Desenvolvimento e implementação de chips dedicados para um novo decodificador de códigos corretores de erros baseado em conjuntos de informação

Abstract

Error-correcting codes are present in almost all modern data communications and data storage systems. Errors during these operations are practically inevitable because of noise and interference in communication channels and degradation of storage media. When topperformance is required, the corresponding algorithms (encoder and decoder) are implemented in hardware. The research project presented in this dissertation, a dedicated chip for a new family of decoders based on information sets, is part of a broad project targeting the development of a new decoder capable of achieving near maximum likelihood decoding (MLD) performance, however with a much simpler hardware, thus demonstrating that the use of this technique (decoding based on information sets), previously prohibitive due to the complexity of the hardware, could now be feasible. Aiming to simplify the hardware, the first step was to modify the original Dorsch algorithm to reduce the number of clock cycles needed to decode a message. The main modifications performed were in the Gauss Jordan elimination procedure and in the number of candidate codewords, which was highly reduced with respect to original Dorsch algorithm. This modified algorithm was first implemented using a hardware description language and evaluated in different FPGA families, where the viability was demonstrated. The algorithm was later implemented in a dedicated chip (ASIC) using CMOS technology in order to complete the demonstration of the feasibility of their implementation, and effective use.