Multiple Primary LED Lamp Colour Controller with Inherent Brightness Limitation

Transcription

1 Technical Note PR-TN 2007/00318 Issued: 05/2007 Multiple Primary LED Lamp Colour Controller with Inherent Brightness Limitation R. Barcena (Univ. Politecnica de Catalunya); B. Ackermann Philips Research Europe Unclassified Koninklijke Philips Electronics N.V. 2007

2 PR-TN 2007/00318 Unclassified Authors address R. Barcena (Univ. Politecnica de Catalunya) B. Ackermann KONINKLIJKE PHILIPS ELECTRONICS NV 2007 All rights reserved. Reproduction or dissemination in whole or in part is prohibited without the prior written consent of the copyright holder. ii Koninklijke Philips Electronics N.V. 2007

3 Unclassified PR-TN 2007/00318 Title: Author(s): Reviewer(s): Multiple Primary LED Lamp Colour Controller with Inherent Brightness Limitation R. Barcena (Univ. Politecnica de Catalunya); B. Ackermann Detlef Raasch, Meiling Schmelzer, Ulrich Schiebel Technical Note: PR-TN 2007/00318 Additional Numbers: Subcategory: Project: IntelliLED - Intelligent LED lamp solutions ( ) Customer: IP&S, BGLE, BU SSL Keywords: Abstract: LED lamps, white LED, color control There is a strong interest in using LEDs for general illumination due to the potential they offer for energy saving, environmental friendliness, new opportunities in lighting design, and control of the intensity, colour, and spatial distribution of light. General illumination requires primarily white light that can be obtained by mixing e.g. the light of red, green, and blue LEDs. This enables also colour adjustability, which is considered to be a most attractive feature of future LED lamps. This master thesis takes as a starting point previous work that focused on developing a setup which permits rapid control prototyping of a red, green and blue LED-based white light source. Using this setup, some improvements and new possibilities have been developed and tested. A brightness control has been developed that limits the maximum brightness of the light source if the LEDs are unable to reach the requested brightness. This problem can appear if the user has set an unreachable brightness. It is aggravated by the aging of the LEDs. The RGB colour control has been modified in order to add more than three primary colours to the light generation process. An RGBA (Red, Green, Blue, Amber) colour control has been set up adding an amber LED. Conclusions: Koninklijke Philips Electronics N.V iii

4

5 Unclassified PR-TN 2007/00318 Contents 1. Introduction RGB colour control RGB colour control theory RGB colour control implementation PWM implementation AM implementation Sensing Calibration System performance RGB colour control improvements Brightness control Theory of control System model System performance RGBA control Theory of control System model System performance Systems integration...75 Acknowledgements...77 References...78 List of figures...80 List of tables...83 Koninklijke Philips Electronics N.V v

6

7 Unclassified PR-TN 2007/ Introduction The use of LEDs for general illumination has become strongly interesting due to the potential they offer for energy saving, environmental friendliness, new opportunities in lighting design and control of the intensity, colour and spatial distribution of light. A lot of applications can take advantage of the unique features that LEDs offer. Other than illumination applications can benefit from the possibility of changing light colours almost instantaneously without changing lamps. In medical applications, their cold light and small size enables new ways of illumination. Used as traffic lights they reduce dramatically maintenance compared with previous solutions. The automobile industry can redefine the size and shape of headlights. Figure 1- LEDs applications Koninklijke Philips Electronics N.V

8 PR-TN 2007/00318 Unclassified LEDs have been available commercially since the early 1960s when General Electric introduced red GaAsP devices based on the pioneering work of Nick Holonyak Jr. The early LEDs were priced at $260 and were available only in low volumes. Since then, the semiconductor industry has dedicated huge efforts on improving features of these devices. The first high volume vendor of light emitting diodes, Montsanto, was so optimistic about the future of LEDs that in 1973 an advertisement was placed in the Wall Street Journal showing an automobile with LED headlights. Nowadays, 30 years later, this target is close to be achieved [6]. Contemporary LEDs have evolved over the past four decades, with remarkable progress in the past decade. They have moved from being indicators to light sources and have begun to replace conventional light sources in a variety of niche applications. LEDs will replace conventional lighting technology for general illumination if researchers are able to improve cost, production and performance limitations. Next to improving the LEDs efficiency and reducing their cost, improved thermal management and colour control are considered to be the key challenges lying ahead. LEDs are light sources with narrow spectra that are typically a few ten nanometres wide. As a consequence of this, they emit light with saturated colours and their colour points in CIE 1931 chromaticity diagram are located close to the perimeter. White light is situated in the centre of the chromaticity diagram (Fig. 2). Figure 2. CIE 1931 Chromaticity Diagram. RGB white light generation. There are two basic ways to obtain white light using LEDs, either colour mixing or using phosphors for down converting the light of ultraviolet or blue LEDs. Down converting can be implemented in a more straightforward way. However, colour mixing enables colour adjustability, which is considered to be a most attractive feature for future LED lamps. Mixing e.g. the light of red, green and blue LEDs, any colour can be created inside the triangle defined by their chromaticity coordinates (Fig. 2). Feedback colour control is needed when using colour mixing according to the poor colour maintenance achieved because of LED characteristics variation. Optical properties of LEDs vary with manufacturing spread, aging, temperature variations and drive current amplitude variations. 8 Koninklijke Philips Electronics N.V. 2007

9 Unclassified PR-TN 2007/00318 This master thesis project is fruit of the interest of the Research Group SSL at Philips Research Laboratories, Aachen, Germany, in the achievement of a deep knowledge in the semiconductor-based solid-state lighting field and its effort to implement new illumination solutions. Figure 3 - Philips Pedestrian LED Luminary Gold if product design award, Philips LED architectural Floodlight IF design award 2006 & The Inner Ring Road Bridge in Bangkok, Thailand, lit up with Philips LED lighting systems. This master thesis project takes as starting point a previous project done by M.Saura [1]. New topologies and control solutions for RGB white light generation had been developed and tested at Philips Research Laboratories, supervised by Dr. Bernd Ackermann. Koninklijke Philips Electronics N.V

10 PR-TN 2007/00318 Unclassified 2. RGB colour control The recent improvements in high-power light-emitting diodes (LED) technology with 100+ lumens per LED chip and efficacy exceeding that of incandescent lamps brings solid-state lighting close to reality. An LED light source made of Red, Green and Blue (RGB) LEDs can provide a compact light source with unique features such as instant colour variability. RGB light sources have the following issues: uniform spatial light mixing and distribution, white colour point maintenance and thermal management. Specifically, the white colour point maintenance is a stringent requirement in many applications. Meeting this requirement is a severe challenge due to the variation in the optical characteristics of the RGB-LEDs with temperature, time, forward current and manufacturing spread of the LED performance. This results in 1) an unacceptable high variability in white light colour point and 2) difficulties in manufacturing reproducible LED lamps. Controlling the colour of a LED lamp using a colour sensor has been investigated at Philips Research Briarcliff, [Muthu 2002, Muthu 2003]. The LED lamp comprises 3 primary colours, usually red, green, and blue (RGB). The colour sensor comprises three light sensors with peak sensitivity in different parts of the visible spectrum, usually also red, green, and blue (RGB) RGB colour control theory The control procedure described in the following emphasizes two considerations: Firstly, the human eye acts as a low pass filter. Actually, it is required that there is no visible flicker. This means that the transfer function of any part of the system that impacts light emission must have a break frequency substantially higher than the break frequency of the low pass filter of the human eye. In order to reproduce the light as perceived by the human eye, a low pass filter is inserted into the feedback path of the colour control system. This can be implemented either in the digital or in the analogue part of the system. The break frequency of this low pass filter has to be substantially lower than the break frequency of the transfer function of any part of the system. As a consequence of this, apart from this low pass filter in the feedback path, all transfer functions can be replaced by their low-frequency approximation for the design of the controller. Secondly, it will usually not be practical to determine individually the transfer functions of the different components of the colour control system. Instead of that, they will be grouped into larger modules, the transfer functions of which are determined in a calibration procedure. It will be sufficient to determine the low-frequency approximation of these transfer functions due to the first consideration. The light to be emitted by a LED lamp is specified by its chromaticity coordinates x and y and its luminous flux Φ lum. From these the tristimulus values X, Y, and Z are calculated. 10 Koninklijke Philips Electronics N.V. 2007

11 Unclassified PR-TN 2007/00318 Y Φlum = (2.1) lm 683 W Y X = x (2.2) y Z Y Y = z = ( 1 x y) (2.3) y y The tristimulus values are grouped into a vector TV (Tristimulus Values). It has to be distinguished between the tristimulus values TV O of the light perceived by the observer, i.e. a person looking at the LED lamp, and the tristimulus values TV S determined in the feedback path of the colour control system. X TV = Y (2.4) Z Ideally, the colour sensor should sense the tristimulus values directly. However, this will not be achieved in practice. The values R, G, and B sensed actually by the colour sensor are grouped into a vector SR (Sensor Readings). R SR = G (2.5) B In a similar way, the control signals for the drivers for the red, green, and blue LEDs will also be grouped into a vector CS (Control Signals). These may be duty cycles for a pulse width modulation control or current amplitudes for an amplitude modulation control. r CS = g (2.6) b Koninklijke Philips Electronics N.V

