IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 59, NO.

3, MARCH 2010 661

Temperature Measurement Technique for Stabilizing

the Light Output of RGB LED Lamps
Xiaohui Qu, Siu-Chung Wong, Senior Member, IEEE, and Chi K. Tse, Fellow, IEEE

AbstractThe efficiency of light-emitting-diode (LED) lights linear current driver [8]. However, in practice, LEDs are often
approaches that of fluorescent lamps. LED light sources find more driven by switching power converters for better efficiency,
applications than conventional light bulbs due to their compact- causing difficulty in measuring signals that were buried in a
ness, lower heat dissipation, and real-time color-changing capa-
bility. Stabilizing the colors of redgreenblue (RGB) LED lights noisy switching environment. This setup also poses engineering
is a challenging task, which includes color light intensity control challenges in correcting color drift due to device and tempera-
using switching-mode power converters, color point maintenance ture variations.
against LED junction temperature change, and limiting LED In this paper, we will focus on color drift due to temperature
device temperature to prolong the LED lifetime. In this paper, we variations. The nonlinearity of the forward driving current to
present a LED junction temperature measurement technique for
a pulsewidth modulation diode forward current controlled RGB
LED light and color output can dramatically be reduced using
LED lighting system. The technique has been automated and can the pulsewidth modulation (PWM) control of the fixed forward
effectively stabilize the color without the need for using expen- current drive, resulting in a one-step calibration of the device
sive feedback systems that involve light sensors. Performance in variation compensation at some chosen light intensity outputs.
terms of chromaticity and luminance stability for a temperature- Lacking a proper method of measuring the junction temperature
compensated RGB LED system will be presented.
of the LED, a feedback system that uses RGB sensors has
Index TermsColor maintenance, light-emitting-diode (LED) been proposed to keep track of the output light intensity to
junction temperature measurement, light-emitting-diode (LED) correct temperature variation, which is impossible to measure
lighting, redgreenblue (RGB) LED, temperature compensation.
[9]. However, the feedback control uses a light intensity sensor
whose value varies as the LED dims down; thus, measuring
I. I NTRODUCTION
errors can be quite large when the intensity gets small. Fur-
URRENTLY, fluorescent lamps represent the most pop-
C ular lighting solution due to their high luminous efficacy
and low running cost. However, because of their limited pro-
thermore, current technology that uses A/D converters, which
have a higher fractional change for a fewer number of voltage
representations, makes the system less accurate for a higher
grammability, fluorescent lamps cannot meet the requirements dimming factor. LED brightness compensation for temperature
of many modern applications. Unlike fluorescent lamps, red and device aging variation has been proposed in [12]. However,
greenblue (RGB) light-emitting diodes (LEDs), with their ca- the work in [12] employs a temperature sensor, which may not
pability to generate instantly different colors and intensities, are give the correct junction temperature of the LEDs, particularly
expected to find many applications in areas such as biomedical when several LEDs share a common package and heat sink. In
apparatus [1], detector systems [2], liquid-crystal-display back- this paper, we propose a simple practical technique for color
lighting [3], [4], and general decorative illuminations. Thus, control by measuring the optical and electrical characteristics
LEDs are expected to become a major kind of light sources in of the LED during the turn-on duration of the PWM forward
the coming decades [5]. driving current. The forward-voltage-to-junction-temperature
LEDs are direct band-gap semiconductor p-n junction variation is roughly linear under a fixed driving current. This
diodes. The band gap of the semiconductor, which was con- property has widely been exploited to sense the ambient tem-
trolled by mixing different proportions of the IIIV elements perature. Thus, temperature compensation can be achieved by
from the periodic table, defines the light-color frequency of the instantly detecting the diode forward voltage with some proper
diode [7]. The band gap of the LED may therefore change from noise reduction techniques.
time to time during operation, where the junction temperature This paper is organized as follows. Section II gives an
changes with power dissipation and the ambient temperature. overview of the RGB LED color-lighting system. A LED
LED junction temperature estimation has been done using a diode junction temperature compensation technique is given in
Section III. Implementation details of the system are given in
Manuscript received September 2, 2008; revised December 5, 2008. First
published September 22, 2009; current version published February 10, 2010.
Section IV. The performance of the proposed compensation
This work was supported in part by Hong Kong RGC General Research Fund technique is evaluated in Section V. Section VI concludes this
under Grant PolyU5306/06E. The Associate Editor coordinating the review paper.
process for this paper was Dr. Juha Kostamovaara.
The authors are with the Department of Electronic and Information Engi-
neering, Hong Kong Polytechnic University, Kowloon, Hong Kong (e-mail:
enscwong@polyu.edu.hk). II. O VERVIEW OF RGB LED C OLOR C ONTROL
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. The data points on the International Commission on Illumi-
Digital Object Identifier 10.1109/TIM.2009.2025983 nation (CIE) 1976 chromaticity diagram in Fig. 1 gives the

