An Approach for Controlled Random Tests with a Given Hamming Distance Generation

This paper addresses the challenges of testing computing systems and their hardware components, especially memory devices. It highlights the limitations of traditional random testing. Such methods often fail to use available information about the system under test and previously generated test patterns. The potential of controlled random testing, which incorporates knowledge of prior patterns, is therefore explored. A class of controlled random tests with a limited number of test patterns is identified and analyzed, including existing standard approaches. The paper introduces a novel measure of dissimilarity between test patterns. This measure is based on calculating Hamming distances for binary patterns after mapping them into different numeral systems, including quaternary, octal, and hexadecimal. We propose a method for generating controlled random tests with a guaranteed minimum Hamming distance. It is based on representing binary patterns as symbols from non-binary numeral systems. In this way, ensuring a specific Hamming distance in the symbolic domain also guarantees at least the same distance in the binary representation. We evaluate the effectiveness of the proposed method through simulations, particularly in the context of memory testing and the detection of multicell faults, i.e., errors caused by interactions between multiple memory cells. This approach can enhance the efficiency and reliability of test procedures in embedded systems, memory diagnostics, and safety-critical applications.