Complexity evaluation of sova based algorithms for decoding of block codes on a sectionalized trellis
Fabrice Labeau, M. Reza Soleymani · International Symposium on Antenna Technology and Applied Electromagnetics · 2002
Turbo decoders are composed of two or more soft-input soft-output (SISO) decoders that exchange reliability information. A classical way of implementing a SISO decoder for a linear block code is to use a trellis-based decoding algorithm, such as MAP, Max-Log-MAP or SOVA [1]. Recently, the idea of sectionalizing the trellis diagram representing a code has been proposed as a means to decrease the decoding complexity associated with the SISO decoders [2-3]. In this paper, we investigate the application of a nonbinary SOVA decoding algorithm to the sectionalized trellis of a binary linear block code. Considering an (n, k) block code C, a minimal bit-level trellis has n sections. A ν-section sectionalized trellis with section boundaries U = {0, h 1 ,…, h v = n} is obtained from the bit-level trellis by removing any stage at a location not in U, and connecting the remaining states by branches whose labels are the labels of the paths originally connecting the nodes in the bit-level trellis (Our notations borrow from [1], so the reader is referred to this book for further details). Optimal sectionalization boundaries can be chosen in order to minimize the complexity in terms of operations and/or memory of a given algorithm for a particular code (such optimal sectionalizations have been derived for MAP [3], Max-Log-MAP [3] and Viterbi [2] decoding).