White light source towards spectrum tunable lighting — A review

Abstract

CIE standard illuminants are important for testing, calibration, printing quality testing, spices color testing and other photometric and calorimetric applications. Currently, fluorescent light sources represent these illuminants giving fixed and broad illumination spectra. Installation of all fluorescent illuminants in a light booth for testing and calibration of light sources becomes difficult and costly. LEDs with their efficient color changing capability can replace all the fluorescent illuminants by a single LED light source producing tunable illumination spectra saving lot of space, cost and power. Implementation of all the standard illuminants with LEDs with an optimum value of CRI and lumen efficacy is a challenging task. Proper control strategies are required to control the peak wavelength shifts for white point stability against junction temperature variations. This review paper highlights the characteristics of a white light source with all the various methods that can be used for controlling junction temperature and selecting an optimal set of LEDs for spectrum tunable lighting.

Full text

White Light Source Towards Spectrum Tunable
Lighting- A Review
Srividya R
Electrical & Electronics Department
Manipal Institute of Technology
Manipal, India
Srividya.r@manipal.edu
Dr.Ciji Pearl Kurian
Electrical & Electronics Department
Manipal Institute of Technology
Manipal, India
ciji.pearl@manipal.edu
Abstract—CIE standard illuminants are important for testing,
calibration, printing quality testing, spices color testing and other
photometric and calorimetric applications. Currently, fluorescent
light sources represent these illuminants giving fixed and broad
illumination spectra. Installation of all fluorescent illuminants in
a light booth for testing and calibration of light sources becomes
difficult and costly. LEDs with their efficient color changing
capability can replace all the fluorescent illuminants by a single
LED light source producing tunable illumination spectra saving
lot of space, cost and power. Implementation of all the standard
illuminants with LEDs with an optimum value of CRI and lumen
efficacy is a challenging task. Proper control strategies are
required to control the peak wavelength shifts for white point
stability against junction temperature variations. This review
paper highlights the characteristics of a white light source with
all the various methods that can be used for controlling junction
temperature and selecting an optimal set of LEDs for spectrum
tunable lighting.
Keywords—CIE standard illuminants,CRI ,photometric and
calorometric applications,illumination spectra,lumen efficacy
I. INTRODUCTION
Fluorescent light sources consume power and release
carbon dioxide. The cost of lighting can be reduced to nearly
half by switching to Solid State Lighting (SSL) [1]. SSL uses
semi-conductor devices such as light emitting diodes (LEDs),
organic light emitting diodes (OLEDs) to produce light rather
than tungsten filaments or gas. SSL has the potential to
improve the efficiency, appearance and quality of lighting [20].
LEDs provide a new way of producing light. LEDs are highly
advantageous with respect to their long life, energy efficiency,
non-toxicity, durability and flexibility. Compared to
conventional light sources that are capable of producing only
one correlated color temperature (CCT) with broad
illumination spectra, LEDs if properly tuned can produce all
CCTs of planckian locus with narrow spectral bands [12,24]. A
CIE illuminant is a spectral characterization that shows the
amount of energy distribution at each wavelength across the
spectrum of white light sources as shown in figure 1. CIE
illuminants are divided into multiple categories or series.
Source A is an incandescent/Tungsten filament lamp; source B
& C represent noon daylight and average daylight and can be
produced by placing a liquid filter in front of A; since earth’s
atmosphere acts as a filter sunlight falling on surface may not
be so accurate and can have variations during the day.
Therefore, D series where introduced which represent natural
or real daylight; F series are approximated by fluorescent lamps
and E represents equal energy radiator producing constant
spectral power distribution (SPD). The light sources that are
represented as a particular standard illuminant are designed and
manufactured to approximate that ideal characterization [2].
Figure 1: Spectral power distribution of CIE illuminants
There are three methods to produce white light with LEDs. (1)
Mixed color LEDs (RGB), (2) blue LED + phosphor and (3)
ultraviolet LED + phosphor. External color control of
phosphor converted LEDs is not possible because their color is
fixed by the materials used in the composition of phosphor.
Compared to phosphor LEDs mixed color LEDs (RGBs) can
be individually controlled by providing separate driving
currents. Thus with proper tuning RGBs are capable of
producing white light of different color temperatures. This
paper is organized as follows: section 2 explains the concepts
behind the generation of white light using RGB LEDs. In
section 3 we discuss the different requirements of a white light
source. Section 4 describes the various techniques that can be
used for control of junction temperature. In section 5 we
discuss the various requirements for optimizing a light source
for spectrum tunable lighting.

