A fast architecture for finding maximum (or minimum) values in a set

High speed architectures for extracting the best (maximum or minimum) values in a given set and their positions is of high importance in many signal processing applications. For example, the search of the two minimum values is an important part in iterative channel decoders for turbo and low-density-parity-check codes. In this paper, a new architecture is proposed to find the g^th best value in an unsorted list of k elements, where the ranking position g can be any integer between 1 and k. The architecture can also be used to find in the assigned set a generic subset of the largest or smallest values. The proposed solution is particularly efficient in terms of hardware complexity and latency when the cardinality of the assigned set k is large and the values are represented on a reduced number of bits n. Synthesis results obtained with a 90-nm CMOS standard cell technology are provided for specific choices of g, k and n. Moreover, the nice properties of regularity and scalability of the proposed architecture are exploited to develop a QCA (quantum cellular automata) based implementation, which achieves lower power consumption or higher speed.