Report on quality metrics related to colour quality
Transcription
Report on quality metrics related to colour quality
Open Access Repository eprint Terms and Conditions: Users may access, download, store, search and print a hard copy of the article. Copying must be limited to making a single printed copy or electronic copies of a reasonable number of individual articles or abstracts. Access is granted for internal research, testing or training purposes or for personal use in accordance with these terms and conditions. Printing for a for-fee-service purpose is prohibited. Title: Report on quality metrics related to colour quality - Deliverable D435 Author(s): Renoux, D. Year: 2013, Funding programme: EMRP A169: Call 2009 Energy Project title: ENG05: Lighting: Metrology for Solid State Lighting EURAMET Secretariat Bundesallee 100 38116 Braunschweig, Germany Phone: +49 531 592-1960 Fax: +49 531 592-1969 secretariat@euramet.org www.euramet.org Report on quality metrics related to colour quality EMRP-ENG05-4.3.5 Version 1.1 D. Renoux Laboratoire National de Métrologie et d’Essais, France A report of the EMRP Joint Research Project Metrology for Solid State Lightning www.m4ssl.npl.co.uk EMRP-ENG05-4.3.5 Version 1.1 Title : Report on quality metrics related to colour quality Reference Version : : EMRP-ENG05-4.3.5 1.1 Date : 02-April-2013 Dissemination level : PUBLIC Author(s) : D. Renoux Laboratoire National de Métrologie et d’Essais, France Keywords : Quality metrics, colour specification, colour rendering, SSL, LED, Abstract : This is a report on quality metrics related to colour quality of artificial white light sources for general lighting. Background information is given on colorimetric parameters used for colour specifications. Then a summary a of the performed review on colour rendering metric ( D311) is presented: principles, approaches and results of implemented relevant metrics are sorted and exposed. The performed subjective experiment of this project and its main results (D312) are described and commented. The last section concludes on quality metrics for colour quality with information on new recommendations/practices, labs and CIE activity, and finally highlights outcomes of the work done for this project. Contact : http://www.m4ssl.npl.co.uk/contact About the EMRP The European Metrology Research Programme (EMRP) is a metrology-focused European programme of coordinated R&D that facilitates closer integration of national research programmes. The EMRP is jointly supported by the European Commission and the participating countries within the European Association of National Metrology Institutes (EURAMET e.V.). The EMRP will ensure collaboration between National Measurement Institutes, reducing duplication and increasing impact. The overall goal of the EMRP is to accelerate innovation and competitiveness in Europe whils t continuing to provide essential support to underpin the quality of our lives. See http://www.emrponline.eu for more information. About the EMRP JRP “Metrology for Solid State Lighting” In the EMRP Joint Research Project (JRP) “Metrology for Solid State Lighting”, the following partners cooperate to create a European infrastructure for the traceable measurement of solid state lighting: VSL (Coordinator), Aalto, CMI, CSIC, EJPD, INRIM, IPQ, LNE, MKEH, NPL, PTB, SMU, SP, Trescal, CCR, TU Ilmenau and Université Paul Sabatier. See http://www.m4ssl.npl.co.uk/ for more information. The research leading to these results has received funding from the European Union on the basis of Decision No 912/2009/EC. - ii - EMRP-ENG05-4.3.5 Version 1.1 SUMMARY The report addresses the colour specification issues of artificial light sources for the purpose of general lighting. The introduction gives the scope of our study: we consider the white lighting for interior applications and we exclude specific applications. We recall the main steps from the introduction of incandescent lamps till today lighting solutions with discharge (fluorescent) and LED-based lighting. We stress that for the fluorescent technology the poor colour rendering slowed down the purchase of consumers and that the official current index for colour rendering gives poor prediction of LED lighting. We need true metrics to provide the industry and the consumer with better quality indices and thus to accelerate the deployment of the very efficient LED technology. The first part of the report presents the background information and the work done for this project. To introduce colour specification of a white light sources a short insight of relevant colorimetry is given. We list the different chromatic diagrams and uniform coordinates spaces and colour appearance models. Those diagrams and spaces are the bases to understand the main colour parameter specification : the Colour Correlated Temperature (CCT), the CCT names, the chromaticity coordinates (x,y), the metamerism index MI, and finally the colorimetric metrics underlying the complex calculation of colour rendering indices. Then the results of the conducted review on colour rendering indices/metrics are summarized. The study sorts out the numerous approaches and methods of colour rendering demonstrating the multidimensional aspect of colour rendering. The recent activity and work of the CIE TC1-69, in charge of colour rendering, is reported with the list of promising candidates for new colour rendering metrics. The prediction results of implemented relevant metrics over a large set of SPDs sorted by types is exposed and commented. The assessment of metrics by field trial is shortly described, subjective and objective results are summed up in a table, as well as correlation coefficients. The metrics giving the best prediction for the conducted subjective are plotted as radar chart for the different subsets of sources types to show the improvement on the current CIE CRI. The second part of the report focuses on quality metrics related to colour rendering with the help of background information given in part one. We start from a definition of a quality metric and stress that the core of the quality metric represented by the colour rendering could not provide absolute information about the colour property of the light source but an information relative to the corresponding reference illuminant. The activity to achieve a new metrics by laboratories and CIE is given in few lines. Recommended specification or practices applying to colour rendering metric of SSL are exposed. We close that part with a synthesis of the findings of this study that we will address as recommendation to the CIE. - III - EMRP-ENG05-4.3.5 Version 1.1 TABLE OF CONTENTS 1. INTRODUCTION ......................................................................................................................3 2. COLOUR SPECIFICATIONS OF A WHITE LIGHT SOURCE ............................................4 2.1. Chromatic diagrams and colour spaces ........................................................................................4 2.2. Correlated Colour Temperature (CCT) and chromaticity.............................................................5 2.3. Colour Metamerism Index ...........................................................................................................7 2.4. Colour Rendering Indices and metrics .........................................................................................7 2.4.1. Introduction .......................................................................................................7 2.4.2. Review of colour rendering indices/metrics ........................................................7 2.4.3. Activity and work of the TC 1-69 of CIE ...........................................................10 2.4.4. Results of implementation of relevant metrics ...................................................11 2.4.5. Metrics assessment by field trials .....................................................................15 3. QUALITY METRICS FOR COLOUR QUALITY..................................................................19 3.1. Quality metric specification.......................................................................................................19 3.2. Synthesis and update on colour quality metrics for lightings ......................................................20 3.2.1. Common specification of colour property of white lighting :.............................20 3.2.2. Activity for new colour rendering metrics.........................................................20 3.3.3. Recommended specification for colour rendering from the ASSIST program.....20 3.3.4. Summary of the work achieved for this task ......................................................21 - IV - EMRP-ENG05-4.3.5 Version 1.1 LIST OF FIGURES Figure 1 : The 1931 CIE chromatic space (x,y) with the Planckian locus and the isotemperature lines ......... 