The widespread use of LEDs in an increasing range of applications presents a growing need to standardize their characterization. Scott Jennato and Greg McKee of Labsphere discuss methods to measure the radiometric intensity and other properties of LEDs.

Experts in the field of light measurement sit on committees such as the International Commission on Illumination (CIE) and the Illumination Engineering Society of North America (IESNA) to establish standard methods or guides which encourage uniformity in the practice and reporting of lamp measurements. For general lighting the methods and equipment are well established, and the standards and measurement guides are periodically revised to remain current with the technology. However, recent growth in the use of LEDs, which is largely due to their efficiency, long life and the multiple colors available, has expanded the focus to include the optical and electrical characterization of such emitters. Radiometric measurement In an LED the energy of the photon is determined primarily by the energy bandgap, where recombination occurs. Since the eye is only sensitive to light with photon energies from 3.1 to 1.6 eV corresponding to wavelengths of 0.40 to 0.78 m compound semiconductor materials are used because they have direct bandgaps and the energy needed to produce photons in the visible spectrum. To convert the wavelength () measured in micrometers to photon energy (E) measured in electron-volts, the relationship = 1.24/E is used. shows some examples of semiconductors and their corresponding photon energies, wavelengths and relative response to the human eye. The characteristics of LEDs, including physical size, flux levels, spectrum, temperature sensitivity and spatial distribution, set them apart from typical light sources which are generally used and measured for their photometric and radiometric quantities. With an LED it is often difficult to achieve a high level of photometric or radiometric measurement accuracy due to uncertainties within measurement equipment and improper test set-up. Averaged LED intensity In an attempt to agree upon a procedure for comparing the intensity of LEDs, the CIE Committee TC 2-34 has published CIE Report 127, which presents the concept of an "averaged" LED intensity. This does not correspond to the physically precise definition of luminous intensity, but relates more to the measurement of illuminance for a fixed geometry. CIE s report recommends two measurement geometries: CIE Standard Condition A and CIE Standard Condition B, which are used to measure average intensities under "near-field" conditions. For averaged LED intensities determined under these conditions, the symbols ILEDA and ILEDB are recommended. Both conditions use a detector with a circular entrance aperture and an area 100 mm2 (corresponding to a diameter of 11.3 mm). The LED is positioned facing the detector aperture, and it is the distance between the front tip of the LED and the detector that constitutes the difference between condition A and B. The distances, solid angle subtended, and the plane angles are given in . The output of one LED is compared with another following the recommendation in CIE Report 127. It should be noted that a number of manufacturers use conditions A and B to specify intensity rather than average intensity. In some cases these values may be the same, but as this is not always the case, caution is advised. Reports for LED flux measurements are still in development. As with general lighting, the best method for measuring the total spectral flux of a lamp and associated colormetric parameters is to use an integrating sphere coupled with a spectrometer. The same applies to LEDs. Flux measurement The radiant flux energy of an LED can be measured using a calibrated spectroradiometer as a function of wavelength. This instrument separates or disperses polychromatic light into its constituent monochromatic components (usually by means of prisms or gratings). The photometric value may then be computed (often using special software) from this measured spectrum. Photometers, on the other hand, use a broadband detector in conjunction with an optical filter in an effort to simulate the spectral luminous efficiency curve of the human eye, commonly referred to as V(), which is referenced using CIE 15.2 Colorimetry Table 2.1. The process involves a change in the detector characteristics which is caused by the absorption of visible photons. The electrical signal generated by the detector is a response to the visible radiation incident on the active area of the detector. This basic difference between spectroradiometers and photometers is extremely important in LED metrology. A disadvantage of photometers is the difficulty in designing a filter that, when combined with a detector, fits the spectral luminous efficiency of the eye accurately. A mismatch is particularly prevalent in the blue portion of the spectrum as a result of the filter materials available. Although corrections can be applied, knowledge of the LED spectral distribution under measurement is required, and the corrections are usually only an approximation. shows the theoretical V() function, the