12 PR-TN 2007/00318 Unclassified Fig. 4 shows the general setup of a LED colour control system using a colour sensor. It is indicated where the signals discussed above occur in the system. As to the tristimulus values, the input signal TV set and the error signal TV err are indicated in addition to the output signals TV O and TV S. TV set + TV err CS G C G D G LED G OSO G LPO (s) - TV S G CAL SR G LPS (s) G S G OSS TV O Figure 4 - Block diagram of a LED colour control system using a colour sensor. The transfer functions depicted in the block diagram of the LED colour control system using a colour sensor (Fig. 4) represent the following parts of the system: G C - controller G D - driver G LED - LEDs G OSO - optical system, from LEDs to observer G LPO (s) - low pass filter observer (human eye) G OSS - optical system, from LEDs to sensor G S - colour sensor G LPS (s) - low pass filter sensor G CAL - calibration matrix Apart from the low pass filters all transfer functions have been replaced by their low frequency limit. The amplitude of the low pass filters at their low frequency limit is assumed to be part of the sensor or the human eye, respectively. Therefore, the low frequency limit of the low pass filter of the sensor is just a unit matrix. Furthermore, there will be a separate low pass filter for the red, green, and blue sensor. G LPS () s LPR = 0 0 () s LP 0 G 0 () s 0 0 LP () s B G LPS ( s 0) = (2.7) The control system parts are grouped into modules for which the transfer function can be easily determined in a calibration procedure. The first one is the transfer function G C2T from the control signals CS to the tristimulus values TV O perceived by the human eye, neglecting the dynamics of the low pass filter of the human eye. G C 2T = G OSO G LED G D (2.8) 12 Koninklijke Philips Electronics N.V. 2007

13 Unclassified PR-TN 2007/00318 TV O = G C 2T CS (2.9) The second one is the transfer function G C2S from the control signals CS to the sensor readings SR, neglecting the dynamics of the low pass filter acting on the sensor signals. G C 2S = G S G OSS G LED G D (2.10) SR = G C 2S CS (2.11) The calibration matrix G CAL can be determined from the requirement that the tristimulus values TV S in the feedback path have to be equal to the tristimulus values TV O perceived by the observer. TV O = TV S (2.12) Inserting (2.11) and (2.12) into (2.9) results in TV S = G C 2T G 1 C 2S SR (2.13) The block diagram Fig. 4 of the colour control system indicates that the tristimulus values determined in the feedback path of the system are linked to the sensor readings by the calibration matrix, which results in C 2T 1 C 2S 1 1 G = G G = G G G (2.14) CAL OSO OSS S Obviously, no calibration is needed if G CAL =1. This requires the optical path from the LEDs to the sensor to be identical to the optical path from the LEDs to the observer, apart from an attenuation k of the total amount of light received. G OSS = k G (2.15) OSO Furthermore, the sensor has to be ideal, i.e. to measure the tristimulus values X, Y, Z directly, apart from compensating for the attenuation k of the total amount of light received. Koninklijke Philips Electronics N.V

14 PR-TN 2007/00318 Unclassified G S = k (2.16) The transfer functions G C2T and G C2S from the control signals to the tristimulus values and sensor readings, respectively, can be determined in a calibration procedure. Three measurements are taken. They are denoted in the following by indices I, II and III. For each measurement the control signal for each LED colour is set to a specific value. Sensor readings are taken and the tristimulus values of the light observed are determined using a spectrometer. The results obtained are stored as outlined in (2.17) to (2.20). ri CS = g I bi X I TV O = YI, Z I RI SR = GI (2.17) BI rii CS = g II bii X II TV O = YII, Z II RII SR = GII (2.18) BII riii CS = g III biii X III TV O = YIII, Z III RIII SR = GIII (2.19) BIII ri rii riii CS = g I g II g III ; bi bii biii X I X II X III TV = YI YII Y O III ; Z I Z II Z III RI RII RIII SR = GI GII GIII (2.20) BI BII BIII (2.9) and (2.20) result in G C 2T = TV CS O 1 (2.21) (2.11) and (2.20) result in 14 Koninklijke Philips Electronics N.V. 2007

15 Unclassified PR-TN 2007/00318 G C 2 S = SR CS 1 (2.22) If for each measurement the control signal for one LED colour is set equal to one and the control signals for the other two LED colours are set equal to zero, then CS = (2.23) and the transfer functions G C2T and G C2S are G = TV C 2T O (2.24) G C S = 2 SR (2.25) Using (2.10), the block diagram of the LED colour control system (Fig. 4) can be simplified as depicted in Fig. 5. TV set + TV err CS SR G C G C2S G LPS (s) G CAL TV S - Figure 5 - Simplified block diagram of the LED colour control system. Furthermore, the LED colour control system can be described using the sensor readings as the control variable by dragging G CAL across the summation point. TV set SR set + SR err CS -1 G CAL G CAL G C G C2S G LPS (s) SR - Figure 6 - Block diagram of the LED colour control system with the sensor readings as control variable. The MIMO (Multiple Input Multiple Output) control is implemented by decoupling the control signals and sensor readings for the different colours. This is achieved by choosing Koninklijke Philips Electronics N.V

16 PR-TN 2007/00318 G C G CAL = G 1 C 2S G dyn () s Unclassified (2.26) The block diagram representation of (2.26) is given in Fig G CAL G dyn (s) G C2S G C = Figure 7 - Block diagram representation of (2.26). As to G C2S it is important to take into account, that in Fig. 6 it represents the real plant that may exhibit nonlinear behaviour, whereas in (2.26) and Fig.7 the inverse of a linear approximation to it is used. For the purpose of control design this difference may be neglected. Then the structure of the system simplifies as depicted in Fig. 8. TV set SR set + SR err -1 G CAL G dyn (s) G LPS (s) SR - Figure 8 - Block diagram of the decoupled LED colour control system. As a consequence of (2.7) the part of the controller described by the transfer function G dyn (s) is a diagonal matrix the elements of which can be designed independent from each other, using e.g. PI controllers as described in [1]. G dyn () s G = dyn, R 0 0 () s G 0 dyn, G 0 () s G dyn 0 0 () s, B (2.27) Fig. 9 then gives the overall block diagram of the decoupled colour control system. TV set + SR err CS G SR set -1-1 CAL G dyn (s) G C2S G D G LED - SR G LPS (s) G S G OSS Figure 9 - Overall block diagram of the decoupled LED colour control system. 16 Koninklijke Philips Electronics N.V. 2007

17 Unclassified PR-TN 2007/ RGB colour control implementation The RGB LED-based light source consists of four different modules. There is one LEDdriving module which provides the currents to the LEDs, one sensor module which measures the amount of light from the LEDs, a control module which adjusts the current of every single LED depending on what is measured on the sensor and a LED lamp comprising red, green and blue LEDs. Figure 10 Colour mixing and control [2]. The setup in the laboratory used to develop the RGB LED lamp comprises the four modules cited before plus an integration sphere where the lamp and sensor are placed, a dspace [13] system, an spectrometer to measure the power spectral flux of the light inside the sphere and to calculate their chromaticity coordinates, two PCs which support the rapid control prototyping software of the dspace system and the spectrometer applications [15], an oscilloscope and power supplies. Figure 11 Research Group SSL at Philips Research Laboratories, Aachen, Germany. Laboratory setup. The role of the control system and the sensor module were outlined at the beginning of this report in chapter 2. RGB colour control. The role of the LED-driving module and what this implies is outlined in the following. The intensity of the light emitted by one LED is virtually directly proportional to the current through it. The Luxeon Star LEDs [11] used in the setup have a maximum DC Koninklijke Philips Electronics N.V

18 PR-TN 2007/00318 Unclassified current of 350mA. It is then necessary to include a subsystem that drives the current through the LEDs and permits to change its value to control the luminous flux of each LED. There are two basic schemes to drive the LEDs, pulse width modulation (PWM) and amplitude modulation (AM). The first one, PWM, is the most common solution adopted nowadays in many electronic devices and almost all microcontrollers include a PWM generator. The second one, AM, is less popular but there have no studies been done which advise against its application. Initially, the PWM solution was implemented by the Research Group SSL at Philips Research Laboratories, Aachen, Germany. The setup is described in Marc Saura s master thesis [1] and it was the starting point for the next improvements of the RGB control. A dspace rapid control prototyping system [13] is used to control the signals for the PWM driver. These signals are passed through a digital to analog converter (DAC) in their last step before being injected into the driver instead of being delivered by a dedicated PWM generator. This means that these signals include all the typical errors in a DAC and they affect the proper response of the system. This was the main reason why the AM solution was adopted for the next setups used in the investigation described in this report. LED Lamp dspace (control) Driver Sensor electronics LEDs Sensors Figure 12 RGB colour control. System overview. The dspace rapid control prototyping is a flexible development system to optimize control designs. It allows controlling the RGB colour control system and changing the specifications easily. 18 Koninklijke Philips Electronics N.V. 2007

19 Unclassified PR-TN 2007/00318 M atla b / S im u link Model dspace Im plem entation Software Real-Time interface Processor / Boards Real-Time Hardware REAL WORLD Control Desk Experiment Software Figure 13 - dspace rapid control prototyping system. RBG colour control models implemented in MATLAB/Simulink [14] are transferred to dspace and implemented into the dspace prototyping hardware using the dspace Real Time Interface (RTI) software. Internally, the system reconfigures itself according to the downloaded model and creates the I/O interface with the outside world. The dspace software also allows developing GUI applications (Control Desk) used to control the experiments PWM driving PWM is thoroughly explained in Marc Saura s thesis report [1].The main characteristics of the system he developed are highlighted below. A PWM signal with variable duty cycle is given from the dspace rapid control prototyping system to the LED drivers. These PWM signals are outputs from the dspace to the lamp. Three duty cycles, red, green and blue are calculated in the dspace system following the ideas explained in the RGB colour control theory paragraph. Figure - 14 Block diagram of the LED colour control system. In Fig. 15 some components of the Simulink model for the PMW colour control system are grouped and related to the RGB colour control theory. Koninklijke Philips Electronics N.V