0018-9456/$26.00 2009 IEEE

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
662 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 59, NO. 3, MARCH 2010

the dimming factor, which is denoted by d = (dr dg db )t and is

given by

d = f T (X) (1)

where f T is a column vector function whose parameters

change with T , which is a column vector of the LED junction
temperatures, i.e., T = (TR TG TB )t . Note that T depends on
both d and the ambient temperature.
Without precise information of temperature change inside
the LED diode junction, the control of the light output from
(1) can only be achieved by a feedback loop that senses the
light outputs from the RGB LEDs and adjusts the control
parameter d to compensate for the light output drifts as defined
from f T [9]. Indirect temperature compensation techniques,
e.g., sensing the temperature of the heat sink on which the
LEDs are mounted [10], [12], may suffer from poor response
to the change in temperature [6]. Although the method in [6]
Fig. 1. Color coordinates of a typical LED with center wavelengths that directly measures the LED forward voltage, it does not use
range from 700 nm to 380 nm. Adjacent dots are separated by 5 nm in center
wavelength. LEDs with center wavelengths of 624 nm (red), 525 nm (green),
PWM control for simplicity. The method of two-diode driving
and 465 nm (blue) can generate color points such as white, light red, light green, voltages [2] may cause excessive error in estimating the diode
and light blue within the triangle. junction temperature and increase the system complexity. Thus,
light sensing is still important to ensure high performance.

