Rachel Schwen and Roy Berns Fall, Design of LED Clusters for Optimal Museum Lighting

Transcription

1 Rachel Schwen and Roy Berns Fall, 2011 Design of LED Clusters for Optimal Museum Lighting

2 Introduction Museum lighting has historically been a field of trade offs. Museum environments have set guidelines for the lighting of fine art that aim to minimize degradation. This is accomplished by eliminating high-energy wavelengths, particularly those in the UV range below 400nm (as defined by CIE 157:2004) or is limited to 75 µw/lumen of ultra-violet radiation. There are also specifications for illuminance. For medium and high responsivity materials, total illuminance is suggested to be at or less than 50 lx (CIE 157:2004). For less responsive pigments, illuminance is suggested to be between 150 and 200 lx (CIE 157:2004, CIBS), and for those materials classified as irresponsive no limit on illumination is specified (NOTE: There are other suggested specifications for lighting in museum environments such as CIBSE, and IES which will not be discussed here). Curators are most interested in preserving artwork for future generations while displaying it in such as way as to allow current viewers to be able to enjoy the full impact of the art. The simple act of meeting these specifications creates the first level of obvious trade offs for curators. I. How much damage is too much for art? Color change due to fading is a major issue when assessing light damage done by a light source. Richardson and Saunders 2007 conducted a study, which indicates that damage as small as 2 ΔE ab units is noticeable and up to 4 ΔE ab indicates a reasonable amount of damage over a 50 to 100 year period. This study only examined a small number of museum staff and utilized computer-generated images to simulate fading to specific degrees. More extensive work on actual paintings with a wider variety of museum professions would help clarify this study. However, this study indicates the importance when evaluating a light source for

3 museum lighting design to ensure that the source has accurate color rendering characteristics so that it is possible to evaluate changes due to light damage. Saunders and Kirby 2008 conducted a study comparing the damage created by several different light sources. They found that Compared to the presence of absence of ultraviolet radiation, the difference between the two daylight sources, or that between the museum fluorescent and tungsten lamps, was very small. This study also makes the important point that the potential deleterious effect on museum objects is not the most important criterion for choosing one over the other. Due to the fact that many LED combinations have very poor color rendering qualities. Druzik and Eshøj 2007 summarized the many trade-offs with museum lighting. Including the need for a communal environment for sharing ideas about conservation techniques, where concerned persons could discuss improving and understanding color rendering metrics, evaluate new lighting technologies and energy conservation standards, and compare damage spectra, as well as asses aesthetic and visual performance of a museum lighting design. II. What total illuminance level should be used to minimize damage while allowing viewers to appreciate the art? Reducing the illuminance of a given exhibit may make painting or drawings difficult to see, or colors difficult to distinguish making the display undesirable to visitors (Crawford 1973). Crawford s study suggests 30 lx is the minimum for color discrimination. Therefore, it is likely that at 50 lx the discrimination of colors would also be difficult. To this end, many studies have been carried out which examine a quantitative metric for the degradation of artwork to justify the use of such a low illuminance level for the specification for museum lighting design. As a result, L S Harrison created a metric called the damage function, which

4 has been adapted by others to reflect annual damage exposer. Discussions pertaining to damage by optical radiation are well summarized in Cuttle Loe et al 1982 state clearly that The current recommendations, particularly those related to lighting, have been set rather arbitrarily, in that the illuminance recommendations have been set as low as was considered reasonably possible without firm evidence that they were necessary. The Loe et al. study indicates that illuminance below 200 lux show a decreased discrimination and quality of the viewed image including both oil and watercolor paintings. III.What correlated color temperature should be used to reduce damage and create a pleasing lighting environment? Many studies have been performed with varying results as to which correlated color temperature (CCT) is preferred when observing fine art. Two articles by Scuello et al 2004 performed a series of studies that found 3600K to 3700K were ideal correlated color temperatures for museum lighting conditions. One experiment was performed using printed postcard reproductions. These postcards were largely blue or red depending on the scene they depicted, as these tend to be problem areas of color rendering for light sources with different CCTs. Also it should be noted that most artist paints are not spectrally similar to conventional printing inks particularly in regard to phthalocyanine cyan ink and ultramarine and cobalt blue artist paints. Cobalt blue and ultramarine both feature a long wavelength tail in the red region of the spectrum, which often impacts how these pigments are rendered, whereas phthalocyanine does not feature this tale and would therefore be more color constant than the equivalent artist pigment. Likewise the addition of black to printing ink adds color constancy to printed materials, which may not be present in actual artwork. The other experiment involved using a Maxwellian-view to evaluate the warm or cool quality of the

