High Speed Decoding of Serial Concatenated Codes

This paper proposes a code design method for a particular class of serial concatenated codes named Systematic with Serially Concatenated Parity (S-SCP) codes. Using a par- ticular family of S-SCP codes based on quasi cyclic code, we propose a code design taking into account both implementation and performance requirements. This design enables a particular scheduling of the decoding process enabling the reduction of the number of parallelized processes required to meet a target throughput. I. INTRODUCTION Turbo-like codes including Low-Density Parity Check (LDPC) and concatenated convolutional codes, have raised a significant interest in the past few year. Hybrid concatenated codes like Hybrid Concatenated Convolutional (HCC) codes, which were first introduced in (1), represent a new class of code construction. A particularly interesting family of code construction from the concatenation of three codes is described in (2) and is labelled Systematic with Serially Concatenated Parity (S-SCP) codes. In this paper we propose to define a sub class of S-SCP codes based on quasi cyclic constituent codes which have interesting properties from the implementation point of view. Scheduling strategies for decoding are then in- vestigated, which aim to optimize throughput while accounting for implementation requirements, such as memory conflicts and pipeline processing. Finally we illustrate the advantages of this code construction through a code design example. II. S-SCP CODES A. Definition S-SCP codes can be viewed either as serially concatenated convolutional codes or structured LDPC codes. This S-SCP structure intends to combine advantages of both Turbo and LDPC codes. The S-SCP structure is built from the concatenation of an outer code, an interleaver and an Inner Parity Generator (IPG) as depicted on Figure 1. The inner parity generator is obtained