6 Figure 2 : SPD colour, number and type of light source ..............................................................................12 Figure 3 : Comparison of CRI 13.3 Ra and CQS Qa 7.5 .............................................................................12 Figure 4: Comparison of CRI 13.3 Ra and CRI-CAM02UCS ....................................................................13 Figure 5: Comparison of CRI 13.3 Ra and MCRI ......................................................................................13 Figure 6: Comparison of metric proposals ...................................................................................................14 Figure 7 : Graph of the LED SPD for the subjective experiment .................................................................15 Figure 8 : Views of the subjective room for the colour rendition experiment ...............................................15 Figure 9 : Radar charts of 6 index/metric proposals .....................................................................................18 LIST OF TABLES Table 1 : Nominal CCT names for white tint with CCT ranges and illuminants ............................................ 5 Table 2: Fidelity and preference attributes.................................................................................................... 9 Table 3 : Metric’s predictions and average subjective scores .......................................................................16 Table 4 : Pearson correlation coefficients ....................................................................................................17 LIST OF SYMBOLS Ri CIE CRI Special index Ra CIE CRI General index -1- EMRP-ENG05-4.3.5 Version 1.1 LIST OF ABBREVIATIONS EMRP European Metrology Research Programme NMI National Measurement Institute JRP Joint Research Project SSL Solid State Lighting GLS General Lighting Services CFL Compact Fluorescent lamps QTH Quartz Tungsten Halogen LED Light Emitting diode OLED Organic Light Emitting diode LED-PC LED Phosphor-Converted CIE Commission Internationnale de l’Eclairage TCS Test Colour Samples CRI Colour Rendering Index CAT Chromatic Adaptation Transform CCT Colour Correlated Temperature CQS Colour Quality Scale CCRI Categorical Colour Rendering Index RCRI ordinal scale based Colour Rendering Index FCI Feeling of Contrast Colour Rendering Index HRI Harmony Rendering Index MCRI Memory Colour Rendering Index GAI Gamut Area Index FDI Fidelity Distortion Index HDI Hue Distortion Index CSI Colour Saturation Index SPD Spectral Power Density -2- EMRP-ENG05-4.3.5 Version 1.1 1. Introduction This report addresses the colour quality specification issues of artificial light sources, for the purpose of General Lighting Services. We will consider the colour quality of white light sources mainly for interior lighting applications such as office lighting, household lighting and public space lighting. We will exclude from this investigation specific interior lighting like medical lighting, scenic lighting or any other lighting applications having specific constraint or dedicated to produce special visual or aesthetic effect. th Artificial lighting began at the end of the 19 century with incandescent bulbs – luminous efficacy around 12 lumen/watt – and a good deal of incandescent lamps, gradually banned from the market, are still in use in offices and in homes. The QTH lamps, similar in technology – and a bit more efficient, luminous efficacy th around 18 lumen/watt - started to be commercialized in the early 20 to naturally and gradually replace the incandescent lamps. A breakthrough in efficiency arrived with fluorescent tubes – efficiency around 80 lumen/watt – introduced in the mid 20th century generating light in a very different way than the formers and also, that is our main concern, with very different optical characteristics. Due to having a base incompatible with household luminaries, these lamps were only used in office and public spaces. Yet this changed with the recent introduction of CFL (Compact Fluorescent lamps) featured with compatible bases and which generate light with similar optical characteristics than fluorescent tubes. The second revolution came up with LED (Light Emitting Diode). The LED technology, well known for a long time as signalling devices and low-level flux coloured light emitters, quickly and sharply gained in optical white luminous flux with the help of recent important research and development. Today these lamps, with efficiency up to 100 lumen/watt and beyond – outperform fluorescent technologies and are starting to be broadly deployed on the market. Once again the way to produce the light is different from the previous, as well as the optical characteristics. We all recall the visual impact of poor colour rendering from the early fluorescent lighting, like those have been installed and still installed in many pedestrian tunnels. This poor colour rendering slowed consumers purchase, and we saw the same phenomenon with the first LED lamps. On other hand, many subjective experiments performed the last decade with LED-based lightings only or with traditional lightings showed that the current colour rendering general index Ra, of the CIE CRI 13.3, was unable to predict the obtained subjective ranking and also provide worse prediction for LED-based lighting in comparison to traditional lighting. Today we need true quality metrics for colour appearance of the traditional (QTH, CFL, FL) and new lighting sources (LED, OLED) to provide the industry as well as the consumers and organisations with better quality criteria. The goal is to promote better LED lighting, and thus accelerate its adoption with the benefit of significantly reducing global energy consumption. The way to specify colour appearance of a light source, with background information, will be presented and commented in this report. Colour rendering metrics, as a major component of colour quality metrics, will be briefly reviewed, see deliverable D311 for detailed information, and analysed through the work done for their assessment and drafted in the deliverable D312. -3- EMRP-ENG05-4.3.5 Version 1.1 2. Colour specifications of a white light source Prior to introduce colorimetric parameters and specification we first recall some key principles and bases of colorimetry. Then we will go trough the relevant CIE colorimetric parameters to characterising white light sources. The last sections will develop the indices and metrics related to the colour rendering quality and investigated in the project, covering the current CIE index and new proposals of colour rendering indices or metrics. The colorimetric parameters and colour spaces presented in the following sub-sections are defined in several CIE in publications [1][2][3][4][5][6] and used in many standards. Colorimetric parameters are derived either directly from the emitted spectrum of the light source or from their interaction with objects through their reflectance spectra. 