5 preferred lighting condition for museum lighting. They concluded that observers prefer lighting that is neither noticeably warm nor cool. Where as, Pinto et al 2006 and Pinto et al 2008 found that 5100K was the ideal CCT for viewing a CRT displaying digitized artwork calculated with a hyperspectral and multispectral imaging system. This study noted that the initial change in average spectra between the digitization and the actual artwork began at 2%. It may be important when using LEDs to examine which CCT is ideal for viewing artwork in a gallery environment. It is also important to account for chromatic adaptation in the evaluation of this question, which was not considered in either study. Manipulating the CCT of a light source was previously difficult to examine due to the limitations of light sources which met the museum requirements. He and Zheng 2010 discussed the possibility of using LED clusters to create a combination that would be tunable to different CCT while maintaining a high color-rendering index. With the addition of broadband white LEDs and three single color LEDs they were able to create clusters with a high CRI and tunable CCT from approximately 2700K to 6100K. One of their combinations also had a high luminous efficacy. This study suggests that LEDs could certainly be used to create a toolbox of lighting design options for museum lighting environments. IV. What options are available for light sources? Traditionally these specifications have limited the light sources available to curators to mostly incandescent lighting. Recently advances in the science of lighting have led to the creation of new light sources that meet museum specifications (CIE 157:2004) and provide additional flexibility in lighting design and presentation. Halogen, and fluorescent blubs have become more widely used in museum lighting as they can now be manufactured with filters that cut UV radiation, the enemy of all curators of historical artifacts (Cuttle 2000). These

6 advances have allowed curators to have greater flexibility in lighting designs and a method to change the lighting characteristics based on the goals they wish to obtain in a given exhibition. For example, Cuttle 2000 proposed the use of MR tungsten-halogen lamp and a MR1620/12 BAB/FL/40 lamp. This work aimed to compare a black body light sources to one with three-bands of radiation. Three prints of famous paintings with different colors used and from different artistic styles were used in a simulated art gallery to test observers on five major categories brightness, clarity, acceptability of overall color appearance, brightness or colorfulness of individual colors and naturalness of individual colors. This study investigated two correlated color temperatures (CCTs) one low and one intermediate. They concluded that once observers had matched for equality of illuminance observers were able to notice a difference between the two sources but they did not show a clear preference for one over the other suggesting that light sources with out a smooth spectral power distribution were acceptable to viewers. However, these new lights create environments are much different than traditional museum environments. These lights have different fundamental characteristics than incandescent blubs. For example, halogen blubs when dimmed emit a more yellow light then at full brightness. In an effort to evaluate metrics for lighting design with a goal of reducing degradation of art work and the best possible color rendering Delgado et al 2011 explored using filters to eliminate regions of the spectrum in order to ascertain if reduced damage might be possible. This work used filtered tungsten-halogens as an analogue to LED sources in an effort to create the best luminous efficacy while preserving high quality color rendering. They came to the basic conclusion that it is possible to optimize filters to reduce damage to art while maintaining high luminous efficacy as well as

7 reasonable CRI values. They cite that their multi-band filter showed more degradation than did the broadband filter. V. What about the use of LEDs? The newest and most promising light source for museum lighting is solid-state lighting, which includes LEDs, light emitting diodes, OLEDs, organic light-emitting diodes, and PLEDs, polymer light-emitting diodes. In this investigation the term LED describes traditional LEDs as well as phosphor coated LEDs. LEDs, if properly designed, can virtually eliminated UV radiation, create more light with less energy use, and can be tailored to meet the needs of curatorial desires. LEDs changed drastically over the last 10 years. One of the major concerns of using LEDs in museum lighting environments is that the traditionally do not exhibit similar color rendering as compared to traditional museum illuminants. The CIE Color Rendering Index (CRI) is often very low in comparison with fluorescent or tungstenhalogen lights. For example, Mahler et al 2008 tested 4 different LEDs clusters and a tungsten-halogen light using the Cercle 32 test for color discrimination. From this testing they concluded, RGB LED light yields poor discrimination and poor color rendering. They also showed a link, for their light sources, between CRI and color discrimination efficiency. This study and many others (van der Burgt and van Kemenade 2010, and Ohno 2004) cite low CRI as a clear down side of the use of LED clusters for lighting but also show that human preference may not be as clearly correlated to this as might have previously been thought. Ohno 2004 evaluated several white light LED clusters with goals to optimize CRI and LER (Luminous Efficacy of Radiation) when compared to conventional lamps. In this investigation of white light LEDs and LED clusters it was noted that chromatic reds maybe

8 difficult colors to properly render. The lack of the red component will drastically change the renderable color gamut of a light source as well as making objects generally appear dull. Six conventional light sources were compared to twelve LED based light sources as well as to typical reference illuminants like CIE D65 and CIE A. The possible use of R (9-12) of the CIE Color Rendering Index could possibly be a more useful tool in evaluating LED lights as these more chromatic colors have the potential to show much greater variation in hue or value. When examining the potential of phosphor-coated LEDs as an alternative to LED cluster lighting this study found that these LEDs along show a significantly reduced LER, which would not be desirable especially as one of the advantages of LED illumination is its energy efficiency. Lastly this study discusses the downsides of using CRI as a metric for the evaluation of a light source and particularly as it applies to LEDs which will be addressed later. VI. What about using LEDs specifically for museum lighting? Several studies have been aimed at using LEDs for museum lighting environments. Work done by Boissard and Fontoynont in 2009 tested the differences between fluorescent, halogen and LED illumination in a series of light booths with the Gretag Macbeth color checker and a print of The Astronomer by Vermeer. The goal of this study was to assess acceptability of the different lighting conditions by observers. They concluded that LEDbased lighting was just as acceptable to viewers as fluorescent of halogen lighting. Berns 2010 investigated the feasibility of using red, green, and blue LED lights to simulate daylight (D65) for use with a wide array of artist pigments. This study examined a set of mixtures of Golden Artist Colors Fluid Acrylics as an analogue for common artist pigments as well as computational derived and real LED combinations that meet the