color coordinates (u , v ) of typical RGB LEDs with wave-

III. T ECHNIQUE OF J UNCTION T EMPERATURE
lengths of 380700 nm for a step size of 5 nm. The diagram
C OMPENSATION AND E XPERIMENTAL M EASUREMENTS
was made in such a way to give an even color perceptible
difference for the distance of two color points, independent Equation (1) can be highly nonlinear if an inappropriate
of the absolute
 positions of the color points. A color distance control method or a working color range or both are used [6]. It
of u v = (u )2 + (v )2 < 0.002 is indistinguishable to is widely known that dimming using PWM is preferred over the
humans. Errors within this small distance are considered very use of amplitude modulation because of the linear relationship
acceptable for most applications. Mixing the light intensity of between the duty cycle d and the LED color light output. We
the three LEDs with wavelengths of 624, 525, and 465 nm can have confirmed this approach by experimental measurement, as
produce all colors within the triangle whose corners are at the shown in Fig. 2(a)(c). The tristimulus X = (X Y Z)t linearly
color coordinates of the RGB LEDs. The color accuracy of light changes with the junction temperature and, hence, with the A/D
that was mixed by the RGB LEDs is therefore dependent on converted values of RGB diode voltages when drivers operate
the following two factors: 1) the stability of the color points of at a stable duty cycle, as shown in Fig. 2(d). In addition, the
the RGB LEDs and 2) the accuracy of the LED-light-dimming tristimulus X linearly changes with the duty cycle at a stable
method. junction temperature. Measurement errors due to switching
The color accuracy at color points of white, light red, light noise and digitization can be reduced by boxcar averaging.
green, and light blue within the triangle in Fig. 1 will be used Therefore, we can simply measure the aforementioned linear
for the evaluation of the proposed compensation technique in correspondence at several duty cycle values, e.g., as shown
Section V. in Fig. 2(a)(c), and accordingly derive the corresponding
We assume that the light stability of color points of LEDs tristimulus at the other duty cycle values. The linearity of the
is largely dependent on the junction temperatures TR , TG , and dependence of RGB lights on the RGB diode voltages makes
TB of the RGB diodes. Similar to other previous studies [3] programming very simple. Moreover, the color points of the
[7], [9], [10], the effects of component aging are ignored here. LEDs with center wavelengths of 624 nm (red), 525 nm (green),
Such effects can be compensated using a similar technique and and 465 nm (blue) in Fig. 1 drifts with the increase in junction
by taking into consideration the thermal history in relation to temperature caused by heat sink temperature are very signifi-
device aging. Alternatively, the aging speed can effectively be cant, as illustrated in Fig. 3, where the color variation u v can
controlled by limiting the LED temperatures below a maximum be as high as 0.026 for a change in heat sink temperature from
threshold. The aging process, thus, is very slow and does not 30 C to 82 C for the blue LED.
affect the result of our proposed temperature compensation We use the Lamina BL-4000 RGB LED light engine [13] to
technique. If desired, it can also be compensated by regular illustrate how essential parameters that were contained in f T of
calibrations using additional LED light measurement fixtures (1) can be found. Each color channel of the RGB LED consists
for specific applications. of a series connection of two corresponding color LEDs. The
An intended light output is usually expressed in terms of the LEDs essentially behave as diodes with a higher cut-in voltage.
so-called tristimulus, which is denoted by X = (X Y Z)t , and At a typical driving current of 350 mA to each color channel,

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
QU et al.: TEMPERATURE MEASUREMENT TECHNIQUE FOR STABILIZING LIGHT OUTPUT OF RGB LED LAMPS 663

Fig. 2. Experimental light tristimulus X versus diode voltage (digital) of (a) red LED light, (b) green LED light, and (c) blue LED light at different duty cycles.
Diamonds, circles, and triangles are data points for X, Y , and Z, respectively, and lines are fitted with the data points. Duty cycles descend from top to bottom
with d = 1.0 to d = 0.4, respectively, for each component in X. Note that Z in X of the red light is almost zero and is not shown. (d) Relationships between the
actual (two diodes in series connection) diode forward voltage and digital temperature Vdi for i = r, g, b. The data points are measured using heat sink temperature
as a varying parameter, which ranges from 20 C to 75 C, in obtaining the corresponding diode forward voltages (junction temperatures).

the LED produces forward diode voltages of 4.5, 6.7, and 7.6 V We will show later in Section IV that the heat sink tempera-
for the RGB channel LEDs, respectively. At this driving cur- ture and junction temperatures indicated by the diode forward
rent, they consume a total electrical power of 6.7 W and provide voltages can broadly vary, depending on the diode working
a total luminous flux of 120 lm. At any forward diode voltage conditions.
below 3 V, the LEDs consume virtually no electrical power. In As explained in Section II, the mixed color is therefore
our experimental setup, a constant current of 330 mA during given as
the turn-on duration of the PWM cycle is applied to all the
RGB LEDs at ambient temperature. We have measured the data Xo Xor Xog Xob
in Fig. 2(a)(c) to obtain X or , X og , and X ob versus the Yo = dr Yor + dg Yog + db Yob (2)
diode forward voltages of the LEDs, i.e., Vdr , Vdg and Vdb , at Zo Zor Zog Zob
(dr dg db ) = (1, 1, 1), (0.8, 0.8, 0.8), (0.6, 0.6, 0.6), and (0.4,
0.4, 0.4). We use the LED heat sink temperature as the variation where
parameter. Sufficient time is allowed for the forward diode
voltages to settle down before taking measurements. Thus, the Xoi i1 Vdi + i2
diode forward voltages (junction temperatures) are measured Yoi = i3 Vdi + i4 (3)
with one-to-one correspondence to the heat sink temperatures. Zoi i5 Vdi + i6