20 PR-TN 2007/00318 Unclassified Figure 15 Simulink model. PWM control scheme. An extra subsystem block that does not appear in the RGB colour control block diagram converts the chromaticity coordinates (x,y) and luminous flux (Y) into tristimulus values. A MAZeT true colour sensor [12] is used for the feedback control of the colour coordinates. The sensor gives a three channel current output signal, one for red, one for green and one for blue, which is conditioned to obtain a voltage signal at good working levels. In absence of light, the sensor, after conditioning, gives a high voltage level around 0.5 volts that drops, when the light increases, close to 0 volts. For the colour control these sensor readings have to be directly proportional to the duty cycle and without offset. In order to accomplish this requirement all the sensor readings are subtracted to the maximum value. Figure 16 MAZeT true colour sensor: MTCSiCS. 20 Koninklijke Philips Electronics N.V. 2007

21 Unclassified PR-TN 2007/00318 A block working as a current limitator is placed before the PWM generator. It limits the maximum levels of the control signal between 0 and 1. The drivers used have been designed to adapt these levels to a current range of 0 to 350mA. The D/A converter board DS2102 is used to give the PWM signals to the LED drivers and the A/D converter board DS2001 is used to read the values from the sensor. As it was said before, a thorough study of the design and performance of the PWM colour control system is described in [1]. [3] can be used as a reference AM driving The main difference between PWM and AM colour control is the way in which the LEDs are driven. In the PWM system, the amplitudes of the signals never change and what is controlled is the time they are on and off. Now, the signals are always on and what is changed is their voltage level in order to control the amplitude of the currents through the LEDs. The first things that have to be changed for the new setup are the drivers. Fig. 17 and Fig. 18 show the scheme of the drivers used. Figure 17 AM driver. Power circuit. Figure 18 AM driver. Signal conditioning. Koninklijke Philips Electronics N.V

22 PR-TN 2007/00318 Unclassified The drivers are designed to work with inputs up to The power circuit has two low dropout voltage regulators, LM317HVT and LM7808CT. The first one is adjusted to 18 volts with two resistors while the second one has a fixed output voltage of 8 volts. Some capacitors are used to stabilize the signal. The main structure of the driver is a voltage follower. The signal injected from the control system is passed through a voltage divider that adjusts its level. Then it is passed through the voltage follower and transformed into a current signal by R4 and R6 resistors in order to drive the transistor present in the output of the circuit. R9 fixes the maximum current that the driver can give. In the circuit of Fig.18, R4 R6 and C4 also serve to counteract the loss of phase margin by feeding the high frequency component of the output signal back to the amplifier s inverting input, thereby preserving phase margin in the overall feedback loop. Capacitive load driving capability is enhanced by using a pull up resistor R2 to V+. Typically a pull up resistor conducting 500 µa or more will significantly improve capacitive load responses. The value of the pull up resistor must be determined based on the current sinking capability of the amplifier with respect to the desired output swing. The open loop gain of the amplifier can also be affected by the pull up resistor. Figure 19 AM driver board. To simplify the appearance of the Simulink model, the system has been divided into the subsystems shown in Fig. 20. There are no differences in the main principles of control between the PWM and the AM scheme. The new Simulink model now includes a feedback loop with a subsystem that reproduces the behaviour of the plant. This permits an easy way to simulate the behaviour of the whole system before implementing it in the dspace rapid control prototyping system. 22 Koninklijke Philips Electronics N.V. 2007

23 Unclassified PR-TN 2007/00318 Figure 20 AM colour control subsystems. As for the PWM system, Simulink model and colour control theory are related in Fig. 21. Figure 21 Simulink model. AM control scheme. Koninklijke Philips Electronics N.V

24 PR-TN 2007/00318 Unclassified Sensing The sensing circuitry uses a MAZeT sensor [12]. It has three photodiodes with sensitivity curves close to the CIE colour-matching functions and gives a current dependant on the amount of light that it receives. The higher the intensity of the light received the more current is provided by the photodiodes. As the signal provided by the sensor is a very low current and the dspace environment works with voltage signals, it is firstly necessary to amplify the signal and secondly to transform it into a voltage. The MTI04 IC is a programmable gain transimpedance amplifier used to accomplish these requirements. Fig.22 shows the circuit used as a conditioning for each channel. Figure 22 Sensing. Conditioning circuit. The circuit is an inverting amplifier plus a low-pass filter. It provides an output of 2.5 volts at darkness conditions. This value drops with increasing light input. The signal to be filtered is around 100Hz and the cut-off frequency of the low-pass filter is set at 1Hz. Figure 23 Sensing. Old sensor & conditioning circuit used in [1]. 24 Koninklijke Philips Electronics N.V. 2007

25 Unclassified PR-TN 2007/00318 The initial implementation of this circuit made by M.Saura [1] was modified to improve its behaviour in terms of interference rejection, easy modification of the gain, size (Fig. 24 and Fig. 25). Figure 24 Sensing. New sensor & conditioning circuit. Figure 25 Sensing. New sensor & conditioning circuit. The gain of the transimpedance amplifier MTI04 can be programmed. Table 1 shows the different levels of gain. This gain is the same for the four channels of the amplifier. Koninklijke Philips Electronics N.V

26 PR-TN 2007/00318 Unclassified Table 1 MTI04 gain. The contribution of each primary colour to the sensor readings is quite different. Green LEDs have the biggest luminous flux of the three primaries used now. On the contrary, blue LEDs have the lowest. The relation between them is around 1/10. If the maximum sensor reading with all LEDs fully on uses the full band of work, then the levels for each primary working individually are acceptable. If this condition is not fulfilled, then the blue channel has a very low variation and a very low signal level and a little variation on it caused by any source of interference can completely change its value. As a consequence of that, selection of the gain is critical. Changing the gain of the MTI04 transimpedance amplifier can result in dividing the sensitivity even by five between two steps of gain. Then it is may not be possible to use the full band of work. It is recommendable to modify the implemented conditioning circuit in order to avoid this problem. It could be solved adding an external potentiometer to adjust the gain. It could be even better if the gain could be adjusted for every channel individually. Price and size restrictions must also be considered that can apply to new implementations. Such investigations of the analog circuitry were not in the scope of this project and were not pursued further Calibration As it was explained in the control theory chapter, the system needs to be calibrated in order to calculate the matrices 1 1 Cal = GC 2SGC 2T G and 1 G C2S where G C2S is the matrix of sensor readings and G C2T the matrix of tristimulus values, both calculated when calibrating. 26 Koninklijke Philips Electronics N.V. 2007

27 Unclassified PR-TN 2007/00318 GC 2 S R = G B r r r R G B g g g R G B b b b GC 2 T X r = Yr Zr X Y Z g g g X Y Z b b b For the nine possible combinations of red, green and blue LEDs and red, green and blue channels of the sensor, each element of the matrix of the sensor readings G C2S is calculated as the value in volts of the difference between the signal delivered, after being conditioned, by the sensor module when the sensor is in dark conditions and when it is illuminated by a LED. For example, the R r value is the difference between the signals for the red channel of the sensor when all the LEDs are off and when the red LED is on. Following the same rule, the G b value is the difference between the signals for the green channel of the sensor when the LEDs are off and when the blue LED is on. The elements of G C2T are the tristimulus values of each LED. Their measurement can not be done directly. The spectrometer gives the chromaticity coordinates of the light and its luminous flux, from which it is possible to obtain the required tristimulus values using these formulas Y Φlum = lm 683 W Y x y Y y Y X = Z = z = ( 1 x y) y The calibration procedure consists of switching fully on one LED each time and measuring the value of the sensor readings, chromaticity coordinates and luminous flux. These calibration measurements are needed to determine which sensor readings correspond to a desired light output and what are its characteristics. The aim of the calibration is to know how each channel of the sensor is affected by a known kind of light. It means to know what the relation is between the control signal and the sensor readings and between the control signal and the tristimulus values. What is what we are measuring (sensor readings) and what we are really seeing (tristimulus values). The values we obtain and the matrices we implement will allow the control to correct the error of perception introduced by the sensor. In an ideal sphere no losses would occur because the light from the LEDs would be perfectly reflected at the walls, but in a real sphere some light is lost. Another calibration process is needed to account for changes in the conditions inside the sphere, like placing the LEDs board, cables, etc. This is achieved with a calibration lamp and software provided with the spectrometer. Koninklijke Philips Electronics N.V

28 PR-TN 2007/00318 Unclassified Figure 26 Sphere plus LEDs and sensor board. The next figures show two different Simulink models used for calibration. They differ as to how the signals that will drive the LEDs are generated. The first diagram was created for an AM modulation while the second one was for a PWM modulation. These two Simulink models permit to switch on each LED individually. Figure 27 AM calibration Simulink model. 28 Koninklijke Philips Electronics N.V. 2007

29 Unclassified PR-TN 2007/00318 Figure 28 PWM calibration Simulink model. The next two figures show the applications from where the measurements are taken. The first one is a capture of a Control Desk window of the dspace system for the PWM modulation. In this case, the duty cycle of the signals that drive the LEDs are controlled. The system automatically presents the sensor readings of each one of the three channels of the sensor, red, green and blue. The window for the AM modulation permits to control the amplitude of the signals in a similar way and also shows the sensor readings. Figure 29 dspace PWM control desk. Koninklijke Philips Electronics N.V