9 requirements for museum lighting, mainly that they exhibit no UV radiation to avoid damage to the art, and that they can be manipulated to low illuminance as well. Psychophysical studies performed on the color rendering of LEDs have mainly focused on color tests with low chroma elements, which do not speak to preference, or specific lighting goals. Vienot et all 2005 discussed color discrimination with three different LED combinations in comparison to a filtered tungsten-halogen bulb using the Munsell DD25 task and the CIE color rendering indices (CRI). As might be expected the light sources with the lowest CRI performed the worst in psychophysical tests and the light source with the highest CRI performed the best. This paper claims, Not only all LED illuminations impair hue discrimination, but the impairment is sever However, this study only examined one such light source and it did not employ any phosphor coated LEDs which feature a more broadband spectrum. Likewise the CRI of this RGB light source was less than 20, which is generally considered to be a very bad light source for color rendering. Therefore, this study s assertion that Our experiment shows that the RGB LED cluster should be avoided. is perhaps true for this specific RGB LED cluster but should not be applied to colored LED clusters on the whole. A further limitation on museums is cost, not only of the initial lighting hardware but also of the long-term energy consumption and replacement of blubs. Likewise, in exhibits where artists materials are very sensitive to the overall lighting level to which they are exposed curators must weigh the risks of each new lighting design option. In order to cut costs, and become more environmentally friendly museums are looking for new light sources whose cost on the whole is less than traditional incandescent lighting. LEDs produce less heat than halogen or tungsten light sources, which would also help preserve the artwork.

10 LEDs also require less energy for equivalent illuminance, which in the long run will cut operational costs for most institutions. LEDs continue to progress in their complexity and utility and if possible museums could greatly benefit from a switch to LED based lighting. Recently the United States Department of Energy has taken a great interest in LED lighting products in Museum lighting environments as part of the Gateway program (US DOE Solid-State Lighting Design GATEWAY program, 2010). The first of these Gateway demonstrations was demonstrated in the Field Museum of Natural History. In this study compared halogen lighting and LED lighting systems in an exhibit where damage was not a major concern. The study asked both patrons, and museum professionals to evaluate the difference between these two lighting conditions. Overall the LED system successfully met the different aspects of museum lighting needs. However, they also made a point of saying it cannot be said that unequivocally that the LEDs are better than halogen systems for museums. As for the Gateway s project specifically for fine art museums a recent article (Brodrick, 2011) summarized pilot studies in the Smithsonian, the Getty Conservation Institute, and the Jordan Schnitzer Museum of Art at University of Oregon. The Smithsonian Institute has been using LED lighting in the American Art Museum. They state that products did not provide a wow factor but that most patrons did not notice the change and the institute saved nearly 75 percent on energy consumption in the four galleries where LED lighting was used. The Getty Conservation Institute is currently performing a long-term study on the potential for light damage with the use of LEDs. This study compares damage induced by LED lighting with that induced by halogen lighting, which the institute found to be comparable. In the Jordan Schnitzer Museum of Art at University of Oregon LED lighting was used on sever different kinds of art as well as different types of LEDs versus a more

11 traditional halogen bulb. Then visitors were asked to choose which lighting condition they preferred for a color painting, a black-and-white photo, and a color chart. Though the halogen lamp in this study received many high marks one LED lamp stood out as a leader in all categories. This museum also cited that the longevity of the LED bulbs made them not only less expensive in energy cost but also less expensive in maintenance and replacement cost, a factor many overlook in the cost evaluation of a switch to LED lighting. Work by Chalmers and Soltic 2010 addressed the use of colored LED clusters first from a theoretical standpoint and later tested their findings by using real LEDs manufactured by Philips Lighting. These real world mixtures were chosen by a process called intelligent spectral design and were field tested at the American Institute for Conservation 2007 Museum Lighting workshop (Cuttle, 2009). The items used for the field-testing were the Gretag Macbeth ColorChecker, a group of anthropological objects, and a reproduction of a Claude Monet oil painting. Subjects in the field test were satisfied with the color rendering of the LED mixtures. This study focuses on optimizing CRI for all of the LED mixtures. This work shows promise in using more narrow-band white light LEDs as potential lighting sources in museum environments. It is also interesting to evaluate different combinations of LEDs to examine the potential of inducing changes in the perception of the art itself. For example Berns 2010 examined maximizing chroma, perhaps with a goal of enlivening faded paintings, or minimizing illuminance or CCT, with the goal of inducing a cave or church effect for alter pieces or frescos. Cuttle 2009 states that lighting designers may appreciate the feedback that an adaptive LED lighting display might provide by allowing user independent control of

