Title

12 – Global LEDs/OLEDs – Summer 2011 www.globalledoled.com

The color of white— color consistency with LeDsRon Steen, Xicato

If the two basic attributes of color and quantity are not correct, the very essence of lighting

is somehow missed.

Global LEDs/OLEDs – Summer 2011 – 13www.globalledoled.com

The color of white—Color consistency with LEDs

Light can be boiled down to two main attributes: amount and color. While this is a very sim-plistic way to look at the com-plex world of lighting, these

two very basic attributes define what most lighting professionals care about. Of course it can be argued many other things come into play, such as aesthetics of beam, con-struction techniques of fixtures, trims, glare, and the list can go on and on. But if the two basic attributes of color and quan-tity are not correct, the very essence of lighting is somehow missed.

We have all heard the phrase “Quality of Light,” and we probably all have a defini-tion in our mind’s eye of what quality light-ing is. I would suspect this definition is in the form of a picture and changes relative to our setting. Quality of light may be a cloud-less blue sky at noon while in the park, a well-lit kitchen where colors for food look good and you can see what you are doing, a dimly lit restaurant which has created the right mood but still provides enough light to read the menu. In any of these examples the definition of what constitutes “Quality” can be derived by the color and the quan-tity. Both topics have a plethora of texts and standards, but when we throw the acronym LED into the conversation, it seems the entire context of the discussion changes, and immediately we start talking about semiconductors, energy efficiency, tech-nology and many other items that are not necessarily core to the quantity and color.

First, let’s address the issue of quantity of light with LEDs. This topic has clearly been the focus of LEDs since their incep-tion, and a great deal has been written on the topic, which is why we will not spend any more time than this paragraph on it. What is important to point out is the “Race to Flux.” It seems the entire focus with LEDs has been, until recently, to get higher lumen packages and higher lumen per Watt (LPW), or efficacy. Clearly the vast majority of government spending for solid state lighting (SSL) has been going into the goal of LEDs being the energy saving and sustainable solution. This promise has been well documented and best shown in what is known as Haitz Law (Figure 1), which clearly shows the trend of LEDs increas-ing in the amount of light produced while price is decreasing, creating a very compel-ling story of Lumens per Dollar and sig-nificant energy savings. LEDs seem to be a panacea to solve all things lighting in every application and, as most lighting profes-sionals know, this is just not the case for many reasons. Two reasons for not going to

LEDs are not providing enough light and not being efficient enough to displace the incumbent technology. In both cases we know the LED is improving and it is simply a matter of time before a solid state solu-tion has the amount of required flux and the correct LPW to carry the day. A third reason for not converting is due to the color and quality of LED light.

Color in and of itself is a highly com-plex topic, and again there are text books and PhD courses assigned to the topic. But when it comes to LEDs, additional nui-sances come into play. The remainder of this article shall attempt to discuss these nuisances and inform the reader of things to beware of when evaluating the color quality of LEDs.

Let us begin with the basics: Light is energy and is expressed as the visible spectrum represented by the colors of the rainbow (Figure 2). The measure of each

discrete color is expressed in terms of nanometers (nm) and referred to as wave-lengths of light. Within the visible spec-trum of light, red, green and blue colors can be combined to create the color pallet.

For a relevant discussion, a basic grounding in lighting metrics is important, and the first lighting metric was provided in the 1840s by Lord Kelvin (Picture 1). Wikipedia says:

The Kelvin is a unit of measure-ment for temperature. It is one of the seven base units in the International System of Units (SI) and is assigned the unit symbol K. The Kelvin scale

1965 1970 1975 1980 1985 1990 1995 2000 2005

Cos

t pe

r lu

men

s (u

S d

olla

rs)

orfl

ux p

er p

acka

ge (

lum

ens)

102

101

100

10-1

10-2

10-3

Figure 1. Haitz’s Law: Every 10 years light output has increased by a factor of 20 while cost per lumen has fallen by a factor of 10

Figure 2. Light is energy and is expressed as the visible spectrum represented by the colors of the rainbow. (Source: Max-Planck-Institut für extraterrestrische Physik)

14 – Global LEDs/OLEDs – Summer 2011 www.globalledoled.com

The color of white—Color consistency with LEDs

is an absolute, thermodynamic tem-perature scale using as its null point absolute zero, the temperature at which all thermal motion ceases in the classical description of thermo-dynamics. The reference point that defines the Kelvin scale is the triple point of water at 273.16K (0.01˚ Celsius). The kelvin is defined as 1/273.16 of the difference between these two reference points.