30 PR-TN 2007/00318 Unclassified The next capture shows how the spectrometer software application presents the spectral power density of the light, the chromaticity coordinates and the luminous flux. Figure 30 Spectral lamp measurement system ver The calibration process is a critical procedure that has a huge effect in the results of the colour control system. A wrong calibration will result in a wrong colour point of work. This procedure is the subject of separate investigations. As the LEDs are affected by its temperature of work, it is also necessary to wait a little time allowing them being warmed up before taking the measurements. Waiting five minutes gives enough good results. 30 Koninklijke Philips Electronics N.V. 2007

31 Unclassified PR-TN 2007/00318 Figure 31 Effect of temperature variation on the spectral emission of a red 1W Luxeon LED [11] System performance The system to be characterized uses the AM colour control solution and it is the starting point for the next studies and improvements of the system performance. To test the system performance, a test point in the chromaticity diagram is set. Figure 32 CIE 1931 Chromaticity Diagram. Selected point of work. Koninklijke Philips Electronics N.V

32 PR-TN 2007/00318 Unclassified In a first approximation, the system is simulated in a non real time setup, using the Simulink model with the new block that implements the behaviour of the plant. The transient response of the system for the requested colour point and brightness of 0.08 is shown in Fig. 33. Figure 33 AM colour control. Simulated transient response. Fig.34 shows the transient response of the system when the brightness is changed from 0.08 to Figure 34 AM colour control. Simulated transient response II. As it can be seen from these two figures, the system is able to find a constant relation between the signals for the red, green and blue LEDs when they reach the steady state. It is also possible to see in both pictures a period where the signals achieve their maximum permitted levels, 0 and 1. If a signal limitator would not have been included in the setup, in a real situation, the LEDs would be damaged. With this limitator, this problem is avoided but then the relation between signals is not constant since the limit is reached. 32 Koninklijke Philips Electronics N.V. 2007

33 Unclassified PR-TN 2007/00318 This causes a variation in the colour point every time the brightness is changed and the limits are reached. As it is seen in Fig. 33, in this particular case, this change in the chromaticity coordinates can last even more than two seconds. When the requested brightness is too high, the control tries to satisfy this requirement increasing the signal for the LEDs and again the maximum value is reached. This means that the colour point obtained is not the one desired. This situation must be avoided in order to preserve the desired colour point. Fig. 35 shows this situation for a requested brightness of Figure 35 AM colour control. Simulated transient response III. The next figure shows the response of the system for some set brightness values. It is interesting to see how these responses have a high over shoot that brings the Am signal to saturation. Figure 36 AM colour control. Simulated transient response IV. Once the simulated colour control model works satisfactorily, it is transferred to the rapid control prototyping system. In the Simulink model used, the plant description block is replaced by the AD and DA converters (Fig. 37). Koninklijke Philips Electronics N.V

34 PR-TN 2007/00318 Unclassified Figure 37 AM colour control. RTI1103 dspace model. The signals captured with an oscilloscope are shown in Fig. 38 and Fig. 39. Figure 38 AM colour control. Transient response I. Figure 39 AM colour control. Transient response II. The main differences with the expected results are two: 34 Koninklijke Philips Electronics N.V. 2007

35 Unclassified PR-TN 2007/00318 The first one is the different final level of the signals. This is caused because the model of the plant used in the simulation does not match perfectly its real behaviour. Some characteristics have been assumed linear but they are not in their whole range of work. The second one is the ripple that appears in these signals. This is caused because of the interferences the sensor receives, as for example external light, changes in temperature or electromagnetic fields. Also the signal from the sensor injected into the control is too low, as it was explained in the sensing chapter, and a little contribution of these interferences force a non-negligible change in the control. LEDs are also affected by these external disturbances and also change their response contributing to this ripple. A detail of the ripple is shown in Fig.40. The cause of this waveform is that the control signal is generated in a D/A converter. Figure 40 AM colour control. Detail of D/A characteristics of the response. As a result of all these differences, the system does not match the exact point in the chromaticity diagram and the difference in xyy coordinates seems to be considerable. At this point a variable transformation from xyy to uv(y) coordinates is necessary. It is considered that a difference less than in these uv coordinates is not detected by the human eye. This means that the human eye cannot distinguish the difference of the colour characteristics of two colours that differ no more than in uv coordinates. This is going to be explained in a following chapter of this report. Fig. 41 shows the spectral power density distribution of the RGB colour point achieved. Figure 41 AM colour control. Spectral power density. Koninklijke Philips Electronics N.V

36 PR-TN 2007/00318 Unclassified The difference between the desired and the obtained colour point in uv coordinates is, Desired: x y Y set set set = 0.33 = 0.33 = 0.05 Obtained: x y Y m m m = = = Error: ( Δu, Δv) = (0.0004,0.0006) The conclusion after the analysis of the system performance, in spite of the differences that appeared between what is desired and what is obtained, is that it is possible to reach a good RGB colour control. However, it would be good to improve the transient state behaviour in order to avoid the saturation of the channels and also the saturation in the steady state when the requested brightness is not reachable. 36 Koninklijke Philips Electronics N.V. 2007

37 Unclassified PR-TN 2007/ RGB colour control improvements Once satisfactory RGB colour control has been achieved, the next step is to add more functionality to the system in order to accomplish some of the requirements that a commercial lamp could have. For example, it is interesting to limit the maximum brightness that can be requested to the system to avoid the change of the colour of the light emitted when the relation among the amplitudes of the signals that control the LEDs are changed. It is also important to prevent the system against the consequences of aging, that can reduce the maximum brightness by even 50% of its nominal value, provoking also the loss of the selected colour of the light. The colour rendering index (CRI) is a metric used to evaluate light sources. The LED industry is working hard to improve the CRI of the LED based lamps so that the technology will be widely accepted for general illumination applications. Because the CRI score is directly related to the spectral power distribution of the light source, it is possible to manipulate the spectrum to produce a higher CRI value. The use of more than three primary colours allows modifying this spectrum and increasing the CRI. Figure 42 - Eye sensitivity function. Koninklijke Philips Electronics N.V

38 PR-TN 2007/00318 Unclassified In the following two different ways to add functionality and to improve the RGB colour control characteristics are going to be studied. The first one tries to control the maximum brightness that can be requested to the system in order to avoid the clamping of the signals that drive the LEDs and the loss of the colour point. It also tries to minimize the effect of aging, another factor that can cause the same effect. In this second case the problem is not the brightness requested but the reduction of it that the LEDs suffer in time. The second one tries to improve the Colour Rendering Index (CRI) reachable by adding a fourth primary colour, i.e. amber. The starting point for these improvements is the system implemented for the RGB colour control in its amplitude modulation solution. At the end of this study both solutions are implemented together Brightness control Aging is one of the most important sources of variation of the maximum brightness. There is no mathematical model for the impact of aging on the lumen output of an LED. Aging can cause the loss of even 50% of brightness in a LED. Aging changes only the amplitude of the power spectral density but not its shape. To maintain the power spectral density, the current through the LEDs must be increased. The maximum current for the LEDs used in this setup is 350mA. If the current exceeds this level, the LEDs can be damaged. To solve the problem of the maximum current, a block in the Simulink model limitates the control signal (voltage) for the drivers. At its maximum level, it allows a current of 350mA. The signals to the LEDs are clamped when their value is out of the work margin of 0 to 1. Brightness control is an important target to accomplish. At present, there are three parameters that have to be fixed to obtain a certain light in terms of colour gamut and intensity. Two of them are xy coordinates and the third one is the brightness. Each of these parameters has a possible region of work. For xy coordinates this space is inside the triangle that the red, green and blue LEDs coordinates describe in the chromaticity diagram. Brightness is the third dimension of this region. Fig.43 shows a possible region of work. It should be noted that this figure is only qualitative, it is just used to illustrate how the work region could be. 38 Koninklijke Philips Electronics N.V. 2007

39 Unclassified PR-TN 2007/00318 Figure 43 Brightness control. xyy region of work. Any set value, i.e. combination of xyy, outside this area is not reachable and causes a control error. This wrong control provokes the saturation of the signals that drive the LEDs. In the particular situation in which the requested xy coordinates are inside the region of work and the brightness is out of it, the control tries to increase the current through the LEDs. Almost one but probably most of the signals to the LEDs will be bigger than the maximum value permitted. This will provoke the clamping of the signals, the change of the needed relation between them and the loss of the desired xy coordinates. It is then necessary to control the requested brightness to the system to avoid this situation, that can appear when aging decreases the maximum brightness reachable for an LED at the beginning of its life or when the set brightness is directly too high Theory of control There are two problems related to the brightness that should be solved, aging and the maximum brightness that can be requested. In a first attempt, a model was tested which tried to solve the effects of aging. It worked properly. However, it was not able to solve the problem of the maximum brightness. It is interesting to explain this solution although it was not useful in the end because it will help us to better understanding the aging problematic. Koninklijke Philips Electronics N.V

