New Method for Evaluating Light Source Color Rendition (IES TM-30-15)

New Method for Evaluating Light Source Color Rendition (IES TM-30-15) IES México XVII Seminario de Iluminación May 18, 2016 Kevin W. Houser, PhD, PE, FIES Professor of Architectural Engineering The Pennsylvania State University khouser@engr.psu.edu Editor-in-Chief LEUKOS, the journal of IES khouser@ies.org

Which do you prefer? 1 2 CCT = 3501 K D uv = 0.0000 R a (CRI)= 50 R 9 = -80 CCT = 3501 K D uv = 0.0000 R a (CRI)= 75 R 9 = 20

Today s Outline Brief overview of CIE CRI Introduction to TM-30-15 IES Method for Evaluating Light Source Color Rendition (Including Excel Software) Familiar source examples and demonstrations Results from recent experiment

CIE CRI (R a ) Test Source Reference Illuminant (approximately) SAME CCT For further reading see CIE 13.3-1995, or Houser K, Mossman M, Smet K, Whitehead L. 2015. Tutorial: Color Rendering and Its Applications in Lighting. LEUKOS. http://dx.doi.org/10.1080/15502724.2014.989802

CIE CRI (R a ) Approximation of Color Samples for R a Color Samples for R 9 R 14 TCS 01 TCS 02 TCS 03 TCS 04 TCS 09 TCS 10 TCS 11 TCS 12 TCS 05 TCS 06 TCS 07 TCS 08 TCS 13 TCS 14

CIE CRI (R a ) R Y GY G BG BP P RP (Illustration Only) +20 GY +10 G Y V* R -10 BG -20 PB P RP U* +30-20 -10 +10 +20 +30

CIE Method for Color Rendering Color Fidelity The accurate rendition of color so that they appear as they would under familiar (reference) illuminants CIE CRI (R a )

CRI = 95, Original Image Original Image courtesy of Randy Burkett Lighting Design

CRI = 80, Desaturated Image Original Image courtesy of Randy Burkett Lighting Design

CRI = 80, Saturated Image (Red Enhanced) Original Image courtesy of Randy Burkett Lighting Design

Original Baseline Original image courtesy of Randy Burkett Lighting Design

CRI = 80 - Hue Shift

CRI = 80 + Hue Shift

CRI = 80 Saturated

CRI = 80 Desaturated

Limitations of Considering Only Fidelity Positive Hue Shift Constant CRI Decrease Saturation Perfect Fidelity CRI = 80 CRI = 80 Increase Saturation Negative Hue Shift

Limitations of Considering Only Fidelity Positive Hue Shift Decrease Saturation Constant CRI One measure is not enough! Perfect Fidelity CRI = 80 CRI = 80 Increase Saturation Negative Hue Shift

CRI (R a ): A measure of average color fidelity. But what about. saturation changes? hue shifts? color discrimination? color preference?

One index is not enough. But how many are needed? And what should they be? Attributes of Color Rendition include: Color Fidelity Color Discrimination Color Preference Tend to be related to saturation, which can be quantified with gamut Sidebar for Further Reading: The more than 25 indices of color rendition that appear in the scientific literature tend to cluster into two categories, those based on comparison to a reference illuminant (i.e., to quantify fidelity), and those related to gamut area (i.e., to quantify increase or decrease in saturation).* * Houser KW, Wei M, David A, Krames MR, Shen XS. Review of Measures for Light-Source Color Rendition and Considerations for a Two-Measure System for Characterizing Color Rendition. Optics Express. 2013; 21(8);10393-10411. http://dx.doi.org/10.1364/oe.21.010393

Two primary motivations for developing the IES Method: 1. The need for an improved measure of color fidelity 2. The need to provide supplementary information about color rendering ability of any given light source

CIE CRI (1965/1974) IES TM-30-15 (2015) CIE 1964 U*V*W* CAM02-UCS (CIECAM02) 8 color samples 99 color samples Medium chroma/lightness Spectral sensitivity varies Munsell samples only Uniform color space coverage Spectral sensitivity neutral Variety of real objects Fidelity Metric Only Fidelity, Gamut, Graphical, Detailed Ref Illuminant Step Function No lower limit for scores Ref Illuminant Continuous (Uses same reference sources, but blended between 4500 K and 5500 K) 0 to 100 scale (fidelity)