II. WHITE LIGHT USING RGB LEDS
An LED is a monochromatic source of light that can
produce only one color depending upon its composition of
materials. When several such LEDs are grouped together as a
luminarie and when forward current and intensity of each LED
is controlled it can behave as a polychromatic source of light
producing millions of colors. RGB additive color model is
used for producing all colors in a fixture or luminarie whereas
subtractive color model is used for reflective surfaces. In
1931, International Commission of Illumination (CIE)
published Chromaticity diagram which defines the entire
range of colors that are visible to an average viewer. The three
color points of primary RGB LEDs is represented by a triangle
as shown in figure 2. Theoretically all color points inside the
triangle can be produced but practically as RGB LEDs
dissipate lot of power and as they are usually preferred to be
controlled digitally it can produce only samples of colors
within the triangle [17]. An 8-bit tricolor LED can generate
approximately 16.7 million colors. The ability of full color
luminaries to produce virtually any color without filters and
other components firmly differentiates LED lighting from
other conventional lighting.
Figure 2: CIE 1931 chromaticity diagram
White point control of RGB LED light sources requires
study of many aspects such as chromaticity coordinates, tri-
stimulus output transformations,color mixing factors and color
point maintenance [3]. Chromaticity coordinates (x,y) describe
the color of the source that lies within the triangle under given
lighting conditions. z coordinate can also be calculated from x
and y values. Tri-stimulus values X,Y,Z show the absolute
amount of three primaries required to obtain the specified
color. They do not correspond to perceptual color [18]. Tri-
stimulus values are quantities based on functions derived from
average data of multiple observers.These values can be
calculated by determining the area under the color matching
functions x(λ),y(λ),z(λ) as shown in figure 3.
X = ∫ P(λ)x(λ)dλ (1)
Y = ∫P(λ)y(λ)dλ (2)
Z = ∫P(λ)z(λ)dλ (3)
Figure 3: Color matching functions
From these tri-stimulus values color point coordinates (x,y,z)
can be calculated using the linear transformations shown
below;
x = X/(X+Y+Z) (4)
y = Y/(X+Y+Z) (5)
z = 1-x-y (6)
Major limitation of CIE (x, y) chromaticity diagram is that the
color space is not uniformly distributed. Green occupies more
space than red and blue, and so it is not easy to differentiate the
colors with their distance. Two commonly used uniform color
spaces for light sources are CIE 1960 (u, v) and CIE 1976
(u’,v’). x-y plane can be transformed to u-v plane using linear
transformations for color difference calculations. It has become
a basic requirement for the color point of the source to lie on or
close to the black body locus for all white light applications.
Black body locus consists of a range of CCTs starting from
2000K to 10000K.
III. REQUIREMENTS OF A WHITE LIGHT SOURCE
The colors of objects that are visible to a viewer depend
upon the spectral distribution of light that is incident upon
them. Each color has a wavelength of its own. If the light
incident is deficient in some wavelengths then those colors will
appear black. Color rendering index (Ra) is a standard measure
used to determine the color rendition properties of a light
source. Higher the value of Ra of a light source the more
natural will be the colors under that source. Sunlight and
incandescent lamps have Ra = 100 which is the maximum
value a light source can have. Lighting industry recommends
Ra above 80 for indoor lighting and Ra above 90 for color
matching tasks. CRI is calculated according to a test procedure
established by CIE. It involves measuring the extent to which a
series of eight standardized color samples as shown in figure 4
differ in appearance when illuminated under a given light
source, relative to the reference source appropriate for the
specific color temperature [22].
Figure 4: 8 standard color samples

The special color rendering index Ri for each test color sample
is calculated as;
Ri = 100 - 4.6(ΔEi) (7)
Where ΔEi is the color difference for each sample between the
two light sources given by;
ΔE = [(ΔL*)2 + (Δu*)2 + (Δv*)2]1/2 (8)
Averaging the 8 Ri values gives the general color rendering
index Ra
Ra = 1/8 

 (9)
For warm light sources with CCT less than 5000K reference
source is an incandescent lamp. For higher CCT sources the
reference is daylight [21]. Recently, CIE has concluded that
CRI can be used for comparison of fluorescent, incandescent
and HID lamps but it cannot effectively predict the color
quality of white light LEDs. Optimizing mixed white light
LEDs with only 8 color samples provides good Ra for some
colors not all. Color quality scale is a metric that is recently
proposed for replacing CRI. CQS provides a measure for
various aspects of color rendering, chromatic discrimination
and viewer preferences taking the spectral properties of LEDs
into account. All eight color samples used in CRI are replaced
with 14 highly saturated color samples by CQS as shown in
figure 5.
