CA 02848855 2014-04-10
1
OPERATION OF A LED LIGHTING SYSTEM AT A TARGET OUTPUT COLOR
USING A COLOR SENSOR
FIELD OF THE INVENTION
The present invention relates to LED lighting systems, and more particularly
concerns a color control method for multi-chromatic LED lighting systems using
a
color sensor.
BACKGROUND
Light-emitting diodes (LED) lighting systems, emitting either white light or
colored
light, are used for numerous applications such as interior and exterior
lighting,
decorative lighting, entertainment and the like. LED lighting systems are
typically
composed of a plurality of individual LED emitters each having a different
narrow
spectral bandwidth. The light output of the overall system is a colorimetric
combination of the light generated by the individual emitters.
The use of LED-based systems for lighting applications provides several
advantages. A major advantage is the superior power conversion efficiency of
LED emitters, which can reach close to 200 lumens per watt ¨ by comparison, a
typical incandescent lamp outputs only around 17 lumens per watt while a
fluorescent lamp provides around 80 lumens per watt. Other advantages of LED
emitters include their long lifetime, achieving around 100 000 hours of
lighting,
and the ability to precisely control the color of the output light. All these
advantages make LED lighting systems very attractive lighting solutions.
There are, however, a few difficulties related to LED-based lighting. Firstly,
the
spectrum of individual LED emitters tends to shift over time, resulting in an
ageing-related drift of the output color. This shift of the output color
evolves in a
complex manner and cannot be predicted by theoretical models. Secondly, the
output color can vary as a function of operation conditions such as the
CA 02848855 2014-04-10
2
temperature of the LED lighting system or the currents that drive one or more
of
the LED emitters.
The quality of the light generated by a LED lighting system affects the
perceived
colors of an illuminated scene: the color rendering property of a LED system
is
therefore a factor to be taken into account. Color rendering can be
characterized
using the CRI (Color Rendering Index), which is a color rendering metric
standardized by the CIE (Commission Internationale de l'Eclairage), or the CQS
(Color Quality Scale), which is an alternative metric proposed by the NIST
(National Institute of Standards and Technology). For example, it is
recognized in
the literature that a CRI of at least 90 is desirable for lighting
applications.
Color rendering metrics are particularly meaningful for LED lighting systems
that
generate white light. A minimum of three primary colors are required for
additive
color synthesis of white light, typically red, green and blue (RGB). Typical
LED
lighting systems with only three LED emitters cannot easily provide white
light
with good color rendering properties. LED-based lighting systems having four
or
more LED emitters with different "primary" colors can be used to reach or to
exceed the CRI threshold of 90, if appropriately controlled. At least four LED
emitters are therefore preferred for quality lighting applications, such as in
museums and for advertisement purposes. However, to obtain the desired results
the LED light system must be carefully controlled to achieve a constant color
output for all values of operation temperature and drive current, for the
lifetime of
the LED lighting system.
It is known in the art to use color feedback during operation of a LED
lighting
system in order to compensate for drifts of the output color. A color detector
is
used to "read" the color of the output light of the LED system and a
correction of
the operating conditions can be applied to compensate for any shift that
arises,
based on the information obtained. The main difficulty associated with
implementing a color feedback scheme is the measurement of the color of the
CA 02848855 2014-04-10
3
LED emitters. The CIE introduced three functions to represent the three color
receptors in an average human eye. The color of the light emitted from a
lighting
system can be represented by three quantities (X, Y, Z) that represent the
integration of a measured spectrum of the light over these functions. A direct
manner of obtaining the color of a light source is therefore to measure the
spectrum of its emitted light and then integrate over the CIE functions to
obtain
the three quantities. However, measuring the spectrum of a lighting system
requires the use of expensive equipment such as a spectrometer. An alternative
is to use a color sensor or colorimeter. Such sensors are typically composed
of
three filtered detectors and are much more affordable than spectrometers. The
filters can be selected to match the three colorimetric functions of the CIE,
in
which the sensor output corresponds to the CIE color coordinates X, Y and Z of
the detected light. In practice, it is however very difficult to obtain
filters that
correspond exactly to the colorimetric functions. Therefore, the values
outputted
by color sensors do not really correspond to the (X,Y,Z) quantities
characterizing
the color of the output light.
There therefore remains a need for a control method for a LED lighting system
that alleviates at least some of the drawbacks above.
SUMMARY
In accordance with one aspect of the invention, there is provided a method for
operating a LED lighting system at a target output color. The LED system has
three or more LED emitters emitting light of different colors combined into a
light
output, each LED emitter being operable at a controllable emitter drive
setting.
The LED lighting system further includes a color sensor.
The method includes the steps of:
a) providing calibration data which includes values for the light output of
the
LED lighting system for a plurality of values of the emitter drive settings of
the emitters. At least one calibration matrix defining a relationship between
CA 02848855 2014-04-10
,
4
measurements obtained from the color sensor of the LED lighting system
and represented by sensor color point coordinates, and absolute color
point coordinates in an absolute color space is also provided;
b) operating the LED emitters at emitter drive settings selected in view of
the
target output color and based on the calibration data;
c) measuring the sensor color point coordinates of the light output using the
color sensor;
d) determining the absolute color point coordinates of the light output based
on the sensor color point coordinates measured at step c) and the at least
one calibration matrix;
e) comparing the absolute color point coordinates determined at step d) to
target color point coordinates representing the target color to determine if
a predetermined matching condition is met. If not, repeating steps c) to e)
using different operation drive settings.
Preferably, the LED lighting system has a same optical configuration during
operation as during a calibration process having provided the at least one
calibration matrix.
In some embodiments, the calibration matrix or matrices relate non-linear
functions of the sensor color point coordinates to the absolute color point
coordinates. The calibration matrices may also include a plurality of emitter
calibration matrices each associated with an individual one of the LED
emitters.
In accordance with another aspect, there is provided a LED lighting system for
operation at a target output color.
The LED lighting system first includes three or more LED emitters emitting
light of
different colors combined into a light output, and a LED driver electrically
connected to each LED emitter. Each LED driver is configured to apply a
controllable emitter drive setting to the corresponding LED emitter.
CA 02848855 2014-04-10
The LED lighting system further includes a color sensor positioned for
measuring
a portion of said light output.
5 A memory is further provided, containing calibration data. The calibration
data
has been obtained from measuring the light output of the LED lighting system
for
a plurality of values of the emitter drive settings. The memory further
contains at
least one calibration matrix defining a relationship between measurements from
the color sensor of the LED lighting system represented by sensor color point
coordinates and absolute color point coordinates in an absolute color space.
The LED lighting system further includes a controller, configured to:
a) control the LED drivers to operate the LED emitters at emitter drive
settings selected in view of the target output color;
b) measure the sensor color point coordinates of the light output using the
color sensor;
c) determine the absolute color point coordinates of the light output based on
the sensor color point coordinates measured at step c) and the at least
one calibration matrix; and
d) compare the absolute color point coordinates determined at step c) to the
target color point coordinates representing said target color to determine if
a predetermined matching condition is met, and, if not, and repeat steps b)
to d) using different operation drive settings.
Other features and advantages of the invention will be better understood upon
reading of preferred embodiments thereof with reference to the appended
drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a perspective view of a multi-chromatic LED lighting system that can
be
controlled in accordance with embodiments of the invention.
