Non-linear test pattern generators for built-in self-test

Abstract

Various linear finite state machines have been widely used as pseudo-random test pattern generators for the built-in self-test of integrated circuits. These generators are inexpensive to use as they have the advantage of low hardware overhead. However, the Geffe generator, a classic type of non-linear finite state machine, has not been frequently studied or used in similar applications. It is known that a Geffe generator, when used as a pattern generator for digital system testing gives improved fault detection. Unfortunately, the area overhead involved is sufficiently high, thus, such a generator becomes impractical for built-in self-test. To solve such problems, we introduce two possible new designs of the Geffe generator. These new designs are based on the generator's original architecture, so they preserve the non-linear structure. Our optimal goal is to achieve a very sharp reduction of the area overhead, and maintain a satisfactory fault detection capability. These two new designs are exercised in the fault simulations of benchmark sequential circuits. The experimental results demonstrate both of our designs can lead to fault coverage, which significantly exceeds the fault coverage of linear machines, and is comparable to the original Geffe generator.