The Kelvin scale is named after the Belfast-born engineer and physi-cist William Thomson, 1st Baron Kelvin (1824-1907), who wrote of the need for an “absolute thermo-metric scale”. Unlike the degree Fahrenheit and degree Celsius, the kelvin is not referred to or typeset as a degree. The kelvin is the pri-mary unit of measurement in the physical sciences, but is often used in conjunction with the degree Celsius, which has the same magnitude. Absolute zero at 0 kelvin is −273.15˚ Celsius.

While Kelvin was searching for a thermo-dynamic scale, a useful by-product was the ability to measure the temperature of fire. As most of us have learned from an early age, the blue part of a flame is the hottest part and the orange part is less hot. Ironically we refer to bluish light as “cool” and orange light as “warm” but suf-fice it to say the Kelvin scale is still used to this day to discuss color temperature of a white light source (Figure 3). As a point of reference our standard incandescent lamp is ~2850 Kelvin and a halogen source is ~3000 Kelvin.

We now fast forward to 1931 where a group of scientists assembled in France and formed an International Commission on Illumination (CIE). Based on a body of research done in the 1920’s, the CIE formed what is known as the 1931 CIE color dia-gram (Figure 4). This diagram is still com-monly used in the industry today and pro-vides a frame of reference to discuss color and color point. Going back to the colors of the rainbow, each wavelength is repre-sented going around the diagram while an

X and Y grid is established to help define a particular color point. This diagram is usu-ally shown with a line which runs the white portion of the color diagram. The line is known as the “black body locus.” Wikipedia defines this locus as: “In physics and color science, the Planckian locus or black body locus is the path or locus that the color of an incandescent black body would take in a particular chromaticity space as the blackbody temperature changes. It goes from deep red at low temperatures through orange, yellowish white, white, and finally bluish white at very high temperatures.” Very much like the flame above, the black body locus serves as a well defined refer-ence point within the lighting community

Another representation on many CIE diagrams is CCT lines. CCT is a correlated color temperature that is simply the corre-lation of the Kelvin temperature (K) super-imposed onto the 1931 color diagram, using the black body locus as the zero point. What becomes important to know

10,000

9,000

8,000

7,000

6,000

5,000

4,000

3,000

2,000

1,000

Coo

l To

nes

neu

tral

To

nes

War

m

Tone

s

HID

Lig

htH

alog

en

Figure 3. Color temperatures in K (degrees Kelvin).

Figure 4. The 1931 CIE color diagram.

black body locus

Figure 5. David MacAdam’s ellipses.

Global LEDs/OLEDs – Summer 2011 – 15www.globalledoled.com

The color of white—Color consistency with LEDs

is that CCT does not define color point but simply defines a color continuum, which is why the CCT is represented with a line rather than a point. Using 3000 CCT as a reference, one can follow the line “north” or upward on the vector and the color moves to the yellow region of the chart just as you can follow the line “south” or downward and the 3000 CCT becomes pink in shade. In both cases the color is defined as 3000 CCT although the white colors are very dif-ferent in appearance.