Figure 5: 14 color samples of CQS
Since certain narrow band spectra of LED sources render few
colors poorly, while rendering all other colors well CIE
chromaticity is also updated in CQS using CIE (L*, a*,b*) for
finding the object color differences. The important point of
CQS is that the output show large deviations strongly.
Therefore the color differences are combined with a root mean
square instead of an average, as is used in CRI. As large
changes in color rendering scores would drastically impact
manufacturers of older technologies, the CQS was scaled in
such a way that the average score for fluorescent lamps is the
same with both the CRI and CQS. CQS is still under
development and has not been widely adopted. The Ra value of
LED light sources depends on the white spectrum [19].
Figure 6: RGB white spectrum
The white spectrum is made of individual LED spectra that are
combined to form the light source. So this spectrum is highly
dependent on the no. of different wavelength LEDs combined
and the selected wavelengths as shown in figure 6. With proper
selection of LED wavelengths RGB LEDs can achieve
required Ra values . To study the color stability of RGB light
sources color temperature and color deviation should be
evaluated. Major limitation of CIE 1931 x-y chromaticity
diagram is the lack of uniform distribution of colors. Green
shades occupy more space compared to red and blue and so it
is not easy to discriminate the difference of two colors with
their distance. Therefore, uniform color space given by CIE
1960 u-v chromaticity diagram can be used where x-y plane is
transformed to u-v plane using linear transformations. Color
deviation can be calculated as
Δuv = [(u - u0)2 + (v - v0)2]1/2 (10)
Where (u, v) are the color coordinates of the light source and
(u0, v0) are the required color coordinates. The value of Δuv
should be as small as possible for a good CRI light source [18].
IV WHITE POINT STABILITY
It is highly difficult and a challenging task to maintain the
white point within acceptable limits for RGB LEDs [13, 16]. It
arises because of changes in output flux, wavelength of LEDs
and changes in LED characteristics that occur with changes in
temperature and ageing [4]. White point can be controlled only
by proper implementation of feedback schemes that can control
the relative contributions of red, green and blue to the white
light. White point gets deviated from the black body locus
mainly due to the variation in junction temperature. The main
causes for junction temperature variation are power dissipation,
device ageing and larger forward current. The effect of this
variation is very crucial as it has the following impacts: (1)
forward voltage reduces, (2) output luminous intensity and flux
reduces, (3) shift in color point and color temperature, (4)
variation in spectra of individual LEDs, (5) peak wavelength
shifts, (6) shift in white point and CCT, (7) variation in SPD
and CRI of the light source. A brief review of all the methods
that have been proposed [7,8,9,10,14,26] for control of this
junction temperature variation using temperature sensors,
luminance sensors, photo detectors and feedback loops are
presented here.
Junction temperature of an LED cannot be directly
measured because of small substrate size. So the temperature of
heat sink on which LEDs are placed is taken as an indirect
measurement. But this measurement will also include thermal
resistance which cannot be avoided. A compensator will
normally consist of a look up table based on empirical results.
To maintain the white color point against variations of junction
temperature, interdependency values of junction temperature
with flux and wavelength of each LED should be known. The
required fluxes of RGB LEDs are calculated based on the
calculated wavelengths at different operating temperatures. The
currents required to produce that flux is also calculated for each
color LED. With all the experimental values obtained a look up
table is formed as compensator and with the help of feedback
loops temperature variations are controlled. Limitation of this
method is that interdependency values of temperature of flux
and wavelengths cannot be precisely known. And this method
cannot correct for changes in LED flux caused due to ageing.