CA 02848855 2014-04-10
6
FIG. 2A to 2D are schematic representations of LED lighting systems according
to embodiments of the invention, respectively including three (FIG. 2A) and
four
(FIG. 2B to 2D) LED emitters.
FIG. 3A is a flow chart of a pre-calibration phase for obtaining calibration
data
and FIG. 3B is a flow chart of a calibration phase for calculating one of more
calibration matrices for a specific color, according to one embodiment.
FIG. 4 is a flow chart of a calibration process to obtain calibration matrices
for
individual LED emitters, according to another embodiment.
FIG. 5 is a flow chart of a control phase of a method for operating a LED
lighting
system according to one embodiment.
FIG. 6 schematically illustrates an example of shifting the target point to
the
geometrical opposite of the measured color.
FIG. 7 shows the results of the color control using a calibration matrix based
on
one color and defining a non-linear relationship.
FIG. 8 presents a comparison of results obtaining using calibration matrices
defining linear (3x4 matrix), and quadratic (3x7 matrix) relationships.
FIG. 9 shows the results of the color control using calibration matrices for
each
individual emitter.
FIG. 10 shows the results of a control process similar to FIG. 9, except
targeting
both illuminants E and D65.
CA 02848855 2014-04-10
7
FIG. 11 shows intensity spectra measured from a RGBA LED operated at
temperatures of 20 C, 50 C and 80 C, respectively.
DETAILED DESCRIPTION OF EMBODIMENTS
In the following description, similar features in the drawings have been given
similar reference numerals and in order to avoid weighing down the figures,
some
elements may not be referred to on some figures if they were already
identified in
preceding figures. It should also be understood herein that the elements of
the
drawings are not necessarily drawn to scale and that the emphasis is instead
being placed upon clearly illustrating the elements and structures of the
present
embodiments.
The present invention generally relates to the control of multi-chromatic LED
(Light-Emitting Diode) lighting systems. LED lighting systems may be used for
numerous applications such as interior and exterior lighting, decorative
lighting,
entertainment and the like. Referring to FIG. 1, a LED lighting system 20 is
shown by way of example. The LED lighting system 20 may include three, four or
more LED emitters 22, each having a different color, controlled by appropriate
control electronics 25. In typical three-emitter embodiments, the LED emitters
22
may for example embody a ROB scheme, the LED lighting system therefore
including a red emitter 22R, a green emitter 22G and a blue emitter 22B. In
the
illustrated example of FIG. 1 a four-emitter embodiment is shown, where the
fourth emitter may typically be a white emitter 22w, therefore embodying a
RGBW color scheme. Although the description below will mostly be applied to
RGB and RGBW embodiments, it will be readily understood that the present
invention may be applied to various color schemes or number of LED emitters.
For example, some four-emitter LED devices use amber (A) or yellow (Y)
emitters instead of white ones, in addition to red (R), green (G) and blue (B)
emitters.
CA 02848855 2014-04-10
8
As known in the art, the resulting light output 21 generated by a LED lighting
system is perceived as a colorimetric combination of the individual light
beams
23R, 23G, 23B and 23w generated by the different LED emitters of the system.
Varying the relative intensities of these light beams therefore provides a
control
of the resulting overall color.
Although the present description refers to LED systems made up of three or
more
LED emitters having different colors, one skilled in the art will understand
that in
practice, a LED system may include a greater number of emitters forming groups
of same colored emitters, for example a group of red emitters, a group or
green
emitters and a group of blue emitters in a RGB scheme. The LED emitters of a
same group may be electrically connected together or operated individually. It
will
be readily understood that in such cases the present method may be applied to
one LED emitter of each group and the remaining LED emitters of the same
group controlled according to the same parameters, or, alternatively,
identical
LED emitters may each be controlled according to the principles explained
herein
without departing from the scope of the present invention.
It will also be understood that referring to "LED emitters of different
colors" is a
shorthand for indicating that the light beams generated by the respective
emitters
have different colors.
Detailed description of the LED lighting system
Referring to FIGs. 2A and 2B, the components of exemplary LED lighting
systems 20 according to embodiments are schematically illustrated. The system
20 includes three or four LED emitters of different colors, here embodied by a
red
emitter 22R1 a green emitter 22G a blue emitter 22B in both embodiments of
FIGs.
2A and 2B, and further including a white emitter 22w in the embodiment of FIG.
2B. A LED emitter is typically embodied by a chip made up of semiconductor
materials doped with impurities, forming a p-n junction. An electrical current
flows
through the junction and it generates light of wavelength determined, among
CA 02848855 2014-04-10
9
other factors, by the band-gap energy of the materials. Each LED emitter may
be
embodied by a "regular" or "direct emission" LED, or by a PCLED (phosphor-
converted LED).
The LED system is configured for operation at a target output color. The
expressions "target output color" or "target color" refer to the color of the
light
which is to be achieved by the LED lighting system, resulting from the
combination of the light beams generated by the individual LED emitters of the
LED lighting system. The target color may be described by color point
coordinates in a given color space, i.e., by a model providing a specific
mathematical representation of colors. Typical color spaces known in the art
include the CIE 1931 XYZ and the CIELAB. CIE 1931 XYZ is historically the
first
attempt to describe colors on the basis of measurements of human color
perception and it is the basis for almost all other color spaces. CIE 1931 XYZ
is
linear in terms of color mixing. This means that a target color can be
expressed
as linear combinations of E primary colors weighted by appropriate
coefficients
Ck. In matrix form:
X X1 Xk C1
Y = = = Yk = : = (1)
Z- target ¨I Z C
-
k _ k_
where X, Y, Z are the tristimulus coordinates of the target color while Xk, Yk
and
Zk are the tristimulus coordinates of each individual LED emitter k. The
CIELAB is
not linear in terms of color mixing but it is more linear than the CIE 1931
XYZ in
terms of color perception. Perceptual linearity means that a change of the
same
amount in the CIELAB coordinates produces a change of about the same visual
importance in the colors represented by those coordinates. Direct and inverse
transformation rules exist among common color spaces, so that any given color
can be expressed univocally in any chosen color space.
CA 02848855 2014-04-10
Both CIE 1931 XYZ and CIELAB color spaces can be described as "absolute"
color spaces in which colors are unambiguous, that is, the coordinates of a
given
color in the space are defined with respect to standard parameters without
reference to external factors. The coordinates of any color in an absolute
color
5 space can, by extension, be referred to as "absolute" color point
coordinates.
Other examples of absolute color spaces include sRGB, Adobe RGB, CIELUV,
and CIE RGB.
Optionally, such as in the illustrated embodiments of FIGs. 2A and 2B, the
10 lighting system 20 includes a user input 26 through which control
parameters can
be provided by the user. Preferably, the user control parameters may include
the
target color, which may be in the form of color point coordinates in a given
color
space or other information allowing deduction of the specific target color
required
by the user. The user control parameters may be provided through knobs, a
keyboard, a mouse, a touchscreen, or any other device providing a suitable
user
interface. It will however be understood that in other variants the target
color may
be preprogrammed, selected or deduced automatically without involving the
intervention of a user.
Other user control parameters may optionally include luminance, Correlated
Color Temperature (CCT), dominant wavelength, saturation, hue, etc.
