Advances in LED color quality are presenting serious challenges to the Color Rendering Index (CRI) and Correlated Color Temperature (CCT). These two metrics, developed in an era of incandescent and fluorescent lamps, are still widely used by the lighting industry to communicate the color performance of all sources. Although their limitations have long been known, the rapid proliferation of LEDs has prompted the development of better metrics to predict their specific color rendering ability.
Understanding CRI and CCT
CRI is particularly unreliable and, say some lighting experts, even irrelevant when applied to LEDs. Although LEDs generally have lower CRIs than conventional sources, such as incandescents, some LEDs have been shown to reproduce color more vividly and attractively—a trait that is particularly desirable to retailers. Moreover, notes Rohit Patil, a color scientist at Xicato in San Jose, Calif., LEDs offer the unique opportunity to “create a custom spectrum of lights” for specific installations, which may prove more useful than an exalted CRI.
Established by the Commission Internationale de l’Eclairage (CIE) in the 1960s, CRI measures a light source’s ability to reveal the intrinsic colors of the objects it illuminates. Testing is done with eight color chips, numbered R1 to R8, and the results are compared to those of a reference source of the same CCT. Sources with a CCT below 5000K are compared against a blackbody radiator—a non-reflective object that, when heated, emits a spectrum of light solely determined by temperature. Sources with a CCT above 5000K are checked against daylight.
Differences in color rendition are evaluated on a scale of zero to 100, with 100 indicating a match (negative CRI numbers are rounded up to zero). A CRI of 80 or above is typically desired for indoor applications—not a difficult feat for incandescents, halogens, and metal halides, which typically have CRIs at or above 90. Many of today’s LEDs are competitive, a notable achievement for a source whose CRIs topped out at 60 or 70 a mere decade ago, says Paul Scheidt, product marketing manager at LED manufacturer Cree.
CCT, the other primary metric, focuses on the tint of white light exhibited by the source. Measured in degrees Kelvin, it relates the color of a white light source’s illumination to the surface temperature of a blackbody radiator. Warm sources have a yellow tint and lower CCT values. Cool sources have a bluish cast and higher CCT values. Candlelight, for example, is rated around 1850K, while daylight exceeds 5000K.
Although CRI and CCT conveniently reduce the complexity of color performance to a single value, “anytime we do that, we lose a lot of information,” Scheidt says. Sources with the same CRI or CCT value can vary widely in appearance and behavior. This is particularly problematic with solid-state lighting, where CRI has not been “very predictive” in the specification of “quality lighting,” says Mark Rea, director of the Lighting Research Center (LRC) in Troy, N.Y.
The Mathematics of Color
Both CRI and CCT are derived through rote mathematic simulation rather than through empirical measurement. CRI testing is calculated on a computing device using a source’s spectral power distribution (SPD), a diagram that depicts the radiant energy a source emits at different wavelengths of visible light—wavelengths of 380 to 780 nanometers—and the spectral reflectance of each color chip. CCT is also computed from the source’s SPD.
The math behind CRI and CCT stems from the CIE colorimetry system. Though not exclusive to lighting, it provides the foundation for all color calculations in the lighting industry today, regardless of source. The system is precise in that it measures color based on spectral characteristics rather than on appearance, which can be more subjective, contextual, and difficult to evaluate. One of the system’s earliest and most commonly used mathematical models is the CIE 1931 color space, which maps all visible color to an x, y graph based on chromaticity. Chromaticity refers to a color’s hue—its dominant wavelength—and saturation, and is expressed in the color space by a pair of coordinates derived from a source’s SPD.
Although the 1931 standard remains in use today, updates have improved its uniformity so that calculated differences between colors are more perceptually accurate. Both CRI and CCT use the CIE 1960 (u, v) color space, but it, too, is considered outdated and the mathematics lacking in rigor, says Michael Royer, a lighting engineer with the Pacific Northwest National Laboratory’s advanced lighting team.