2.1. Chromatic diagrams and colour spaces CIE 1931 (x,y) chromatic diagram The human vision, under well-lit condition, can be seen as the combination of three photoreceptors stimulated by light : the LMS cones receptors covering respectively a Short, Medium and Long wavelength range in the visible spectrum. CIE defined two sets of colour-matching functions (CMFs), the second set ( (?), (?), (?)) ,having only positive values, was adopted for convenient application. The tristimulus values (X,Y,Z) of a stimulus are then obtained from the colour stimulus function P(?) of the light source and the CMFs using the formula [f1], with k a normalizing constant to match absolute photometric quantity on Y. X = K ? P(? ). (? )d? ; Y = K ? P(? ). (? )d? ; Z = K ? P(? ). (? )d? ; [f1] The 1931 chromaticity coordinates (x,y) are calculated from the the tristimulus values (X,Y,Z) using the formula [f2] : x= X/( X+Y+Z) ; y=Y/(X+Y+Z); [f2] After the definition of the CIE 1931 (x,y) chromatic diagram, related to the 1931 standard colorimetric observer, the CIE defined successively several uniform chromatic diagrams and uniform colour spaces with associated colour difference formula. All the coordinates of the new spaces can be calculated from the tristimulus value (X,Y,Z) of the stimulus, the list is given hereafter : CIE two-dimensional UCS chromatic diagram and extension to three-dimensional uniform colour space § the two-dimensional CIE 1960 (u,v) UCS diagram (obsolete – but in use for CCT computation) § the three-dimensional CIE 1964 uniform coordinate space (W*,U*,V*) – (obsolete for colorimetric specification, but still in use for the CRI calculation – see section 2.4). § the two-dimensional CIE 1976 UCS diagram (u’,v’) (in use [7]) § the three-dimensional CIELUV or CIE 1976 (L*,u*,v*) uniform coordinate space (in use) § the three-dimensional CIELAB or CIE 1976 (L*,a*,b*) uniform coordinate space (most popular) -4- EMRP-ENG05-4.3.5 Version 1.1 CIE Colour Appearance Model (CIECAM) § CIECAM97s [4] § CIECAM02 [5] § CAM02-UCS [11] The CAM models are used to predict colour appearance attributes for distinct viewing conditions, the inputs of CIECAM are the tristimulus values of the stimulus (X,Y,Z) , the tristimulus values of an adapting white point, the adapting background, and the surround luminance information. More complex visual phenomena are embeded in these models like chromatic adaptation transform (CAT), parametric degree of adaptation, and non-linear cones responses. Li et al derived an uniform space, called CAM02-UCS, from the CIECAM02 and proposed an updated CIE colour rendition index (CAM02UCS-CRI) whose principle is the same than the CIE CRI 13.3 but implemented in this new UCS. 2.2. Correlated Colour Temperature (CCT) and chromaticity The CCT is one of the most popular and used parameter to specify a light source, natural or artificial like of daylight phases and light emitted from bodies elevated at very high temperature, or any white artificial light sources The colour temperature of a light source is defined by the temperature of the closest Planckian radiator in the (u,v) chromatic space. Note : u=u’ and v= 2v’/3 with (u, v’) the coordinates of the CIE 1976 UCS. Today the common ways to communicate to consumer quantitative or qualitative information about the spectral content of light sources are nominal CCT names and CCT in Kelvin. Nominal CCT names were introduced in the ANSI standard C78.376 – “Specification for the chromaticity of fluorescent lamps” [8]. In the following table 1 some examples are given with CCT ranges in Kelvin and well known illuminants and CIE reference illuminants, notice that there is no agreement on CCT ranges. Nominal CCT names CCT ranges Illuminant examples CIE reference illuminants Warm white < 3000 K Incandescent (2600- 2900 K) A QTH (2700 – 3000 K) ‘Neutral’ white [3000-4000 K] Discharge lamps Cool white, Daylight > 4000 K Discharge lamps, direct and indirect daylight, cloudy daylight (8000 K), xenon/mercury arcs 6500 K B (4874 K) , C (6774 K), E, D50, D55, D65, D75 Table 1 : Nominal CCT names for white tint with CCT ranges and illuminants It can be seen from the figure 1 that the CCT can relatively well express the balance between the reddish or bluish tint of the light source, the Planckian locus following roughly the blue-red axis, but does not enable to express the greenish or magenta shade of the light source for a given CCT. -5- EMRP-ENG05-4.3.5 Version 1.1 A more accurate information is given by chromaticity coordinates, (x,y) or (u’,v’) but generally reserved for professional or industrial specifications like LED chromaticity binning for lamps or luminaires manufacturers and are not meaningful characteristics for consumers, buyers, retailers. We later will see that colour rendering indices and other developed indices for colour rendering or colour quality could rely upon only on the difference with a reference illuminant and therefore indices could not reflect important information about colour rendering or property of the light source. For instance a daylight simulator and an incandescent lamp could obtain high and equal CIE CRI but we all know that they have very different colour rendering or property. Figure 1 : The 1931 CIE chromatic space (x,y) with the Planckian locus and the isotemperature lines (reprinted from wikipedia.org) With no better meaningful parameter to reflect the spectral content of a light source, we must keep the CCT specification as a colour quality metric parameter and colour specification for common users. The 1931 chromaticity coordinates (x,y) ou (u’,v’) are used for colour specification in standard requirements and for professional usage along the lighting chain. One other important specification is the distance Duv , in the CIE 1976 chromatic diagram (u,v), of the test light source from the Planckian locus. This distance can be expressed with a signed value to indicate if the light source is above (+) or below (-) the Planckian locus. The ANSI_NEMA_ANSLG C78.377 “Specification for the chromaticity of Solid-State-Lighting Products” was the first published standard specifying range of chromaticities for SSL. Eight CCT and Duv targets are defined with tolerance. Background information is given with the equivalent tolerance quadrangle in (x,y) or (u’,v’) space. These quadrangles are approximately 7-MacAdams ellipse [10]. One MacAdams ellipse corresponding to the perceptual threshold of colour difference. The specification of colour rendering is based on the general index Ra of CIE CRI 13.3 Ra. -6- EMRP-ENG05-4.3.5 Version 1.1 2.3. Colour Metamerism Index Metamerism phenomenon is the matching of apparent colour, under a given illuminants, of objects having different spectral power distributions, such objects are named metamers with respect to that illuminant. The CIE special colour metamerism index is defined in publication 15.2 [1] and its supplement [2]. It is calculated from the colour-difference index ?E of the two metamers under a specified reference illuminant . The CIE Metamerism Index (MI) of a light source is derived by calculating the mean colour distance of 5 metamers in the visible spectrum and in 3 the ultraviolet range, computed either in the CIELAB or CIELUV colour space. The metamerism index is generally used to qualify daylight simulators. But this property to preserve mesmerism has been many times quoted as a relevant property of colour rendering quality but is rarely used to specify light sources for GLS. 2.4. Colour Rendering Indices and metrics 2.4.1. Introduction The study and assessment of colour rendering indices and metrics, current and proposals, were the major concern of the research achieved to address the perception of colour quality of light sources : (1) a large review and implementation of proposals of colour rendering indices/metrics has been performed (D311) and (2) a real-life colour experiment was conducted with 43 individuals subjected to 9 lights sources (D312). We present in this report the synthesis of the outputs of these deliverables, details and complete references can be found in the related reports. 