The lighting system 20 further includes a LED driver 24 electrically connected
to
each LED emitter 22. The illustrated embodiment of FIG. 2A therefore includes
three LED drivers 24R, 24G, 24B while the embodiment of FIG. 2B further
includes
a fourth LED driver 24w. The LED drivers 24 may be embodied by any device or
combination of devices that can be configured to apply a controllable drive
setting
to the corresponding LED emitter. It will be readily understood that the
intensity of
the light generated by a LED emitter can be changed through a control of its
driving conditions. Controlling the drive conditions of LED emitters is
typically
CA 02848855 2014-04-10
11
achieved by acting on the time-averaged forward current injected in the LED
emitter.
In some embodiments the LED emitters 22 are controlled according to a PWM
(Pulse Width Modulation) scheme. In this case the drive setting may be a
current
modulation duty cycle, that is, the duty cycle of a periodic current waveform
having constant predetermined maximum and minimum current values, the
minimum current value being possibly an absence of current, i.e., a zero
current
value. Variants of PWM are known in the literature and may use a fixed or
variable modulation frequency, constant or variable current values and complex
waveforms. In PMW embodiments, each LED driver 24 includes for example a
modulated current source with controllable duty cycle.
In other embodiments the LED emitters may be driven according to a Constant
Current (CC) regulation method where the drive setting would be embodied by a
constant current value. In CC regulation embodiments, each LED driver 24
includes for example a continuous current source with controllable current
amplitude.
Other driving methods, such as pulse frequency modulation, pulse density
modulation or the like are also known in the art and considered to be within
the
scope of the present description.
The lighting system 20 may include a temperature determining module
configured to measure, calculate or estimate the operation temperature of the
LED lighting system 20. The operation temperature can be representative of
temperature values measured at various locations in the system. The
temperature determining module may include a device or devices, such as a
thermocouple, a thermistor or other appropriate sensor for measuring the
temperature at one or more locations in the systems. For example, in the
embodiment shown in FIG. 2B the LED emitters 22R, 22G, 22B and 22w are
CA 02848855 2014-04-10
12
shown mounted on a heat sink 29, and a temperature sensor 28 is shown also
coupled to the heat sink to measure variations in the heat extracted by the
heat
sink 29. In another variant (not shown), each individual LED emitter may be
mounted on a separate heat sink and a temperature sensor provided on each
heat sink.
In another embodiment the temperature determining module may include a
junction voltage meter (not shown) connected to each LED emitter in order to
measure the corresponding junction voltage drop. The junction voltage drop may
be used to determine the junction temperature of each LED emitter. As a LED
emitter dissipates heat when lit, the junction temperature depends on a number
of parameters including the injected current, the junction voltage drop, the
environment temperature and the efficiency of dissipation of the heat flowing
from
the junction to the environment. Since each LED emitter 22 can be operated
under different drive conditions, the junction temperature may vary from
emitter
to emitter within a same LED lighting system 20. The individual junction
temperature value of each LED emitter of a LED lighting system can be used to
estimate a value for the operation temperature of the whole system.
Still referring to FIGs. 2A and 2B, the lighting system 20 further includes a
controller 30. The controller 30 may be embodied by a microcontroller, a
processor, an electronic circuit or by any other device or combination of
devices
providing the processing/computing power required to perform the tasks
described below. The controller 30 is configured to execute the steps of the
method according to embodiments of the invention, which will be described
further below.
The lighting system 20 further includes a memory 32 containing calibration
data
for each LED emitter. The calibration data may include values for the light
output
of the LED lighting system for a plurality of values of the drive settings of
the
emitters, the light output being measured using both the color sensor and a
CA 02848855 2014-04-10
13
spectrometer, as will also be explained further below. The memory may be
embodied by any device or combination of devices apt to store the calibration
data, such as a random-access memory (RAM), a programmable or non
programmable read-only memory (ROM), a solid-state memory, an universal
serial bus (USB) flash drive, a hard-disk drive, a magnetic tape, an optical
disk or
the like.
Although the controller 30, memory 32 and LED drivers 24R, 24G, 24B and 24w
are shown in FIGs. 2A and 2B as parts of a same group of control electronics
25,
it will be readily understood that these components may be arranged in a
variety
of configurations without departing from the scope of the invention.
Referring to FIGs. 1, 2C and 2D the LED lighting system 20 further includes a
color sensor 27. The expression "color sensor" is understood to refer to a
device
capable of measuring light intensities included within at least three distinct
portions of the visible spectrum, corresponding to three primary colors, such
as
red (R), green (G), and blue (B). In some embodiments, the color sensor may be
based on optically-filtered detection of light to obtain light detection
values
associated with the different colors. Typically, color sensors include one or
more
wide-bandpass light detectors, that is, detectors that output a signal
amplitude
related to the intensity of light incident thereon irrespectively of its
wavelength,
and a plurality of optical filters disposed in the path of the incoming light
before it
reaches the light detectors. In this manner, the color sensor 27 can output
color-
specific values without the need for more complex and expensive spectrometers.
It will therefore be readily understood that the color sensor may be embodied
by
a variety of known devices such as RGB sensors or colorimeters (in which the
filters approximately match the CIE 1931 color matching functions). Very low
resolution spectrometers, measuring only a few spectral bands in the visible
wavelength range, can also be considered as color sensors.
CA 02848855 2014-04-10
14
The color sensor is preferably arranged so as to receive a portion of the
combined light outputs of the LED emitters for monitoring purposes. For
example,
FIG. 2C illustrates an embodiment where the color sensor 27 is placed in front
of
the LED emitters 22R, 22G, 22B and 22w, within the light output 21 at a great
enough distance to collect light that is well mixed. Referring to FIG. 2D, two
additional possible configurations are shown. In one such configuration, the
sensor 27 may be placed directly beside the LED emitters, so as to collect
light
backscattered from an output window 31, a conditioning lens or other optical
element placed in front of the LED emitters. In an alternative configuration,
if the
light from the LED emitter is guided by an optical component in its path, such
as
is typically the case with a plastic output window, then the color sensor 27
may
be positioned at an extremity of this output window 31, to receive a part of
the
guided light. In all of the illustrated embodiments, the color sensor is
connected
to the controller 30 to provide data signals thereto. Of course, one skilled
in the
art will readily understand that a variety of other configurations
incorporating a
color sensor within a LED lighting system are possible without departing from
the
scope of the invention.
Method for operating a LED system
In accordance with embodiments of the present invention, there is provided a
method for operating a LED lighting system at a target output color, the LED
system having three or more LED emitters emitting light of different colors
combined into a light output. As explained above, each LED emitter is operable
at a controllable emitter drive setting, for example a constant current value
in a
CC driving scheme, or a current modulation duty cycle in a PWM driving scheme.
The LED lighting system further includes a color sensor.
Although the method is described herein as applied to lighting systems such as
those shown in FIGs. 2A to 2D, it will be readily understood that other
embodiments of the present method may be used to control LED lighting systems
having different configurations.
CA 02848855 2014-04-10
,
Calibration data and calibration matrices
The method first includes a step of providing calibration data and one or more
calibration matrices.
5
The calibration data includes values for the light output of the LED lighting
system measured for a plurality of values of the drive settings of said
emitters.
