Cost-effective LFSR synthesis for optimal pseudoexhaustive BIST test sets

Abstract

A large variety of methods have been proposed for the generation of pseudoexhaustive test sets for the built-in selftest of combinational circuits. The test pattern generator, in most all of the methods, is some form of linear feedback register (LFSR). In this work, we concentrate on a simple LFSR structure, known as LFSRISR, and show that particular orderings of the LFSR cells can significantly reduce the test set size. In addition, we show for the first time that an LFSRISR designed with a particular cell ordering and the allowance of a marginal number of additional cells guarantees pseudoexhaustive test sets of the minimum size 2", where w is the maximum input dependency limit of the circuit under test. Extensive experimentation on benchmark circuits and comparisons with the hardware overhead of other methods indicate the advantage of this approach.

New Test Pattern Generators for the BIST Pseudo-Exhaustive Testing based on Coding Theory Principles

“…The experimental results indicate the efficiency of the presented approach. The results also show that the insertion of additional segmentation cells required by the method outlined in [26] is not required.…”

“…The test set lengths designed for all combinational benchmark circuits are optimal (minimal) test set lengths (the degree of p(x), w, equals k) without requiring the insertion of segmentation cells. From the results in [26], additional segmentation cells are required to find an applicable primitive polynomial.…”

“…A simple LFSR/SR, which is not capable of generating the optimal test set length because of fixed assignment of residues, may become feasible through reassigning the residues. From the results in [26], additional segmentation cells are required to find an applicable primitive polynomial. This paper introduces an efficient approach to design a generator that generates a pseudo-exhaustive test pattern set.…”

“…Sometimes, the required search time to generate the minimal test set is large, or the search procedure requires the test set length of greater than the minimal test set. Dimitrios Kagaris and Spyros Tragoudas [26] have suggested a permuted LFSR/SR which is a simple LFSR/SR that drives a permuted set of inputs. A simple LFSR/SR, which is not capable of generating the optimal test set length because of fixed assignment of residues, may become feasible through reassigning the residues.…”

“…Examples are syndrome driver counters (SDCs) [19], constant-weight counters (CWCs) [20], condensed LFSRs [21], cyclic LFSRs [22], combined LFSR and XOR gates (LFSR/XORs) [23][24], combined LFSR and shift register (simple LFSR/SRs) [25], permuted LFSR/SRs [26], and convolved LFSR/SRs [15,[27][28]. It has been found the PETPGs based on the universal test set method require longer test lengths than the PETPGs based on the output-specific test set method [4,[27][28].…”