2.4.2. Review of colour rendering indices/metrics We sorted out the different approaches and implementations of the current index and proposals of new index/metric for colour rendition. There is no universal definition of the colour quality of a light source but several rendering properties of light source have been identified with a related metric proposal, the metric quantifying the quality of the properties. These different properties are related to the different aspects of the colour appearance of objects. The relevance of the properties could vary depending on the targeted application : expectation of individuals are quite different with the environment, for instance a living room and a office room. We expose in the following subsections the main “trade-offs” and key points issued from the review of colour rendering indices or metrics. These main trade-offs address almost all key points of a colour rendering metric, revealing the dimension of colour rendering of light sources. We close this section with the activity of the CIE TC1-69 for a new colour rendering metric. -7- EMRP-ENG05-4.3.5 Version 1.1 A- Fidelity versus preference Colour properties roughly belong into two categories related to: § the ability to preserve the appearance of objects lighted by the assessed source, § the ability to enhance the appearance of objects lighted by the assessed source. A common designation for the metrics belonging to the first category is “fidelity metric” and for the metrics belonging to the second category is “preference metric”. We can extend the definition of fidelity metric by a metric quantifying the preservation of a given set of colour appearance attributes and likewise the definition of a preference metric is a metric quantifying the enhancement of a given set of colour appearance attributes. Colour appearance attributes for fidelity and preference A table of some visual colour attributes is given below with an attempt to sort them under fidelity attribute or preference attribute. There is no strict definition or consensus about this classification for all the attributes. The table hereafter is similar to that one presented by F. Viennot during a TC CIE meeting [10]. We can be confuse about the meaning and the use of an attribute either for fidelity or preference assessment. K. Smet reported in a paper that the subjectively rated naturalness attribute and the subjectively rated fidelity attribute over a set of experiments, performed by different laboratories, were negatively correlated, or with converse trends. Since the publication of the CRI CIE 13.3 in 1974 many papers outlined that perceived naturalness is enhanced if the colour appearance of some common objects is more colourful. (base of the flattery index). Then naturalness should not be a strictly a fidelity attribute, but if we define naturalness as “the colour match with the internal representation we have of natural or familiar objects” we should say that naturalness addresses fidelity. J. Schanda et Al reported that the harmony coefficient they derived, supposedly maximum for the reference light, could be greater for the test light source than the reference light and then the harmony index was expressed as an absolute difference of harmony coefficients. We can wonder if colour harmony should be considered as a “preference” or “fidelity” attribute since the test light source could have an enhanced harmony. The basic fidelity metric is calculated from the colour-difference index ?E between the test light and the reference light on a set of test colour samples. By extension a fidelity metric can rely on the absolute difference of an attribute between the test light and the reference light on a set of colour samples, quantifying the preservation of the attribute by the test light. A preference metric can be calculated as a signed difference of attributes between the test light source and a reference light source. But a preference metric can be calculated as a fidelity metric but discounting colour shifts in the computation when the effect of these colour shift enhance appearance (gain of contrast, gain of chroma, ect) as it is done for the Colour quality Scale general index CQS Qa. -8- EMRP-ENG05-4.3.5 Version 1.1 Fidelity Preference Preservation of visual attribute Enhancement of visual attribute colour consistency vividness, colourfulness, visual clarity, preservation of metamerism attractiveness colour discrimination naturalness (?) naturalness (?), memory colours (?), colour harmony (?) colour harmony (?) memory colour (?) colour categorisation Table 2: Fidelity and preference attributes B - Reference illuminants versus memory colour of objects Both fidelity metric or preference metric can be implemented by comparison of the test light with reference illuminants. Reference illuminants are defined by their spectrum. Memory/Reference objects are defined by their reflectance spectrum and by their optimal coordinates in a chromatic space or a by a distribution of a perceptual characteristic in a chromatic space. It is needed to proceed to subjective experiments to establish these memory objects. The best examples are the chromaticity-coordinates increments added to the reference illuminants of the “flattery index” and the similarity functions of the “memory colour rendering index” (MCRI). Note on reference illuminants : Since the publication of CIE CRI 13.3 the referent illuminants are the CIE phases daylight (CCT >5000 K) and Planckian/Blackbody radiator/emitter (CCT < 5000K). Daylight has always been considered as the ideal colour rendering light source providing a large colour variety with a lot of shades and excellent discrimination of all colours and surely contribute the more to built up our colour memory or internal representation of natural objects. This is less true for the Planckian radiator, but since the advent of artificial light incandescent/QTH lamps, whose spectra is very close to Planckian radiator, are the most familiar sources. C - Continuous rating versus ranking Peter Brodogi and Al develop an ordinal scale based colour rendering index (RCRI), to meet the user’s expectation in term of specification to provide him with simple information about the colour quality. They claim that customer is not able to state on colour quality equality from close indices and unable to fix difference thresholds while comparing indices. D – Statistical binary colour test versus representative colour test Zukauskas et Al developed a statistical approach to qualify colour rendering, rather than working on a limited set of test colour samples they developed an analysis on a 1269 Munsell Mat colour Atlas. The -9- EMRP-ENG05-4.3.5 Version 1.1 method compute the following indices with given tolerances, in the (L,x,y) space, applied on differences with the reference light : § CFI: counts of TCS rendered with high fidelity, § CSI: counts of TCS rendered with an increased saturation, § HDI: counts of TCS rendered with an important hue distortion. E- One component metric versus multi-component metric New proposals combine several metrics/indices to take into account several aspects of colour rendering. A proposal of a fidelity metric combined and a preference metric is given by S.Rea et AL who recommended to use the CRI in conjunction with the GAI (Gamut Area Index). Other indices were suggested to supplement the CRI, or a fidelity metric, like the gamut indices, the Feeling of Contrast Index FCI , the Harmony Rendering Index (HRI). It is not communally admitted that the increase of gamut well represents an increase of colour discrimination or increase of chroma for all colour shades. Zukauskas et Al recommended a 4-dimension colour quality-metric that can be used as simple communication tools to the end user. The four components of this metric are the CCT, the colour fidelity index CFI, the colour saturation index CSI and the hue distortion index HDI (see point D above). Many researchers involved in the development of colour rendering metrics claim that it is not possible to specify the colour rendering properties of light sources with a single parameter, some recommend to use separate indices for different task and different environment. Others are in favour to keep one index with simple calculation for applicability. 