In some implementations, the values for the light output may have been
10 measured by a spectrometer. As known in the art, spectrometers are devices
providing the spectral intensity profile of a light beam. Since recording
spectra
can require significant processing and storage capacity, it can be preferable
to
provide these values of the calibration data in the form of absolute color
point
coordinates in an absolute color space. In accordance with one embodiment, the
15 absolute color point coordinates may be tristimulus coordinates X, Y and
Z in the
CIE 1931 XYZ color space. As mentioned above, the tristimulus coordinates are
defined relative to color matching functions related to the perception of
colors by
the photoreceptors, or cones, of the human eye. By definition, the Y
coordinate
corresponds to the luminance, Z is nearly equal to blue stimulation and X is a
mix
of cone response curves chosen to be non-negative.
From a recorded light spectrum S(A) the tristimulus coordinates are calculated
as
follows:
X = lc, I S (A) = CMFx (2) d 2
A
Y = kef S (2) = CMFy (A) c 1 2
(2)
A
Z = k, 1 S (A) = CMFz (2)d),
A
where kc is a constant, CMFx, CMFy and CMFz are the color matching functions
specified by the CIE and X represents the light wavelength.
CA 02848855 2014-04-10
16
In other implementations, a calibrated colorimeter could also be used in place
of
the spectrometer to record the tristimulus coordinates for a plurality of
values of
the emitter drive settings.
The calibration data may therefore include the tristimulus coordinates Xi,, Yn
and
Zn measured for the light output of the LED lighting system for each one of N
values of the emitter drive settings. As explained below, this information may
be
used during operation of the LED system to determine initial drive settings
for the
emitters in view of the target color. Of course, the use of the standard CIE
1931
XYZ color space is shown here by way of example only, and in other
embodiments any other convention allowing the calculation of absolute color
point coordinates from the recorded spectra could be used.
The one or more calibration matrices define a relationship between
measurements obtained from the color sensor of the LED lighting system, these
measurements being represented by sensor color point coordinates, and
absolute color point coordinates in an absolute color space. In some
embodiments, the calibration matrices are pre-calculated as part of a
calibration
procedure that also involves measuring the light output of the LED lighting
system with a spectrometer to obtain the calibration data. Such a calibration
procedure further involves measuring the light output of the LED lighting
system
using the color sensor, and using both sets of data to calculate the
calibration
matrix or matrices. It should be noted that the color sensor used to obtain
the
data to generate the calibration matrices is the same color sensor providing
feedback during operation of the LED lighting system. It has been found that
significant variations can be observed in the outputs from different color
sensors
for a same light beam being measured, and that the calibration matrices should
be device-specific in order to provide the desired degree of exactitude.
Furthermore, it is preferred that the optical configuration used during the
operation of the LED system be the same as the optical configuration used
CA 02848855 2014-04-10
17
during the calibration process for measurements taken by the optical sensor.
All
the same components are therefore preferably used, i.e., the same LED
emitters,
the same color sensor, as well as the same additional optics such as lenses,
mirrors and the like, all in a same position relative to one another.
As mentioned above, the measurements obtained from the color sensor of the
LED lighting system are represented by sensor color point coordinates, that
is,
expressed as sensor color point coordinates in a sensor color space. For
example, when the color sensor is embodied by a RGB sensor, the output
thereof is expressed as R, G and B values for the light output.
Preferably, the relationships defined by the calibration matrix or matrices
are
calculated to be exact for at least one color point. Examples of such
relationships
will be provided further below in the context of two exemplary embodiments of
the invention.
Calibration matrix for a specific color
As mentioned above, the calibration matrix or matrices provide a
transformation
from the sensor color space, specific to the color sensor of the lighting
system, to
an absolute color space. In the example below, the sensor color space is
embodied by a RGB space and the absolute color space by the standard CIE
1931 XYZ color space, although it will be readily understood that other
conventions may be used without departing from the scope of the invention.
A common practice known in the art is to relate RGB and XYZ color spaces by a
single linear matrix transformation. However, this practice introduces errors
as
the actual relationship between the two color spaces is not exactly linear.
Furthermore, it can be shown that the calibration is highly dependent on the
colors used to calculate the calibration matrix.
CA 02848855 2014-04-10
18
In view of the above, in one embodiment a calibration matrix relating non-
linear
functions of the sensor color point coordinates to the absolute color point
coordinates may be provided. The calibration matrix may relate a system sensor
vector, including the non-linear functions of the sensor color points
coordinates,
to a color vector, including the absolute color point coordinates. For
example, the
non-linear functions of the sensor color point coordinates may include second
order terms of the sensor color point coordinates, which, in one simple form
can
be mathematically expressed as:
R -
G
FB
yi = A i
(3)
[Z] R2
G2
-B2-
where the left-hand side vector is the color vector containing the absolute
color
point coordinates X, Y and Z, A is a 3x7 calibration matrix and the right-hand
side
vector is the sensor vector containing both first- and second-order terms of
the
sensor color point coordinates R, G and B. Optionally, the sensor vector may
further include a constant term compensating for dark current effects, such as
the
"1" in the middle of the sensor vector of equation (3) above. One skilled in
the art
will readily understand that although a color sensor measuring three different
colors is presented herein, the same principles would apply to sensors having
a
greater number of channels.
In some variants of this embodiment, a single calibration matrix may be
provided
for the LED lighting system, preferably associated with an average operation
temperature of the system. In other variants, a plurality of calibration
matrices
may be provided, each associated with a different operation temperature of the
LED lighting system.
CA 02848855 2014-04-10
19
In some embodiments, the calibration matrix A is obtained from a network of
test
color points spanning a test region surrounding a calibration color point.
To obtain the elements of the calibration matrix, calibration data for several
color
points can be used. For example, for a minimum of seven color points, equation
(3) can be rewritten as follows:
Ri R2 = = = Rn-
G1 G2 = = = Gn
Xi X2 == = Xn
{ Bi B2 ... Bn
yl y2 . .. yn = A 1 1 ... 1
R4
2'1 Z2 ¨ Zn R? R=== (4)
G? G1 '" ,
_B? 131 ¨ 134_
where the indices 1 to n represent the different color points used for the
calibration. In one embodiment, a solution for the calibration matrix A can be
calculated using:
-R1 R2 ¨ Rn-+
Gi G2 = = = Gn
Xi X2 === Xn Bi B2 ... Bn
A = yi. y2 ... yn 1 1 ... 1
1
i (5)
Zi. Z2 ¨ Zn Ri R2 ¨ Rii
G? GI
_13? ¨ 1317._
where the superscript + refers to the Moore-Penrose pseudo-inverse defined by:
A 1 = (NT N)-1NT (6)
where N is a matrix having more columns than lines. Other solutions to the
system of equation (5) can alternatively be used to calculate the calibration
matrix A, as will be readily understood by one skilled in the art.
CA 02848855 2014-04-10
The choice of the color points in the network of test color points used to
calculate
the calibration matrix can have a significant impact of the performance of the
color sensor. It can be shown that using points that are close to one single
color
will give a calibration that is exact only for this specific color. A network
defining a
5 broader color sample is therefore preferred ¨ however, if the test color
points are
too far apart, the precision of the calibration will be lesser for the entire
test
region. The test color points are therefore preferably selected to strike a
balance
between these competing requirements. It can also be shown that the test
region
should preferably span the entire luminous region to be controlled by the
color
10 sensor. For example, the selected points may have color values distributed
between -20<a*<20 and -20<b*<20 in the CIELab diagram, and/or a luminosity
range corresponding to the luminance range of the target application, such as
Y
values between 5 cd/m2 and 30 cd/m2 in one implementation. Finally,
simulations
have shown that a minimum of thirty test color points are preferably used to
15 calculate the calibration matrix. This allows the calibration matrix to
better
represent the average variations observed during the operation of the lighting
system.