IES Method for Color Rendition High Level Average Values Fidelity Index (R f ) Gamut Index (R g ) Core Calculation Engine Modern Color Science New Color Samples Graphical Representations Color Vector Graphic Color Distortion Graphic Detailed Values Skin Fidelity (R f,skin ) Fidelity by Hue (R f# ) Chroma Shift by Hue (R c# ) Fidelity by Sample (R f,ces# )

IES Method for Color Rendition Color Fidelity Color Gamut Graphics The accurate rendition of color so that they appear as they would under familiar (reference) illuminants Fidelity Index (R f ) (0-100) The average level of saturation relative to familiar (reference) illuminants. Gamut Index (R g ) ~60-140 when R f > 60 Visual description of hue and saturation changes. Color Vector Graphic

b' Fidelity Index: R f 40 30 20 10 0-10 -20 Average similarity in appearance of test and reference sources Analogous to CIE R a, greater accuracy Scores of 0 to 100 Scale similar to CIE R a, but high scores harder to achieve Equal weight to all directions of shift Should not be expected to correlate with any single object color -30-40 -40-30 -20-10 0 10 20 30 40 a' Reference Source Test Source [Flattened to 2D]

b' b' Relative Gamut Index: R g 40 6 5 4 3 40 6 5 4 3 30 20 7 2 30 20 7 2 10 0 8 1 10 0 8 1-10 9 16-10 9 16-20 -20-30 10-40 11 12 13 14-40 -30-20 -10 0 10 20 30 40 a' Reference Source Test Source 15-30 10-40 11 12 13 14-40 -30-20 -10 0 10 20 30 40 a' Reference Source Test Source 15

b' Relative Gamut Index: R g R g = 100 A t A r 40 30 20 7 6 5 4 3 2 R g > 100: Average increase in saturation R g < 100: Average decrease in saturation 10 0 8 1-10 9 16-20 -30 10-40 11 12 13 14-40 -30-20 -10 0 10 20 30 40 a' Reference Source Test Source 15

Theoretical Example Original Desaturated Red-Enhanced CRI = 95 CRI = 80 CRI = 80 R f = 93 R f = 78 R f = 78 R g = 100 R g = 90 R g = 110 Original Image courtesy of Randy Burkett Lighting Design

Theoretical Example Average values can hide important information! Original Desaturated Red-Enhanced CRI = 95 CRI = 80 CRI = 80 This is limitation of CIE R a, and IES R f and R g R f = 93 R f = 78 R f = 78 R g = 100 R g = 90 R g = 110 Image courtesy of Randy Burkett Lighting Design

Color Vector Graphic Gamut is not a dimension of perception. It is best interpreted with reference to a complementary graphic. 380 430 480 530 580 630 680 730 780 Color Vector Graphic Color Distortion Graphic R f = 81 R g = 101 CCT = 2496 K R a = 88 (Source No. 286)

b' Color Vector Graphic COLOR VECTOR GRAPHIC 40 30 20 7 CES CHROMATICITY COMPARISON 6 5 4 3 2 10 0-10 8 9 1 16-20 -30 10-40 11 12 13 14-40 -30-20 -10 0 10 20 30 40 a' Reference Source Test Source 15

b' Color Vector Graphic 40 30 20 10 0-10 7 8 9 CES CHROMATICITY COMPARISON 6 5 4 3 2 1 16-20 -30 10-40 11 12 13 14-40 -30-20 -10 0 10 20 30 40 a' Reference Source Test Source 15

b' Color Vector Graphic 40 30 20 10 0-10 7 8 9 CES CHROMATICITY COMPARISON 6 5 4 3 2 1 16 Increased Saturation Decreased Saturation Hue Shift -20-30 10-40 11 12 13 14-40 -30-20 -10 0 10 20 30 40 a' Reference Source Test Source 15

Theoretical Example Original Desaturated Red-Enhanced CRI = 95 CRI = 80 CRI = 80 R f = 93 R f = 78 R f = 78 R g = 100 R g = 90 R g = 110 Original Image courtesy of Randy Burkett Lighting Design

Excel TM-30-15 Calculation Tool

CES01 CES04 CES07 CES10 CES13 CES16 CES19 CES22 CES25 CES28 CES31 CES34 CES37 CES40 CES43 CES46 CES49 CES52 CES55 CES58 CES61 CES64 CES67 CES70 CES73 CES76 CES79 CES82 CES85 CES88 CES91 CES94 CES97 Halogen (MR16) TM-30 Library Source No. 80 R f = 99 R g = 99 R f,skin = 99 R f,h1 = 98 R cs,h1 = -1% R a = 99 R 9 = 93 CCT = 2988 K D uv = 0.0010 LER = 180 Fidelity Index by Sample, R f,cesi 100 90 80 70 60 50 40 30 20 10 0

