Surface micromachining generally offers more design freedom than related technologies, and it is the technology of choice for most microelectromechanical applications that require multi-level structures. However, the design flexibility that surface micromachining offers is not without limitations. In addition to determining how to fabricate devices in a planar world, the designer also needs to consider issues such as film quality, thickness, residual stress, topography propagation, stringers, processing limitations, and concerns about surface adhesion [1]. Only a few years ago, these were the types of issues that limited design complexity. As the technology improved, the number of mechanical layers available to the designer became the dominant constraint on system functionality. In response, we developed a 5-level polysilicon fabrication technology [2] that offers an unprecedented level of microelectromechanical complexity with simultaneous increases in system yield and robustness. This paper outlines the application that was the driving force behind this work and describes the first devices specifically designed for and fabricated in this technology. The 5-level fabrication technology developed to support this program is known as SUMMiT-V. Four mechanical layers of polysilicon referred to as polyl, poly2, poly3, and poly4 are fabricated above a polyO electrical interconnect and ground plane layer [2,4]. PolyOmore » is 0.3 pm thick, polyl is 1.0 pm, poly 2 is 1.5 pm, and both poly3 and poly4 are 2.25 pm. All films except polyl and poly2 are separated by 2-pm thick depositions of sacrificial oxide. A 0.5-m sacrificial oxide between polyl and poly2 typically defines the clearance between close mating parts such as hubs and hinges. This entire stack is built on a single crystal substrate with a dielectric foundation of 0.8 pm of nitride over 0.63 m of oxide. Seventeen drawing layer are combined to generate the 14 photolithographic masks used to pattern these films during a 240-step fabrication sequence. Mirror Operation To become operational, both mirrors must be driven up to a 45 degree angle. In this position, optical energy entering through an opening in the substrate beneath one mirror [5] is redirected to the second mirror, then down through another substrate opening and onto the target receiver. Each mirror is actuated through a chain of gears driven by a mirror control engine. This chain incorporates a series of gear reduction units that significantly increase drive torque and positional resolution. Also included in this chain are two gears that are not coupled to each other (see figure 3). This prevents the mirror control engine from driving the rack that actuates the mirror. To complete the drive train, two additional gears must be inserted between the interrupted gear pair [4]. The coupling gears that perform this function are shown in figure 4. Both of these gears are fabricated on a plate that moves towards the interrupted pair of gears as the discrimination sequence The plate onto which the coupling gears are fabricated is attached to the left end of the maze rack, so it moves as the rack moves. If the wrong path is taken at any of the 24 decision points in the maze, the coupling gears will not move far enough to complete the mirror gear chain, and the mirror can never be operated. Thus, this is a single attempt device with more than 16 million possible code sequences.« less