Limitations of CRI
Beyond numbers, one longstanding criticism of CRI has been the pastel appearance of the eight test colors, which “are not representative of the world,” says Julian Carey, senior director of marketing at LED phosphors manufacturer Intematix Corp. Seven additional color patches, named R9 to R15, have been introduced and include a saturated red, yellow, green, and blue, as well as two skin tones and a green representative for vegetation. However, these patches are not applied to the calculation of CRI and are only recommended for supplemental information.
Equally problematic, CRI is an average of the color shifts on the eight test colors. Consequently, an LED product with standout performance on some test colors and poor rendering on others still achieves a high rating. To better inform specifications, some LED manufacturers are publishing the individual values of R1 through R15.
CRI is often mistaken as an indicator of how pleasantly colors will be rendered. In fact, it functions more as a fidelity index. Performance is rated with respect to a reference source—either a blackbody radiator or daylight—which is considered the gold standard. But this could be misleading too. “What if I can create a light source that does much better in rendering than the reference source?” Patil asks. It would likely be penalized, he says, even if “colors appear more colorful than under the reference source.”
Given the strides made in phosphor-converted white LEDs, which account for the majority of LEDs used in architectural lighting applications, it may be time for a new reference source. Whereas early LEDs relied on a yellow phosphor to absorb energy from a blue diode and produce white light (often with a bluish tinge), advancements in phosphor compositions now allow the manipulation of spectral content and therefore color rendition. Some companies have replaced the blue LED with one in the near-violet region to produce a fuller, more continuous spectrum and thus colors that are more vivid and whites that are more nuanced. A fuller spectrum, however, might come at the expense of energy efficiency, as more phosphor requires more energy to convert the blue LED into white light.
Solid-state lighting is also challenging the adequacy of CCT, but it’s not the first to do so. Like other sources, such as high-intensity discharge lamps, LEDs with the same CCT can differ vastly in chromaticity and, therefore, appearance. One may have a greenish cast, while another may seem slightly pink.
While conventional wisdom suggests that light sources that cleave close to the blackbody locus appear whiter, the LRC has found otherwise. In residential applications, most people prefer whiter light than the warm output of incandescents, says Rea, who co-wrote the 2013 paper “Class A Color Designation for Light Sources Used in General Illumination” in the Journal of Light and Visual Environment. When mapped in the color space, the chromaticity of perceived whiteness follows an irregular path that goes both above and below the blackbody locus.
CCT can be particularly ineffective in assessing phosphor-converted LEDs. Manufacturers sort LEDs into bins based on the CIE 1931 color space, Scheidt says. Bin size, which refers to the area of tolerance for chromaticity differences, is measured in units of SDCM (standard deviation of color matching), or MacAdam ellipse steps. The latter takes its name from color scientist David MacAdam, who discovered that chromaticity shifts undetectable by the human eye fell within an ellipse on the 1931 color space.
The ANSI C78.377-2008 LED binning standard defines one bin size as a seven-step MacAdam ellipse. This, Patil says, is “huge” and may account for the criticism of early LEDs as having poor uniformity. “Manufacturers were making LED sources that fell into [one] bin, but looked really different.” When installed side by side, they can produce a rainbow effect. Improvements and innovations in manufacturing have enabled some companies to put LEDs in bins as small as one or two SDCMs. A difference of three SDCMs is noticeable by the majority of the population, Patil says.
Although bin specification is typically the purview of luminaire manufacturers, designers should know the manufacturer’s tolerance for initial color consistency. A difference of three SDCMs, for example, may become even more pronounced as the number of luminaires increases. Responsible manufacturers publish this information, says Steve Landau, Xicato’s director of marketing communications; if they don’t, he says, “that’s a big red flag.”
Given CRI’s uneven history, several new metrics and classifications have been proposed that address the indexes’ limitations, particularly as solid-state lighting gains a larger market share.
The National Institute of Standards and Technology’s Color Quality Scale (CQS) offers improvements on multiple fronts. It is a fidelity metric that results in a single-number rating, but tests with a broader range of colors (15 instead of eight) that are higher in chroma and saturation than R1 to R15. Color preference is also considered, Patil says. “If a light source makes colors appear more colorful than does the reference source,” he says, “it will have a higher number.” To penalize color distortion, the CQS imposes an upper limit on saturation that, if exceeded, will lower a source’s rating.
