Architecture and implementation for a high reliability long-term mission space computer

Typical spacecraft computer performance requirements are discussed, along with the potential architectural implementation options. A comparison of basic redundancy schemes and their merits in terms of long-term reliability, error detection, error correction, power, and weight is presented. The basic approach to obtaining high long-term (several years and more) reliability requires intelligent allocation and use of redundant assets. A fully cross-strapped module level redundancy implementation is selected as the overall optimum approach for a contemporary spacecraft application. This architecture provided the best performance in terms of long-term reliability given the slate of technology at the time the basic development was done in the early 1980s. The redundant assets (modules) should be unpowered since unpowered circuits demonstrate considerably lower failure rates than their powered counterparts. It is shown that an optimum exists for the number of internally cross-strapped modules needed to achieve the stated long-term mission reliability. Various aspects of the implementation are discussed, including component selection, radiation hardness, memory sparing, memory error detection/correction, internal cross-strapping, and fault detection