In feedback control of LED flux photodiodes are used to
measure the flux and using simple feedback loops the values
are sent to controllers [23]. The controllers have to simply
maintain the flux to roughly maintain the white point. Since
flux variations are different for red, green and blue three
photodiodes with three separate feedback loops can be used for
white point control. This method will correct for LED ageing
and variation of LED flux with temperature [6]. But since
controller is going to adjust the fluxes of three colors in such a
way that roughly white point can be maintained wavelength
changes of individual LEDs will occur. It does not correct for
shift in wavelength with temperature. In feedback of color
coordinates method white light is controlled directly by
measuring their color coordinates. Apart from the normal RGB
sensors used for measuring intensities of LEDs special RGB
sensors are used whose spectral responses match the CIE 1931
color matching functions. These sensors consist of photodiodes
with optical filter which can directly give the values of (X, Y,
Z) color coordinates. A high degree of accuracy can be
achieved with this method as white point will be controlled
directly by using tri-stimulus values rather than controlling it
using parameters like junction temperature and luminous flux.
Combined temperature and flux feedback is an
improved method combining the advantages of both
temperature and flux feedback. So this method can be used to
control temperature variation correcting for errors in
wavelengths and LED ageing. This technique is proved to be
efficient in controlling luminous intensity against variations in
junction temperature and input power [11, 25].
In multivariable robust control, the design of ET and
EP sub-models as in figure 7, 8 can also be validated by design
of experiments full factorial method. This method shows
which of the two independent variables (junction temp, input
power) are having significant effect on luminous intensity
taken as dependent variable. From the graphical plots of figure
9 and 10 all the experimental values are taken as shown in
tabulation 1. Analysis of each model is done separately using
MINITAB.
Figure 7: Multivariable robust control [11]
Figure 8: Electrical-thermal-luminous model [11]
Table 1: Experimental values of ETR and EPR models [11]
Figure 9: ET model [11] Figure 10: EP model [11]
Factorial analysis:
Table 2: Full factorial design of ET model
Table 3: Full factorial design of EP model
The number of runs depends on no. of factors and their
levels given by the relation:
N=LF (11)
Where N is the no. of runs or experiments need to be
performed; L is the levels and F are the factors. Junction
temperature and input power are the 2 factors with two levels
each making a combination of 4 runs + 1 run for center point
as shown in table 2 and 3 for ET and EP models respectively.
TR
(°C)
EPR
(Cd/%PR)
PR
(%)
ᶲPR
(Cd)
40
14.200
50
1180
40
14.200
70
1600
40
14.200
90
2000
45
14.899
50
1160
45
14.899
70
1520
45
14.899
90
1900
50
15.572
50
1106
50
15.572
70
1499
50
15.572
90
1871
55
16.237
50
1080
55
16.237
70
1404
55
16.237
90
1719
60
16.911
50
1033
60
16.911
70
1398
60
16.911
90
1742
65
17.650
50
1000
65
17.650
0
1353
65
17.650
90
1610
PR
(%)
ETR
(Cd/°C)
TR
(°C)
ᶲTR
(Cd)
90
-12.88
40
2000
90
-12.88
50
1871
90
-12.88
60
1742
90
-12.88
70
1613
90
-12.88
80
1484
90
-12.88
90
1356
70
-10.09
40
1600
70
-10.09
50
1499
70
-10.09
60
1398
70
-10.09
70
1297
70
-10.09
80
1196
70
-10.09
90
1095
50
-7.32
40
1180
50
-7.32
50
1106
50
-7.32
60
1033
50
-7.32
70
960
50
-7.32
80
887
50
-7.32
90
814
Run
Order
Center
Point
Blocks
Input
power
Junction
temperature
Luminous
intensity
1
0
1
70
65
1350
2
1
1
90
40
2000
3
1
1
50
90
814
4
1
1
50
40
1180
5
1
1
90
90
1356
Run
Order
Center
Point
Blocks
Input
power
Junction
temperature
Luminous
intensity
1
1
1
50
65.0
1000.0
2
1
1
50
40.0
1180.0
3
1
1
90
40.0
2000.0
4
1
1
90
65.0
1610.0
5
0
1
70
52.5
1499.1

65.052.540.0
2000
1800
1600
1400
1200
1000
temperature
Mean
50 Corner
70 Center
90 Corner
power Point Type
Interaction Plot for intensity
Data Means
Simulation results
The results of regression and variance analysis are shown
below. It shows that both junction temperature and input
power are having significant effect on variation of luminous
intensity in thermal model and in case of luminous model only
power is having significant influence on the variation of
luminous intensity. The level of significance considered is
0.05. R-Sq. (Coefficient of determination) shows the
percentage of which variation in intensity is well explained by
this factorial model. Main effect and interaction effects of the
two independent variables on the dependent variable can be
clearly seen from the graphs of figures 13, 14, 15 and 16.