12 intensity and color temperature. This would give curators a toolbox of LED lighting designs to work with for different exhibition purposes. Methods To this end, an optimization of white-light LED clusters has been conducted. This study investigated combinations of colored LEDs to produce white-light for potential use in museum lighting environments. In order to access the trade offs and requirements impacting museum lighting a matrix, provided in Table I has been created to prioritize the criteria that would one would face when using LEDs for museum lighting. This technique, as outlined in Michalski and Rossi-Doria 2010, allows for a graphical representation of trade-offs to be assigned a value of importance with which the light sources can be compared to one another to access how well they meet the necessary criteria for the museum environment. Table I. Values are assigned to each criterion with 1 being the highest priority and 5 being the lowest priority. UV damage Overall illuminance Geometry of Lighting Color Rendering Color Temperature Can I see the art? X X X Will I be able to see the art in the future? X X X Do I have an aesthetic X X X X appreciation? What information was X X X the artist trying to convey? Are the colors well X represented? Overall priority In order to ascertain which lights would meet the above criterion first actual colored LEDs had to be procured. Spectral power distribution data were obtained from NIST for 125 visible range colored LEDs available on the market from a variety of manufacturers. The goal of this study was to create an LED equivalent to the CIE standard illuminant D65, as it is a standard daylight environment for color discrimination. It is possible to use this method for any chosen

13 illuminant including black-body illuminants by simply changing this reference illuminant. In order to calculate colorimetric properties of these light sources the spectral power distributions had to be interpolated to fit the same range as the 1931 CIE standard observer. It is then possible to calculate XYZ, tristimulus values, for all LEDs and illuminant D65 from reflectance and 2- degree 1931 observer. The LED data were normalized to a maximum illuminance value of one, so that their addition is relative to the same equiluminance point. Viénot et al 2011 also use this technique when calculating their light sources. X = 683 Y = 683 L! x! Δλ! L! y! Δλ! Z = 683 L! z! Δλ! Where L is the spectral power distribution of one LED light source, xyz are the CIE standard observer, and the Δλ is 1nm. Calculate xyz from XYZ for CIE illuminant D65 X x = X + Y + Z Y y = X + Y + Z Z z = X + Y + Z A ratio weighting of the three colored LEDs that will match D65 chromaticities is determined. β! β! β! = X! X! X!!! Y! Y! Y! Z! Z! Z! x! y! L! L! z! y! L! where: 1 = L! Subscripts n represents chromaticities of D65, Subscripts 1, 2, and 3 represent different colored LEDs for each possible 3-LED mixture.

14 The ratio or weighting values are determined by matrix math and are therefore unconstrained values. It is essential to eliminate all combinations that produce any negative β-value, as these values will not create realizable mixes, because negative light is not a realizable concept. These values can be any non-negative number as if the value surpasses one it is possible to add more of one colored LED to the mix to create more than all the light of one blub. Likewise these three LED lights are not required to be Red, Green and Blue as a traditional additive mixing might suggest. The lights were only required to be in different enough regions of the spectrum as to be classified under a different hue name. This allows us to calculate the spectral power distribution of combinations of three LEDs. spd = β! L!! + β! L!! + β! L!! Where L 1, L 2, and L 3 are the spectral power distribution of three colored LED lights and the coefficients are the ratios of these lights that achieve CIE D65 chromaticities. From this spectral power distribution it is possible to calculate color-rendering properties, as these are all calculated a color space it is not necessary to normalize the spectral power distribution. In order to evaluate the color rendering properties of these mixtures it is important to use conventional color rendering metrics (including CRI, CQS, and LER) as well as metrics more suited to an art environment. Discussion: Based on the criteria that the LED combinations be realizable there were 64,827 possible combinations of three LEDs that achieved the chromaticity of CIE illuminant D65. In order to find the combination that most reliably replicated color in different situations color rendering qualities were evaluated.

15 I. Color Rendering Metrics There are many color rendering metrics created for the evaluation of light sources. The need for a metric to describe the color rendering properties of a light source for museum lighting began as far back as the 1950s by Crawford. There are two basic classifications for color rendering metrics those that create visualizations of the color rendering quality of a light source and those that create a numeric index of the quality of the color rendering of a light source. The most common and popularly accepted color rendering metric for light sources is the Color Rendering Index and because of this fact this metric is used as a basis for comparison for these new light sources. Of late many additional color rendering metrics have been invented including the Color Quality Scale (CQS) created by Davis and Ohno 2006 which is also included in the evaluations as it uses highly chromatic colors to examine the color rendering quality of light sources. Luminous Efficacy of Radiation is another tool used to examine how well a light source converts electrical energy into light energy. This tool is useful in that it allows one to examine how well the light source emits light, which is important if the bulb needs to be energy efficient and cost effective. As this is also an industry standard our light sources will also be evaluated by this metric. Very recently a color rendering metric has been created that relies on memory colors (Smett). Though this metric may be very useful in some applications it will not be used to evaluate these light sources as this study is more concerned about the accurate rendering of the light and not the aesthetic rendering of the light source. Among the visualization metrics is the gamut area index created by Rea et al This technique examines the possible colors that each light source can display. This

