A butterfly structure for rate k/n convolutional codes

Abstract

In this paper, we propose a butterfly structure for Viterbi decoder which works for convolutional codes of all rates R = k/n. It provides an efficient way to find the inherent symmetry of trellis branches. By exploiting the symmetry, only a part of branch metric need to be computed and the others can be derived from the computed branches. Consequently, the computational complexity of Viterbi decoder can be reduced significantly with no error performance loss. In the best case, the butterfly structure can reduce the branch metric computation by a factor of 4