Novel Approach for LED Luminous Intensity Measurement

Transcription

1 Novel Approach for LED Luminous Intensity Measurement Ron Rykowski Hubert Kostal, Ph.D. * Radiant Imaging, Inc., Main Street NE, Duvall, WA, ABSTRACT Light emitting diodes (LEDs) are being utilized as the light source in increasingly complex and sophisticated products, including flat panel displays, surgical lamps and even digital projectors. These applications place extreme demands on LED performance, which, for both the developer and manufacturer, translate into the need to precisely characterize and control source output, specifically color and luminous intensity distribution characteristics. Unfortunately, the traditional methods for performing luminous intensity and colorimetric measurements of LEDs suffer from several significant drawbacks. In particular, spot photometers and radiometers only sample a very limited amount of source output and operate very slowly. The latter factor can be an important consideration, even in research settings, because LED output is often not stable over time, especially during warm-up or in the presence of temperature or input power fluctuations. Thus, a long data acquisition period can make an instrument report spatial output variations that don t really exist. Now, new instrumentation based on the Imaging Sphere enables rapid, high spatial resolution measurement of LED color and luminous intensity over an entire hemisphere. This paper reviews the parameters typically utilized to characterize LEDs, explores Imaging Sphere operation, and compares the results of Imaging Sphere measurements with highly accurate reference data from a goniophotometer. Keywords: LED, light emitting diode, luminous intensity, colorimetry, photometry, radiometry, Imaging Sphere, output metrology, output characterization, color measurement 1. INTRODUCTION For most of the time since its commercial introduction over 40 years ago, the LED has largely been used as simple indicator light or numeric display backlight source. The vast majority of these illumination applications are relatively undemanding in terms of LED output requirements. Specifically, they don t need high accuracy or consistency in terms of LED wavelength (color) or spatial output characteristics. Table 1. Representative luminaire and display applications for LEDs Architectural luminaires Signs Displays Backlights Instrument panels Projectors Medical and scientific instruments Task lighting Automotive tail lamps and headlamps Traffic signals This situation has changed dramatically over the past few years. This is primarily due to the development of higher output LEDs and the expanding need for low-cost, large size displays; display backlights; keyboard and control pad illumination; and low-cost lighting systems. These factors have fueled the demand for LED sources, and also placed * kostal@radiantimaging.com, phone +1 (425) , fax +1 (425) , Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 1

2 much more stringent demands on their optical performance. This, in turn, has created a need for accurate, high speed, low cost methods for characterizing LED output for manufacturers, systems integrators, and end-users. 2. LED MEASUREMENT PARAMETERS AND REQUIREMENTS For luminaire and display applications of LEDs, the two primary optical parameters of interest are color and brightness, both as a function of angle. There are a number of ways these characteristics can be defined and quantified, depending upon the needs of the particular application Defining Color Probably the most obvious way to represent the spectral properties of a light source is to plot its output power as a function of wavelength. This is precisely what is measured by a spectroradiometer. Once obtained, the spectral power distribution of a light source can be used to derive a number of other quantities, including dominant wavelength, full width half maximum (FWHM) spectral bandwidth, color rendering index (CRI) and correlated color temperature (CCT). However, the one thing not immediately apparent when examining a plot of spectral power distribution is the color and brightness that the source would appear to the human eye. Fig. 1. The CIE 1931 Color Chart represents all possible colors with two coordinates (CIE x and CIE y). In order to measure color appearance, it is necessary to define a calibrated color space that is well correlated with the human visual response. For example, the CIE color space uses three tristimulus values to quantify the apparent color of a source or surface, and then plots two quantities (e.g. CIE x and CIE y) derived from the tristimulus values. This approach allows all possible colors to be represented in a two-dimensional plot because brightness is ignored (in other words, all shades of a given color are plotted at a single point) and because the three color coordinates (CIE x, CIE y and CIE z) are normalized. This enables the third coordinate to be calculated given the other two. In addition to chromaticity coordinates, the parameters of dominant wavelength and CCT can also be calculated from the tristimulus values. While tristumulus values can be calculated from the spectral power distribution, it is not always convenient or necessary to measure LED sources using a spectroradiometer, especially when spatially resolved (color as a function of angle) information is desired. Tristimulus values can also be measured more directly by using a set of color filters whose Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 2