16 method does not create a direct comparison between two light sources but instead creates a visualization of the available colors a light sources could render. A. E 00 The most important evaluation of these light sources was an evaluation of E 00, C * ab, and H * ab for both the conventional Gretag Macbeth color checker and a custom designed color checker for artist paints. These evaluations allow a more accurate picture of how the light source will impact a given art object. The custom designed paint target, created with Golden artist matte acrylic paints, was developed for evaluation of spectral cameras for fine art reproduction. It is composed of 54 colored patches which include high and low chroma color circles as well as two gray scales one with titanium white and the other with Titan bluff to simulate aged lead white paint, a series of metameric grey pairs, and analogues for dark and light skin colors. The target is intended to cover a wide range of colors, which might be utilized by artists in a given painting. Figure 1. Photo of the actual MCSL paint target

17 B. Color Rendering Index The most common color rending metric is the Color Rendering Index (CRI), which evaluates the color rendering of a light source by comparing it to a standard illuminant in this case D65 in W * U * V * space. This method was formalized in 1995 from previous work done by Nickerson in This is done by comparing fourteen different Munsell colored samples the first eight of which are referred to collectively as R a, which represent near neutral patches and the remaining of which represent more saturated colors (red, yellow, green blue), flesh tones (white) and leaf green under both illuminants and calculating E i. For each color patch a Special Color Rendering Indices R i is calculated by R! = E! for i = 1 to 14 The General Color Rendering Index referred to as R a is calculated by R! = An illuminant with perfect color rendering would have a CRI of 100, which represents zero difference between test patches under both illuminants.!!!! R! 8 It has been noted that CRI is not always the best metric for determining the color rendering qualities of a light source. CRI only functions well with the color fidelity of standard light sources, and does not include aspects of color quality such as color preference of chromatic discrimination (Boissard and Fontoynont 2009). Many indications that the current CRI method does not give fully reliable data for white light composed from narrow band colored LEDs (van der Burgt and van Kemenade 2010). Cuttle 2009 also states that CRI is inadequate for assessing the suitability of

18 light sources for museums and art galleries. Some noticeable drawbacks include: the reliance on just the first eight medium-saturated colors which is a noted area of problems in W * U * V * space, the limitation of not maintaining the directionality of the color shifts, the use of the outdated, minimally used, and non-perceptually uniform W * U * V * space, and the limitation of not accounting for the shift in chromaticity coordinates across the Planckian locus (Ohno, 2004). It is clear that the limitations of CRI are great, however, it is still the accepted standard for color rendering in the lighting community. C. Color Quality Scale These for these reasons NIST developed a new metric for evaluating colorrendering properties. This new metric is called Color Quality Scale (CQS) (Davis and Ohno, 2006). This metric utilizes fifteen saturated Munsell samples, which were selected to have high chroma and that encompass the entire hue circle at approximately even spacing. This metric is based in CIELAB space and does not penalize a light source for increases in chroma; an issue many LED light sources face. CQS also penalizes light sources that have very high and very low CCT, as these light sources tend to be very colored light sources. CRI relies on an average to calculate ΔE, where as, CQS calculates ΔE with root-mean-square (RMS), which guarantees that large hue shifts of any sample impacts the overall CQS score of the light source. This metric is also limited to between Lawrence Taplin 10/5/11 8:20 AM Comment [1]: RMS distinction what does this mean magnitude? zero and one hundred to it is easily comparable to CRI. D. Luminous Efficacy of Radiation The lighting industry also relies heavily on a metric referred to as the luminance efficacy of radiation or LER, which evaluates the energy efficiency of a light source. This

19 represents the ratio of luminous flux to radiant flux produced by a given light source. It is calculated by LER = 683 lm W!! V(λ)S(λ)dλ S(λ)dλ Where S (λ) is the spectral power distribution of the light source and V (λ) is the 1924 CIE standard photometric observer. II. Light Sources for Specific Purposes A. Arbitrary Example of a Poor Light Source for all metrics It is clear from this work that the vast majority of combinations will have poor color rendering qualities for all color rendering metrics. An example of a poor color rendering illuminant is given below to illustrate how the visualizations help one to evaluate the color rendering properties of a given light source. Table II. Listing of common color quality metrics for the new LED light source and CIE illuminant D65 for comparison Color Quality metric Poor rendering LED cluster CIE illuminant D65 Ra CQS LER wavelength(nm) Figure 2. The computational spectral power distribution of a poor LED mixture for color rendering.

20 It is clear from the spectral power distribution that this light source creates a D65 metamer by using a large portion of red light. Therefore one could surmise that this light source would likely tint achromatic samples slightly red. Figure 3 below shows this clearly. Though it is important to note that the grey corner patches to not significantly change in color. Likewise both grey-scales maintain close to neutral color, both white patches are still white for both the reference D65 light source and the new poor color rendering light source. It is clear from Figure 4 that this illuminant not only changes the hue of most of the color patches including the grey metamer pairs but it also changes the chroma of most of the patches to be more chromatic. This increase in chroma is typical of many LED clusters. These same visualizations will be performed for all the illuminants discussed in this paper as well as visualizations for the Gretag Macbeth Color Checker. Figure 3. This is a rendering of the Munsell Color Science Lab paint target with CIE illuminant D65 on the top left triangle and the poor LED illuminant in the bottom right triangle rendered in srgb space.