Figure 11: Regression and variance analysis of ET model in MINITAB
Figure 12: Regression and variance analysis of EP model in MINITAB
Figure 13: Main effect of Electrical-thermal model
Figure 14: Interaction effects of Electrical-thermal model
Figure 15: main effect of Electrical-luminous model
Figure 16: Interaction effects of Electrical-luminous model
V REQUIREMENTS FOR OPTIMIZATION OF LIGHT
SOURCE
To optimize the white light source for producing tunable
spectra with LEDs we need to select an optimal number of
LEDs with appropriate wavelengths so that all wavelengths
that are present in standard illuminant models are present in
our light source. Because as mentioned earlier if the
wavelengths reflected by a surface are absent in the source
then the surface will appear dark or gray not colored. So a
light source spectrum to have good CRI should have all
wavelengths that are present in a standard source. Therefore,
selection of an optimization algorithm for selecting the
appropriate LED set from a large number of LEDs is very
important. C C Wu [2] has proposed an optimal pruning
process overcoming the drawbacks of negative – pruning
process. In negative pruning process LEDs are selected from a
large set using negative coefficient criterion alone. But new
907050
1800
1700
1600
1500
1400
1300
1200
1100
1000
65.052.540.0
power
Mean
temperature
Corner
Center
Point Type
Main Effects Plot for intensity
Data Means
907050
1700
1600
1500
1400
1300
1200
1100
1000
906540
power
Mean
temperature
Corner
Center
Point Type
Main effect
906540
2000
1800
1600
1400
1200
1000
800
temperature
Mean
50 Corner
70 Center
90 Corner
power Point Type
Interaction Plot for Intensity
Data Means

pruning process makes the selection by considering negative
synthesis coefficients and synthesis error. In full-search
method, LEDs are selected randomly and combinations of
LED sets are made from which an optimal LED set is selected
with least synthesis error. This new pruning process is said to
follow the same methods with reduced number of
computations. As LEDs with broader Full width half
maximums (FWHMs) have more negative synthesis
coefficients many no. of times pruning process has to be done.
Therefore, for representing a white light source as CIE
standard illuminant for photometric and calorimetric
applications with very small color differences the LED set
should consist of large number of narrow band LEDs without
any phosphor coated LEDs.
CRI and luminous efficacy of an n-band light source
depends on the peak wavelengths of individual LEDs, number
of bands, bandwidth of each of the LED used and shape of the
power distribution function within each band. Luminous
efficacy can be increased by increasing the no. of bands but
improvement in CRI will not be much [5, 15]. CRI can be
improved by expanding the narrow band spectra by peak
wavelength shifts without affecting the shape of SPD. But still
longer wavelength shifts (red) affect efficacy and CRI more
than shorter wavelength shifts (blue and green). So, there is
always a tradeoff between the improvement of CRI and
luminous efficacy of a light source.
CONCLUSION
A brief discussion of the various parameters involved in the
generation of white light using RGB LEDs is made along with
specifying the various techniques for white point stability. A
small analytical analysis validating the results of [11] is also
done using DOE. White point stability is very important in
characterizing a light source as a standard illuminant.
REFERENCES
[1] N. Pousset, B. Rougié et al.,” Impact of current supply on LED
color”,Lighting Research and Technology, vol 42,pp 371–
383,2010
[2] C-C Wu,N-C Hu et al., ”Optimal pruning for selecting LEDs to
synthesize tunable illumination spectra”,Lighting Research
Technology, vol 44,pp 484-497,2012
[3] Andrew Chalmers, Snjezana Soltic, ”Towards the Optimum Light
Source Spectrum”,Advances in optoelectronics, vol 2010,ID
596825,9 pages,2010
[4] Z. Vaitonis A. Stonkus A et al.,” Effect of junction temperature
oscillation on output characteristics of a light-emitting diode under
pulse-width and pulse-frequency modulation driving modes”,IET
Optoelectronics, vol 6(1),pp 52-56,2012
[5] Snjezana Soltic, Andrew N. Chalmers, “Influence of Peak
wavelengths on Properties of Mixed-LED white-Light Sources”,
Advances in optoelectronics, vol 2010, Article ID 437564,8
pages,2010
[6] S. K. Ng,K. H. Loo et al.,” Color Control in RGB Driver System
Applicable to LED of All Ageing Conditions”,IEEE annual
conference of Industrial electronics society ,pp2186 - 2191 ,2012
[7] XiaohuiQu, Siu-Chung Wong et al., “Temperature Measurement
Technique for Stabilizing the Light Output of RGB LED Lamps”,
IEEE transactions on instrumentation and measurement, vol.