CES01 CES04 CES07 CES10 CES13 CES16 CES19 CES22 CES25 CES28 CES31 CES34 CES37 CES40 CES43 CES46 CES49 CES52 CES55 CES58 CES61 CES64 CES67 CES70 CES73 CES76 CES79 CES82 CES85 CES88 CES91 CES94 CES97 High Pressure Sodium TM-30 Library Source No. 56 R f = 32 R g = 61 R f,skin = 34 R f,h1 = 5 R cs,h1 = -48% R a = 17 R 9 = -225 CCT = 1971 K D uv = -0.0001 LER = 382 Fidelity Index by Sample, R f,cesi 100 90 80 70 60 50 40 30 20 10 0

CES01 CES04 CES07 CES10 CES13 CES16 CES19 CES22 CES25 CES28 CES31 CES34 CES37 CES40 CES43 CES46 CES49 CES52 CES55 CES58 CES61 CES64 CES67 CES70 CES73 CES76 CES79 CES82 CES85 CES88 CES91 CES94 CES97 Neodymium Incandescent TM-30 Library Source No. 88 R f = 86 R g = 109 R f,skin = 84 R f,h1 = 78 R cs,h1 = 11% R a = 77 R 9 = 15 CCT = 2756 K D uv = -0.0048 LER = 136 Fidelity Index by Sample, R f,cesi 100 90 80 70 60 50 40 30 20 10 0

CES01 CES04 CES07 CES10 CES13 CES16 CES19 CES22 CES25 CES28 CES31 CES34 CES37 CES40 CES43 CES46 CES49 CES52 CES55 CES58 CES61 CES64 CES67 CES70 CES73 CES76 CES79 CES82 CES85 CES88 CES91 CES94 CES97 Linear Fluorescent F32T8/835 TM-30 Library Source No. 38 R f = 75 R g = 99 R f,skin = 84 R f,h1 = 74 R cs,h1 = -12% R a = 79 R 9 = -5 CCT = 3563 K D uv = -0.0002 LER = 349 Fidelity Index by Sample, R f,cesi 100 90 80 70 60 50 40 30 20 10 0

CES01 CES04 CES07 CES10 CES13 CES16 CES19 CES22 CES25 CES28 CES31 CES34 CES37 CES40 CES43 CES46 CES49 CES52 CES55 CES58 CES61 CES64 CES67 CES70 CES73 CES76 CES79 CES82 CES85 CES88 CES91 CES94 CES97 PC White LED TM-30 Library Source No. 184 R f = 81 R g = 94 R f,skin = 86 R f,h1 = 75 R cs,h1 = -13% R a = 81 R 9 = 0 CCT = 3429 K D uv = 0.0001 LER = 332 Fidelity Index by Sample, R f,cesi 100 90 80 70 60 50 40 30 20 10 0

CES01 CES04 CES07 CES10 CES13 CES16 CES19 CES22 CES25 CES28 CES31 CES34 CES37 CES40 CES43 CES46 CES49 CES52 CES55 CES58 CES61 CES64 CES67 CES70 CES73 CES76 CES79 CES82 CES85 CES88 CES91 CES94 CES97 Hybrid LED (PC+Red) TM-30 Library Source No. 92 R f = 89 R g = 105 R f,skin = 97 R f,h1 = 91 R cs,h1 = -1% R a = 94 R 9 = 89 CCT = 2776 K D uv = 0.0011 LER = 336 Fidelity Index by Sample, R f,cesi 100 90 80 70 60 50 40 30 20 10 0

CES01 CES04 CES07 CES10 CES13 CES16 CES19 CES22 CES25 CES28 CES31 CES34 CES37 CES40 CES43 CES46 CES49 CES52 CES55 CES58 CES61 CES64 CES67 CES70 CES73 CES76 CES79 CES82 CES85 CES88 CES91 CES94 CES97 RGB LED TM-30 Library Source No. 108 R f = 80 R g = 114 R f,skin = 81 R f,h1 = 70 R cs,h1 = 15% R a = 71 R 9 = -27 CCT = 3906 K D uv = 0.0027 LER = 299 Fidelity Index by Sample, R f,cesi 100 90 80 70 60 50 40 30 20 10 0

Existing Sources IES TM-30 R g 140 130 120 110 100 90 Phosphor LED Color Mixed LED Hybrid LED Standard Halogen Filtered Halogen Triphosphor Fluorescent, 7XX Triphosphor Fluorescent, 8XX Triphosphor Fluorescent, 9XX Metal Halide 80 70 60 50 60 70 80 90 100 IES TM-30 R f