21 change of Lab for MCSL color checker b* 0 L* a* Figure 4. These arrows represent the change in direction and magnitude from CIE illuminant D65 to the poor LED illuminant in CIELAB space for each colored patch on the Munsell Color Science Lab paint target. B. The Optimal Light Source for all metrics Using these methods an optimal light source was found that exhibited both good conventional color-rendering qualities and which reduced E 00, from a reference light source of CIE illuminant D65, for both the Gretag Macbeth color checker and Munsell Color Science Lab s newly created paint color checker. This illuminant utilizes a Yellow phosphor LED, Green LED, and Blue LED in specific ratios to obtain 6503K CCT and the chromaticities of D C* ab Best set of three colored LEDs for color rendering wavelength (nm) Figure 5. The computational spectral power distribution of the optimal 3 LED cluster for color rendering.

22 It is important to note that the spectral power distribution shown in Figure 5 has very little intensity at high energy (400nm and shorter wavelengths). This is an important characteristic to use in museum lighting environments. Likewise, the intensity in the near infrared range is very low as well which implies low heat production. Figure 6. This is a rendering of the Munsell Color Science Lab paint target with CIE illuminant D65 on the top left triangle and optimal LED illuminant on the bottom right triangle rendered in srgb space. change of Lab for MCSL color checker b* 0 L* a* C* ab Figure 7. These arrows represent the change in direction and magnitude from CIE illuminant D65 to the optimal LED illuminant in CIELAB space for each colored patch on the Munsell Color Science Lab paint target.

23 Figure 8. This is a rendering of the Gretag Macbeth color checker with CIE illuminant D65 on the top left triangle and optimal LED illuminant on the bottom right triangle rendered in srgb space. change of Lab for MacBeth color checker b* 0 L* a* C* ab Figure 9. These arrows represent the change in direction and magnitude from CIE illuminant D65 to the optimial LED illuminant in CIELAB space for each colored patch on the Gretag Macbeth color checker. Figures 6 and 8 demonstrate a direct comparison of CIE illuminant D65 (top triangle) and the new illuminant (bottom triangle). The largest differences for both targets are for the high chroma red color patch. This is likely due to the absence of a distinctly red LED from the optimized light source, thereby making red patches appear slightly less red.

24 Figures 5 and 7 demonstrate the difference between CIE illuminant D65 and the optimized light source when rendering both the new Munsell Color Science Lab target and the traditional Gretag Macbeth color checker respectively. Neither target shows a significant changes in chroma, lightness, or hue for any color. There are also not significant trends to the variations in color. Table III Examines the traditional color rendering qualities of this light source. For comparison CIE illuminant D65 is also displayed. Table III. Listing of common color quality metrics for the new LED light source and CIE illuminant D65 for comparison Color Quality metric Best color rendering LED cluster CIE illuminant D65 Ra CQS LER C. Light Source for increasing Chroma Another goal of these LED evaluations was to create illuminants, which would artificially induce an increase or decrease in chroma. This would artificially fade in the case of paintings where the artist may have intended a work of art to be in a lighting environment that is dim or enhance a painting that may have faded. Recent work have illustrated that it is possible to create LED clusters to enhance faded objects (Viénot 2011, Žukauskas et al 2009, Žukauskas et al 2010, Žukauskas et al 2011a, and Žukauskas et al 2011b) as well as dull objects which may have been intended for a dim environment. Viénot does not attempt to quantify this change with a metric but relies of visualizations and imaging. Žukauskas et al 2009 created of yet another set of color rendering metrics Color saturation index (CSI), Color dulling index (CDI), Color Fidelity Index (CFI), and Hue distortion index (HDI), which examine the saturation level and color rendering quality of solid-state light sources. These metrics depend on

25 MacAdam ellipses to evaluate the color rendering properties of a given light source while comparing the reflectance of Munsell color chips. For the purposes of this study, three LED combinations were created which either increased or decreased chroma in CIELAB space while maintaining as small a hue angle shift as was possible for both the Gretag Macbeth color checker and the new MCSL target. Due to the large shifts in chroma the traditional color rendering metrics do not perform well in this task as they require the color rendering of light sources to be as close to a reference light source like CIE illuminant D65 or CIE illuminant A as possible. Therefore those these metrics have been calculated but do not indicated the quality of these light sources. It is important to ensure that light sources enhanced or reduced chroma in as uniform a way as possible. In order to achieve this a covariance matrix was calculate from the twelve maximum chroma samples on the new MCSL target for both CIE illuminant D65 and for the new illuminants. If the covariance of a * and b * for CIE illuminant D65 and the new light source were either both positive or both negative these light sources are said to decrease or increase chroma respectively. However, this alone does not address the uniformity of the chroma shift over the whole hue circle. In order to address this issue the magnitude of the difference between the covariance of a * and b * between CIE illuminant D65 and the new light source should be as small as possible. In order to further evaluate the magnitude of this change Principle Components Analysis was performed on all the applicable new illuminants and was compared to that of illuminant D65. These ellipsoids were all centered at the same location and therefore it is only necessary to evaluate the change in angle, which evaluates if the new light source changes the direction of the major and minor axes. The