CQS uses a color space that is more uniform and its calculations are more rigorous than those for CRI, Scheidt says. CQS also factors in extreme color temperature, which impairs a source’s ability to render color, and takes a root-mean-square of the color shifts of all 15 test colors rather than an average. This ensures that poor performance on a few samples is given proper weight. CQS also rates sources on a scale of zero to 100, but negative scores are not possible, unlike in CRI.
Though CQS has not been adopted as a standard yet, it is receiving much interest. The system is being used by many in the industry, says Yoshi Ohno, NIST Fellow, Sensor Science Division, who helped develop the scale. It is also under consideration by CIE technical committee TC 1-91, which is tasked with recommending color quality metrics.
The LRC recently proposed a certification of white light called Class A color. Intended as a communication tool for non-lighting professionals, the Class A designation is given to a source only after it has fulfilled four requirements: it has a CRI that is 80 or higher; the chromaticity must fall along a line of preferred tint, established through research; the chromaticity must fall within areas of roughly four-step MacAdam ellipses; and its gamut area index (GAI) should be between 80 and 100.
GAI, which measures color saturation or vividness, is derived from a light source’s SPD and the same eight test colors that determine CRI. Calculations are done on a uniform CIE color space to produce chromaticity coordinates that form a polygon. The enclosed area is the gamut area. A larger area generally means a higher index and more saturated colors. Unlike CRI, GAI is not a fidelity metric, and an index greater than 100 is possible.
For the specification community, the LRC’s Rea recommends GAI as a secondary metric to CRI. Research has shown it can influence light source preference. In tests, neodymium lamps, which have a lower CRI than incandescents but a higher GAI, tend to fare better in color rendering. For retail applications, LED products with a GAI of 130 or more can enhance merchandise’s appeal by making colors “really pop,” Intematix’s Carey says.
The IES Color Metrics Task Group plans to finalize work on new color metrics this fall, after which they will undergo several rounds of approval by the Color Committee, Technical Review Council, and the Board of Directors, the PNNL’s Royer says. The effort, which Royer is chairing, will incorporate aspects of CQS and have both a fidelity metric and a gamut area metric. Fifteen new test colors—different from R1 to R15—will cover a full range of hues and saturations, and the calculation methods and color spaces will be updated.
In addition to TC 1-91, the CIE has created a technical committee, TC 1-90, to develop a fidelity index to replace CRI. The CIE is also contemplating an update to its color-matching functions. Direct measurement of spectral cone sensitivities has revealed inaccuracies in the color matching functions (CMFs)—which determine a source’s chromaticity coordinates—especially in the blue region. This “has significant ramifications for the LED industry,” Xicato’s Patil says. With blue diodes as the starting point for phosphor-converted LEDs, rectifying CMF shortcomings may lead to more accurate assessments of LED color performance.
The accuracy of an index, no matter how great, should never replace an actual mock-up of a light source. The degree of color rendering and white light needed is specific to the particular application. However, metrics that keep apace with LED technology can help lighting designers to better bridge the gap between measured color and perceived color.
A list of introductory articles that discuss color perception and rendering as they relate to the lighting industry.
“LED Color Mixing: Basics and Background,” by Cree, 2014. Available at: cree.com/xlamp_app_notes/color_mixing.
“Defining the Color Characteristics of White LEDs,” by Steven Keeping, April 23, 2013. Available at Digi-Key Corp.: bit.ly/1qArLLo.
“LED Color Characteristics,” by the U.S. Department of Energy, Office of Energy Efficiency & Renewable Energy, Building Technologies Program, January 2012. Available at: 1.usa.gov/1lO7AWI.
Note: This article has been updated since first publication to state that correlated color temperature (CCT) can be particularly ineffective in assessing phosphor-converted LEDs. The previous version of the article and the print version of the article incorrectly stated that CCT can be effective in assessing phosphor-converted LEDs.