relative spectral flux output of a blue LED, the relative spectral distribution of a typical tungsten incandescent lamp, and the typical response of a photopic detector. At a wavelength of 470 nm, a typical photopic detector with an f1 response of 4% can exhibit spectral mismatches as large as a factor of 2 between the V() function and the detector response. If an incandescent source or a source of similar spectral content is measured, the correction for the slope of the photopic detector to the V() curve is minimal, since the light is continuous and there is relatively little light in the blue portion of the spectrum in relation to the higher wavelengths. A mismatch in the response curves results in only a slight error of the measured photometric value. LEDs, however, have a completely different spectral power distribution, which tends to be a narrowband Gaus-sian distribution with a specific peak and FWHM of a couple of tens of nanometers. The relatively poor match of the photopic detector to the V() function can result in large deviations in the measured photometric quantities. This is particularly true for blue and red LEDs. Errors exceeding 100% are not unusual for blue LEDs (figure 2). Within the production environment, 20 000 or more LED die on a 2 inch wafer are commonplace. After the fabrication process, some or all of the LED die on the wafer are characterized for optical and electrical output (based on manufacturer specifications). During the measurement process, bad die are usually "inked out" and good die are "binned" and sold as die, or made into lamps. When it is time to sell the LED die, a brighter product translates into more money. Another variable observed closely during the characterization process is color. The eye is very sensitive to color and it is not unheard of for it to determine the differences in two LEDs based on a 23 nm spread in dominant wavelength, especially from blue to yellow. Since few end-users want their product to have a multitude of colors, accuracy in dominant wavelength is critical. Once its quality has been verified, the die is made into a lamp by attaching a lead- frame assembly, and is then fixed with an epoxy lens. After the LED lamp is constructed, most manufacturers characterize 100% of the LED lamps optically and electrically. Based on the sheer volume of LEDs manufactured on a daily basis, there is a great need for accurate high-speed radiometric measurements. A useful tool for this is an integrating sphere coupled to a CCD-based spectroradiometer attached to an autoprober or an in-line lamp tester. The integrating sphere The function of an integrating sphere in this application is to gather the light output from the die or lamp. Since the light is being spatially integrated, the sphere removes any concerns about accurate alignment of the light source to the aperture of the integrating sphere. The theory of the integrating sphere assumes that the interior surface is perfectly diffusing and has spatially uniform reflectance. The radiant exchange from diffuse surface to diffuse surface integrates the light, resulting in equal radiance at any point on the sphere wall. Theoretically, light received on any area of the sphere-wall surface is directly proportional to light introduced to the sphere. When a spectroradiometer is used to detect the indirect illuminance of the source, the user must consider the possible errors that may contribute to the measurement results, including but not limited to non-linearity, directional and positional effects, detector instability, reference and sample instability and noise. The size of the sphere and the spherewall coating reflectance affect the throughput. Higher reflectance provides better sphere responsivity, and the sphere is less subject to errors caused by the the differences in the angular intensity distributions of the LEDs. The smaller the integrating sphere the greater the throughput, but the smaller sphere is more sensitive to angular intensity distributions of the LEDs. As the diameter of the sphere increases, the throughput decreases and the sphere becomes more difficult to mount on an autoprober or other piece of manufacturing equipment. However, the sphere becomes increasingly less sensitive to angular intensity distributions. Three main factors help decide the optimum size of the sphere. These include the size of the LEDs, self-absorption of the LED sockets and LEDs (if applicable), and finally the sphere responsivity. The first two factors tend towards making the sphere as large as possible, while the responsivity limits the sphere size. To minimize absorption effects, the sphere may be calibrated with a spectral flux standard, with the light source assembly in the sphere. The standard lamp remains in the sphere throughout the testing process and self-absorption errors are essentially negated. Self-absorption, however, is not to be confused with near-field absorption. The system responsivity is a function of the sphere throughput and spectrograph and optical coupling of the two. Careful consideration in the design of and the use of high-speed LED test and measurement equipment is essential to achieve valid measurement results.