering Society (IES) published TM-30-15, IES Method for Evaluating Light Source Color Rendition (TM-30). Following the publication of TM-30, DOE has been part of an effort to educate members of the lighting industry about the new method, including hosting three webinars on the material. Several other presentations are also available, as well as a set of frequently asked questions. In 2018, IES TM-30 was updated and qualified as an American National Standards Institute (ANSI) standard.

TM-30, described in a DOE Technology Fact Sheet, provides a comprehensive system for characterizing how a light source will make colors appear, but it does not indicate how a light source should make colors appear in different situations. Since 2015, DOE has conducted research to measure subjective evaluations and preferences for different color rendition conditions, with a goal of providing lighting manufacturers and specifiers with more complete information to effectively balance color quality and energy efficiency in a variety of lighting applications. DOE has produced several peer-reviewed journal articles related to TM-30, which are summarized below.


Analysis of Color Rendition Specification Criteria (March 2019, Light-Emitting Devices, Materials, and Applications)

Methods for evaluating light source color rendition have recently undergone substantial changes. This article explores the ability of color rendition specification criteria to capture preferences for lighting color quality. For a compilation of five recent psychophysical studies on perceptions of colors, recently proposed specification criteria using ANSI/IES TM-30-18 substantially outperformed all currently used specification criteria in identifying preferred lighting conditions. To understand the consequences of changing color rendition specification criteria, the performance of a set of 484 commercially-available SPDs was evaluated.

Chart showing performance of different color rendition specifications for predicting the rank order of color preference for 166 SPDs evaluated in experiments conducted at PNNL.

A Vector Field Color Rendition Model for Characterizing Color Shifts and Metameric Mismatch (February 2019, LEUKOS)

This article describes a way to distinguish between two distinct components of light-source-induced color shifts, base color shift and metameric color shift, to provide a more complete understanding of color rendition. Working within the existing framework of IES TM-30-18 and CIE 224:2017, it shows that base color shift varies smoothly with location in color space in a pattern that is determined by the SPD of the light source. Patterns of smooth variation can often be approximated well with a low-order polynomial function. Here, a vector field model is presented, based on a second order polynomial function. The polynomial coefficients are adjusted, for a given light source, to provide a least squares fit to the calculated color shifts of a standard set of color samples. The adequacy of this model was verified by comparing it to another approach for characterizing base color shift that is based on discretization of color space and a much larger set of color samples. Once the vector field model of base color shifts for a given light source is determined, the metameric color shifts can be calculated from the residuals, and the distribution of those shifts can be statistically summarized. Based on this information, a Metameric Uncertainty Index (Rt) is proposed to provide new information about a light source. In particular, it can be used to estimate the likelihood of noticeable metameric mismatches induced by a given light source, which could lead to improved predictions of the perceived color quality of light.

Four charts showing correlation between Rf and Rt for the large SPD set; examples of two SPDs having a same value of Rf, but different values of Rt; and corresponding vector field maps for these SPDs.
(a) Correlation between Rf and Rt for the large SPD set. (b) Examples of two SPDs having a same value of Rf , but different values of Rt. (c) and (d) show corresponding vector field maps for these SPDs. 

Comparing Measures of Gamut Area  (November 2018, LEUKOS)

This article examines how the color sample set, color space, and other calculation elements influence the quantification of gamut area. The IES TM-30-18 Gamut Index (Rg) serves as a baseline, with comparisons made to several other measures documented in scientific literature and 12 new measures formulated for this analysis using various components of existing measures. The results demonstrate that changes in the color sample set, color space, and calculation procedure can all lead to substantial differences in light source performance characterizations.

It is impossible to determine the relative “accuracy” of any given measure outright, because gamut area is not directly correlated with any subjective quality of an illuminated environment. However, the utility of different approaches was considered based on the merits of individual components of the gamut area calculation and based on the ability of a measure to provide useful information within a complete system for evaluating color rendition. For gamut area measures, it is important to have a reasonably uniform distribution of color samples (or averaged coordinates) across hue angle—avoiding exclusive use of high-chroma samples—with sufficient quantity to ensure robustness but enough difference to avoid incidents of the hue-angle order of the samples varying between the test and reference conditions. It is also important to use a modern, uniform color space that is suitable for the quantification of color appearance and color difference.