Demonstration Live demonstration of SPDs realized using the 7-channel ETC D22, illustrating variations in IES R f, R g, and CIE R a and R 9. The below observations can be partially understood by examination of the 8 slides that follow. What to look for? Source 7 Source 1 Very close to reference conditions Despite a CIE R a of 50 (and IES R f of 64), many would not find this source objectionable because of the manner in which it enhances gamut. 1 vs 2 Despite a 25 point difference in CIE R a, the higher CIE R a source is clearly less desirable 2 vs 3 Despite similar CIE R a and R 9, color rendering is very different. The IES R f and R g measures, plus the graphics, provide more appropriate information. 3 vs 4 Both have similar IES R f and R g, but render objects differently. This illustrates the importance of the color vector graphic. 4 vs 5 Same IES R f, but source 5 has an IES R g that is greater than source 4 by 15 points. Note that source 5 appears more desirable than source 4, despite the fact that CIE R a is 11 points lower. Red rendition is very different, even though CIE R 9 is similar for both. 5 vs 6 IES R f is approximately 80 for both, but source 6 has IES R g of 87 (desaturating), versus IES R g of 115 (saturating) for source 5. Note that CIE R a is higher for the desaturating source (which makes object appear less desirable), yielding a result that is different than many would expect.

Human Judgements of Color Rendition Vary with Average Fidelity, Average Gamut, and Gamut Shape Michael Royer, Pacific Northwest National Laboratory Andrea Wilkerson, Pacific Northwest National Laboratory Minchen Wei, Hong Kong Polytechnic University Kevin Houser, Penn State University Robert Davis, Pacific Northwest National Laboratory Funding Royer, Wilkerson, and Davis supported by U.S. Department of Energy Laboratory Directed Research and Development (LDRD) award Houser subcontracted by Pacific Northwest National Laboratory. Wei supported by stipend through Penn State, with later stages supported by Hong Kong Polytechnic.

Relate judgements of color quality to TM-30 measures 140 130 120 110 R g 100 90 80 70 60 50 60 70 80 90 100 Color Vector Graphic analyses and plots courtesy of Tony Esposito, PhD R f Candidate, Penn State University. Based on simulations using 11-channel LED Cube. Goals Hypotheses Methods Results Discussion Conclusions

a priori hypotheses 1. As Rf increases, color would be judged as more normal. 2. As Rg increases, color would be judged as more saturated. 3. Higher levels of Rg would be more preferred than lower levels of Rg. 4. Higher levels of red saturation would be preferred. Goals Hypotheses Methods Results Discussion Conclusions

Apparatus and Test Space Goals Hypotheses Methods Results Discussion Conclusions

Independent Variables: Rf, Rg, and Gamut Shape 130 10 120 10 18 110 8 16 23 26 R g 100 6 14 22 26 90 4 12 20 2 80 2 70 60 70 80 90 100 R f Goals Hypotheses Methods Results Discussion Conclusions

Independent Variables: Rf, Rg, and Gamut Shape 130 120 110 10 9 8 7 18 17 16 15 23 24 R g 100 6 5 14 13 22 21 26 25 90 4 3 12 11 20 19 80 2 1 70 60 70 80 90 100 R f Goals Hypotheses Methods Results Discussion Conclusions

Independent Variables: Rf, Rg, and Gamut Shape 130 13 120 110 R g 100 90 14 13 14 80 70 60 70 80 90 100 R f Goals Hypotheses Methods Results Discussion Conclusions

Participants 28 participants (12 male, 16 female) 19 to 65 years of age (mean male = 44, mean female = 38) Ishihara 24 plate test revealed one redgreen deficient male, not excluded. Goals Hypotheses Methods Results Discussion Conclusions

Dependent Measures a priori hypotheses 1. As Rf increases, color would be judged as more normal. 2. As Rg increases, color would be judged as more saturated. 3. Higher levels of Rg would be more preferred than lower levels of Rg. 4. Higher levels of red saturation would be preferred. Normal Saturated Like 1 2 3 4 5 6 7 8 Shifted 1 2 3 4 5 6 7 8 Dull 1 2 3 4 5 6 7 8 Dislike Goals Hypotheses Methods Results Discussion Conclusions

Procedures Pre-Experimental Preparation Informed Consent Ishihara Color Screening White Lab Coat Experimental Trials 3 practice trials (2 announced) 26 experimental trials (walk and look within room, await cue from experimenter, complete survey, step out, exchange survey forms, repeat). Goals Hypotheses Methods Results Discussion Conclusions