59(3), pp 661 – 670, 2010
[8] Folkert D, Roscam Abbing et al., ”Light-Emitting Diode Junction-
Temperature Sensing using Differential Voltage/Current
Measurements”,IEEE Sensors conference, pp 861 – 864,2011
[9] K. H. Loo, Wai-Keung Lun et al., “On Driving Techniques for
LEDs: Toward a Generalized Methodology”, IEEE transactions on
power electronics, vol. 24(12),pp 2967-2976,2009
[10] W.A. Rodrigues, L.M.F. Morais et al., “Comparative analysis of
power LEDs dimming methods”, IEEE annual conference of
Industrial electronics society, pp 2907 – 2912, 2011
[11] Fu-Cheng Wang, Chun-Wen Tang et al., “Multivariable Robust
Control for a Red–Green–Blue LED Lighting System”,IEEE
transactions on power electronics, vol 25(2),pp 417 – 428,2010
[12] S. W. Brown, J. P. Rice et al., “Spectrally Tunable Sources for
Advanced Radiometric Applications”, Journal of Research of the
National Institute of Standards and Technology, vol 111(5),pp
401-410,2006
[13] K. Liang, W. Li et al., “Color measurement for RGB white LEDs
in solid-state lighting using a BDJ photo detector”,International
journal of displays,vol. 30,pp 107-113,2009
[14] Tai-Ping Sun, Chia-Hung Wang,”Specially Designed Driver
Circuits to Stabilize LED Light Output Without a Photo
detector”,IEEE transactions on power electronics, vol. 27(9),pp
4140- 4152, 2012
[15] Zhang Lei, et al.,“Color rendering and luminous efficacy of
trichromatic and tetra chromatic LED-based white
LEDs”,Microelectronics Journal, vol 38,pp 1-6,2007
[16] Savitha G. Kini, Abhilash Prabhu, “Simulation Study on
Controller Design for Color Mixed White Light LED Panel for
Studio Lighting”,IEEE conference on industrial electronics and
applications, pp 1582 – 1587,2012
[17] Subramanian Muthu, Frank J. P. Schuurmans et al.,”Red, Green,
and Blue LEDs forWhite Light Illumination”,IEEE journal on
selected topics in quantum electronics, vol. 8(2),pp 333-338, 2002
[18] Yoshi Ohno, Wendy Davis, “Rationale of Color Quality Scale”,
Article by National Institute of standards and technology,2010
[19] Lee Boon Hooi, ”Understanding RGB LED mixing ratios to realize
optimal color in signs and displays”, LEDs magazine, pp 67-72,
2013
[20] Daniel A. Steigerwald, Jerome C. Bhat et al., ”Illumination With
Solid State Lighting Technology”,IEEE journal on selected topics
in quantum electronics, vol. 8(2),pp 310-320, 2002
[21] Subramanian Muthu, Frank J. P. Schuurmans et al.,“Red, green
and blue LED based white light generation: Issues and control”,
IEEE conference Industry applications, vol. (1), pp 327 –
333,2002
[22] P.Deurenenberg, C.Hoelen et al.,“Achieving color point stability in
RGB multi-chip LED modules using various color control loops”,
Proc. SPIE,vol. 5941,pp 63-74,2005
[23] S.Robinson and I.Ashdown,”Polychromatic optical feedback
control,stability and dimming,” Proc.SPIE,vol.6337, pp. 633714-1-
633714-10,2006
[24] K. Lim,J.C.Lee,G.Panotopulos et al.,”Illumination and color
management in solid state lighting,” in proc. IEEE Industrial
application conference,$1th IAS Annual meeting,Tampa,Fl,pp
2616-2620,2006
[25] B.J.Haung and C.W Tang,”Thermal-electrical-luminous model of
multi-chip polychromatic LED luminarie,” Appl. Therm.Engg.,
vol. 29,no.16,pp.3366-3373,2009
[26] S.Y.Lee,J.W.kwon et al.,”New design and application of high
efficiency LED driving system for RGB-LED backlight in LCD
display”, in Proc. 37th IEEE Power Electron. Spec. Conference, pp
1-5,2006