Chart image showing the comparison of (a', b') CAM02-UCS coordinates for five color sample sets.
Comparison of (a', b') CAM02-UCS coordinates for five color sample sets. The illuminant is CIE D50.

Comparing Measures of Average Color Fidelity (December 2017, LEUKOS)

The introduction of new measures of color rendition, especially IES TM-30-15, has stirred debate within the lighting industry on the relative merits of the tools, as well as the amount of difference between the new tools and prior tools, such as CIE Ra. This article focuses on comparing three measures of average color fidelity: IES Rf, CIE Ra, and CIE Rf. Using a large set of commercially-available, experimental, and theoretical spectral power distributions (SPDs), the analysis contrasts past efforts to make similar comparisons using smaller or more focused datasets. It highlights the interactive effect of gamut shape and color space non-uniformity, which results in a range of IES Rf values of at least to 50 to 86 for SPDs having a CIE Ra value of 80. It also examines how these differences can be overlooked in psychophysical experiments relying on a small number of SPDs, which can present misleading findings on the value and meaning of the measures. When considering the results, it is important to remember that average color fidelity is only one aspect of color rendition.

Chart showing the comparison of CIE Rf and CIE Ra for the large SPD set.
Comparison of CIE Rf and CIE Ra for the large SPD set, showing only data points where both measures are greater than or equal to 50. This includes 2,303 SPDs (203 commercial, 136 experimental, 1,964 theoretical).


Chroma Shift and Gamut Shape: Going Beyond Average Color Fidelity and Gamut Area (October 2017, LEUKOS)

Though sometimes referred to as a two-measure system for evaluating color rendition, IES TM-30-15 includes key components that go beyond the two high-level average values, Fidelity Index (IES Rf) and Gamut Index (IES Rg). This article focuses on the Color Vector Graphic and Local Chroma Shift (IES Rcs,hj), discussing the calculation methods for these evaluation tools and providing context for the interpretation of the values. We illustrate why and how the Color Vector Graphic and Local Chroma Shift values capture information about color rendition that is impossible to describe with average measures (such as CIE Ra, IES Rf, or IES Rg), but that is pertinent to more completely quantifying color rendition, and to understanding human evaluations of color quality in the built environment. We also present alternatives for quantifying the Color Vector Graphic and Local Chroma Shift values, which can inform the development of future measures.

IES TM-30-15 Color Vector Graphics for two light sources with spectral power distributions added to the center of each graphic.
IES TM-30-15 Color Vector Graphics for two light sources with spectral power distributions added to the center of each graphic.

Human Perceptions of Color Rendition at Different Chromaticities (August 2017, Lighting Research & Technology)

An experiment was conducted to evaluate how perceptions of a light source’s color quality depend upon color rendition and chromaticity. Thirty-four participants each evaluated 50 lighting scenes in a 3.7 m by 5.5 m room filled with objects. The lighting scenes included five chromaticity groups, with 10 systematically-varied color rendition conditions repeated in each group. Participants, who chromatically adapted to each chromaticity group, were asked to rate each scene on eight point scales for saturated-dull, normal-shifted, and like-dislike (preference), as well as choosing whether they found the scenes to be acceptable or unacceptable.

Mean preference ratings for each of the color rendition conditions, and the mean preference ratings for all conditions in each chromaticity group.
Mean preference ratings for each of the color rendition conditions, and the mean preference ratings for all conditions in each chromaticity group.

The findings suggest that color rendition perceptions can vary with chromaticity, with an interactive effect of CCT and Duv. The same IES TM-30-15 measures—Rf, Rcs,h16, and Rg—could be used to effectively model perceptions within each chromaticity group, and provided suitable performance for the overall set of 50 conditions. The differences in ratings between the chromaticity groups were substantially smaller than the range in ratings for the 10 color rendition conditions within each group, allowing the same acceptability-based criteria of IES Rf ≥ 75, IES Rg ≥ 98, and -7% ≤ IES Rcs,h16 ≤ 15% to be applied to all chromaticity groups.