2.4.3. Activity and work of the TC 1-69 of CIE The TC well acknowledges that the CIE CRI 13.3 is based on outdated colorimetric metrics and gives poor prediction for LED lighting. New indices are considered by the TC, not based only on “colour fidelity” but also taking into account “colour preference”. The CIE 2010 activity report included also the following indices as promising candidates : FCI , RCRI , HRI and CCRI. At the last workshop on colour quality held at Hangzhou, September 2012, the program included the presentation by experts of following indices as the most promising CRIs : CQS, nCRI, GAI and the MCRI. From press release, LED magazine Feb. 2012, we know that no consensus was yet found, and that some members were in favour to quickly adopt the CQS and other members wanting a more precise measure of colour rendering. The final version of nCRI specification, which a refinement of the CRI-CAM02UCS, has just been distributed to TC 1-69 members in December 2012. We just include at the time of this report drafting a recent paper from YF Chou et AL. [21] describing the final refinements of the CRI-CAM02UCS to form the nCRI. The new index is calculated with 273 TCS broken down in 4 groups: 90 colour constant , 90 colour inconstant, 90 reflectance difference, 3 complexion. The colorimetric observer is the 10° observer, average colour difference is calculated as the RMS of CIECAM02UCS colour differences and general index as exponential function of this colour difference as done for the CQS Qa calculation. - 10 - EMRP-ENG05-4.3.5 Version 1.1 2.4.4. Results of implementation of relevant metrics The results are those obtained and reported for deliverable D311. The aim was to compare different colour rendering indices /metrics in conjunction with specific implementations on a set of 122 SPDs representative of all lighting technologies. Fourteen indices/metrics have been implemented. The selection of metrics took into account the CIE TC1-69 recommendations for relevant metrics to complement and supplement the current index for colour rendering specification, we added a statistical method (CFI/CSI/HDI) in our investigation, the results are then given for the following indices/metrics: § CRI Ra : current CIE CRI 13.3 general index [2], § CQS Qa : proposal Colour Quality Scale general index [12], § MCRI Sa : Memory Colour rendering Index general index [18] , § CRI-CAM02UCS : updated CRI Ra with the CAM02-UCS [11], § CRI Ra96 : CIE proposal to update the CRI Ra [3], § RCRI : ordinal scale Colour Rendering Index [19], § CCRI : Colour Category Rendering Index [16], § HRI : Harmony Rendering Index [17], § FCI : Feeling of Contrast Index [14], § GAI : Gamut Area Index [13], § CQS Qg : Gamut Area Scale [12], § CFI stat : Colour Fidelity Index – statistical [15], § Combination “X“ + “Y”: combination using the mean value of the indices “X” and “Y”. SPDs data base The collection has been established from the SPDs of author’s Excel ® spreadsheet, CIE publications for standard illuminants, and LNE’s measurements. The collection can be broken down in the following main subsets: § 7 SPDs of incandescent and halogen lamps, and planckian radiator with or without optical filter § 49 SPDs of fluorescent tubes and compact lamps, including some CIE standards (Fn, F3.n) § 5 SPDs of miscellaneous lamps (HMI, Mercury arc, Xenon arc) § 9 SPDs of HPS lamps, including some CIE standards (HPn) § 52 SPDs of 3 subsets of LED lamp types: phosphors converted (PC), phosphors converted with NUV excitation (NUV), and LED clusters. - 11 - EMRP-ENG05-4.3.5 Version 1.1 Numerical results The graphical prediction of implemented metrics are plotted versus the n° of SPD clustered by type of light source as presented in figure n°2. The following graphs are excerpts of the deliverable D311 for some indices/metrics considered as good candidates for a new CRI and do not summarize all the results. SPD categories 0 7 14 21 28 35 42 49 56 63 70 77 84 91 98 105 112 119 SPD N° QTH/incandescent Fluorescent HMI/Hg/Xe HPS LED clusters LED PC LED PC NUV Figure 2 : SPD colour, number and type of light source Comparison CIE Ra 13.3 / CQS 7.5 Value 120 100 80 CIE Ra 13.3 60 Qa 40 20 0 0 10 20 30 40 50 60 70 80 90 100 110 120 SPD N° Figure 3 : Comparison of CRI 13.3 Ra and CQS Qa 7.5 - 12 - EMRP-ENG05-4.3.5 Version 1.1 Comparison CIE Ra 13.3 / CRI CAM02-UCS Value 120 100 80 CIE Ra 13.3 60 CRI-CAMUCS 40 20 0 0 10 20 30 40 50 60 70 80 90 100 110 120 SPD N° Figure 4: Comparison of CRI 13.3 Ra and CRI-CAM02UCS Comparison CIE Ra 13.3 / MCRI Value 120 100 80 CIE Ra 13.3 60 MCRI 40 20 0 0 10 20 30 40 50 60 70 80 90 100 110 120 SPD N° Figure 5: Comparison of CRI 13.3 Ra and MCRI - 13 - EMRP-ENG05-4.3.5 Version 1.1 Comparison of metric proposals Value 120 GAI+CRI 100 CQS + GAS 80 60 CRICAMUCS 40 MCRI 20 0 0 10 20 30 40 50 60 70 80 90 100 110 120 SPD N° Figure 6: Comparison of metric proposals Comments on the obtained results implemented indices/metrics The results of indices/metrics applied on 122 SPDs do not assess them, since three is no comparison with subjective ratings, but rather the results give the general behaviour of indices/metrics globally and against the source type and enable to perform comparison between them. For example on figure 5 we can see that LED clusters obtain very high MCRI values in comparison to CRI Ra, and that for continuous spectra (QTH, fluorescent, LED NUV) the MCRI values are systematically smaller for high values of CRI Ra. We can deduce that MCRI increases or decreases with the chroma of the light source and that most of LED sources have better prediction than traditional sources. Other comments, on the presented graphs in this report, are that CRI-CAMUCS is very similar to CRI Ra but with increased values when CRI Ra is low for LED clusters. We observe the same phenomenon for CQS Qa but with superior increase for LED clusters. On figure 6 we can compare the combined metric “CRI Ra+GAI” and “CQS Qa + GAS” and observe that the Qa+GAS gives much greater predictions for LED sources. A global finding of the D311 is that there is a good correlation between the 14 implemented test metrics for fluorescent sources, with an average correlation coefficient equal to 0.90. This average correlation coefficient drops to 0.57 for LED sources but stays greater than 0.70 for the fidelity metrics : CRI Ra, CRI-CAM02UCS, CQS Qa and CIE CRI Ra96. Combined metric “fidelity + gamut” and MCRI obtain low correlation with the average set of metrics, “CRI+FCI” obtains the lower coefficient of correlation of 0.18 with the whole set, indicating a likely big change in colour rendition prediction with theses metrics. Other specific findings were brought on specific implementation. The methods for the reference source determination, CCT and the closest reference in a uniform chromatic diagram, performed for the computation of CIE Ra96. The CCRI