40 PR-TN 2007/00318 Unclassified The way to avoid the problems related to the brightness is to know which is its maximum value reachable and to not permit a request larger than it. Unfortunately, this is not possible to accomplish during the manufacturing process because brightness depends on many external factors as for example manufacturing spread, temperature or aging. It is also not possible to fix a working range for the brightness due to the wide range of variation it has. Aging can cause a variation of even 50% of the LEDs initial brightness value and no mathematical model is known for it. It is then necessary to know the maximum brightness at every single moment during the LEDs operation. Sensor readings are the measurement of the light emitted by LEDs. When the light emitted decreases, sensor readings also decrease their value and the RGB colour control system tries to compensate this by increasing the current through the LEDs. This reduction of the light intensity is by definition a reduction of the brightness. When the brightness of the LEDs decreases too much, the control tries to increase the current through the LEDs over the maximum level allowed trying to maintain the set brightness. For fixed xy coordinates, the brightness is maximum when one of the control signals almost reaches its maximum level without going over it. As the first step to obtain a desired kind of light, xy coordinates and brightness (Y) are set. These three parameters are then transformed into tristimulus values and passed through G cal -1 to obtain the set sensor readings (SR set ). This transformation from xyy to SR set is bidirectional and it could be done in the inverse direction. This means that it is possible to calculate the brightness of the light using the sensor readings. If we know when we have the maximum brightness and how to obtain its value from the sensor readings, then the problem of controlling the brightness is solved. When the maximum signal for one LED is achieved, the brightness will be calculated from the sensor readings and it will be set as the maximum value for the brightness. In the future, aging will decrease the maximum brightness that the LEDs can achieve, so it will be mandatory to reduce the new Y set allowing the system to detect a new future saturation of the control signals. The control procedure will detect the achievement of the maximum value for the control signals, will calculate the brightness value and will set as the maximum Y we can request, for example, 90% of it. The solution adopted can solve perfectly the effects of the aging in the RGB colour control sytem. However, this solution does not work properly when the problem is not the aging but the brightness set manually. To modify manually the set brightness means to introduce a step in the control system. The variation of the control signals is faster than the dynamics of the measuring system. It is necessary that the variation of the control signals is slower than the measuring system allowing it to take the measurements of the sensor readings before the signals saturate. The problem is not if the first signal saturates. The problem is that the other signals are still increasing their value and the brightness and they are also modifying the xy coordinates. To solve the problem of the step caused by the manual set of the desired brightness, the main idea of the brightness control must be changed completely. The control signals for the LEDs (Am signals) have the information of when the system reaches its maximum level of brightness. It is not mandatory to know at which level of brightness this occurs, it is only necessary to not allow them to go over it. Hence it is possible to base the control 40 Koninklijke Philips Electronics N.V. 2007

41 Unclassified PR-TN 2007/00318 of the maximum brightness on the information that the Am signals provide. They are then used as a feedback to control the set brightness (Y set ). As it was shown in a previous chapter, the RGB colour control system presents an overshoot that can bring the transient response of the system to saturation despite it can work properly in the steady state. This is caused because any change of Y set means to inject a step in the control. To avoid this behaviour, the control of the Y set is necessary. To prevent going over the maximum and under the minimum values of the Am signals, these minimum and maximum values will be used as asymptotes for the limits of the working region of these Am signals. Y set will be controlled to eliminate the step to the control system. Y set will be increased or decreased using these formulas, Y Y = Y = Y calc calc ( Yset Ycalc ) ninc ( Yset Ycalc ) ninc + ( e e )( Am Am) m inc max ( Ycalc Yset ) ndec ( Ycalc Yset ) ndec + ( e e )( Am Am) mdec min where Y calc is the previous value of Y, Am max and Am min are the maximum and minimum values of the Am signals and the asymptotes for the curves, Am is the actual value of these signals and n inc, m inc, n dec and m dec are parameters that allow to control the speed and the overshoot of the Am response. The difference (Y set -Y calc ) controls the increment of Y. When this difference is close to zero, the increment applied to Y is also close to zero. If the difference is negative, the increment is then also negative. This allows the system to reach the Y set controlling how it is reached with the n parameters. Fig. 44 shows the curve for the increment that is added to Y calc to which the next formula corresponds, ( Yset Ycalc ) ninc ( Yset Ycalc ) ninc ( e e ) Koninklijke Philips Electronics N.V

42 PR-TN 2007/00318 Unclassified Figure 44 Factor of increment. Fig. 45 shows how Yset is theoretically reached depending on the selection of n. Figure 45 - Y set. The m parameter present on the second half of the formula will have a similar effect on the set brightness. Both parameters will be selected in order to obtain a certain transient response in terms of speed and maximum allowed overshoot. The decrement of Y, when Y set is smaller than the actual Y, follows the same rules seen before but in the opposite direction of correction. The value of Y set affects directly the value of the increment. The bigger Y set, the bigger the increment will be. Two different values of the reachable Y will present a different 42 Koninklijke Philips Electronics N.V. 2007

43 Unclassified PR-TN 2007/00318 behaviour in spite of that the final Am value will be the same. In the worst case the system will be even unstable and will oscillate. To improve and try to avoid this behaviour, the maximum value for Y set allowed will be the result of the sum of the brightness of every single LED working at its maximum current. When the Am signal is close to its maximum or minimum permitted value, the difference (Am max -Am) or (Am min -Am) also reduces the increment or decrement to add to Y calc. It will be close to zero when the Am signals are close to the limits of the working region pre-established allowing describing the asymptotic behaviour for them. Am is the highest value among the Am signals when Y is increased and it is the lowest one when it is decreased. Aging can cause the reduction of even 50% of the maximum brightness reachable. To compensate this loss of brightness, the system will try to maintain Y set increasing the current through the LEDs. This will be done increasing the Am signals. If these signals go over the limit of the working region, then the colour of the light will change. To avoid this behaviour, Y set must be reduced accordingly. In this situation, (Am max -Am) will be negative, helping the increment for Y to be also negative and as a consequence the new Y for the control system will be reduced. The time constant for the aging is high enough to allow the control to reduce Y set before the colour of the light varies. This permits to work with the entire available brightness unlike it occurs with the solution of control explained at the beginning of this chapter. In spite of this, a 10% of security margin is adopted to avoid other sources of interferences that can affect the brightness System model The starting point for the implementation of the brightness control solution is the RGB colour control diagram in its amplitude modulation version. Figure 46 AM RGB colour control. Simulink diagram. In the Fig. 47 the modifications introduced in the Simulink model are shown. The set brightness (Y set ) is now passed through the Brightness control block where it is modified as it was explained in the theory chapter. The Am signals are the feedback for the control procedure. These signals are measured before passing through the Signals level adapter Koninklijke Philips Electronics N.V

44 PR-TN 2007/00318 Unclassified block. It is important to know their real value before being clamped when they are out of the working area in order to select among them the biggest or smallest one. Figure 47 AM RGB colour control with brightness limitation. Simulink diagram modifications. The brightness control block is shown in Fig.48. Figure 48 Brightness control block. All the functional subsystems have been highlighted for the easy understanding of the control system. In the maximum brightness detector, Y set is compared with the maximum brightness that can be set. If it is bigger than this maximum value, Y max is passed as the set brightness. 44 Koninklijke Philips Electronics N.V. 2007

45 Unclassified PR-TN 2007/00318 The maximum brightness detector includes the maximum brightness calculator (Fig. 49), where the maximum brightness (Y max ) that can be set is calculated as the sum of the brightness of every single LED when the maximum current allowed flows through them. In the calibration procedure of the RGB colour control system, the LEDs work at 50% of their maximum current. The brightness achieved in this procedure is then half of the theoretical maximum value. The brightness is supposed to be directly proportional to the current through the LEDs. Then, it is necessary to multiply these measured values of brightness by a factor of two to obtain the brightness at full current. This is not fully correct because in the operation limits of the LEDs the brightness does not follow this rule and does not increase its value in the same way. However, this difference will provide us a little margin to guarantee the possibility to fix Y set at the maximum level that the system can achieve. The limitation of Y set is needed because the factor of correction that has to be applied in the Y calculation depends directly on the difference between the Y set and the actual Y delivered to the RGB colour control. If Y set is too high, the system can be unstable, as it was explained in the control theory chapter. Figure 49 Maximum brightness calculator. The brightness memory block provides the feedback for the calculation of the correction factor (increment or decrement). The output of the brightness control block is the value of Y that must be injected to the RGB colour control. It is obtained as the sum of the previous value of brightness, Y calc, and the correction factor, Y inc. The result, Y, is passed through a memory block and becomes Y calc for the next calculus. Y inc is calculated in the correction factor block. In this block the two possible corrections, increment and decrement, are calculated according to the inputs Y set, Y calc and Am in its two versions, Am big and Am small. The correction factor block (Fig.50) implements the formulas for the increment and decrement values. The parameters n inc, m inc, n dec and m dec can be adjusted internally to modify the control law. Koninklijke Philips Electronics N.V

46 PR-TN 2007/00318 Unclassified Figure 50 Correction factor block. As outputs, the correction factor block delivers two values, one for the increment and another for the decrement of Y. The set brightness change detector will decide which of them will be used. This subsystem memorizes the last value of brightness set when the Y set is changed. This change is detected by the detect change block. This block detects any event in its input and delivers a short pulse in its output. This pulse changes the position on the selector block and the last Y sent to the control system is then passed through it. This selector has also a closed loop between its output and one of its inputs with a memory block which memorizes the previous value passed. This new memorized value is then compared with the new Y set requested. Depending on the result of the comparison, the subsystem will decide if the value for the correction has to be an increment or a decrement. Am selector (Fig.51) chooses among the Am signals the biggest and the smallest one. These two signals are used later to calculate the increment and decrement values as outlined before. Am signals are passed through a memory block before the Am selector block. Simulink forces to do it in this way to permit the use of the feedback loops. 46 Koninklijke Philips Electronics N.V. 2007