The above requirements for a good calibration do not include the placements of
20 the individual points within the network. In most embodiments, as long
as the test
color points spawn a large enough area, they can be located at random.
The calibration data and the data used to calculate the calibration matrix can
be
obtained through a suitable calibration process. In accordance with one
embodiment of the invention, such a calibration process may include the two
phases described below, that is, a pre-calibration phase providing a
relationship
between the drive settings applied to the LED emitters and the output color of
the
lighting system, and a calibration phase where calibration data for the
network of
test points is obtained and the calibration matrix is calculated.
CA 02848855 2014-04-10
21
Referring to FIG. 3A, the pre-calibration phase 100 of an implementation of
the
calibration process is illustrated. The pre-calibration phase 100 first
involves a
step 102 of selecting the number E of LED emitters in the lighting system. For
example, in the case of a RGB system such as shown in FIG. 2A, E would be set
to 3. A number of calibration values are to be obtained for each emitter, as
tallied
by an emitter counter k first set to 1 (step 104). For a given emitter, a
number N
of drive settings, for example the driving current of the corresponding
emitter, is
selected (step 106), and a current counter n set to 1 (step 108). For each
increment of the current counter n, the corresponding drive current is applied
(step 110) to the corresponding emitter, and a spectrum is measured with a
spectrometer (step 112). The absolute color coordinates corresponding to each
measured spectrum can be calculated using the CIE colorimetric functions and
equation (2) above. The current counter n is incremented (step 114) and a next
spectrum is measured until the maximum number of currents N has been
reached. Once the absolute color coordinates have been obtained for all the
current values for a given emitter, the emitter counter k is in turn
incremented
(step 116) until all of the emitters have been processed.
As will be seen from the above description, the end of the pre-calibration
stage
may provide the calibration data relating the light output of the LED lighting
system, when individual LED emitters are lit, to a plurality of values of the
drive
settings of the emitters. The pre-calibration phase is preferably performed at
an
average operation temperature, so that the average spectral shift of the LEDs
is
measured. Although this data will be used to determine a starting value for
the
drive settings in view of the target color during the operation of the
lighting
system, precision is not required since the present method provides fine
tuning of
the drive settings through the use of the color sensor, as explained below.
Referring to FIG. 3B, there is shown a flow chart of the calibration phase 120
of
the calibration process. In this phase, values for the light output of the LED
CA 02848855 2014-04-10
22
lighting system using both the device-specific color sensor and the
spectrometer
are obtained, and this information is used to calculate the calibration
matrix.
Optionally, in the case where calibration matrices are desired for different
temperature values, the corresponding number F of calibration temperatures,
and
therefore, the number of calibration matrices to calculate, is selected as a
preliminary step 122. The provision of one or more calibration matrices may be
determined based on the expected range of the operation temperature of a
particular LED lighting system. By way of reference, FIG. 11 illustrates the
shift in
the spectra of a RGBA LED with temperature. It can be seen that the shift is
non-
linear and that it varies from emitter to emitter. If only one calibration
matrix is
used, the temperature can be selected in the middle of the operation
temperature
range, for example 50 C for operations between 20 C and 80 C. Otherwise,
representative values spanning the range of temperatures at which the lighting
system is to be operated may be selected. A temperature counter i is set to 1
(124), and the corresponding temperature Ti is set 126. The steps that follow
are
performed for each desired value of Ti, the temperature counter i being
incremented (step 128) after each iteration.
Although not mentioned in the flow chart of FIG. 3B, if it has not yet been
done
the calibration process may include a step of selecting the specific
calibration
color point for which the calibration will be performed, which is preferably
the
same or close to the target color for at which the LED lighting system is to
be
operated. It should be noted that various colors close to the selected one
will also
be controllable by the system within an acceptable degree of precision.
The calibration next involves generating a network of test color points
spanning a
test region surrounding the calibration color point (130). Preferably, the
network
includes a number N of at least 30 points tallied by color point counter
n=1...30.
The distribution of the test color points around the calibration color point
is
CA 02848855 2014-04-10
23
preferably uniform, and may be random or selected according to a specific
pattern.
The calibration process next includes calculating the drive settings needed to
attain each test color point (132). This is preferably accomplished using the
calibration data obtained through the pre-calibration process above. The drive
settings are calculated by solving the matrix system of equation (1),
reproduced
here for convenience:
X X1 Xk
Y Yk = : 9 (7)
- ¨ target ¨ Z C
k_ _ k_
where X, Y, Z are the absolute color coordinates of the target color, Xk, Yk
and Zk
are the absolute color coordinates of each individual LED emitter k (in "fully
on"
condition) and Ck are weight coefficients which can take a value between 0 and
1. The Ck coefficients are related to the emitters drive settings: a
coefficient set to
1 correspond to a fully lit up emitter, a coefficient set to 0 corresponds to
an
emitter in off condition and intermediate values correspond to a
proportionally
dimmed emitter, with luminance Yk reduced by a factor Ck.
In the illustrated example, the drive settings are embodied by current values
applied to each LED emitter. In embodiments where the LED lighting system
includes only three LED emitters, for example R, G and B emitters, the
solution of
the matrix system of equation (7) is unique, that is, a single set of Ck
coefficients,
and therefore of drive settings, corresponds to each target color point. If
the
lighting system contains more than three LED emitters, however, an infinite
number of solutions exist for each color point. Various schemes are known to
those skilled in the art in order to select a preferred or optimal solution. A
common choice is to use the Moore-Penrose pseudo-inverse to find a solution.
Another approach could be to find the solution that optimizes a parameter
related
CA 02848855 2014-04-10
24
to a color rendering metric, such as a Color Rendering Index (CRI) or a Color
Quality Scale (CQS).
Starting with the first color point n=1 (134), the calibration process next
involves
applying (136) the drive settings associated with this color point to all the
LED
emitters of the lighting system. The light output of the LED lighting system
is
measured (138) using both the spectrometer and the color sensor. The color
point counter n is incremented (140) and steps 136 and 138 are repeated for
each of the N test color points. Values for all the Xn, Yn and Zõ on the one
hand,
and Rn, Gn and B,, on the other hand, are therefore obtained.
Next, the calibration matrix A associated with the applied temperature is
calculated (142). This may for example be accomplished using equation (5) and
the values for all the Xõ, Yn and Zn and R,,, G,, and Bn obtained from the
previous
step.
At the end of the process, the calibration matrix or matrices and calibration
data
for the light output of the LED lighting system measured using a spectronieter
in
the calibration phase are stored in the memory, for use during the operation
of
the LED lighting system.
Calibration matrices for individual LED emitters
A LED lighting system can typically be characterized by its color gamut, that
is,
the range of colors it can produce. In some implementations, it may be
desirable
to provide a calibration for a LED lighting system that is valid and precise
for the
entire gamut of the lighting system, or at least within a portion thereof
larger than
the colors close to a specific calibration color point such as provided for in
the
previous section.
It can be shown that by using a single calibration matrix defining a linear or
non-
linear relationship between the absolute color coordinates and the sensor
color
CA 02848855 2014-04-10
coordinates, the obtained calibration is precise within a limited color gamut.
This
can be observed from the following considerations.