method gives interesting results increasing prediction for LED sources but also increasing the predictions for HPS sources – this was corrected with a small modification of the main formulae. We assessed also the impact of the TCS using the metric set or a Macbeth Digital Colour Checker or the 1296 Matte Colour Chips Munsell Atlas. We observed significant differences on the general index depending on the set of TCS and also variables differences between metrics on special indices (see D311 for details). - 14 - EMRP-ENG05-4.3.5 Version 1.1 2.4.5. Metrics assessment by field trials A real living room has been built to carry out a subjective experiment in a common and real environment. The test room has been furnished and decorated with many coloured natural and artificial objects. A special lighting system has been designed: it enables to uniformly light the test room and to quickly select a lighting source out of 12 and changing only the spectrum of the illumination in the test room. The experiment has been performed with 43 people and nine lighting sources, including 6 led-based lightings, broken down into two CCT domains centred around 2700 K and 5000 K. 1 LED WW 2700 0,9 Spectral density 0,8 LED WR 2700k 0,7 0,6 LED RGBY 2700K LED RGB 5000k LED NUV 5000k LED CW 5000k 0,5 0,4 0,3 0,2 0,1 0 380 480 580 680 780 Wavelenght (nm) Figure 7 : Graph of the LED SPD for the subjective experiment Figure 8 : Views of the subjective room for the colour rendition experiment - 15 - EMRP-ENG05-4.3.5 Version 1.1 The index/metric predictions with the average subjective scores for preference are given in the table 3 below for each of the nine lighting. We can notice that the subjective scores of the cold sources vary in a large interval (87 to 40) while the warm sources vary in a smaller interval (78 to 97) with two equal scores (88) and one score very close (91). Light sources FL 5000K LED NUV 5000K LED CW 5000K LED RGB 5000K LED WR 2700K LED WW 2700 RGBY 2700K HAL 2700K CFL 2700K CRI Ra 93,70 98,10 70,68 35,58 88,56 82,78 76,20 99,70 82,00 CQS Qa 96,45 99,10 71,33 62,89 90,48 79,40 79,06 96,91 75,78 MCRI Sa 92,71 91,90 75,87 95,36 94,38 88,36 95,22 94,32 82,90 CRI-CAMUCS 94,30 98,51 71,02 49,72 86,79 78,82 80,27 99,02 75,97 RCRI 100,00 100,00 56,09 56,09 98,00 74,40 80,90 100,00 74,40 CCRI 91,96 94,63 73,92 55,06 82,75 77,07 71,93 86,06 77,58 CRI Ra96 95,21 98,52 71,30 47,50 88,23 82,02 80,79 99,28 79,49 HRI 97,92 99,36 104,51 66,93 95,16 101,01 92,16 100,11 99,32 FCI 111,09 108,59 86,23 178,79 138,31 118,83 148,20 123,00 116,31 GAI 97,71 90,14 78,39 134,56 63,44 48,35 57,57 48,93 49,02 CQS Qg 103,41 100,84 88,54 139,64 109,28 97,29 110,69 97,36 98,20 CRI + GAI 95,71 94,12 74,54 85,07 76,00 65,57 66,89 74,32 65,51 CQS Qa + Qg 99,93 99,97 79,94 101,27 99,88 88,35 94,88 97,14 86,99 Ra96+ CCRI 93,59 96,58 72,61 51,28 85,49 79,55 76,36 92,67 78,54 CFI stat 94,25 100,00 16,71 11,11 52,72 33,57 24,90 100,00 30,42 Subj. Preference 85,35 86,74 63,95 40,23 88,14 77,91 88,14 97,44 90,93 Table 3 : Metric’s predictions and average subjective scores The correlation, between the predictions of fifteen metrics and the subjective scores of preference , have been computed using the Pearson and Spearman coefficients of correlation and considering different rounding of subjective and objective data. These correlation coefficients have been computed on the whole set of lightings and the five possible subsets with respect to the CCT (cold/warm) and the technology (all technology/LED technology). From these correlations, with the subjective scores, a set of best metrics in term of prediction of “preference” has been drawn and is illustrated with radar charts for all subsets of lighting sources (see figure 9). We are well aware that the number of samples is too low to correctly state on some subsets, but the goal here is to see the trend. The results of tables of Pearson (see table 4) and Spearman coefficients of correlation wit subjective scores (see D312) obtained from our experiment can be summarized as follow : § the fidelity metrics are well correlated to “cold sources”, the current metric CRI Ra fails for warm sources and new fidelity metrics and CQS Qa perform better for warm sources, § the gamut metrics and FCI do not correlate well for “cold sources” and “warm sources” but correlate very well for “warm LED”, § the MCRI also correlates very well for “warm LED” but fails for all other source subsets, - 16 - EMRP-ENG05-4.3.5 Version 1.1 § other metrics like CCRI or HRI, based on specific properties of colour rendering, produce interesting results for some cases but globally do not perform better than the current CRI. § statistical method, like CFI which is based on fidelity, can also give interesting results and better results than the current index for some cases. § combined metrics (CRI+GAI, Qa+Qg) do not correlate in general, excepted for warm LED and warm sources. Metrics all sources cold sources warm source LED sources cold LED warm LED mean rank CRI Ra 0.912 0.998 0.576 0.923 0.998 -0.044 0.727 6 CQS Qa 0.762 0.961 0.581 0.839 0.948 0.500 0.765 4 MCRI 0.130 0.016 0.235 0.162 -0.159 0.991 0.229 12 CRI-CAMUCS 0.862 0.996 0.634 0.921 0.996 0.596 0.834 1 RCRI 0.783 0.894 0.602 0.842 0.860 0.725 0.784 3 CCRI 0.776 0.999 0.531 0.812 0.999 0.052 0.695 7 CRI Ra96 0.892 0.998 0.612 0.937 0.998 0.381 0.803 2 HRI 0.699 0.782 -0.075 0.682 0.791 -0.945 0.322 9 FCI -0.413 -0.712 0.020 -0.379 -0.731 0.941 -0.212 13 GAI -0.761 -0.690 0.006 -0.797 -0.757 0.945 -0.342 15 CQS Qg -0.628 -0.715 -0.041 -0.577 -0.740 0.991 -0.285 14 CRI + GAI -0.191 0.613 0.486 -0.222 0.463 0.577 0.287 10 CQS Qa + Qg 0.056 0.106 0.430 0.067 -0.054 0.910 0.252 11 Ra96+ CCRI 0.851 0.999 0.594 0.885 0.999 0.064 0.732 5 CFI stat 0.609 0.916 0.654 0.618 0.889 0.202 0.648 8 Table 4 : Pearson correlation coefficients A general observation can be drawn from this specific subjective experiment and applies for the preference attribute : the indices /metrics which reward very high chroma can fails, a too high-chroma could lead to deprecated perceived quality, but on other hand the metrics which reward increased chroma seem to better rank warm LED sources having close subjective scores of preference. We know that is observation just relies on a set of 3 warm LED-based sources, one quadric-chromatic , one phosphor-converted and one phosphor-converted with a red LED, representing all type of LED-based warm lights but with just small differences of rated preference between two of them. We also know that the RGB 5000K presented a very strong saturation and was somehow atypical for a general lighting application. Nevertheless it is interesting to have such cases for general validation, but to better understand how metric perform for prediction it is more subjective experiments with more test cases. The result of Pearson correlation of six metrics are plotted using radar charts (see figure 9) with the current CRI Ra to better visualize the improvement of the metric/index on the different subsets. Even if the correlation with the warm lights sources subjective preference is not high we can see the gain against the current CRI predictions. - 17 - EMRP-ENG05-4.3.5 Version 1.1 CRICAM UCS Pearson correlation all sources Pearson correlation all sources CRI Ra warm LED cold sources cold LED warm source CRI Ra warm LED cold sources cold LED warm source LED sources LED sources CRI Ra96 Pearson correlation all sources cold sources cold LED warm source all sources CRI Ra warm LED cold sources cold LED warm source LED sources LED sources Ra96+ CCRI all sources CQS Qa Pearson correlation CRI Ra warm LED Pearson correlation RCRI CFI stat Pearson correlation CRI Ra CRI Ra all sources warm LED cold sources cold LED warm LED cold sources cold LED warm source warm source LED sources LED sources Figure 9 : Radar charts of 6 