47 Unclassified PR-TN 2007/00318 Figure 51 - Am selector block System performance As for the RGB colour control implementation, the first step to analyze the system performance will be to work in a simulated environment and later the model of the system will be transferred to the rapid control prototyping system. The set point selected for the RGB colour control implementation is maintained to compare both solutions and highlight the improvements of the new brightness control. The response of the system for the set point x = y = Y = is shown in Fig.52, Koninklijke Philips Electronics N.V

48 PR-TN 2007/00318 Unclassified Figure 52 AM RGB colour control. Brightness control. Simulated transient response. and the response for the RGB colour control before controlling the brightness is depicted in Fig. 33 as it was shown in a previous chapter. Figure 33 AM colour control. Simulated transient response. As it can be seen (Fig. 52), the overshoot has disappeared and the relation between the Am signals is never lost. This means that the colour of the light does not change during the transient response unlike it did before. Without controlling the brightness, the system reaches its final value 1.5 seconds after being turned on. Now, a longer time is needed to reach the same value. However, 0.5 seconds after being turned on, the value reached differs by approximately 15% from its final value, after 1.5 seconds it is close to 5% and in 4 seconds around 1.5%. This difference can be accepted as a good response of the system because of the characteristics of the human eye. The human eye is not too sensitive to brightness changes. However, it can detect little changes in colour. Therefore, it is appropriate to say that the response of the system has been improved. 48 Koninklijke Philips Electronics N.V. 2007

49 Unclassified PR-TN 2007/00318 The next two figures are used to show how the response of the system for some set brightness was before limiting its set value and how it has changed after limiting it. Fig. 53 is also interesting to show how the new system implementation has a different response in terms of speed and over shoot depending on the set brightness in spite of the work point (x,y) not having been changed. Figure 53 AM RGB colour control. Brightness control. Simulated transient response II. Figure 36 AM colour control. Simulated transient response IV. Fig. 54 shows the response of the system with brightness control and Fig. 34 without it when the brightness is changed from 0.08 to The same behaviour as for increasing Y set can be observed. Koninklijke Philips Electronics N.V

50 PR-TN 2007/00318 Unclassified Figure 54 AM RGB colour control. Brightness control. Simulated transient response III. Figure 34 AM colour control. Simulated transient response II. Fig. 55 shows the curve described by the brightness delivered to the colour control system. It should be noted that this curve is obtained for a specific value of the parameters n and m, brightness and working point. 50 Koninklijke Philips Electronics N.V. 2007

51 Unclassified PR-TN 2007/00318 Figure 55 AM RGB colour control. Set brightness. The response of the system can be modified by changing the values of the parameters m and n that are present in the formula that control the brightness. With these two parameters it is possible to control the transient response of the Am signals. The next figures show different system responses for some values of m and n and the set point x=0.33, y=0.33, Y=0.05. Figure 56 AM RGB colour control. AM and brightness transient responses. n modification. Koninklijke Philips Electronics N.V

52 PR-TN 2007/00318 Unclassified Figure 57 AM RGB colour control. AM and brightness transient responses. m modification. As it can be seen, both m and n control the speed of the system response and the magnitude of the overshoot. This response is also influenced by its proximity to the maximum Am value. In the future, it could be interesting to control these two parameters allowing them to vary according to the value of the set brightness, what can make the response of the system more uniform independent of this set brightness. Once the desired results have been obtained in the simulation environment, the model of the system is transferred to the rapid control prototyping system. Figures 58 and 59 show the shape of the transient response of the system at different brightness values. This shape depends basically on two values: The first one is the difference between the requested brightness and the brightness requested previously, where we are and where we want to arrive. In this case, Fig. 58 shows the rise up of the Am signals starting with the system off, brightness equal to zero, and Fig. 59 shows the inverse situation, starting from Y setmax. The second one is the distance between which is going to be the final value of the Am signals and the maximum permitted value for them. How far we are from the Am max. 52 Koninklijke Philips Electronics N.V. 2007

53 Unclassified PR-TN 2007/00318 Figure 58 AM RGB colour control. AM transient response I. Figure 59 AM RGB colour control. AM transient response II. In both figures it is possible to see a little overshoot for the lower difference of brightness requested that disappears when it is increased. This effect is a consequence of the second value explained previously. This value brings the system to an over damped response when it decreases. To this point, the behaviour of the system is as what we obtained as a result when we used the Simulink model of the plant before. The next figures show how the system response changes with the m and n parameters. Here only the situation in which the system starts from a set brightness of zero is shown. All the other possible scenarios, as for example the decrement of Y set, will show a similar behaviour. Koninklijke Philips Electronics N.V

54 PR-TN 2007/00318 Unclassified Figure 60 AM RGB colour control. Transient response. Variation of n. Figure 61 AM colour control. Transient response. Variation of m RGBA control A high lumens-per-watt value does not necessarily mean that the quality of the light is also good. The notion of colour quality can be a subjective measure. The colour rendering index is a unit of measure that defines how well colours are rendered by different illumination conditions in comparison to a standard. The CRI has been widely adopted and used by the lighting industry to characterize the quality of a light source [4]. Light sources are compared to a reference with the same colour temperature and scored based on the colour shift from a palette of base colours. 54 Koninklijke Philips Electronics N.V. 2007

55 Unclassified PR-TN 2007/00318 Figure 62 RGBA control. CRI improvement. Colour rendering for different light sources. LED lamps comprising more than three primary colours exhibit an increased colour gamut and improved CRI. This is the reason why it is interesting to modify the system to allow it to control more than three primary colours. Figure 63 RGBA board prototype. Because the CRI score is directly related to the spectral power distribution of the light source, it is possible to manipulate the spectrum to produce a higher CRI value. In our implementation a fourth amber LED is used to improve the CRI. Koninklijke Philips Electronics N.V

56 PR-TN 2007/00318 Unclassified Theory of control If more than 3 primary colours are mixed then the control procedure outlined in chapter 2 has to be modified. This will be discussed in the following for mixing 4 primary colours, i.e. red, green, blue, and amber (RGBA). Whilst there are still 3 sensor readings as specified by (2.5), there have to be now 4 control signals instead of the 3 control signals specified in (2.6). r g CS = (3.1) b a As a consequence of this, the matrices G C2T as defined in (2.9) and G C2S as defined in (2.11) will no longer be quadratic but matrices with 3 rows and 4 columns. Therefore, it is no longer possible to determine their inverse which would be needed in (2.13), (2.14), and (2.26). The control of this LED lamp with 4 primary colours will be implemented by reducing it to the control of a LED lamp with 3 primary colours dealt with in chapter 2. This is achieved by introducing a fixed relation between the control signals of 2 of the 4 primary colours. In the following, the control signals r and a for the drivers for the red and amber LED will be derived from a common control signal ra and a parameter k defining their ratio. ra CS = g (3.2) b 0 k 1 k < 1/ 2: k > 1/ 2: r = ra a = ra a = 2 k ra r = 2 ( 1 k) ra (3.3) Using this fixed relation between 2 control signals, the control procedure outlined in chapter 2 can be applied for each value of k. Making k a variable parameter of the control procedure that can be adjusted in a feed forward scheme allows to exploit fully the colour gamut of the LED lamp and may be used to optimize its colour rendering. In order to do so, the transfer functions G CAL -1 and G C2S -1 (cf. Fig. 2.6) have to be known for each k. Interestingly, G dyn (s) does not depend on k since the control is designed such that the sensor signals are decoupled. 56 Koninklijke Philips Electronics N.V. 2007

57 Unclassified PR-TN 2007/00318 The transfer functions G -1 CAL and G -1 C2S can be determined as a function of k by executing the calibration procedure outlined in chapter 2 for several values of k and interpolating in between. k is defined piecewise on the intervals 0 k 1/ 2 and 1/ 2 k 1. Therefore, when considered as functions of k, the elements of the transfer functions G -1 CAL and G -1 C2S are expected to exhibit discontinuities at k = 1/ 2 and these two intervals will be dealt with separately. Then the most simple procedure would be to determine the transfer functions G -1 CAL and G -1 C2S for k=0, k=1/2, and k=1 and interpolate linearly in between in the intervals 0 k 1/ 2 and 1/ 2 k 1, respectively. When the light of two LEDs is mixed, the resultant light has its chromaticity coordinates placed in the line that joins their chromaticity coordinates. The final point in this line depends on the luminous flux of the LEDs and also depends on their chromaticity coordinates. It will vary according to the LEDs used. The exact characteristics of the mixed light can only be obtained measuring them with an oscilloscope. However, this can not be done in a real environment. It is not practical to measure these characteristics every time the mixing relation between the LEDs is changed. Because of this, it is necessary to use a procedure which will estimate these values automatically according to a previous calibration. The solution adopted at this moment only takes care of the changes because of the modification of the k parameter presented before. This parameter will change the signal that drives the LEDs. This will affect the luminous flux of the LEDs. Other influences as for example temperature variation or aging are not taken into account. Measurements of the mixed light are taken at some k values with the spectrometer. Results are shown below and in Fig. 64. These values will be used later to check how accurate the procedure is. AMBER-RED LEDS (measured) Am 0,5 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0,6960 0,6756 0,6603 0,6507 0,6450 0,6417 0,6359 0,6275 0,6154 0,5964 0,5672 y 0,3034 0,3238 0,339 0,3484 0,3542 0,3574 0,3632 0,3715 0,3837 0,4024 0,4314 Y 0,0134 0,0166 0,0199 0,0226 0,0248 0,0264 0,0247 0,0226 0,0201 0,0174 0,0146 Table 2 Spectrometer measurements of the mixed light of amber-red LEDs. Koninklijke Philips Electronics N.V