Taking the RGBW LED lighting system of FIG. 2B by way of example, the
5 different absolute color coordinates of the lighting system can be
developed for
every LED emitter as follows:
X = + Xy + Xb Xw
Y = Yr Yg Yb + Yw (8)
Z = Zr Zg Zb Zw
where r, g, b and w are indices representing the four LED emitters. The same
principle can be applied to represent the different sensor color point
coordinates
10 outputted by the color sensor, such that
R = Rr + Rg + Rb + Rw
G = Gr + Gg + Gb + Gw (9)
B = Br + Bg + Bb + Bw
Traditional calibration procedures relate RGB values to absolute color
coordinates linearly through a transformation matrix A, such that
[Xyl [aa11 aa12 aa131 [RGi
(10)
[z] [a31 a32 a33] [B]
15 By combining equations (9) and (10) the X coordinate may be represented
as:
X = all (Rr + Rg + Rb + Rw) + a12 (G + Gg + Gb + Gw)
(11)
+ ai3 (Br + By Bb Bw)
Taking under consideration equation (8) it can be assumed that, for the red
emitter:
X. -= aiiRr + a12Gr + ai3Br (12)
CA 02848855 2014-04-10
26
And so forth for all the LED emitters. Repeating this procedure for the Yr and
Zr
absolute color coordinates, the relationship between the absolute color point
coordinates and the sensor color coordinates for the red emitter can be
expressed as:
dri a12 a131 rri
Yr = a21 a22 a23 Gr
(13)
4 a31 a32 a33 Br
The same form would apply for the other three LED emitters G, B and W.
It can be observed from the above considerations that the same matrix is used
to
relate the different color spaces for each emitter. However, in order to
correctly
represent the behavior of the LED emitters this transformation should vary
from
one LED emitter to another. The linear transformation approach can therefore
be
viewed as an average of all the individual matrices of the different emitters.
More
specifically, if A is the transformation matrix used in Equation (10), such
that:
1-X R
Y=
A FG
(14)
[Z] [B
then the average matrix A can be expressed in terms of the individual
transformation matrices for the r g b and w emitters Ak:
IB
FRIT
)k Ak {Gk G
(15)
Bk
A =
R2 + G2 + B2
where k=r, g, b, w is an index representing one of the LED emitters, Ak is the
emitter calibration matrix associated with the LED emitter represented by
index k,
Rk, Gk and Bk are the sensor color point coordinates for light from the LED
emitter
represented by index k, the superscript T represents vector transpose and R, G
and B represent the output coordinates of the color sensor for the light
output of
the entire system.
CA 02848855 2014-04-10
27
Therefore, in accordance with another embodiment, the one or more calibration
matrices may include a plurality of emitter calibration matrices, each
associated
with an individual LED emitter.
Preferably, each emitter calibration matrix defines a relationship between an
emitter sensor vector comprising the sensor color point coordinates for the
associated individual one of the LED emitters, and an emitter color vector
comprising the absolute color point coordinates for the associated individual
one
of the LED emitters. Mathematically, this relationship corresponds to:
Fri
Ar
Ygl= Afirg
gi
Zr Br
(16)
Xb Rbi
ybi= Ab[Gb r R
y:1=Aw[G:i
zb Bb [Zvi [Bwi
Or, in compacted form:
[Xk R1
Ykl= Ak[Gk
(17)
Zk Bk
where k=r, g, b, w is an index representing one of the LED emitters, Ak is the
emitter calibration matrix associated with the LED emitter represented by
index k,
Rk, Gk and Bk are the sensor color point coordinates for light from the LED
emitter
represented by index k, and Xk, Yk and Zk are the absolute color point
coordinates for the light from the LED emitter represented by index k.
The overall relation between the two color spaces can be written as:
[
ryi= Ar[RG +ri ri+Ab[RG:1 + Aw
(18)
A9 (i Z] [Br
9 1Bbi iBwi
Accordingly, the X coordinate can therefore be expressed as follows:
CA 02848855 2014-04-10
28
X = aiRr + afiRg + 41Rb + arifi + 42G, + af2G9 + all2Gb + ar2G14,
(19)
+ 43Br + 4389 + a3Bb + div3Bw
with c171;in for k=r, g, b, w representing the matrix element of the different
emitter
calibration matrices. A similar expression will readily be developed for the Y
and
Z absolute color coordinates.
It will be readily seen that equation (19) contains a lot more information
than the
previous case expressed by equation (11). It will also be understood that
using
this approach, the values obtained for the calibration will be valid for any
target
color, throughout the gamut of the lighting system. This may be intuitively
understood by considering the case where a color approaches the edge of the
gamut where only one emitter is turned on, the matrix that will be used is the
one
of the emitter - the absolute color point coordinates for the light associated
with
the other LED emitters will tend to zero, and only the terms of the
calibration
matrix associated with the lit emitter will contribute to the color
conversion.
In one variant, each emitter sensor vector may further include a constant term
compensating for dark current effects, such as for example:
R
Xk k
kF 1
Ykl A GB
(20)
Zk :
1
The constant term compensates for the dark current, that is, the signal
outputted
by the corresponding emitter when it is turned "off'.
In another variant, each emitter calibration matrix may relate an emitter
sensor
vector comprising non-linear functions of the sensor color point coordinates
to a
color vector comprising the absolute color point coordinates. The non-linear
functions of the sensor color point coordinates may for example include second
order terms, such as:
CA 02848855 2014-04-10
29
Gk
Xk Bk
Ykl= Ak 1
(21)
Zk
_B
Such an embodiment may be particularly advantageous for LED emitters having
large spectral variations with their drive settings. It will be readily
understood that
the relationship defined by the emitter calibration matrices may take other
mathematical forms without departing from the scope of the invention.
As with the previously described embodiment, a suitable calibration process is
preferably performed to obtain the calibration data, i.e., providing values
for the
light output of the LED lighting system for a plurality of values of the
emitter drive
settings, and calculate the emitter calibration matrices.
Referring to FIG. 4, a flow chart of a calibration process 200 according to
one
embodiment is illustrated.
Optionally, in the case where emitter calibration matrices are desired for
different
temperature values, the corresponding number F of calibration temperatures,
and
therefore, the number of sets of emitter calibration matrices to calculate, is
selected as a preliminary step 202. If only one set of emitter calibration
matrices
is used, the calibration temperature may be selected as the average operation
temperature, for example 50 C for operation between 20 C and 80 C. Otherwise,
representative values spanning the range of temperatures at which the lighting
system is to be operated may be selected ¨ for example 30 C, 50 C and 70 C. A
temperature counter i is set to 1 (204), and the corresponding temperature Ti
is
set (206). The steps that follow are performed for each desired value of Ti,
the
temperature counter i being incremented (step 228) after each iteration.
CA 02848855 2014-04-10
,
The calibration process 200 then involves a step 208 of selecting the number E
of LED emitters in the lighting system. For example, in the case of a RGB
system
such as shown in FIG. 2A, E would be set to 3. A number of calibration values
are to be obtained for each emitter, as tallied by an emitter counter k first
set to 1
5 (step 210).
For a given emitter, a number N of drive settings, for example the driving
current
of the corresponding emitter, is selected (step 212), and a current counter n
set
to 1 (step 214). For each increment of the current counter n, the
corresponding
10 drive current is applied (step 216) to the corresponding emitter, and
the light
output of the lighting system is measured (218), using both spectrometer and
color sensor. The absolute color coordinates corresponding to each spectrum
measured by the spectrometer can be calculated using the CIE colorimetric
functions and equation (2) above. The current counter n is incremented (step
15 224) until the maximum number of currents N has been reached.