Approximately one decade later, with the advent of film, Kodak Company became very interested in color. In the early 1940’s, Kodak tasked David MacAdam to map the recently created 1931 color space and try and determine when the human can see the difference in color. MacAdam car-ried out a series of experiments where he showed different color swatches to a group of people and came to a determination of when the observer could see the differ-ence in color. The experiments resulted in a series of ellipses (Figure 5). The size of the ellipse varies throughout the color space and shows how the human eye has differ-ent sensitivities at different color points. For example, a significant change in X and Y color point is required within the green region prior to 50% of the observing pop-ulation being able to see the difference in color while very little movement in X, Y in the purple region will yield the ability to recognize differences. While this particu-lar metric is not commonly used, it is quite handy as a reference. It should be noted that MacAdam’s data is under pressure for being valid as only 200 subjects were tested, and there was no real age, gender or ethnic con-siderations. Be that as it may, the definition serves as a point to discuss the ability to see the difference in color, which becomes the real issue when we begin to discuss white color differences with White LEDs.

We have all probably experienced seeing differences in color from source to source within a built environment, espe-cially within office buildings, when we can witness bluish or greenish or pinkish hues from fluorescent tubes. Picture 2 is of an installation using compact fluores-cent lamps and obviously highlights the issues with color point consistency even within incumbent technology. Most light-ing designers know to specify a particular lamp manufacturer to avoid this problem. Individual lamp manufacturers have con-trolled color point within “tolerable” speci-fications, but then again there are only a few big players in the conventional lamp market. Each large manufacturer would

claim a portion of the color space specifica-tion, and as long as an individual built envi-ronment was populated with a single man-ufacturer, everything would be fine. After time, though, as alternative lamps become stocked and installed, the color consistency issue arises.

This problem is exacerbated with LEDs due to the inherent variation within the LED process. Not only is there color variation between manufacturers, there is significant color variation within a single manufacturing run. With LEDs, the playing

field is not limited to just a few major players but is occupied by hundreds of manufacturers making LED products.

The factors affecting LEDs color differences arise from three main areas: the wave length of the LED, the formu-lation of the phosphor and the coating tech-niques. Of course white light can be created using RGB or RGBA arrays of LEDs, but the predomi-nant method for creat-ing white LEDs is with a blue LED and a phosphor coating (Figure 6). We will discuss the phosphor conversion approach rel-ative to color issues.

The blue LED, or the “pump” color, has a varia-tion band of about +/- 5 nanometers. The phos-phor has variation in the conversion characteris-tics. Some of these vari-ables include particle size

Picture 2. Example of CFL color inconsistency.

White light

phosphor

InGaN LED

yellow Light

blue Light

blueChip 1 phosphor

+

Figure 6. White LEDs are created using a blue LED and a phosphor coating.

Picture 3. The phosphor is a powder that needs to be mixed with a binding material, such as a clear silicone, to be dis-pensed over the top of the chip.

16 – Global LEDs/OLEDs – Summer 2011 www.globalledoled.com

The color of white—Color consistency with LEDs

and density. The phosphor is a powder that needs to be mixed with a binding material, such as a clear silicone, to be dispensed over the top of the chip (Picture 3). This is now the other variable, the mixture of the phos-phor within the binding matrix, and the amount of phosphor material that is dis-pensed over the LED. Varying amounts of phosphor will change the conversion char-acteristics of the blue light into white.

The variation issues mentioned above all compound and have created the need for the LED manufacturer to create “bins” of product. A bin is simply a sorting of the LEDs that come off the end of the assem-bly line. The bins are traditionally sorted on three variables: flux (amount of light), forward voltage (vF—amount of voltage it takes to turn on the LED) and color. At the end of the line, each LED package is measured for the aforementioned charac-teristics and put into a bin. When product is purchased, bin codes can be specified to assure you are receiving what you want. The big problem with this approach is the inherent nature of variation. The LED man-ufacturer cannot guarantee which parts are going to be coming off the end of the line, and the buyer does not have a predict-able source of supply if extremely tight bin codes are required by the user.