index/metric proposals - 18 - EMRP-ENG05-4.3.5 Version 1.1 2.5. Measurement and processing of spectral data We did not investigate the methods of measurement of spectral data for this WP related to perceived quality of SSL. Radiometric measurement and use of spectro-radiometer are addressed in WP1. The reference documents we use are the CIE publication 15:2004, the CIE publication n°63 – 1984 and the IESNA LM-79-08. Actually the most useful standard for SSL is the IESNA-LM79. We performed in a previous study [22], at LNE-CNAM, an uncertainty analysis on chromaticies (x,y), CCT and CRI using the Monte-Carlo method of propagation of uncertainty and following the recommendation of the supplement 1 to ISO/BIPM GUM (guide for uncertainty measurement) [20]. For a reference setup based on a single monochromator/PMT the expanded uncertainties on special indices Ri were comprised between 0.2 and 1.0 for 5 typical LED lightings. With a CCD spectro-radiomer, Minoltat CS-2000, the expanded uncertainties on special indices were comprised between 0.4and 1.6 for 5 typical LED lightings. The maximal expanded uncertainties for the CIE CRI Ra was 0.2 for the reference setup and was 0.4 with the CS-2000 for the same 5 typical LED lightings. The study also revealed that the calculation of (x,y), CCT and CRI with different implementations (commercial software, diffused excel spreadsheets from different authors, CIE program) can yield to significant discrepancies up to 0.9 on the special index R9. The differences on Ra were smaller than 0.3. 3. Quality metrics for colour quality 3.1. Quality metric specification The goal of a quality metric is to provide users (not only end-users : manufacturer, integrator, retailers, consumers) with a reliable and meaningful information about the quality of a product. The delivered Information could be a numerical value (95%), a class (class A, B, C), or an ordinal category : for instance “excellent”, “very good”, “good”, “fair”, ”poor”. The minimum requirement for a basic quality metric could be: § § § Define the set of parameters that can be scaled to quality factors or performance criteria, Specify measurement standards or measurement methods for these parameters, Provide guidance to interpret quality metric specifications in term of absolute quality and comparative quality of the product. For colour rendering most of the indices/metrics are by definition the core of a colour quality metric, they specify a method to compute, from the source spectrum, a numerical value, generally scaled between 0 and 100, representing the quality of the colour rendering of the light source. It should be recall here that this value is for most of the metrics, at least the fidelity metrics, a relative value expressing an average colour shift with the colours rendered by a reference illuminant of the same CCT. The properties of the reference illuminants vary considerably with the CCT, for instance number of colour shades, colour discrimination, colour identification or naming, and only the CCT specifies this reference illuminant. Special indices provide the same kind of information, the magnitude of colour shift, TCS by TCS. The only factor dealing with the absolute rendering is the CCT weighting factor of the CQS, this factor accounts for the drop of gamut for low CTT of the test light source. - 19 - EMRP-ENG05-4.3.5 Version 1.1 3.2. Synthesis and update on colour quality metrics for lightings Usually the colour quality specification of a lighting source is a part of the colour property specification. The worldwide official general index of colour rendering quality is the CIE CRI 13.3 Ra. The TC1-69 recognized that the CIE CRI 13.3 is based on outdated colorimetric metrics and give poor prediction for LED lighting. But industry would not replace it before CIE TC1-69 endorses new metrics and publishes them. 3.2.1. Common specification of colour property of white lighting : 1. CCT names : “warm white”, “(neutral)white”, “cool white”, “daylight” 2. CCT value in Kevin (K) and distance from spectrum locus Duv 3. Chromaticity (x,y) or (u’,v’) 4. Relative spectral density (graph) 5. Colour rendering quality: CRI 13.3 Ra and special indices (or just the Ra/R9) or other see 3.3.3. Photometric measurement protocol and chromaticity specification for SSL can be respectively found in references [9] and [7]. The reference [9] is not a protocol for measurement of chromaticity but describes the measurement of photometric quantities by means of spectroradiometers and then could be applied to tristimulus determination of TCS for colour rendering . It is interesting also to note that in reference [7] Duv tolerances are given for the defined CCT ranges, these tolerances should limit the low values of the index of colour rendering. Other developments aiming at traceable measurement of SSL with spectroradiometers can be found the WP1 (D131 to D134) of the ENG05 project. 3.2.2. Activity for new colour rendering metrics Many laboratories have been working on new metrics for colour rendition and published proposals. The CIE 2010 activity report listed the following indices as promising candidates : CQS, GAI , MCRI, FCI, RCRI , HRI and CCRI supported by 17 (+/-) research reports from 10 labs/groups from 7 different countries. At the last workshop on colour quality held at Hangzhou, September 2012, the program included the presentation by experts of the following indices as the most promising CRIs : CQS, nCRI, GAI and the MCRI. From press publication, LED magazine, we learnt that the TC1-69 was close to recommend a dual standard, one being the CQS. According to J. Schanda answering the article, is was demonstrated that prediction of current CRI and CQS with a given TCS could change from high to low prediction with a real metameric samle of that given TCS. Since then the final version of the new metric, called nCRI, has just been distributed to TC1-69 members in December 2012. This new metric is a refinement of the CRI-CAM02UCS which uses the 10° observer, several elaborated sets of TCS with metameric and “reflectance difference” TCS sets and features, likewise CQS, new formula for the general and special indices (RMS value, exponential form). 3.3.3. Recommended specification for colour rendering from the ASSIST program The Alliance for Solid-State Illumination Systems and Technologies (ASSIST), an US initiative gathering researchers, manufacturers, and government organizations, established in 2002 by the Lighting Research Center (LRC) published guidelines and recommendations. One interesting paper [22] is downloadable “The Class A Color Designation for Light Sources » by M. S. Rea, & J. P. Freyssinier , this paper finalizes the - 20 - EMRP-ENG05-4.3.5 Version 1.1 approach already described in a previous paper “Color “Rendering: a tale of two metrics” [13]. The author argue that the currents metrics used by industry, CCT and CRI, is not meaningful for consumers and propose a “class A color” designation for general illumination : white illumination with CRI = 80 and 80= GAI =100. 