3 transmission response corresponds to the tristimulus curves. This is the approach used in imaging colorimetry, which mates a CCD camera with the CIE matched color filters to directly measure chromaticity coordinate data for each pixel in the image Luminous Intensity and Luminance Spatial characteristics of a source such as a LED are defined by the luminous intensity and luminance, two terms that are sometimes confused with each other. Luminous intensity is a measure of light source output as a function of angle. Luminous intensity is a far field measurement; that is, it treats the source as a structureless point. Conversely, luminance is defined to be the luminous flux emitted per unit solid angle per unit area of a source. As such, it is a measurement of the spatial and angular distribution of source output, as opposed to just angular distribution. Luminance is measured in the near field; that is, at a distance from the source small enough so that any source structure is apparent. Luminous intensity can be extrapolated from a luminance measurement. In terms of LED applications, accurate luminance data is necessary for performing package design, or when optics are used in the near field, such as with automotive headlamp reflectors, flashlights reflectors and display backlights. Luminous intensity data is appropriate for most illumination and display (LED sign) applications, and to assess the far field performance of sources such as headlamps and tail lights. 3. TRADITIONAL SOURCE MEASUREMENT TECHNIQUES Various instruments for measuring the luminance, luminous intensity and color characteristics of LEDs have been available for many years. Unfortunately, none of these instruments provides the combinations of characteristics necessary to satisfy the needs of the majority of development and production applications, where both speed and detail of measurement are required. In particular, the most critical considerations in selecting an instrument for most measurements of LEDs are the ability to provide the necessary data, measurement speed, measurement accuracy and cost. Measurement speed is important for two reasons. First, because any instrument should be able to support throughput rates consistent with the rest of the manufacturing process, and therefore not be a bottleneck in the production line. The second is because LED output is often not temporally stable, especially when first powered on (warm up). As a result, a long data acquisition time can make an instrument report output variations that don t really exist. The most commonly used instruments for quantifying LED output characteristics are the integrating sphere, the goniophotometer and the source imaging goniometer (SIG). The characteristics of each are worth examining Integrating Sphere An integrating sphere is a hollow sphere whose inner surface has a diffuse, high reflectance coating. Source light is introduced through an entrance port, and multiple reflections from the inner surface of the sphere cause this input to become homogenized to an extremely high degree. Output is then is measured either with a photodetector or spectrometer at an exit port. Because all spatial information is lost with an integrating sphere, they are most useful for measuring total flux, integrated spectrum and integrated chromaticity coordinates. They cannot directly provide information on luminance, luminous intensity or color as a function of angle. The main advantages of the integrating sphere are rapid measurement speed and low cost. A typical application for an integrating sphere would be binning of LEDs for center wavelength Goniophotometer In a goniophotometer, a source is rotated relative to a detector, and measurements of source output are recorded as a function of angle. A simple photodetector is utilized for luminous intensity measurements or a spectroradiometer can be employed to obtain color data. Depending upon the needs of the application, the measurements may be performed Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 3

4 over a single arc or in two dimensions. Two configurations are both utilized, one in which the source is rotated and detector is held stationary, and the opposite setup where the detector is moved while the source is stationary. Depending upon the type of detection system used, goniophotometers are capable of obtaining high accuracy, far field measurements. A typical use of goniophotometer derived data is to model the source in optical design software for R&D applications. The major limitations of goniophotometers for LED measurement are relatively slow measurement speed, which forces waiting for the LED warm-up interval which may be 30 minutes to an hour, and the need to operate in a dark room to eliminate ambient light. In addition, the goniophotometer is generally slow taking many minutes to perform a full LED characterization because measurements are needed at many angles - and can yield incomplete data. Specifically, they do not generally provide comprehensive sampling of source output over all angles. Fig. 2. Schematic representation of a goniophotometer that allows LED output to be measured over 4π steradians Source Imaging Goniometer A SIG is similar to a goniophotometer in that the source is mounted on a two axis goniometer, enabling 360 of rotation in two axes. But the detector in a SIG is a CCD camera which acquires an image of the source at numerous angles, rather than simply measuring flux or spectral distribution. Furthermore, the SIG can also obtain colorimetric data by taking a series of images at each angle through CIE calibrated color filters. Thus, the SIG directly measures luminance and near field chromaticity. The primary application for the SIG is generation of detailed source models for use with optical and illumination design software. Because the SIG provides both high accuracy and comprehensive spatial coverage (up to 4π steradians), it can be considered the gold standard of reference to which measurements from other instruments are referenced. However, the slow measurement speed of the SIG, together with its cost and the large file sizes that it generates make it unsuitable for general production applications where only far-field data is needed. In addition, accurate measurement with a SIG requires that the LED warm-up period be accounted for and that measurements take place in a dark room to eliminate ambient light. 4. THE IMAGING SPHERE FOR LUMINOUS INTENSITY MEASUREMENT The Imaging Sphere was developed specifically to address the limitations of these other technologies and deliver rapid, accurate measurements of LED luminous intensity and color (as a function of angle). The Imaging Sphere comprises light tight enclosure which contains the hemispherical measurement chamber and camera, thus enabling its use in production environments having high levels of ambient light. System software provides the capabilities for sophisticated data analysis by development personnel, as well as a programming environment that enables the creation of simplified, menu driven, pass/fail testing routines for use by production line personnel. Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 4