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
664 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 59, NO. 3, MARCH 2010

Fig. 3. Evaluation of color points of the LEDs with center wavelengths of 624 nm (red), 525 nm (green), and 465 nm (blue), which were marked with small
diamonds and change with different heat sink temperatures for a constant-current-driven (a) red, (b) blue, and (c) green LEDs. Lines are least square fits of data
points using second-degree polynomials. (d) Corresponding drifts of color points relative to the color point at a heat sink temperature of 30 C. Note that the color
space of the CIE1976 uniform chromaticity scale is used here to give an even perceptual color difference.

with i = r, g, b and ij being the coefficients of least square fits 9X

v = . (5)
in Fig. 2(a)(c). X + 15Y + 3Z
X or , X og , and X ob are linearly independent; thus, it is
always possible to determine a unique d such that (2) is The whole process can automatically be done with the help of
satisfied. If X or , X og , and X ob are temperature invariant and a well-calibrated visible-light spectrometer.
d is well controlled, the light-color output will be very stable,
as given in (2). However, the three color bases X or , X og , and
IV. I MPLEMENTATION OF RGB LED J UNCTION
X ob are temperature dependent, as given in (3). Using (2) and
T EMPERATURE M EASUREMENT T ECHNIQUE
(3), the desired color can be achieved. This technique will be
IN L IGHT -C OLOR S TABILIZATION
implemented in Section IV and evaluated in Section V.
The transformation of (X, Y, Z) to other CIE color space The Lamina BL-4000 RGB LED light engine is selected for
systems is monotonic. For instance, the relation between XY Z detailed study and illustration. We use switching converters as
(tristimulus) and u v (chromaticity) is simply LED drivers to improve the power efficiency. Each RGB light
channel is separately driven by a current-controlled switch-
4X ing converter, as shown in Fig. 4. The switching converter
u = (4)
X + 15Y + 3Z essentially behaves as a current source whose magnitude is

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
QU et al.: TEMPERATURE MEASUREMENT TECHNIQUE FOR STABILIZING LIGHT OUTPUT OF RGB LED LAMPS 665

Fig. 4. Switching converter with PWM control as the constant current driver.

Fig. 6. LED lamp with the LED microcontroller system, which is composed
Fig. 5. Schematic of the RGB LED color control system. of three PWM current drivers and PIC18F1320 control circuitries.

controlled by a PWM signal. Our converter is modified from

a control IC LT1510 designed for battery chargers that operate
at a switching frequency of 200 kHz. Component values have
been optimized for fast transient of the output current under
a PWM gate pulse VPWM , which is applied at a frequency of
200 Hz to the metaloxidesemiconductor transistor, as shown
in Fig. 4.
Three identical circuits, each resembling Fig. 4, are used
for driving the RGB LED channels, the schematic of which
is shown in Fig. 5. The LED drivers are controlled by the
microcontroller PIC18F1320. Fig. 6 shows a photo of the LED
with the microcontroller and power converters. Fig. 7 shows
typical waveforms of the forward voltage across the RGB
diodes under a constant current at 330 mA, which is pulsewidth
modulated for brightness control. The power flow of the buck
switching power converter in Fig. 4 is unidirectional. When
the current source is turned on, the diode voltage abruptly
increases. When the current source is turned off, the small
ripple filtering capacitor of 0.1 F discharges its energy to the
diode. The diode voltage rapidly drops when it is above the Fig. 7. LED diode forward voltage waveforms. Upper trace: red LED.
diode cut-in voltage, where the diode current is at its high level. Middle trace: green LED. Lower trace: blue LED.