Once all the data related to a given emitter has been obtained, the emitter
calibration matrix Ak for the current emitter k can be calculated (222). For
example, the relationship defined by equation (20) above can be expressed as:
Ri R2 ¨ Rn
X1 X2 ¨ Xn
[
yi y2 yn = A G1 G G
2 ¨ n
Zi Z2 ¨ Zn
k Bi B2(22)
1 1 1 1
where the indices 1..n represent the N different currents used during the
previous
steps of the calibration, and the calibration matrix Ak can now be calculated
using:
X1 X2 ¨ xn
[
v R1 R
G1 G2 Gn
BZ Z2 ¨ Z ni 1 B 22
A = == .. .= RE inn +
k 111 Y2 '''
' n =.(23)
1 1 1 1
CA 02848855 2014-04-10
,
,
31
where the superscript + refers to the Moore-Penrose pseudo-inverse as defined
in equation (6) above. Of course, those skilled in the art will readily
understand
that other solutions to the system can also be used to calculate the emitter
calibration matrix Ak=
Once the absolute and sensor color coordinates have been obtained and the
emitter calibration matrix been calculated for all the current values for a
given
emitter, the emitter counter k is then incremented (step 226) until all of the
emitters have been processed. The emitter calibration matrices, the absolute
color coordinates and the sensor color coordinates for all the drive settings
used
during the calibration process are kept in memory. As will be explained
further
below, the calibration matrices and the sensor color coordinates will be used
to
obtain the absolute color coordinates X, Y and Z from the measured sensor
color
coordinates R, G and B during the control phase, while the calibration data
for
different drive settings will be used to estimate the currents to apply to the
different LED emitters to achieve the desired target color.
Control phase
Referring to FIG. 5, there is shown an algorithm of a control phase 300 of the
method for operating a LED lighting system at a target output color according
to
one embodiment.
The method generally includes operating the LED emitters at drive settings
selected in view of the target output color and based on the calibration data
and
calibration matrix or matrices.
In the illustrated embodiment, this first involves selecting (step 302) color
coordinates of the target color. The color coordinates are preferably in an
absolute color space, such as the CIE 1931 XYZ or CIELAB color spaces.
Optionally, if calibration matrices are available for several temperatures,
the
operation temperature T that will be considered in the process is also
selected.
CA 02848855 2014-04-10
32
This step may for example involve measuring the junction temperature of the
LED emitter or the temperature at a specific location within the LED lighting
system, and identifying the calibration matrix or matrices that have been
calculated for the calibration temperature closer to the desired operation
temperature.
A first target color point is then set (304) to obtain the desired target
color. The
target color points may be expressed as coordinates in the same absolute color
space as used for the calibration data, for ease of reference. Preferably, the
first
target color point is chosen to correspond to the same absolute color
coordinates
as the target color, or to the closest color point for which an entry exists
in the
calibration data.
The current or other drive settings to apply to each LED in order to obtain a
light
output having a color generally corresponding to the first target color point
is then
calculated (306). This process is the same as used for the calculation of the
drive
settings used during step (132) of the calibration phase of FIG. 3B. If the
LED
lighting system includes only three LED emitters, then there exists a single
set of
drive settings corresponding to the target color point, which can be directly
extracted from the calibration data. If the LED lighting system includes a
greater
number of LED emitters, then an infinite number of solutions to the system of
equations (20) theoretically exist and one should be chosen according to a
predetermined criterion. For example, the Moore-Penrose pseudo-inverse may
be used or a solution can be determined in view of optimizing a parameter
related to a color rendering metric, such as a CRI or a CQS. If the
calibration
matrices have been calculated using calibration points chosen according to a
particular solution for the sets of drive settings, then the same approach is
preferably used during step 306 of the control phase.
CA 02848855 2014-04-10
33
Once the drive settings, here current values for each LED emitter, have been
determined based on the target color point, these drive settings are applied
to the
LED emitters (308).
The method next involves measuring the sensor color point coordinates of the
light output using the color sensor. In the present embodiment, R, G and B
values for the overall light output of the LED lighting system are therefore
obtained.
The method next involves (step 312) determining the absolute color point
coordinates of the light output of the LED lighting system. This determination
is
based on the sensor color point coordinates R, G and B measured at the
previous step, and on the calibration matrix or matrices associated with the
operation temperature. In one exemplary embodiment, the absolute color point
coordinates are expressed as X, Y and Z coordinates in the CIE 1931 XYZ color
space.
The details of the calculations performed when applying the calibration
matrices
depend on the approach used to calculate the calibration matrices themselves.
In the case of the first embodiment described above, where there is one
calibration matrix associated with a given operation temperature relating non-
linear functions of the sensor color points coordinates to the absolute color
point
coordinates, the determining of step 312 may be embodied by a simple matrix
operation. In one implementation, a system sensor vector containing both first
and second order terms of the measured sensor color point coordinates R, G,
and B is built, and multiplied with the calibration matrix, for example
according to
equation (3) above. Of course, the system sensor vector may take other forms
containing non-linear functions of the sensor color point coordinates without
departing from the scope of the present invention. If calibration matrices are
available for different temperatures, the matrix that was obtained at the
CA 02848855 2014-04-10
34
calibration temperature that is closest to the operation temperature is
preferably
used. In some implementations, if the operation temperature is between two
calibration temperatures, the absolute color point coordinates may be
calculated
from a linear interpolation using the calibration matrices associated with
both
closest operation temperatures.
In the second embodiment described above, each calibration matrix is
associated
with an individual LED emitter of the lighting system, and the individual
contribution Rk, Gk and Bk of each LED emitter to the overall sensor color
point
coordinates R, G, and B therefore needs to be obtained.
In one implementation, the determining of the absolute color point coordinates
of
the light output may include a substep of estimating or measuring the sensor
color point coordinates of the light emitted from each LED emitter.
In variants where the LED emitters are driven using the PWM driving scheme,
this may be accomplished by offsetting in time the pulses from the different
LED
emitters, so that the individual LED emitter contributions to the overall
light output
can be deduced from color sensor measurements taken at different moments
over the PWM duty cycle (see for example U.S. Patent No. 7,397,205).
Another approach could include driving each LED emitter with currents
modulated at different frequencies and using a Fourier transform on the color
sensor data to deduce the relative contribution from each frequency,
containing
the color information from each emitter (see U.S. Patent No. 8,159,150).
Estimating the individual contributions from each emitter to the measured
sensor
color point coordinates may alternatively be performed based on calibration
information. For example, in embodiments using the calibration process of FIG.
4, the calibration data may include values for the sensor color coordinates at
each value of the drive settings for which calibration is performed, and this
CA 02848855 2014-04-10
information may be used to interpolate the RGB values for each LED emitter at
the applied currents. The interpolated RGB values for each emitter are then
added and compared to the sensor color point coordinates measured by the color
sensor. One skilled in the art will readily understand that the two values
will not
5 closely match, as there is a spectral shift if the junction temperatures
differ from
those used during the calibration phase, or as a result of aging. To take this
shift
into consideration, an error factor can be attributed in proportion to the
relative
intensities of the emitters. To find the relative intensities, the Y value may
be
interpolated at the applied current using the calibration data and then
repeating
10 this step for every emitter. The individual Rk, Gk and Bk values for each
LED
emitter are then corrected accordingly.