26 magnitude of the major and minor axes also needs to be evaluated with the goal that the magnitude of difference of the major and minor axes should be as close to zero as possible. With this criterion light sources, which increase or decrease in chroma can be examined and evaluated as only changing chroma while maintaining the hue of the original samples. Increased Chroma LED mixture spectral power distribution wavelength (nm) Figure 10. The computational spectral power distribution of the optimal 3 LED mixture that enhances chroma. Figure 11. This is a rendering of the Munsell Color Science Lab paint target with CIE illuminant D65 on the top left triangle and enhancing chroma LED illuminant on the bottom right triangle rendered in srgb space.

27 change of Lab for MCSL color checker b* 0 L* a* C* ab Figure 12. These arrows represent the change in direction and magnitude from CIE illuminant D65 to the enhancing chroma LED illuminant in CIELAB space for each colored patch on the Munsell Color Science Lab paint target. From figure 12 it is clear that the yellow color patch and light blue color patch show a slight decrease in chroma but that all other colors show an increase in chroma with out a significant change in hue as is clear in the rendering of the MCSL paint target. From figure 11 it is interesting to note that the row of metameric greys show a large amount of variation, which is not seen in either the titanium white or Titan bluff (to simulate aged lead white paint) grey scales. Figure 13. This is a rendering of the Gretag Macbeth color checker with CIE illuminant D65 on the top left triangle and enhancing chroma LED illuminant on the bottom right triangle rendered in srgb space.

28 change of Lab for MacBeth color checker b* 0 L* a* C* ab Figure 14. These arrows represent the change in direction and magnitude from CIE illuminant D65 to the enhancing chroma LED illuminant in CIELAB space for each colored patch on the Gretag Macbeth color checker. Figure 13 demonstrates changes in chroma that are less apparent in the Gretag Macbeth color checker than in the MCSL paint target. However, in Figure 14 it is clear that the yellow color patch now increases in chroma and the blue color patches do not decrease in chroma as they had in the MCSL paint target. For comparison Table IV shows the best 3LED light source, CIE illuminant D65 and the new increase of chroma LED combinations. It is clear that the color rendering qualities of this source perform worse, with the exception of LER, which performs better. This demonstrates that this light source does indeed change the colors displayed but does not reduce efficacy. This light source is not intended to render colors exactly as they appear under a reference light source but to enhance their chroma without a significant change in hue. Table IV. Listing of common color quality metrics for the increasing chroma LED light source and CIE D65 for comparison Color Quality metric Increasing chroma New light source CIE illuminant D65 Ra CQS LER

29 D. Light Source for Decreasing Chroma The other goal of this process was to create a lighting design, which would decrease chroma or artificially fade a painting. This lighting design would be valuable for only specific situations in which artist s intent may be a goal. For example, perhaps an artist creates an alter piece for a church where the lighting condition is typically candle light, and now this alter piece is being displayed in a museum where the lighting makes the colors appear much brighter than they were originally intended. This lighting design would allow an exhibitor to create lighting closer to the intended lighting condition. This might also be helpful when displaying cave paintings for anthropological displays. x 10 3 Decrease Chroma LED combo wavelength (nm) Figure 15. The computational spectral power distribution of the optimal 3 LED cluster that decreases chroma.

30 Figure 16. This is a rendering of the Munsell Color Science Lab paint target with CIE illuminant D65 on the top left triangle and decreasing chroma LED illuminant in the bottom right triangle rendered in srgb space. From figure 15 it is clear that the spectral power distribution of this light is centered toward the green and yellow portions of the spectrum. Likewise from Figure 16 and 17 that this light source decreases chroma of all colors with the exception of blues and yellows. As most paper yellows with age this aspect maybe less of a concern. The dramatic decreases in red and green should perceptually fade most paintings. However, psychophysical testing will be required to see if these differences are in fact noticeable. change of Lab for MCSL color checker b* 0 L* a* C* ab Figure 17. These arrows represent the change in direction and magnitude from CIE illuminant D65 to the decreasing chroma LED illuminant in CIELAB space for each colored patch on the Munsell Color Science Lab paint target.

31 Figure 18. This is a rendering of the Gretag Macbeth color checker with CIE illuminant D65 on the top left triangle and decreasing chroma LED illuminant in the bottom right triangle rendered in srgb space. change of Lab for MacBeth color checker b* 0 L* a* C* ab Figure 19. These arrows represent the change in direction and magnitude from CIE illuminant D65 to the decreasing chroma LED illuminant in CIELAB space for each colored patch on the Gretag Macbeth color checker. From Figures 18 and 19 it is clear that the increases in chroma in the yellow and blue color patches do not hold for the Gretag Macbeth color checker. This may indicate that only certain yellow and blue paint samples will increase in chroma. Again psychophysical testing will be required to test if this light sources performs well perceptually. The color rendering qualities