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
666 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 59, NO. 3, MARCH 2010

Fig. 8. Functional block level and gain adjustment of Fig. 5.

Fig. 10. Comparison of digital diode forward voltage Vdi and heat sink
temperature. The sampled digital diode forward voltages are shown as dots,
and the lines that were superimposed with the dots are calculated using (9). The
LEDs are driven by a PWM current of 330 mA and a duty cycle of d = 0.5.
The heat sink temperature is kept at 25 C, whereas a step change from d = 0.5
to d = 0.8 is applied to the blue channel.

Fig. 9. Comparison of digital diode forward voltage Vdi and heat sink
temperature. The LEDs are driven by a PWM current of 330 mA and a duty
cycle of d = 0.8. The heat sink temperature is externally controlled for a step
change from 25 C to 35 C at t = 9 s.

When the diode voltage falls below the diode cut-in voltage,
the diode current becomes very small. Thus, we observe a slow
decay of the diode voltage after the current source is turned
off. However, a charge of 0.1 F is small; thus, it negligibly
contributes to the brightness of the LEDs. After the current
source is turned on, sufficient time is allowed for the forward
diode voltages to settle down. Then, the voltage during the turn-
on duration is detected and converted to a digital format suitable Fig. 11. Automatic measurement and calibration suite.
for PIC18F1320 programming.
During the calibration phase, the digital format of the scaled
simple integrator formula, i.e.,
diode turn-on voltages, as shown in Fig. 2(d), is sent from
the PIC18F1320 to a personal computer through the RS232
Vdi,current = Vdi,measured + (1 )Vdi,previous (9)
connection at a rate of 200 samples per second. A voltage
scaling circuit in Fig. 8 is used for the functional block level
where the current value of the digital temperature Vdi,current is
and gain adjustment. The scaling relationships are given as
taken as the weighted value of the measured one Vdi,measured
and the previously updated one Vdi,previous , and is essen-
Vdr = 6328.2Vr 22533 (6) tially an integration constant that is adjusted to smoothen the
Vdg = 3286.9Vg 17968 (7) averaged waveform. In continuous time, (9) can be written
[14] as
Vdb = 4577.6Vb 28059 (8)
dVdi (t) ln(1 )
where Vi (i = r, g, b) is the actual diode turn-on voltage = (Vdi (t) Vdi,measured ) (10)
dt
(in volts).
The digital temperature Vdi (i = r, g, b) that was boxcar where is the time between measurements. Taking expecta-
averaged for noise reduction is further smoothened using a tion of the aforementioned linear equation gives a noisefree

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
QU et al.: TEMPERATURE MEASUREMENT TECHNIQUE FOR STABILIZING LIGHT OUTPUT OF RGB LED LAMPS 667

Fig. 12. Evaluation of output color points, which were marked as small diamonds and change with different heat sink temperatures for set-point colors at
(a) light red, (b) light blue, (c) light green, and (d) D65 white, marked as small triangles. Lines are least square fits of data points using second-degree
polynomials.