3.3.4. Summary of the work achieved for this task We conducted a large review of all proposals for colour rendering quality. There are numerous proposals putting into question all the principles of the current rendering metric, the CIE CRI 13.3. This review showed that the concept of “colour quality rendering” is multi-dimensional, each dimension could reflect a specific property of colour rendering - most of them can be sorted out under the fidelity properties category and the preference properties category. The fidelity properties are related to the preservation of colour appearance of objects under the test light source while the preference properties are related to the enhancement of the colour appearance of objects under the test light source, the enhancement being necessarily a departure from fidelity, those two aspects cannot be achieved at the same time. In both two approaches, fidelity or preference, there is an underlying question: what the is the reference light. Preservation attributes can be derived from the absolute colour shift between the test light and the reference light, and preference attribute derived from some specific signed difference of the test light to the reference light – like a increase of chroma in favour of subjective preference. Most of methods/principles of current CRI are given alternative methods: ordinal scale against continuous scale, statistical description rather than limited test colours, colour naming or categorization rather average colour shifts. All the proposal use updated chromatic diagrams or colorimetric metrics. One the most original metric is the MCRI using “colour memory” of objects rather than reference illuminants [18], and is based on similarity functions, subjectively built, in a relevant chromatic diagram. The well recognized property of LED-based lighting is the enhancement of chroma/saturation of colour. Several new proposals attempt to take it into account and derived index/metric supplementing or complementing the CRI. Two solutions are proposed/recommended to take into account this increase of chroma (1) to supplement the CRI with the GAI (Lighting Research Centre metric [13][20]), (2) to complement the CRI discounting in the colour difference index calculation the contribution of the increase of chroma (NIST - CQS [12]). The computation and application, over a broad set of spectra representing all technologies, of relevant metrics, including those listed by CIE as promising candidates, enable us to watch the behaviour of each of current and proposals of index/metric. It confirmed that the new metrics do not bring so much to traditional lights and that the greatest discrepancies are obtained for LED-based lightings. We observed that new based-fidelity metrics (including CQS) increase the low scores of many LED-clusters or LED-PC lighting sources and that gamut effectively increase for these light sources, with very high values reached for the FCI. The MCRI has a special behaviour not balanced between traditional/LED-based lighting. The dependence with the set and choice of TCS is clearly highlighted. We set up a subjective experiment in a real living room, with LED-based and traditional lighting, to assess metrics and to learn about the subjective “preference” and its relationship to other subjective attributes like perceived “fidelity”, “naturalness”, ”vividness” and “chart colour quality”. All attributes co-vary in the same direction, and may be if specific relationship can be observed for the appearance a limited set of objects, as found in other subjective experiments, they likely could not apply on a large set of objects and materials. The main results of the subjective experiment is (1) that fidelity metrics perform quite well for the “cold sources”, (2) the current CRI fails (correlation=0) to predicting the “warm led”, (3) the fidelity metrics, - 21 - EMRP-ENG05-4.3.5 Version 1.1 including CQS, with updated colorimetric metric perform better for “warm sources” and “warm led sources” than the current CRI, and (4) gamut and MCRI perform very well to predicting preference for “warm led sources” but not for “warm source” and completely fails for “cold sources”. As a result combined metric “fidelity+gamut” do not give good prediction on the whole set of lightings “cold an warm”. The ordinal scale based CRI – RCRI derived forom the CRI-CAM02UCS – seems to improve predictions for “warm LED” but obtains lower correlation for “cold LED”. The HRI, based on harmony, provides a high negative correlation for “warm LED” and a null correlation for “warm sources”. The CRI-CAMUCS globally gets the greatest correlation coefficients, but the spearman correlation for “warm sources” is still low (0,35). This subjective experiment was not conducted to give the “truth” on colour rendering metrics but was developed as a facility to assess metrics in a real environment and with real lightings, with no bias or artefacts, simply asking people to rate their preference in those real conditions. LNE still continues to investigate colour rendering metrics and colour quality specification through its own program of R&D, testing activities for consumer’s organisations, customers, and future projects. The last discussions at the CIE meetings turned around the CQS and the nCRI, a refinement of CRICAM02UCS whose specification has been recently distributed. The CIE is not a international regulation body but the most recognized organization for lighting specifications and probably the industry will follow the new metrics, single or dual, endorsed by CIE. Then our contribution will be to address a recommendation to CIE outlining the outcomes of this study and to inquire for the new metric nCRI implementation details for further assessment. - 22 - EMRP-ENG05-4.3.5 Version 1.1 References [1] CIE publication 15: 2004 – Colorimetry – third edition. [2] CIE 13.3: 1995, Publ. No. 109-1995 - Method of measuring and specifying colour rendering properties of light sources. [3] CIE Collection 1999 research note: colour rendering, TC 1-33 closing remarks [4] CIECAM97s - The CIE 1997 interim colour appearance model Publication No. 131 - 1998 [5] CIECAM02 - A Color appearance model for color management systems. Publ. 159 -2003 [6] CIE 177 - 2007 - Colour rendering of white LED light sources [7] ANSI_NEMA_ANSLG C78.377-2008-Specification for the chromaticity of Solid-State-Lighting Products. [8] ANSI_NEMA_ANSLG C78.376-2007 – Specification for the chromaticity of fluorescent lamps [9] IESNA LM-79 – Electrical and Photometric measurements of Solid-State Lighting Products [10] Two families of colour rendering judgements. F. Viénnot - CIE,TC 1-69 Budapest, June 2009 [11] C.Li et Al. “Assessing Colour Rendering Properties of Daylight Sources part II: a new colour rendering index: CRI-CAM02UCS – University of Leeds, 2011. [12] W. Davis, Y. Ohno, 2010, “Color quality scale,” Opt. Eng. 49(3), 033602, 16 pages. [13] M.S. Rea, J.P. Freyssinier-Nova, 2007, “Color Rendering: a tale of two metrics”, Col. Res. Appl. 33, pp 192-202. [14] K. Hashimoto, T. Yano, M. Shimizu, Y. Nayatani, “New method for specifying color-rendering properties of light sources based on feeling of contrast”, Col. Res. and Appl. 32, 361-371, (2007). [15] A.Zukauskas et al., “Statistical Approach to colour quality of solid-state lamps”, IEEE journal of selected topics in quantum electronics, vol n°6 (2009). [16] H. Yaguchi, Y. Takahashi, S. Shioiri, “A proposal of color rendering index based on categorical color names,” Int. Lightning Congress Istanbul, Vol. II, (2001) [17] F. Szabó, P. Bodrogi, J. Schanda, “A colour harmony rendering index based on predictions of colour harmony impression,” Lighting Res. Technol. 41, pp 165-182, (2009). [18] K. Smet,W.R. Ryckaert, M.R. Pointer, G. Deconinck, P. Hanselaer, “Colour Appearance rating of familiar real objects,” ,” Col. Res. Appl. 36, pp 192-200, (2011). [19] P. Bodrogi, S. Bruckner, T. Q. Khanh, “Ordinal scale based description of Colour Rendering,” Col. Res. Appl. 36, pp. 272-285, (2010). [20] N. Pousset, B. Rougié and A. Razet - Uncertainty evaluation for measurement of LED colour by Monte Carlo simulations Metrologia vol. 46, n° 6, p. 704-718, 2009. [21] Yi-Fan Chou, Ronnier Luo, Janos Schanda, Peter Custi, Ferenc Szabo, G Savari – Recent developments in Colour Rendering Indices and their impacts in viewing graphic printed materials. [22] M. S. Rea, & J. P. Freyssinier « The Class A Color Designation for Light Sources » Lighting Research Center, Rensselaer Polytechnic Institute, Troy, NY, USA http://www.lrc.rpi.edu/education/outreachEducation/pdf/CLE6/3-Rea.pdf - 23 -