58 PR-TN 2007/00318 Unclassified Figure 64 Chromaticity coordinates of amber-red light. Also the light characteristics of each LED are individually measured at some Am values. Results are shown below. Later, some of these values will be used as the values for the calibration procedure. RED LED Am 0,5 0,4 0,3 0,2 0,1 0 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0,6961 0,6961 0,6961 0,6961 0,6961 0,6961 0,6958 0,6956 0,6954 0, y 0,3034 0,3034 0,3034 0,3034 0,3034 0,3034 0,3035 0,3036 0,3038 0, Y 0,0134 0,0134 0,0134 0,0134 0,0134 0,0134 0,0112 0,0085 0,006 0,003 0 AMBER LED Am 0 0,1 0,2 0,3 0,4 0,5 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0 0,5545 0,5567 0,5601 0,5636 0,5672 0,5672 0,5672 0,5672 0,5672 0,5672 y 0 0,4433 0'4414 0,4382 0,4348 0,4314 0,4314 0,4314 0,4314 0,4314 0,4314 Y 0 0,0036 0,0075 0,0105 0,0129 0,0146 0,0146 0,0146 0,0146 0,0146 0,0146 Table 3 Individual spectrometer measurements of amber LED and red LED lights. 58 Koninklijke Philips Electronics N.V. 2007

59 Unclassified PR-TN 2007/00318 Once defined the real behaviour of the LEDs we are going to work with in terms of chromaticity coordinates and luminous flux for any of the fixed k values, the mixed light will be then calculated according to next formulas, x = i i Y xi y Yi y i i i Yi i y = z =1 x y Yi y i i where x i, y i and Y i are the chromaticity coordinates and the luminous flux of each of the LEDs involved in the light generation. The chromaticity coordinates and luminous flux of one light can be calculated as the addition of some individual lights, in this case the light of some LEDs. The next table shows the supposed characteristics of the light using the previous formulas for the same k values used before when they were measured with the spectrometer. AMBER-RED LEDS (calculated) Am 0,5 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0,6961 0,6741 0,6575 0,6481 0,6428 0,6401 0,6341 0,6252 0,6147 0,5962 0,5672 y 0,3034 0,3252 0,3416 0,3509 0,3562 0,359 0,3648 0,3736 0,3842 0,4023 0,4314 Y 0,0134 0,017 0,0209 0,024 0,0263 0,028 0,0258 0,0231 0,0207 0,0177 0,0146 Table 4 Calculated brightness and chromaticity coordinates of amber-red mixed light using individual spectrometer measurements of each LED. The results obtained do not match exactly with what was measured using the spectrometer. This difference can be assumed not important if there is not a big difference between light characteristics. It is assumed that a change in light characteristics can not be observed by the human eye if this variation is smaller than uv< when using uv coordinates, however, according to a 4-step MacAdam ellipse, a difference of can be accepted. A methodology was created by MacAdam in 1943 for mathematically constructing ellipses about target points (somewhat useful to lamp industry and now part of ANSI standards). The goal of the original research was to determine a series of boundaries around several colour targets (x, y coordinate) on the CIE chromaticity diagram, illustrating how much one can stray from the target (along various colour axes) before perceiving a difference from the target colour [10]. MacAdam ellipses are described as having steps, which really means standard deviations. If a large sample of the population were used (which it was not) and if a Koninklijke Philips Electronics N.V

60 PR-TN 2007/00318 Unclassified trained observer could reliably repeat his observations (which he can not), then the steps would translate to probabilities for the general population as follows: 1 sd = % of the general, colour-normal population 2 sd = % 3 sd = % Any point on the boundary of a 1-step ellipse, drawn around a target, represents 1 standard deviation from the target. Note that this also means that drawing a line through the target from that point, thereby creating a point on the opposite boundary, the 2 boundary points will be 2 standard deviations from one another. Any point on the boundary of a 2-step ellipse represents 2 standard deviations from the target. For a 3- step ellipse, the boundary represents 3 standard deviations from the target, and so on. ANSI recommends that lamp manufacturers stay within a 4-step MacAdam ellipse. This means that, given a certain target point on the CIE diagram, lamp manufacturers are given a fairly wide range of perceptible differences. Consider that a point on the boundary of a 4-step ellipse is 8 standard deviations from a point on the opposite side of that same boundary. 60 Koninklijke Philips Electronics N.V. 2007

61 Unclassified PR-TN 2007/00318 Figure 65 MacAdam ellipses. As it was said before, uv coordinates are going to be used to check the accuracy of the calculated characteristics. A difference smaller of between measured and calculated light will indicate that the approximation is good enough (4-step ellipse). Light characteristics can be calculated as the combination of two LEDs with the previous formulas. The following formulas carry out the coordinate s transformation and the calculation of the deviation uv in colour point (u,v) from the reference (u r,v r ). Koninklijke Philips Electronics N.V

62 PR-TN 2007/00318 Unclassified u = ( X 4X + 15Y + 3Z) v = ( X 6Y + 15Y + 3Z) Δ uv = 2 ( u ur ) + ( v vr ) 2 In the next three tables can be seen the deviations between the measured light by the spectrometer, the reference, and the expected light using formulas. AMBER-RED LEDS (measured) Am 0,5 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0,6960 0,6756 0,6603 0,6507 0,6450 0,6417 0,6359 0,6275 0,6154 0,5964 0,5672 y 0,3034 0,3238 0,339 0,3484 0,3542 0,3574 0,3632 0,3715 0,3837 0,4024 0,4314 Y 0,0134 0,0166 0,0199 0,0226 0,0248 0,0264 0,0247 0,0226 0,0201 0,0174 0,0146 Table 2 Spectrometer measurements of the mixed light of amber-red LEDs. AMBER-RED LEDS (calculated) Am 0,5 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0,6961 0,6741 0,6575 0,6481 0,6428 0,6401 0,6341 0,6252 0,6147 0,5962 0,5672 y 0,3034 0,3252 0,3416 0,3509 0,3562 0,359 0,3648 0,3736 0,3842 0,4023 0,4314 Y 0,0134 0,017 0,0209 0,024 0,0263 0,028 0,0258 0,0231 0,0207 0,0177 0,0146 Table 4 Calculated brightness and chromaticity coordinates of amber-red mixed light using individual spectrometer measurements of each LED. uv 0 0,0028 0,0049 0,0044 0,0036 0,0027 0,0028 0,0034 0,0009 6E-05 0 Table 5 Deviation between measured and calculated characteristics of amber-red mixed light. At this point, the obtained error is the difference between how we think the mixed light is, starting from the characteristics of two single LEDs, and how the spectrometer says it is. This error is smaller than ANSI recommends in the whole range measured. However, despite we have been working with the real characteristics of the LEDs, previously measured with the spectrometer, the error is not null. This means that the method employed introduces some kind of errors. This was also seen when the calibration procedure was explained in a previous chapter. Here appears again the same problematic when dealing with LEDs. Their characteristics have many sources of interferences that 62 Koninklijke Philips Electronics N.V. 2007

63 Unclassified PR-TN 2007/00318 can vary the measurements. It is then necessary to be critical with these results and even more with the ones we are going to obtain when interpolating. Considering good results the ones obtained before, it is possible to assert that we can know how the mixed light is going to be if the lights it comes from are known. The next step will be to be able to know how these lights are, which the characteristics of the LEDs are, and using the previous procedure to calculate the mixed light characteristics. It is neither possible nor practical to measure every combination of LEDs. It is then necessary to interpolate between pre-established values. The selected values for interpolation are those for Am=0.1 and Am=0.5 of each LED. The variations of LED characteristics are considered linear when varying Am. However, this is not true at all and it is mandatory to use only the linear section to avoid later errors. That is why not the whole Am range of work is used. The upper 0.5 limit is imposed by that used when calibrating, while 0.1 is used because measurements at lower Am values cause wrong spectrometer results related to the little light generated by the LED. This means that the effective range of work of k is from 0.1 to 0.9, which corresponds to full red LED (Am r =0.5) and 20% amber LED (Am a =0.1) and 20% red LED (Am r =0.1) and full amber LED (Am a =0.5). Just to point, this limited range of Am to 0.5 is later extended to Am=1 so the maximum brightness of the LEDs is accessible. Talking here about Am is always related to the calibration procedure and not to the normal working stage. To check how the interpolation works, the mixed light is calculated for the same k values used before. Results are shown in next table. RED LED interpolated Am 0,5 0,4 0,3 0,2 0,1 0 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0,6961 0,6961 0,6961 0,6961 0,6961 0,6961 0,6957 0,6954 0,6950 0, y 0,3034 0,3034 0,3034 0,3034 0,3034 0,3034 0,3035 0,3035 0,3036 0, Y 0,0134 0,0134 0,0134 0,0134 0,0134 0,0134 0,0108 0,0082 0,0056 0,003 0 AMBER LED interpolated Am 0 0,1 0,2 0,3 0,4 0,5 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0 0,5545 0,5577 0,5608 0,5640 0,5672 0,5672 0,5672 0,5672 0,5672 0,5672 y 0 0,4433 0'4403 0,4374 0,4344 0,4314 0,4314 0,4314 0,4314 0,4314 0,4314 Y 0 0,0036 0,0064 0,0091 0,0119 0,0146 0,0146 0,0146 0,0146 0,0146 0,0146 Table 6 Amber-red LED characteristics as a result of interpolation (blue) and spectrometer measurements (black). Koninklijke Philips Electronics N.V