equation, i.e., of the RGB LEDs broadly vary, although the heat sink is
maintained at a temperature of 25 C. The digital temperatures,
dVdi (t) ln(1 )   which are shown as points in Fig. 10, are sample points that
= Vdi (t) Vdi,measured (11) were taken from the A/D circuitry of the microcontroller. The
dt
digital noises are effectively filtered out by (9), giving solid
which gives Vdi (t) Vdi,measured = Vo exp((t/ )), where lines in Fig. 10 that essentially follow the trends of the noisy
= (/ ln(1 )), and Vo is the disturbance amplitude. points at the expense of a small time delay.
Here, we use = 0.05 to give a response time constant of The computer is also connected to a spectroradiometric sys-
0.1 s, as shown in Fig. 10. tem from Oceanoptics Inc. to measure the absolute or relative
Then, Vdi is graphically compared with the measured diode irradiance of LED lights [11] through a universal-serial-bus
voltages. As illustrated in Fig. 9, the waveforms of the heat local bus connection, as shown in Fig. 11. Then, the Spectra-
sink temperature and diode forward voltages indicate that the Suite software calculates tristimulus XY Z, hue u v w , and
temperatures of the heat sink and RGB LED junctions broadly other correlated color parameters from the spectrum distrib-
vary during transient and converging to different values at a ution. Note that the spectroradiometric system can easily be
steady state, depending on the values of duty cycles and the heat programmed to automate the measurements. The 18 unknown
sink temperature. Fig. 10 shows that the digital temperatures coefficients in (3) can be calculated from the data that were

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
668 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 59, NO. 3, MARCH 2010

Fig. 13. Evaluation of output brightness difference in the Y value, which Fig. 14. Evaluation of output color difference using u v , which corresponds
corresponds to Fig. 12. Data points that were shown as triangles, squares, and to Fig. 12. Data points that were shown as triangles, squares, circles, and
circles correspond to D65 white, light blue, and light red, respectively, with the diamonds correspond to D65 white, light blue, light red, and light green,
set point at Y = 2600. Data points that were shown as diamonds correspond respectively. Lines are least square fits of data points using second-degree
to light green, with the set point at Y = 2000 and y-axis at the right. Lines are polynomials.
least square fits of data points using second-degree polynomials.
at (u , v ) = (0.196, 0.469), (0.32, 0.51), (0.12, 0.53), and
obtained from this computer-controlled system. The algorithm (0.17, 0.42), respectively, as shown in Fig. 1. For about a
is given as follows. For each duty cycle value equal to 1, 0.8, 50 C change in temperature of the heat sink, the position
0.6, and 0.4, we measure the tristimulus XY Z that corresponds of color point on the CIE 1976 u v plane and the brightness
to the digital diode forward voltage as the LED heat sink in Y value are plotted in Figs. 12 and 13, respectively. The
is heated up from room temperature to about 100 C. The maximum color difference is 0.0041, which has been recorded
measurement is done for each of the RGB LEDs. The results for D65 white, as shown in Fig. 14. To show the effectiveness
are shown earlier in Fig. 2(a)(c). of the proposed method in dimming operation, the brightness
By curvefitting in Fig. 2(a)(c), the coefficients can be ob- (Y value) and chromaticity coordinates for three brightness
tained as values of D65 white have been evaluated and reported in Fig. 15
for different heat sink temperatures. The measurement also
Xor 2.2421 Vdr 6832.1 employs the aforementioned spectroradiometric setup under the
Yor = 1.0281 Vdr 3183.5 (12) same condition so that the experimental results are consistent

Zor 0.0 Vdr 0.0 and reliable. The error is mainly due to switching noise, and the
precision of the duty cycle can be made very small by using

Xog
0.020625 Vdg + 818.925 boxcar averaging, digital filtering, and a high-bit-resolution
Y
og 0.232025 Vdg + 1695.625
=
(13) timer of the microcontroller. Users can simply input the desired
Zog 0.039175 Vdg + 96.865 color tristimulus values, and then, the microcontroller system
can automatically adjust the duty cycles of RGB LED drivers

Xob
0.07005 Vdb + 618.8 for stable color light intensity output without the use of any
Y =
0.01385 Vdb + 590.725 .

