The authors present a method for fully testing chips designed using synthesis and silicon compilation. The method is targeted for a multiprocessor architecture that implements low-speed to medium-speed signal-processing algorithms. By taking advantage of the specific properties of the architecture, the method allows a chip to be partitioned into several functional units. The authors use the C-test concept instead of the traditional automatic test-pattern generation to derive a compact set of test vectors. The fault model covers both the stuck-at class and part of the transistor stuck-open and stuck-closed cases. For large units with embedded memory, the authors adopt a self-test approach.
展开▼