5 4.1. Imaging Sphere Design and Operation The optical path of the Imaging Sphere includes a coated, diffuse, low reflectance hemisphere, a curved secondary mirror and a CCD-based imaging photometer or colorimeter essentially this acts as a reversed fish-eye lens to look at the LED from multiple angles at once. The hemisphere is attached to a flat, non-reflecting baseplate having a small aperture at its center. This entire assembly is contained in a light tight enclosure, meaning that ambient light is not an issue. In operation, the LED under test is positioned at the aperture into the hemispherical chamber. Light from the LED strikes the inner surface of the coated hemisphere, which is essentially a curved screen. The convex mirror at the base of the chamber enables the camera to image the entire inner surface of the hemisphere in a single exposure. This image contains all the information necessary to reconstruct the angular intensity profile of the illumination over nearly 2π steradians. The angular resolution of the instrument is determined by the camera s image sensor, and is typically <0.5. Fig. 3. A) The main functional optical elements of the Imaging Sphere. B) An LED under test illuminates virtually the entire inner surface of the Imaging Sphere. C) The mirror acts as a fisheye lens enabling the camera to view the entire inner surface of the Imaging Sphere and capture 2π steradians of source output in a single exposure. The other major component of the Imaging Sphere is a PC-based software application that controls all camera functions, acquires image data, and then factors in corrections based on various previously performed calibrations. The software may also be used to control other part handling equipment to more fully automate testing and to integrate the system with other production line equipment. Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 5

6 The Imaging Sphere system software also allows the user to view and analyze the image data in a variety of ways. A true color view of the output can be viewed directly as a radar plot, and derived parameters such as luminance, CIE coordinates, L*a*b* coordinates, and tristimulus values can be displayed as radar plots, cross-section graphs, and 2D or 3D Iso-Plots for the entire image or for specific areas of interest. Color information can also be plotted on CIE color charts. Fig. 4. A typical Imaging Sphere measurement of luminous intensity shown as a radar plot in false color Imaging Sphere Calibration While the Imaging Sphere is conceptually simple, using it to produce data having a high accuracy and signal-to-noise ratio requires calibrations to reduce the effects of electronic and optical errors. The most important of these calibrations are background noise subtraction, image distortion removal, and system flat field calibration. These calibrations are performed once, during manufacturing. They can be repeated, if needed, in the field this is most likely to be desired for R&D or calibrated measurement service applications Background Noise Removal Background noise subtraction is necessary because secondary reflections from any given point in the Imaging Sphere produce background noise in other parts of the image. Fortunately, it turns out that the diffuse nature of the Imaging Sphere inner coating causes secondary (and higher order) reflections to create a uniform level of background noise throughout the image that is entirely independent of the particular intensity pattern on the dome. Thus, the background noise in any given image is simply a constant component of the signal which can be subtracted out once its level has been determined. The Imaging Sphere contains a baffle that prevents source light from impinging directly on the curved mirror. This baffle creates a small shadowed area on the edge of the image, and the only light that reaches this region is background noise. Calibrating out background noise is therefore accomplished by measuring the average light level in this area, and then subtracting this from the overall image. Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 6