(14) other devices, e.g., sensors.
ob
Zob 0.406475 Vdb + 3429.75

These coefficients are then used for a test run of color accuracy VI. C ONCLUSION
in the computer. If the color accuracy is within tolerance, they Color control is an important issue in the design and man-
will be programmed into the PIC18F1320 for stand-alone real- ufacture of LED lighting systems. Due to device variation,
time control of light colors. aging, and sensing nonlinearity, achieving color precision and
standardization for a large number of LEDs is often a difficult
task. The use of the measured LED diode turn-on voltage
V. E VALUATION
during the ON state of PWM periods has been proposed to
Four color points are chosen to evaluate the performance stabilize the LED light color. The implementation technique
of the proposed method. We examine the color and bright- has been outlined and verified by some experimental data. The
ness change with temperature at four different points on the technique can effectively control the color, dramatically reduce
color coordinatesD65 white, light red, light green, and light the complexity, and eliminate the need for using expensive
bluewith corresponding CIE 1976 chromaticity coordinates feedback systems that involve light sensors.

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
QU et al.: TEMPERATURE MEASUREMENT TECHNIQUE FOR STABILIZING LIGHT OUTPUT OF RGB LED LAMPS 669

Fig. 15. Evaluation of (a) output brightness, (b) color coordinate, and (c) u v at three set points of D65 white, which change with different heat
sink temperatures. Data points that were shown as triangles, circles, and diamonds correspond to set points at Y = 1600, Y = 2600, and Y = 4200,
respectively.

R EFERENCES IEEE J. Sel. Topics Quantum Electron., vol. 8, no. 2, pp. 310320,
[1] A. E. Moe, N. Banani, L. A. Lee, B. Marquardt, and D. M. Wilson, Mar./Apr. 2002.
Enhanced fluorescence emission using a programmable reconfigurable [8] E. M. Sa, F. L. M. Antunes, and A. J. Perin, Junction temperature
LED-array-based light source, in Proc. Int. Conf. Eng. Med. estimation for high-power light-emitting diodes, in Proc. IEEE Int.
Biol. Soc., Sep. 2004, vol. 3, pp. 20902093. Symp. Ind. Electron., Aug. 2007, pp. 30303035.
[2] K. Saucke, G. Pausch, J. Stein, H.-G. Ortlepp, and P. Schotanus, Sta- [9] S. Muthu, F. J. Schuurmans, and M. D. Pashley, Red, green, and
bilizing scintillation detector systems with pulsed LEDs: A method blue LED based white-light generation: Issues and control, in Proc.
to derive the LED temperature from pulse height spectra, IEEE Trans. Ind. Appl. Conf., Oct. 2002, pp. 327333.
Nucl. Sci., vol. 52, no. 6, pp. 31603165, Dec. 2005. [10] S. Muthu and J. Gaines, Red, green, and blue LED based white-light
[3] C. C. Chen, C-Y. Wu, and T.-F. Wu, Sequential color LED back- source: Implementation challenges and control design, in Proc. Ind.
light driving system for LCD panels, IEEE Trans. Power Electron., Appl. Conf., Oct. 2003, pp. 515522.
vol. 22, no. 3, pp. 919925, May 2007. [11] Oceanoptics Inc., Nov. 2008. [Online]. Available: http://www.
[4] H. J. Chiu and S. J. Cheng, LED backlight driving system for oceanoptics.com/
large-scale LCD panels, IEEE Trans. Ind. Electron., vol. 54, no. 5, [12] L. De Oto and T. E. Clary, LED brightness compensation
pp. 27512760, Oct. 2007. system and methods, Eur. Patent 190 158 7A2, Mar. 19,
[5] F. Richard, Light-Emitting Diodes: A Guide to the Technology and Its 2008.
Implications. Bracknell, U.K.: BSRIA, 2005. [13] Lamina Lighting Inc., Nov. 2008. [Online]. Available: http://www.
[6] G. D. Youle, Light source control system, U.S. Patent 7 067 993, laminaceramics.com/
Jun. 27, 2006. [14] V. Misra, W. Gong, and D. Towsley, Fluid-based analysis of
[7] D. A. Steigerwald, J. C. Bhat, D. Collins, R. M. Fletcher, M. O. Holcomb, a network of AQM routers supporting TCP flows: With application
and M. J. Ludowise, Illumination with solid-state lighting technology, to RED, in Proc. ACM SIGCOMM, 2000, pp. 151160.

