From cell phones to automotive dashboards, illuminated displays have become commonplace. Historically, they were designed with rudimentary software tools, and, consequently, many prototype iterations were required. But in the past decade, the explosive growth in the market has fueled the development of software packages specifically constructed for illumination system design. These programs rely on ray tracing. They propagate light rays through-out the system and take into account such issues as refraction, reflection and scattering at each interface and medium. Because conventional optical design programs also use ray tracing, there may be confusion between these two types of software. However, optical and illumination system design programs are not interchangeable. They use different approaches in the way they model systems, perform ray tracing and provide output. There are also important differences in the user interface for the two types of software, reflecting the different tasks required. Optical vs. illumination design In virtually all optical design programs, the object is modeled as a point source, or a series of point sources at positions on and off an optical axis. The program creates a predetermined set of rays -- a ray fan -- emanating from each of these points. In contrast, advanced illumination design programs model light sources as points, surfaces or volumes, or any combination of these elements. The spatial and angular distribution of a source's output, as well as its spectral characteristics, are specified. Moreover, structural elements of the source -- the glass envelope of a bulb or the heat sink in an LED, for example -- may also be included. The generation of a Monte Carlo ray set is based on this source model. This is a random, rather than predetermined, set of rays that emanates from throughout the source in accordance with its distribution characteristics. In typical optical system design software, nonoptical surfaces such as edges and bevels, and mechanical constructs such as mounts, usually are either ignored or only partially represented. Ray tracing involves propagating a ray from one optical surface to the next in sequence along the optical axis; when a ray misses a surface, is directed out of the system or hits a nonoptical surface, such as a component edge, it may be ignored or may generate an error. However, illumination design packages use models that are reminiscent of mechanical computer-aided design (CAD) software. Each component is represented as a complete, three-dimensional object, with sides and edges. Other elements, such as mechanical mounts, housings, mount- ing hardware and circuit boards, also can be included in the system. The optical properties of all these surfaces are specified. In some programs, surfaces may be defined as nonuniform, containing arrays of bumps, grooves or paint spots that may vary in density with position. Nor are there fixed limits on the number of inputs and outputs. For example, multiple light sources might be used, or a single source might illuminate several points in a display. The procedures and user interface employed to create this type of 3-D model are similar to those used in mechanical CAD software. Usually a toolbox of common shapes is provided, and Boolean operations (unions, intersections) enable the user to quickly create complex elements. A visual, interactive interface is important in this process because even many commonly encountered shapes are quite difficult to describe mathematically. Once it has been created, object orientation and position in 3-D space are controlled without reference to an optical axis. It is also important to note that ray tracing in illumination software is nonsequential; there is no predetermined order in which a ray is expected to encounter surfaces. Thus, a ray may exit the source, be reflected toward the source by a lens-retaining ring, then be reflected toward the first lens by the source envelope, and so on. Just about any interaction between light and any of the modeled elements is possible. With most conventional optical systems, specifying system performance using parameters such as spot diagrams, point spread functions, aberration curves and modulation transfer functions is straightforward. Commercial software packages generate and plot these functions in a readily accessible format, and high-performance programs provide design optimization algorithms. This is possible in large part because it's relatively simple to define what constitutes good performance for most imaging systems -- for example, minimum spot size or wavefront distortion -- and because there are a limited number of system parameters to be varied. Specifying and optimizing illumination system performance is not as easy. Illumination design programs typically provide angular and spatial distribution plots of light energy, usually calibrated in physical units such as lumens, lux and watts per square meter. The primary purpose of these plots is to deliver an accurate representation of what the human eye would see when viewing an actual system -- in other words, brightness or change in brightness as a function of viewing angle. This process must be interactive so that the designer can rapidly make changes to the system and see the results immediately. Software iterations are required because optimization is a more open-ended process in illumination system design. There are many reasons for this; for instance, performance metrics are typically application-specific, and many more system parameters may be varied to achieve the desired output. Currently, no commercial illumination design programs offer a general optimization capability, but this is an active area of research and the future looks promising. With these factors in mind, the best way to understand the usefulness and potential of illumination design software is to examine some typical applications. Applications Backlit liquid crystal displays (LCDs) are found in numerous consumer electronic devices, such as cell phones, personal digital assistants and laptop computers. LEDs light most smaller LCDs, while a larger display