2510332 - February 2009
The integrator function spatially redistributes the image of the arc from a highly peaked distribution to a more uniform, flat-topped distribution, resulting in relatively flat spatial distribution of light on the screen (see Figure 7). As lamps become more etendue matched to the devices, they become more spatially nonuniform at the focus of the reflector, so some kind of integration technique always is used on DLP products today. Although frequently used in LCD products, lens-array type integrators (fly-eye) are not a good choice for DLP projectors except in cases where absolute minimal path length is required. A rod integrator, solid or hollow, is the best choice for reasons described in the following paragraphs. Types and characteristics of various integration techniques are:
Lens arrays. Lens arrays typically are two molded lens-array plates spaced a certain distance apart. Typically, they are used with a parabolic reflector in near-collimated space to facilitate design. This is convenient for three-panel LCD or DLP products that do not require a focus through a sequential color wheel, but not for single-panel DLP products. Typically, lens arrays are less efficient than a rod-type integrator for several reasons, most having to do with manufacturing techniques. Most are molded glass, requiring small drafts between lens elements in order to release from the mold. This area represents lost light. This loss is repeated for every lens element on every plate. Then, the array on the first plate must accurately align with the array on the second plate; any misalignment causes further losses. Then, each image formed by the lens pair from each plate must accurately image to the device array. It is not possible to align each image formed by each pair independently because they are molded together into plates, the image must be large enough that any tolerances in position of the device array relative to the lens array (and vice-versa) will be accounted for. All of these tolerance buildups result in larger losses relative to a single rod image. And finally, the quality of collimation of the incoming light beam determines the amount of crosstalk between lens pairs. When a skew ray from one lens enters the adjacent lens (crosstalk) instead of the one directly paired with it, this ray is lost to overfill.
Solid-glass-rod integrators. These are commonly used, but increasingly are being replaced by hollow mirror-integrator tunnels. They both work on the same principle of creating reflections inside the rod to spatially randomize the input to a more uniform output, without changing the numerical aperture. The solid glass rod does this by total internal reflection off the glass/air sides. The number of reflections inside the rod is a function of the index of the glass, the numerical aperture of the input, and the cross section and length of the rod. It is somewhat challenging to mount an unclad glass rod in a system, because every point of contact with the wall creates TIR failure where light exits the rod, causing throughput loss as well as a concentrated thermal load at the point of contact. Also, the rod entrance and exit faces should be AR coated without spilling over to the sides, which can be a relatively expensive process. Since the output face also is in focus on the device (and therefore screen), any imperfection or dust particle that settles on this face will be imaged to the screen and appear as image defects. Also, the number of reflections per unit length of rod is lower in glass than in air due to the index of the glass, requiring a longer length of rod for the same amount of integration. Finally, safety bevels on the faces usually are required to avoid chipping, but increase overfill losses in proportion to their size.
Theoretically, TIR is more efficient than a mirror reflection, so, for relatively long rod lengths and/or high-powered systems, glass (or fused silica or quartz) is still the preferred choice due to efficiency and thermal effects. Exit-face dust protection can be provided by abutting
May not be reproduced without permission from Texas Instruments Copyright 2009 Texas Instruments Incorporated