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.
670 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 59, NO. 3, MARCH 2010

Xiaohui Qu received the B.Sc. and M.Sc. degrees Chi K. Tse (M90SM97F06) received the
in electrical engineering from the Nanjing University B.Eng. (with first-class honors) degree in electrical
of Aeronautics and Astronautics, Nanjing, China, in engineering and the Ph.D. degree from the University
2003 and 2006, respectively. She is currently work- of Melbourne, Melbourne, Vic., Australia, in 1987
ing toward the Ph.D. degree in power electronics and 1991, respectively.
in the Department of Electronic and Information He is currently the Chair, a Professor, and the
Engineering, Hong Kong Polytechnic University, Head of the Department of Electronic and Informa-
Kowloon, Hong Kong. tion Engineering, Hong Kong Polytechnic Univer-
From April to August 2006, she was an electrical sity, Kowloon, Hong Kong. His research interests
engineer with the Power System Center, Tyco Elec- include power electronics, complex networks, and
tronics Corporation, Shanghai, China. Her research nonlinear systems. He is the author of Linear Circuit
interests include light-color control for LED lighting applications, power- Analysis (Addison-Wesley 1998) and Complex Behavior of Switching Power
factor-corrected switching regulators, and resonant ballast converters. Converters (CRC Press, 2003) and a coauthor of Chaos-Based Digital Commu-
nication Systems (Springer-Verlag, 2003) and Chaotic Signal Reconstruction
with Applications to Chaos-Based Communications (World Scientific, 2007).
He is a coholder of a U.S. patent and two pending patents. He is currently
the Associate Editor for the International Journal of Systems Science and the
International Journal of Circuit Theory and Applications and a Guest Editor of
a few other journals.
Dr. Tse is the Editor-in-Chief of the IEEE CIRCUITS AND SYSTEMS
SOCIETY NEWSLETTER and the Associate Editor for the IEEE
Siu-Chung Wong (M02SM09) received the B.Sc. TRANSACTIONS ON CIRCUITS AND SYSTEMSPART I: REGULAR PAPERS
degree in physics from the University of Hong Kong, and the IEEE CIRCUITS AND SYSTEMS MAGAZINE. Since 1999, he has been
Hong Kong, in 1986, the M.Phil. degree in elec- the Associate Editor for the IEEE TRANSACTIONS ON POWER ELECTRONICS.
tronics from the Chinese University of Hong Kong, From 1999 to 2001, he was the Associate Editor for the IEEE TRANSACTIONS
Shatin, Hong Kong, in 1989, and the Ph.D. degree ON C IRCUITS AND S YSTEMS P ART I: F UNDAMENTAL T HEORY AND
from the University of Southampton, Southampton, APPLICATIONS. In 2005, he was an IEEE Distinguished Lecturer. He is the
U.K., in 1997. recipient of the L.R. East Prize from the Institution of Engineers, Australia, in
He is currently an Assistant Professor with the 1987, the IEEE TRANSACTIONS ON POWER ELECTRONICS Prize Paper Award
Department of Electronic and Information Engineer- in 2001, and the International Journal of Circuit Theory and Applications Best
ing, Hong Kong Polytechnic University, Kowloon, Paper Award in 2003. In 2007, he received the distinguished International
Hong Kong. His research interests include power Research Fellowship from the University of Calgary, Calgary, AB, Canada.
electronics and Internet technologies. While with Hong Kong Polytechnic University, he received two Presidents
Dr. Wong is a member of the Electrical College, Institution of Engineers, Awards for Achievement in Research, the Facultys Best Researcher Award,
Australia. the Research Grant Achievement Award, and a few other teaching awards.

Authorized licensd use limted to: IE Xplore. Downlade on May 13,20 at 1:453 UTC from IE Xplore. Restricon aply.