may employ cold-cathode fluorescent sources. Primary design objectives for LCD backlights are to couple the maximum amount of light into the specified viewing angle, to achieve the desired falloff in brightness as a function of viewing angle, to deliver uniform illumination over the entire display and, for full-color displays, to achieve uniform color. The latter can be a challenge with LED-based systems because full color is usually achieved by combining light from red, green and blue sources. Moreover, because LCDs use polarized light, it is helpful if the design software adequately models polarization effects. LED backlights typically employ light pipes to combine the output of multiple sources and to transform it into a uniform line. The output from a light pipe is directed into the edge of a diffuser that sits underneath the display and redirects light toward the user. However, the flux in the diffuser decreases with distance from the input light pipe, so the diffuser response must vary accordingly. To accomplish this, a matrix of paint spots with variable spacing is placed on the back of the diffuser. However, producing this paint spot pattern presents some production difficulties and doesn't deliver optimum efficiency. A better solution would be to mold a series of bumps directly into the diffuser. Until recently this approach was more difficult to model, but advances in illumination design software should make it much more attractive to manufacturers (Figure 1). Figure 1. In large LCDs, light from a cold-cathode fluorescent lamp (CFL) is introduced into the edge of the display. An array of bumps in a diffuser placed behind the display evenly channels illumination toward the viewer's eyes. Data projection Another key application is in digital data projectors. Over the past decade, these have steadily become smaller, brighter and less expensive, while increasing in output resolution. Improvements in illumination system design have been a key component in this evolution. The main design goals in projector illumination are to deliver high source-collection efficiency and high output uniformity in a compact package. These goals necessitate a different approach from traditional projection optics, such as overhead or movie projectors, which typically utilize a condenser system that trades off collection efficiency against uniformity. To achieve both, digital data projectors usually rely on a pair of lamps with either a mixing rod or lenslet arrays. A mixing rod is a long light pipe used to homogenize source output. Relay optics transfer its output to a spatial light modulator. Numerous mixing rod design forms have been investigated. The goal is to achieve maximum uniformity in minimum length. In terms of software capabilities, the ability to accurately model light pipe performance and investigate the effects of real-world tolerancing considerations, such as misalignments, is critical to meeting these goals. Today, the most commonly used solution is a rod with a rectangular cross section that matches the shape of the modulator. Usually, a length-to-width ratio of about 10:1 is required to achieve the necessary uniformity; shrinking this length and reducing overall system cost are the present focus of most design efforts. Lenslet arrays are rectangular arrangements of small lenses. These can provide a package-size advantage over light pipes, although they generally have a higher initial tooling cost (Figure 2). Figure 2. Pairs of lenslet arrays can be used to homogenize the source output in digital data projectors. This approach results in a compact package. Some projector designs utilize polarized light and dichroic filters. Software must therefore be able to model these components and must include effects such as the angular dependency of dichroic coatings. In the automotive world, backlit controls, gauges and displays are the prevailing standard. A major consideration is the total cost of the sources because a typical car interior has 50 or more light sources. Anything that can reduce this count has a tremendous cost impact for the manufacturer. A typical automotive illumination system feeds an incandescent or LED source into several light pipes. For example, a single source might light an entire row of buttons on a sound system or a climate control center (Figure 3). Sometimes the controls or indicators being lighted may have variable position; e.g., speedometer needles or volume control knobs. Here, total internal reflection, rather than paint spots or bumps, redirects the light. Because of the tremendous diversity of shapes and layouts encountered, there is no single design form for these tasks, complicating the designer's job. Furthermore, there may be significant manufacturing variations in the parts, such as surface smoothness in light pipes. Consequently, it is still necessary to construct prototypes, and a major goal of the design process is to minimize the number required. Figure 3. In these front and back views of automotive radio knobs, a single source evenly illuminates several graphics. All of this makes the design of automotive illumination systems a very interactive process. After creating an initial form, the designer must be able to rapidly alter and view the results to investigate various scenarios. It is also critical to model surrounding mechanical mounts in this process because they constrain light pipe geometry and can alter the system output. In summary, illumination design is now performed with software tools specifically developed for this task. These tools must provide high-quality visualization, interactive ray tracing to enable real-time design decisions and connectivity with other tools such as CAD programs. They must also help to shorten the overall design cycle. This software has already played a key role in the development of a number of consumer products. As illumination design software becomes even more powerful and easier to use, it should deliver more cost-effective solutions and find greater applicability. Meet the author Bruce Irving is manager of international distribution for Optical Research Associates in Westborough, Mass.