Preference varied systematically. Higher levels of Rg were generally preferred to lower levels of Rg. p = 0.000 Dislike 130 Model r 2 = 0.68 a priori hypotheses 5.5 5.0 4.5 4.0 IES TM-30 R g 120 110 100 90 1 2 3 3. Higher levels of Rg would be more preferred than lower levels of Rg. 4. Higher levels of red saturation would be preferred. 3.5 80 Like 70 60 70 80 90 100 IES TM-30 R f p = 0.042 Goals Hypotheses Methods Results Discussion Conclusions

Preference varied systematically. Higher levels red saturation were preferred. 1 2 3 (These aren t necessarily the most preferred sources possible, just the most preferred sources from this experiment). Goals Hypotheses Methods Results Discussion Conclusions

Same fidelity and gamut, but different gamut shape, can lead to significantly different preference. p = 0.000 Dislike 5.5 130 120 Model r 2 = 0.68 5.0 4.5 4.0 IES TM-30 R g 110 100 90 3.5 80 Like 70 60 70 80 90 100 IES TM-30 R f p = 0.042 Goals Hypotheses Methods Results Discussion Conclusions

Same fidelity and gamut, but different gamut shape, can lead to significantly different preference. Goals Hypotheses Methods Results Discussion Conclusions

Preference increased with red-saturation, with limits. Mean Preference Rating 8 7 Dislike y = 85.457x 3 + 12.746x 2-9.6207x + 4.1387 R² = 0.8132 6 5 4 3 2 Like 1-30% -20% -10% 0% 10% 20% 30% Hue Bin 16 Chroma Shift (R cs,h16 ) Goals Hypotheses Methods Results Discussion Conclusions

Participant Preference Rating Post-hoc modeling of preference 7 6 Less Liked 5 4 3 2 More Liked R² = 0.9355 2 3 4 5 6 7 TM-30 Model Predicted Preference Rating Best Model for Preference: Like-Dislike = 7.396-0.0408(R f ) + 103.4(R cs,h163 ) - 9.949(R cs,h16 ) Goals Hypotheses Methods Results Discussion Conclusions

What about existing light sources? 50% 40% 30% 20% 10% Experimental Preferred Zone* R cs,h16 0% -10% -20% -30% -40% -50% Goals Hypotheses Methods Results Discussion Conclusions

What about existing light sources? IES TM-30 R g 140 130 120 110 100 90 Experimental Preferred Zone* Phosphor LED Color Mixed LED Hybrid LED Standard Halogen Filtered Halogen Triphosphor Fluorescent, 7XX Triphosphor Fluorescent, 8XX Triphosphor Fluorescent, 9XX Metal Halide 80 70 60 50 60 70 80 90 100 IES TM-30 R f Goals Hypotheses Methods Results Discussion Conclusions

Results from this small study When combined in regression models, the TM-30 measures demonstrated excellent correlation with participant evaluations Preference model r 2 = 0.94 Normalness model r 2 = 0.83 (not discussed) Saturation model r 2 = 0.95 (not discussed) Because of gamut shape, visually detectable differences in R f and R g are as little as zero points. Sources that increased saturation in red were liked (These data suggest R CS,H16 of about 2% to 16%) Commercially available sources are unlikely to be optimized for preference (at least partially due to the lack of appropriate optimization tools) Goals Hypotheses Methods Results Discussion Conclusions

Acknowledgments The slides in this presentation include images, ideas, and contributions from: Randy Burkett, Randy Burkett Lighting Design Tony Esposito, Penn State University Michael Royer, Pacific Northwest National Laboratory The experiment was performed by: Michael Royer, Pacific Northwest National Laboratory (Principal Investigator and Lead Author) Andrea Wilkerson, Pacific Northwest National Laboratory Minchen Wei, Hong Kong Polytechnic Kevin Houser, Penn State University Robert Davis, Pacific Northwest National Laboratory

Kevin W. Houser, PhD, PE, FIES Professor of Architectural Engineering The Pennsylvania State University 104 Engineering Unit A University Park, PA 16801 USA Phone: (814)863-3555 Email: khouser@engr.psu.edu Additional TM-30-15 Collaborators: Resources http://www.personal.psu.edu/kwh101/tm30/main.htm Dr. Dale Tiller Dr. Xin Hu Dr. Bill Thornton Dr. Steve Fotios Mr. Mike Royer 80