Because of this variation relative to color, NEMA, ANSI and the US DoE teamed up to create a standard for color binning LEDs. The standard is ANSI C78.377 (Figure 7 and Table 1). This stan-dard basically uses the intersection point of a particular CCT vector with the black

body locus and draws a box around the intersection point. For 3000 CCT the box spans approximately 7 MacAdam Ellipses. The reasons for the specification are two-fold: 1) The current CFL specification is a 7 step MacAdam Ellipse and 2) if the stan-dard were any tighter the yields for the LED manufactures would fall and prices would potentially go up. Since the standard has been created, the major LED manufactur-ers have mostly followed the basic boxes set out by ANSI. Each of these boxes have then been subdivided into color bins. Figure 8 is a snap shot of typical binning structures from LED manufacturers. In these exam-

ples, Manufacturer A has chosen a single bin strategy of only using the ANSI stan-dard. Manufacturer D on the other hand has subdivided each ANSI box into 18 dif-ferent boxes. It would seem manufacturer D would be a good selection by simply speci-fying a single box where the color point is desired, but as mentioned before, it is diffi-cult if not impossible for the LED manufac-turer to guarantee steady supply of a spe-cific bin selection throughout the lifecycle of a product. Another complexity arises if a “bin per project” strategy is employed. This strategy requires the ability to track exactly which color bin was used for the project

Figure 7. Graphical representation of the chromaticity specification of SSL products in Table 1, on the CIE (x,y) chromaticity diagram. (Source: American National Standard Lighting Group)

Nominal cct1

target cct and tolerance (K)

target Duv and tolerance

2700 K 2725 ± 145 0.000 ± 0.006

3000 K 3045 ± 175 0.000 ± 0.006

3500 K 3465 ± 245 0.000 ± 0.006

4000 K 3985 ± 275 0.001 ± 0.006

4500 K 4503 ± 243 0.001 ± 0.006

5000 K 5028 ± 283 0.002 ± 0.006

5700 K 5665 ± 355 0.002 ± 0.006

6500 K 6530 ± 510 0.003 ± 0.006

flexible CCT (2700-6500 K)

T2 ± ΔT3 Duv4 ± 0.006

1 Six of the nominal CCTs correspond to those in the fluorescent lamp speci-fication [2]: 2700 K, 3000 K (Warm White), 3500 K (White), 4100 K (Cool White), 5000 K, and 6500 K (Daylight), respectively.2 T is chosen to be at 100 K steps (2800, 2900, …., 6400 K), excluding those eight nominal CCTs listed.3 ΔT is given by ΔT = 0.0000108×T2 +0.0262×T +8.4 Duv is given by Duv = 57700× (1/T)2 −44.6× (1/T) + 0.0085.

Table 1. Nominal CCT categories. (Source: ANSLG.)

Figure 8. Typical binning structures from LED manufacturers. (Source: Philips, Osram, Cree, Nichia web-based data sheets.)

18 – Global LEDs/OLEDs – Summer 2011 www.globalledoled.com

The color of white—Color consistency with LEDs

and any replacement products would have to be matched to the color bin if a replace-ment product is required.

So why is all this color binning stuff required if a standard has been created to say “what is good enough.” If we take the work of David MacAdam as being valid, then this metric can be applied to make the determination of what is good enough. Figure 9 shows how the MacAdam ellipse’s can be viewed. Just like rings of the trees show their age, each MacAdam ellipse around a particular center point can define the magnitude of noticeable color differ-ence. Thus, we refer to these as the number of steps of ellipse. Another term often used is Standard Deviation Color Match (SDCM). The ANSI requirement at 3000K is approximately a 7-step MacAdam Ellipse. If we take MacAdam at his word, then with anything beyond one ellipse we can begin to see the shift in chromaticity. The question is how far we have to go before it becomes objectionable. Every light pattern shown in Picture 4 is within the definitional confines of what is considered 3000 Kelvin or CCT by the ANSI standard. While the camera does not represent the magnitude as well as the human eye, clearly we can see the dif-ference in color between the right and left

beam in each picture. Using the 7-step pic-ture as an example, the pattern on the left is running north on the 3000 CCT vector line towards the yellow region while the pattern on the right is running south towards the pink region of the 1931 CIE Diagram. If the color point were to run “Northwest” in the 3000 CCT ANSI box from the center point, the light would appear green in tone rela-tive to a point in the “Southeastern” corner appearing reddish.