64 PR-TN 2007/00318 Unclassified AMBER-RED LEDS (measured) Am 0,5 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0,6960 0,6756 0,6603 0,6507 0,6450 0,6417 0,6359 0,6275 0,6154 0,5964 0,5672 y 0,3034 0,3238 0,339 0,3484 0,3542 0,3574 0,3632 0,3715 0,3837 0,4024 0,4314 Y 0,0134 0,0166 0,0199 0,0226 0,0248 0,0264 0,0247 0,0226 0,0201 0,0174 0,0146 Table 2 Spectrometer measurements of the mixed light of amber-red LEDs. AMBER-RED LEDS (calculated) Am 0,5 k 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 x 0,6961 0,6741 0,6620 0,6527 0,6456 0,6401 0,6330 0,6241 0,6124 0,5962 0,5672 y 0,3034 0,3252 0,3371 0,3464 0,3535 0,359 0,3659 0,3747 0,3862 0,4023 0,4314 Y 0,0134 0,0170 0,0198 0,0225 0,0253 0,028 0,0254 0,0229 0,0203 0,0177 0,0146 Table 7 Calculated brightness and chromaticity coordinates of amber-red mixed light using individual interpolated characteristics of each LED. uv 0 0,0028 0,0032 0,0035 0,0011 0,0027 0,0045 0,0051 0,0041 6E-05 0 Table 8 Deviation between measured and calculated by interpolation characteristics of amber-red mixed light. Using interpolation, uv is still smaller than the maximum recommended. The method used gives good results and is useful to predict how the light characteristics we are dealing with are. In spite of that, the error is uncontrolled and very dependant on the calibration procedure. Improving this procedure it is supposed to be possible to improve the uv. It is possible to reduce the number of effective LEDs from four to three with the procedure presented in this chapter in order to be able to use the formulas presented in the RGB colour control theory chapter. Summarizing, from now on the calibration procedure to follow when RGBA colour control is used will consist of firstly measuring the characteristics of the four LEDs at Am equal to 0.5 and secondly measuring the characteristics of the red and amber LEDs at Am equal to 0.1 in order to be able to interpolate later. When the system will be running, a calibration module will calculate the third LED characteristics from the stored data of red and amber LEDs according to the value of k and then the calibration matrices G CAL -1 and G C2S System model As for the brightness control solution, the starting point for the implementation of the RGBA control is the RGB colour control diagram in its amplitude modulation version. 64 Koninklijke Philips Electronics N.V. 2007

65 Unclassified PR-TN 2007/00318 Figure 46 AM RGB colour control. Simulink diagram. Fig. 66 shows the modifications introduced into the Simulink model. Figure 66 AM RGBA colour control. Simulink diagram. The new model has four main elements, a slider gain named RA which controls the relation between the red and amber LEDs as it was seen in the theory chapter, a block named RA Limits which establishes the range of proportionality available for the mix of the LEDs, another block named RA to R-A which splits one AM control signal into two AM signals, one for the red LED and another one for the amber LED according to the RA value selected, and finally, one block which recalculates the matrices G CAL -1 and G C2S -1. The proportion in which the red and amber LEDs are mixed is limited. The minimum relation between them is 100%-20% in both combinations, red-amber or amber-red. This limit is used to avoid the problems that appear working at lower relations when calculating the calibration matrices. In this Simulink model, the colour control system calculates the calibration matrices depending on which the relation between red and amber LEDs is. Remember that both Koninklijke Philips Electronics N.V

66 PR-TN 2007/00318 Unclassified LEDs will be dealt as a unique LED. It is supposed that the characteristics of a LED in terms of luminous flux and chromaticity coordinates are linear in its whole range of work. However, that is not real at all. In the limits of the work region, when the signal that drives the LED is too small or close to one, these characteristics are not linear and this results in a wrong estimation of the resultant LED and, consequently, and in a wrong calculation of the calibration matrices. The RA Limits block is presented in next figure. RA is limited between 0 and 1 and the block allows any value between 0.1 and 0.9. If RA is smaller than 0.1, the output is fixed to 0 and if it is bigger than 0.9 it is fixed to 1. Figure 67 RA Limits block. The RA to R-A block applies the relation presented in the theory chapter (note that k is renamed in the Simulink model by RA). 0 k 1 k < 1/ 2: k > 1/ 2: r = ra a = ra a = 2 k ra r = 2 ( 1 k) ra 66 Koninklijke Philips Electronics N.V. 2007

67 Unclassified PR-TN 2007/00318 Figure 68 RA to R-A block. The calculation of the calibration matrices is now partially automated. The calibration process described at the beginning of this report has been somewhat modified. Following the new calibration recommendations suggested by the investigations described before, tristimulus values of the four LEDs are measured at 50% of the LED maximum driving signal. This allows reducing the non-linearity of the LEDs at their maximum power. Sensor readings are also measured and extended to the work region (multiplied by two). It is still necessary to measure the tristimulus values of the red and amber LEDs at 20% of the previous driving signal. It is 10% of their maximum power. With these last measurements and the ones at 50% of the red and amber LEDs is calculated, depending on the RA value, a new pattern of tristimulus values that correspond to a new redamber LED. Remember that it was mandatory to reduce the number of LEDs in order to be able to use the expressions presented in the RGB control theory. Koninklijke Philips Electronics N.V

68 PR-TN 2007/00318 Unclassified Figure 69 CIE 1931 Chromaticity diagram. LEDs reduction. The automation of the process refers to the automatic interpolation between the tristimulus values of the red and amber LEDs at 50% and 10% according to RA value and the procurement of the new LED and subsequent calculation of the matrices. Previously it was made by hand. The next figures show what is inside the block. Its functionality has already been explained. Here the formulas are implemented that calculate each element of the G CAL -1 and G C2S -1 matrices. 68 Koninklijke Philips Electronics N.V. 2007

69 Unclassified PR-TN 2007/00318 Figure 70 - GCALinv & GC2Sinv matrices calculation block and subsystems. Koninklijke Philips Electronics N.V

70 PR-TN 2007/00318 Unclassified Adding a fourth LED also enforces other changes in the Simulink model. The modelled plant used in simulation now accounts for the contribution of the new LED. Figure 71 Plant model of the system used in simulation development mode System performance The same point of work used testing the RGB colour control is now used. x = 0.33 y = Y = RA is set to 0.3, what means a relation 1/0.6 between red and amber LEDs. As it can be seen in the next figures, the final amplitudes of the driving signals have changed from those we obtained previously, however their shape is similar. 70 Koninklijke Philips Electronics N.V. 2007

71 Unclassified PR-TN 2007/00318 Figure 72 RGBA colour control. Transient response I. Figure 73 RGBA colour control. Transient response II. Fig. 74 shows how the driving signals change when RA is changed from 0, RGB colour control, to 1, GBA colour control in steps of 0.1. It is interesting to see that the GBA control is not stable at these rates and how the amber signal saturates. This occurs because the requested brightness is not reachable. The amber LED has less brightness than the red LED. This problem would be solved using the brightness control presented in the previous chapter. Koninklijke Philips Electronics N.V

72 PR-TN 2007/00318 Unclassified Figure 74 RGBA colour control. Simulated transient and steady state response for some RA values. Once the model is tested by simulation obtaining good results, it is transferred to the dspace rapid control prototyping. The next figure is a capture of the driving signals in the dspace environment. 72 Koninklijke Philips Electronics N.V. 2007

73 Unclassified PR-TN 2007/00318 Figure 75 RGBA colour control. AM transient response. With RGBA control, it is interesting to test the improvement of the CRI, the major aim of the modifications introduced in the Simulink model. It was said previously that the CRI could be improved controlling the spectral power density of the light generated. The next figures show the spectral power density before and after the addition of the fourth LED. It is easy to observe that the gap between the green and red peeks is now filled by the amber LED. Therefore, it is possible to assert that the CRI can be changed and improved. Figure 76 Spectral power density of resultant light using RGB and RGBA colour control. Next figure shows a series of captures of the spectral power density for different values of RA, starting with RA=1 and finishing with RA=0.1. Koninklijke Philips Electronics N.V

74 PR-TN 2007/00318 Unclassified Figure 77 Spectral power density for some RA values. The next table shows the colour points and brightness values obtained with a spectrometer for the previous combinations of LEDs. Table 9 Brightness and chromaticity coordinates of RBGA mixed light. uv coordinates give us a reference to confirm whether the colour point obtained is good enough. As it was explained in a previous chapter of this report, the UV error must be lower than if it is desired that the human eye does not observe any difference in the colour of the light. The worst obtained case from the table has an error of Δ( u, v) < (0.008,0.003) This error is bigger than it is desirable; however, it is quite close to an acceptable value. This increasing of the error is probably caused by the calibration process, where red and amber LEDs are measured at power levels too close to zero. As it was explained in the calibration process, at these rates the characteristics of the LEDs are not as linear as they were supposed to be. 74 Koninklijke Philips Electronics N.V. 2007

75 Unclassified PR-TN 2007/ Systems integration At this point, both modifications of the original RGB colour control, brightness control and RGBA colour control are implemented together. These modifications are independent from each other and only a few extra modifications are needed. Fig. 78 shows the resultant Simulink model. Figure 78 Multiple primary LED lamp colour controller with inherent brightness limitation. The system performance follows the behaviour found previously. As example, Fig.79 shows the response of the system when a lower brightness is requested from the maximum brightness reachable by the system for a colour point (x,y)=(0.33,0.33) and RA=0.3 (red fully on and amber at 60% of its maximum brightness). Koninklijke Philips Electronics N.V

76 PR-TN 2007/00318 Unclassified Figure 79 RGBA colour control. AM transient response for some RA values. 76 Koninklijke Philips Electronics N.V. 2007