7 Distortion Removal The configuration of the Imaging Sphere gives rise to two types of geometrical distortion in its final image. The first occurs because the camera views the inner surface of the hemisphere off-axis. The second is due to the fact that the hemispherical shape of the Imaging Sphere is being mapped on to a flat, square or rectangular CCD. Distortion removal is accomplished by fitting each individual Imaging Sphere instrument with a specialized dome containing an array of small sources. The actual position of each of these sources is known with high precision. An image of this calibration dome is acquired by the system. The system software then determines the image transformations required in order to make the raw image data of the light sources correspond with their known locations. This specific transformation is then stored by the system software and used to process all subsequent images Flat Field Calibration Flat field calibration involves removing all color and luminance errors introduced by the system itself. Sources of such error might be non-uniformities in the Imaging Sphere diffuse coating, or inhomogeneities in the neutral density filters or color filters. Flat field calibration is performed by measuring a reference source whose characteristics are known to a high degree of accuracy. Then, the system produces a calibration file that matches the measurements provided by the Imaging Sphere with those obtained from the reference instrument. 5. ASSESSING IMAGING SPHERE ACCURACY In order to gauge the performance of the Imaging Sphere under real world conditions, the system was used to measure four different LEDs, and the results were compared with those derived from measurements of the same devices made with a SIG Experimental Setup The instruments used for this comparison were the Radiant Imaging SIG-300 and the Radiant Imaging IS-LI (Imaging Sphere for Luminous Intensity). Both the SIG-300 and IS-LI were outfitted with a Radiant Imaging PM-1603F-1 camera. This is an imaging colorimeter that utilizes a 512x512 full frame, scientific CCD sensor. The sensor is thermoelectrically cooled, and specified to deliver 16-bit dynamic range data. The PM-1603F-1 acquires color data by making a sequence of three exposures through filters that are well matched with the CIE x, y and z curves. Testing with the SIG-300 was performed in a dark room, and blackout curtains were used to eliminate ambient light. Measurements with the IS-LI were performed with ambient light present because the instrument itself is light tight. Four different LEDs were measured. These were a Luxeon model LXHL-LW5C, and Super Bright LEDs model numbers RL5-W10015, RL5-G5023, RL5-B2545. The Luxeon LXHL-LW5C LED is a 5 watt white light LED designed to provide high output levels for illumination applications such as flashlights, task lights, surgical lights and airport taxiway lights. It is rated to provide a total luminous flux of 120 lumens. The Super Bright LEDs RL5-W10015 is also a white light LED with a specified luminous intensity of 10 cd The Super Bright LEDs RL5-G5023 is a green LED (nominal peak wavelength 524 nm), with a specified luminous intensity of 5 cd. The Super Bright LEDs RL5-B2545 is a blue LED (nominal peak wavelength 472 nm), with a specified luminous intensity of 2.5 cd Method One measurement was acquired for each of the four LEDs on each of the instruments. Measurements on the SIG-300 were performed after each LED was sufficiently warmed up to reach a stable output intensity. This is necessary because obtaining a full set of source measurements with the SIG-300 can require over an hour. There was no device warm up period before measurements with the IS-LI. This is because the IS-LI captures all of its data in a single exposure. Therefore, even if there are fluctuations in output during the brief exposure period (a few seconds), they are Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 7

8 registered simultaneously over the entire image, and do not lead to spurious readings of spatial non-uniformity. In addition, the relative spatial distribution of LED luminous intensity output does not vary during the warm-up period. System software supplied with the IS-LI automatically applies all the necessary calibrations and converts raw image data into luminous intensity. The IS-LI software was then used to export this data into a format readable by other data analysis programs. This enabled further mathematical analysis and comparison with the results of the SIG-300 measurements. Similarly, near field data captured by the SIG-300 was extrapolated into the far field using that instrument s software (SIG 1.0), and the results were exported into Excel. For both instruments, data was exported in one degree increments in both inclination and azimuth angles. The SIG-300 itself has previously demonstrated high agreement with spectroradiometers, and is therefore considered to be the benchmark reference in all the analysis that follows Results Figure 5 gives provides a quick qualitative comparison between the data derived from the SIG-300 and IS-LI. Specifically, it shows a radar plot of luminous intensity (false colored) for the blue Super Bright LEDs RL5 B2545 obtained with each instrument. This LED was chosen because it has a substantial non-uniformity in output, making it easy to visually spot any similarities or differences between the two measurements. The region displayed in the two graphics is just the central potion of the output distribution (about 35 half angle) of the LED. Again, zooming in on this portion of the output distribution, where the nonuniformity is most pronounced, makes it easier to visually compare the results from the two instruments. Fig. 5. Image detail for the Super Bright LEDs RL5 B2545 (blue LED) showing matching of fine scale luminous intensity detail between the SIG-300 and Imaging Sphere measurements (false color display to show luminous intensity variations). Figure 6 shows vertical and horizontal cross sectional plots of normalized luminous intensity for the Luxeon LXHL-LW5C LED as measured by both of the instruments. The data is normalized because there is no way to ensure that the LED produced the exact same output during the two different measurements (due primarily to temperature and supply current variations). Clearly, there is a very good match between the readings. The FWHM (full-width halfmaximum) determined from the IS-LI data is 124, and from the SIG-300 it is Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 8

