LAMP: Controllability, Observability, and Maintenance Engineering Technique (COMET)

Recent advances in integrated-circuit technology offer the circuit and system designers many opportunities to explore new, lowcost, high-performance design techniques. The increased operational speed and the logic complexity of many medium-scale-integration (MSI) and large-scale-integration (LSI) designs, however, also present acute problems in factory testing and field maintenance. For 1505 factory testing, it becomes increasingly difficult to diagnose faults in an LSI package, partly owing to equipment packaging constraints and partly to inadequate fault isolation technique (s) for nonclassical and/or multiple faults. In field maintenance, while the isolation of faults to a component or chip level is unimportant, the problem of quickly, and automatically, detecting and recovering from faults is further compounded by increases in circuit size and complexity. In addition, the cost of using fault-simulation techniques to generate data for the trouble-location manual1 ( T L M ) may become economically prohibitive, especially for large systems. Finally, the problem of accurately updating trouble-location data whenever circuit or design changes occur remains important but unresolved. Many designers of fault-tolerant computer systems have studied these problems.2 Some have proposed design approaches with built-in automatic-fault-detection hardware.2,3 Others have explored diagnosable design concepts purely from a structural standpoint based on graph-theoretical techniques.4'5 Unfortunately, the search for practical methods of generating (and updating) T L M data for large systems has been largely unsuccessful.