Human Perceptions of Color Rendition Vary with Average Fidelity, Average Gamut, and Gamut Shape
(August 2016, Lighting Research & Technology)

An experiment was conducted to evaluate how subjective impressions of a light source’s color quality depend upon the details of the shifts it causes in the color appearance of illuminated objects. Twenty-eight participants each evaluated 26 lighting conditions in a 3.1 m by 3.7 m room filled with objects selected to cover a range of hue, saturation, and lightness. TM-30 Fidelity Index (Rf) values ranged from 64 to 93, Gamut Index (Rg) values from 79 to 117, and Hue Angle Bin 1 Chroma Shift (Rcs,h1) values from -19% to 26%. All lighting conditions had the same nominal illuminance and chromaticity.

Color Rendition-image
Color rendition testing room.

Participants were asked to rate each condition on eight-point scales for saturated-dull, normal-shifted, and like-dislike, as well as classifying the condition as one of saturated, dull, normal, or shifted. The findings suggest that gamut shape is more important than average gamut area for modeling human preference, with red playing a more important role than other hues. Average fidelity alone is a weak predictor of human perception, especially CIE Ra (i.e., CRI). Nine of the top 12 rated products had a CIE Ra value of 73 or less, which indicates that the criteria of CIE Ra ≥ 80 may be excluding many preferred light sources.

The Role of Presented Objects in Deriving Color Preference Criteria from Psychophysical Studies
(December 2016, Leukos)

Although it is a critical component of any measure of color rendition, a standardized set of color samples can seldom perfectly match a real space or a real set of observed objects. This means there will always be some level of mismatch between predicted and observed color shifts. This article explores how the color distortions of three object sets that could be used in experiments compare to the color distortions predicted using the color evaluation samples of TM-30. The experimental object sets include those from a recent experiment, a set of produce (10 fruits and vegetables), and the X-Rite ColorChecker Classic. Specifiers should carefully consider how average measures of color rendition are applied to real spaces, and experimenters should carefully select experimental objects to avoid mischaracterizations.

Color coordinates for the four object sets evaluated.
Color coordinates for the four object sets evaluated.

What is the Reference? An examination of alternatives to the reference sources used in IES TM-30-15
(December 2016, Leukos)

This paper documents the role of the reference illuminant in the IES TM-30-15 method for evaluating color rendition. TM-30-15 relies on a relative reference scheme; that is, the reference illuminant and test source always have the same correlated color temperature (CCT). The reference illuminant is a Planckian radiator, model of daylight, or combination of those two, depending on the exact CCT of the test source. Three alternative reference schemes were considered: 1) either using all Planckian radiators or all daylight models, while maintaining the CCT match of the test and reference; 2) using only one of ten possible illuminants (Planckian, daylight, or equal energy), regardless of the CCT of the test source; 3) using an off-Planckian reference illuminant (i.e., a source with a negative Duv).

No reference scheme is inherently superior to another, with differences in metric values largely a result of small differences in gamut shape of the reference alternatives. While using any of the alternative schemes is more reasonable in the TM-30-15 evaluation framework than it was with the CIE Test Color Method, the differences still ultimately manifest only as changes in interpretation of the results. Reference illuminants are employed in TM-30 to provide a familiar point of comparison, not to establish an ideal source.

Comparison of CIE D Series Illuminants and Planckian radiation at the same CCTs.
Comparison of CIE D Series Illuminants and Planckian radiation at the same CCTs.

Development of the IES Method for Evaluating the Color Rendition of Light Sources
(June 2015, Optics Express)

A two-measure system for evaluating light sources’ color rendition has been developed that builds upon conceptual progress of numerous researchers over the last two decades. The system quantifies the color fidelity and color gamut (change in object chroma) of a light source in comparison to a reference illuminant. The calculations are based on a newly developed set of reflectance data from real samples uniformly distributed in color space (thereby fairly representing all colors) and in wavelength space (thereby precluding artificial optimization of the color rendition scores by spectral engineering). The color fidelity score Rf is an improved version of the CIE color rendering index. The color gamut score Ris an improved version of the Gamut Area Index. In combination, they provide two complementary assessments to guide the optimization of future light sources. This method summarizes the findings of the Color Metric Task Group of the Illuminating Engineering Society of North America (IES). It is adopted in the upcoming IES TM-30-2015, and is proposed for consideration with CIE.

Metrics and Test Methods