9 Normalized Intensity Normalized Intensity LXHL-LW5C IS vs SIG Vertical Cross Section LXHL-LW5C IS vs SIG Horizontal Cross Section IS SIG IS SIG Inclination (degrees) Inclination (degrees) Fig. 6. Vertical and horizontal cross sectional plots of normalized luminous intensity for the Luxeon LXHL-LW5C LED as measured by the IS-LI and SIG-300. Figure 7 plots the difference in luminous intensity readings between the IS-LI and SIG-300 as a percentage of the maximum value obtained at each point along arcs at four different azimuthal angles. In other words, if, at angle X, the readings from the two instruments were 0.80 and 0.78, this would be an error of 2.5%. If, at angle Y, the readings from the two instruments were 0.04 and 0.02, the error would be 50%. Again, the data show an excellent level of agreement. Fig. 7. Difference in luminous intensity readings for the Luxeon LXHL-LW5C between the IS-LI and SIG-300 as a percentage of the maximum value obtained at each point. Luminous intensity measurements of the Super Bright LEDs RL5-W10015 are presented in Figure 8. As before, these show an excellent match, which is further confirmed by a plot of the difference in the readings as a percentage of maximum value at each point. Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 9

10 Normalized Intens Normalized Intens RL5-W10015 IS vs SIG Azimuth 0 RL5-W10015 IS vs SIG Azimuth IS SIG 0.20 IS SIG Inclination (degrees) Inclination (degrees) Fig. 8. Vertical and horizontal cross sectional plots of normalized luminous intensity for the Super Bright LEDs RL5-W10015 as measured by the IS-LI and SIG-300. Fig. 9. Difference in luminous intensity readings for the Super Bright LEDs RL5-W10015 between the IS-LI and SIG-300 as a percentage of the maximum value obtained at each point. Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 10

11 Delta Delta Values of FWHM for the three Super Bright LEDs devices were derived from the luminous intensity curves. These values, together with the absolute difference (in degrees) are listed in the table. These values show good agreement. Table 2. Measured and nominal values for FWHM for three of the measured LEDs. LED SIG-300 FWHM (degrees) IS-LI FWHM (degrees) Difference (degrees) RL5-W RL5-G RL5-B Table 3. Measured values for peak emission location for three of the measured LEDs. LED SIG-300 Peak Location (degrees) IS-LI Peak Location (degrees) Peak Location Difference (degrees) Inclination Azimuth Inclination Azimuth RL5-W RL5-G RL5-B Measurements of the locations of the peaks in output intensity from the three Super Bright LEDs emitters were also compared. Values from the IS-LI and SIG-300 agree very well, here, and are essentially within the noise level created by the inability to position each LED in the instruments with perfect consistency with respect to fiducial marks. It should also be noted that the RL5-W10015 and RL5-G5023 both have relatively narrow output distributions, therefore any mechanical positioning error will affect the pointing number more dramatically, in terms of percentage. Finally, color measurements obtained with the IS-LI were compared with those from the SIG-300. Specifically, the difference in the two CIE color coordinates (CIE x and CIE y) measured by each instrument was plotted as a function of angle, over two azimuthal arcs. Again, a very high level of agreement was found. Chromaticity Difference Azimuth 0 Chromaticity Difference Azimuth Inclination (degrees) Cx Cy Inclination (degrees) Cx Cy Fig. 10. Difference s in chromaticity coordinates (CIE x and CIE y) for two different azimuthal arcs for the LXHL-LW5C as measured by the IS-LI and SIG-300. Delta = Cx(SIG-300)-Cx(IS-LI) or Cy(SIG-300)-Cy(IS-LI). Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 11

12 6. CONCLUSION The results of testing clearly show excellent agreement between the IS-LI measurements and the SIG-300 reference. The differences in IS-LI measurements of luminous intensity relative to the SIG-300 are often less than 1%. Similarly, the agreement between the two instruments when used to measure LED output peak location is on the order of 1 or less. Therefore, it can be concluded that the IS-LI can deliver accurate measurements of LED luminous intensity and color far-field distributions. The IS-LI is able to acquire a far-field measurement of LED source output in a matter of seconds. This enables the IS-LI to obtain accurate readings without a LED warm up period. Furthermore, the light tight IS-LI can be used with high levels of ambient light. These factors, together with its compact size, this makes the IS-LI easy to use in an R&D laboratory or to integrate into a production line environment. Thus, it is seen that Imaging Sphere technology provides a practical, robust means for assessing LED output characteristics that is well correlated with established reference instruments. Preprint of Radiant Imaging Presentation at Photonics West 2008 (22 January 2008) Page 12