Analog decoding of tail-biting convolutional codes on graphs

Abstract

A general method to develop Tanner graphs from tail-biting convolutional codes (CC) is proposed. Recursive systematic convolutional (RSC) and non-RSC codes are considered consistently and it is shown that the elimination of redundant states leads to a graph with low complexity. In addition the graphical representation is extended to derive the condition for which the tail-biting termination is valid. This analysis also leads to a unique graph applicable for decoding of both RSC and non-RSC codes. This graph is realized by exploiting the analog decoding scheme and MOS transistors. The circuit-level simulation is performed and the effect of important design parameters such as decoding latency, consumption and input dynamic range are considered.