As we would assume, the closer we approach to 1 MacAdam ellipse, the less color difference there is. The picture labeled

1x2 Step represents 2 beams with a color point spatial differential of 1 Ellipse in the Y direction and 2 Ellipse’s in the X direc-tion. In field research by Xicato, the 1x2 step spatial difference is the maximum allowed variation. Additionally there is a generally accepted rule of thumb which defines less than 2 MacAdam Ellipses are required when adjacent fixtures are ren-dered against a white wall and less than 4 MacAdam Ellipses are acceptable when lighting a multi-colored scene such as pro-duce in a grocery store. In most cases the 7 step as accepted in the ANSI standard is considered unacceptable by lighting pro-fessionals but indeed may be good enough in non-critical areas. Way finding, path-way and street lighting may be an example where a 7 MacAdam Ellipse tolerance could be accepted. As it is often said in lighting “It is all about the application.” Color consis-tency is no different.

Everything discussed so far has been about color point consistency of a prod-uct coming “out of the box.” The other important attribute is how well the prod-uct maintains color overtime. So far, the LED lighting world has been worried about lumen maintenance, and a standard has been issued by IES known as LM-80. This standard lays out a test protocol to measure how well an LED holds the lumens over time but mentions nothing about holding color over time.

If we buy into the original premise that light is about quantity and color, then our

Figure 9. Each MacAdam ellipse around a particular center point can define the magnitude of noticeable color difference.

Picture 4. Each of these light patterns is within the definitional confines of what is considered 3000 Kelvin or CCT by the ANSI standard.

Figure 10. The mapping of color movement over time where each color grouping of points represents a device under test (DUT).

www.globalledoled.com Global LEDs/OLEDs – Summer 2011 – 19

The color of white—Color consistency with LEDs

current lifetime standards are only address-ing one attribute of the two required compo-nents, and after all isn’t the LED story about energy efficiency and long life to make the sustainability story believable? Many in the LED industry are coming to the realization that color maintenance may be the limit-ing factor relative to lifetime. To confound the issue, while there are accepted models to extrapolate lumen depreciation to 50k hours of operation, there are no known statistical models to predict color spatial movement over time. Figure 10 shows the mapping of color movement over time where each color grouping of points repre-sents a device under test (DUT). Each LED in the data set is performing well within 1 MacAdam Ellipse relative to itself and stay-ing mostly within the 1x2 MacAdam step box previously referenced and well within the 4 step MacAdam ellipse reference.

In closing, color consistency and quality of color with LEDs are two critical compo-nents to the acceptance of the light source across all applications. This article has only addressed the color consistency portion of this subject but has not addressed the rea-sons to select a particular CCT, broached the issues with spectral power distribu-tions, outlined the pros and cons of com-peting LED architectures relative to color and has not addressed the very meaty sub-ject of CRI (color rendering index) which has been much maligned in recent years. If the lighting professional does not ask the tough questions about LED color consis-tency, many consumers and end customers may be disappointed in the results. Some in the LED field have said the American public is not willing to pay for quality of light, and this becomes a very interesting premise. Is the real question, “How long will the American public purchase poor quality of light?” Since lighting is in transformation from the vacuum tube to solid state, the market will tell us all what is good enough.

Ron Steen is a veteran of the LED world and started playing with LEDs in 1996 while working at General Motors. Ron success-fully launched the first full function LED tail lamp on the Cadillac DeVille in 1999 and did pioneering work with LED headlamps. Ron moved to the general lighting field with Philips in 2004 and was director of product management solid state lighting systems and drivers prior to his current position as VP of business development NA with Xicato, which provides LED modules to fixture manufac-tures with a focus on light quality without compromise.

WWW.IES-ENERGIA.COM