32 given in IV show very poor performance for this light source both CRI and CQS penalize light sources which decrease chroma where as only CRI penalize light sources which increase chroma. However, this light source again increases in efficacy, implying that this light source does have some value for the purpose for which it was designed. Table V. Listing of common color quality metrics for the new LED light source and CIE illuminant D65 for comparison Color Quality metric Decreasing Increasing chroma New light source CIE illuminant D65 Chroma Ra CQS LER Conclusions: This investigation has shown there are certainly possibilities for colored LED combinations to achieve accurate color rendering. There are also possibilities of using colored LED combinations to create novel lighting designs for specific display purposes. These sources have little to no UV radiation as well as a limited IR radiation, which makes them ideal for museum lighting environments. It will be interesting to evaluate this technique with other CCTs in the future and to examine if CCT does in fact impact observer desirability. As these light sources were all designed to match most closely to CIE illuminant D65, it is possible that it will be more difficult to find LED combinations at other CCTs, which have good color rendering qualities. There are certainly more avenues to investigate LED clusters. Then next step would be to investigate higher order (4,5 or 6) combinations. It may also increase the variability and performance of LED clusters to consider adding white phosphor coated LEDs to evaluation process (if high quality data is available for these sources).

33 It will be important to create a lighting environment in the future where psychophysical testing can be performed to test if the computational projections match observer opinion. It would be vital to not only test curators but also to test museum patrons to determine if these lighting conditions were indeed ideal for these uses. Works Cited: Related to LEDs and museum lighting: Berns RS. Designing White-Light LED Lighting for the Display of Art: A Feasibility Study. Color Research and Application 2010; online Boissard S, Fontoynont M. Optimization of LED-based Light Blendings or Object Presentation. Color Research and Application 2009; 34:4, Brodrick J. Can Museums Measure Up? Gateway Roundup. July Chalmers A, Soltic S. Towards the Optimum Light Source Spectrum. Advances in OptoElectronics. 2010; 2010:596825, 1-9 Cuttle C. New Opportunities for LEDs in Museum Lighting. In Proceedings of the Professional Lighting Design Convention October 2009; Berlin, Germany; US DOE Solid-State Lighting Design GATEWAY program. Demonstration Assessment of Light-Emitting Diode (LED) Accent Lighting at the Field Museum in Chicago IL. 2010:PNNL Related to LEDs as light sources: van der Burgt P, van Kemenade J. About Color Rendition of Light Sources: The Balance Between Simplicity and Accuracy. Color Research and Application 2010; 35:2, He G, Zheng L. Color Temperature Tunable White-Light Light-Emitting Diode Clusters with High Color Rendering Index. Applied Optics 2010; 49:2, Mahler E, Ezrati J.J, Viénot F. Testing LED Lighting for Colour Discrimination and Colour Rendering. Color Research and Application 2008; 34:1, Ohno Y. Color Rendering and Luminous Efficacy of White LED Spectra. Fourth International Conference on Solid State Lighting; 2004; Denver Colorado; SPIE; Bellingham, Washington. Proc. of SPIE; 5530, Ohno Y.Spectral Design Considerations for White LED Color Rendering. Optical Engineering :11,

34 Vienot F, Mahler E, Serreault L, and Margalith H. Discriminating Colours under LED illumination. In Proceedings of 10 th Congress of the International Colour Association, 2005; Granada, Spain; Žukauskas A,Vaicekauskas R, and Shur MS. Colour-rendition Properties of Solid-State Lamps. Journal of Physics D:Applied Physics 2010; 43:354006, Žukauskas A,Vaicekauskas R, and Shur MS. Solid-state lamps with optimized color saturation ability. Optics Express 2011a; 18:3, Related to Museum lighting: CIE Technical Report 157: 2004 Control of Damage to Museum Objects by Optical Radiation Crawford BH. Measurement of Color Rendering Tolerances. Journal of the Optical Society of America 1959; 49:12, Crawford BH. Colour Rendition and Museum Lighting. Studies in Conservation 1960; 5:2, Crawford BH. Further Investigations of Colour Rendering, and the classification of Light Sources. Studies in Conservation 1961; 6:2 and 3, Crawford BH. The Color Rendering Properties of Illuminants: the Application od Psychophysical measurements to their evaluation. British Journal of Applied Physics 1963; 14:6, Crawford BH. Just Perceptible Color Differences in Relation to Level of Illumination. Studies in Conservation 1973; 18:4, Cuttle C. A Proposal to Reduce the Exposure to Light of Museum Objects Without Reducing Illuminance or the Level of Visual Satisfaction of Museum Visitors. Journal of the American Institute for Conservation 2000; 39:2, Cuttle C. Lighting works of Art for Exhibition and Conservation. Lighting Research and Technology 1988; 20:2, Delgado M, Dirk C, Druzik J and WestFall N. Lighting the World s Treasures: Approaches to Safer Museum Lighting. Color Research and Application 2011; 36:4, Druzik J, Eshøj B. Museum Lighting: Its Past and Future Development. Presented at Conference on Museum Microclimates. 2007; Copenhagen, Denmark; Loe DL, Rowlands E, and Watson NF. Preferred Lighting Conditions for the Display of Oil and Watercolour Paintings. Lighting Research and Technology 1982; 14:4,