The determining of the absolute color point coordinates of the light output
preferably includes a next substep of calculating the absolute color point
15 coordinates of the light from each LED emitter Xk, Yk and Zk based on
the sensor
color point coordinates of the light emitted from the corresponding LED
emitter
Rk, Gk and Bk and the corresponding emitter calibration matrix Ak. In one
embodiment, this may be performed by applying equation (20), or equation (21),
or the like. A substep of calculating the absolute color point coordinates of
the
20 light output as the sum of the absolute color point coordinates of the
light emitted
from each of the LED emitters is then performed, according to:
Xkl
ixyl = I[yk
(24)
z
which is equivalent to equation (18) above.
As with the previous embodiment, if sets of emitter calibration matrices are
available for different temperatures, the emitter calibration matrices that
were
obtained at the calibration temperature that is closest to the operation
temperature are preferably used. In some implementations, if the operation
CA 02848855 2014-04-10
36
temperature is between two calibration temperatures, absolute color point
coordinates may be calculated from a linear interpolation using the sets of
emitter
calibration matrices associated with both closest operation temperatures.
The control phase 300 of the operation method finally includes step of
comparing
(314) the absolute color point coordinates determined at the previous step to
the
target color point coordinates representing the target color, to determine if
a
predetermined matching condition is met. The comparison is performed in a
predefined color space, defining a comparison color space. Of course, if both
coordinates are not provided in the same color space, a step of converting one
set of coordinates is performed. For example, in one embodiment the target
color
may be expressed in color point coordinates in the CIELab color space, and the
determined absolute color point coordinates obtained in the CIE 1931 XYZ color
space are converted to CIELab before a comparison is made.
The comparison above may employ a predefined color difference formula valid in
the comparison color space. For example, if the CIELab color space is used as
comparison color space, one may calculate the color error by employing the
CIE76 AE:b color difference formula, which is defined as the Euclidean
distance
between the measured color point and the target color point in the CIELab
space:
AEa*b = .j(atarget ¨ ameasured)2 + (btarget ¨ bmeasured)2 + (Ltarget ¨
Lmeasured)2
(25)
With this color difference formula, a difference of less than 2 will not be
detected
by an observer, so the matching condition in such an embodiment may be
expressed as A E a* b 5_ 2.
If the obtained color is within reasonable proximity to the target color, then
the
matching condition is achieved and the currently applied settings are
maintained.
CA 02848855 2014-04-10
37
If the matching condition is not considered achieved, however, different drive
settings should be applied and tested according to the procedure detailed
above.
In one embodiment, the target color point based on which the new drive
settings
are determined may be selected as the geometrical opposite, in the
corresponding absolute color space, to the target point used in the previous
iteration (step 316). An example of the target color point selection is
illustrated in
FIG. 6. It is to note that the target color itself is not changed, only the
target point
used for the control phase. New operation drive settings are calculated (306)
and
applied (308) based on the new target color point and the remainder of the
process is repeated to determine if these new drive settings allow achieving
the
desired target color.
Experimental results
Implementations of the method and system described above were tested using a
four-LED model XLamp MC-E RGBW Cold White available from Cree Inc. The
color sensor used for both the calibration process and the control process was
a
JENCOLOR True Color MTCS sensor (tradename) from MAZeT GmbH.
In a first series of experiments, the calibration approach for a single color
using a
calibration matrix providing the relationship of equation (3) was used. FIG. 7
shows results of the control process for an operation temperature ranging from
20 C to 75 C and luminosity ranging from Y=5 to Y=30. The small circle
represents AEa*b=1, which is considered the maximum color variation for a
light
source observed directly, while the outer circle represents AE;b=2.3, which is
considered the maximum target color variation for a surface illuminated by the
light source. These target color error values correspond to the limit of human
perception to color error [M. Mahy, L. Van Eycken, and A. Oosterlinck,
"Evaluation of uniform color spaces developed after the adoption of CIELAB and
CIELUV," Color Res. Appl. 19, 105-121 (1994)]. The gray-scale bar represents
luminosity. The target color corresponds to illuminant D65. An average error
of
CA 02848855 2014-04-10
38
AE;b=0.49 was obtained with a maximum of AE;b=1.06 with an average
luminosity error of 0.51% and a maximum of 4.04%.
FIG. 8 presents a comparison of the use of calibration matrices for a specific
color respectively defining linear (3x4 matrix) and quadratic (3x7 matrix)
relationships between the absolute color point coordinates and the sensor
color
point coordinates. The circle represents the condition AE;b=1. Points were
taken
at temperatures ranging from 20 C to 75 C. It can be seen that the quadratic
transformation gives the best results, but also that this approach provide
best
results for target colors within a limited range of the calibration color
point.
Another set of experiments was performed using the same lighting system, but
using a calibration process that provides calibration matrices for each
individual
emitter. FIG. 9 shows the results of this control process in similar
conditions as
with the previous experiments, that is for operation temperatures ranging from
C to 75 C, luminosity ranging from Y=5 to Y=30 and a target color
corresponding to illuminant 065. The circle represents AE;b=1. An average
error
of AE*ab = 0.69 was obtained with a maximum of AE*ab=2.38, with an average
luminosity error of 0.56% and a maximum of 3.80%. The maximum error is
20 clearly off target. However, it will be noticed that the points that
give the largest
errors are the ones with the lowest and highest luminosities, the former being
affected by noise in the color detection and the latter being influenced by
emitter
spectral shifts due to variations of the junction temperature. Within a more
limited
luminosity region, every point fits within the acceptance circle.
FIG. 10 shows results for two color controls targeting illuminants E and D65.
No
additional calibration is required to control a different color. An average
error of
AE*ab =0.71 was obtained with a maximum of AE*ab=2.38 with an average
luminosity error of 0.76% and a maximum of 2.45%. It can be observed that
there
is almost no difference in error between the two targets. Table 1 present the
results for targeting different CCTs.
CA 02848855 2014-04-10
39
CCT Average Maximum error Average Y Maximum Y
(K) error(Akth) (AE.b) error (/0) error (%)
2500 1.58 3.36 0.83 3.28
3000 1.23 3.39 0.79 3.03
3500 1.18 3.68 0.92 6.40
4000 0.99 3.26 0.83 5.48
4500 0.93 2.92 0.76 4.16
5000 0.80 2.11 0.59 2.21
5500 1.06 2.50 0.50 2.05
6000 1.26 4.38 0.72 5.22
6500 1.18 4.74 0.57 3.31
7000 1.19 4.24 0.51 2.29
7500 1.20 4.27 0.65 3.93
Table 1: Results of the control phase targeting different CCTs
A slight decrease in performances can be observed compared to using D65 as
the target color, but the errors are still within acceptable limits
considering that
the greatest error contributions come from the high and low luminosities and
the
much larger range of validity of the calibration with respect to the previous
example, in which illuminant A (CCT of 2856 K) could not be reached.
By comparing the results for the two calibration approaches disclosed above,
it
appears that calibrating for a single color using a non-linear relationship
provides
better precision, but within a limited range, whereas using different
calibration
matrices for each LED emitter has no such range limitation. Both approaches
could therefore be of interest depending on the circumstances of a particular
implementation.
Of course, numerous modifications could be made to the embodiments described
above without departing from the scope of the invention as defined in the
appended claims.
