Toward the implementation of analog LDPC decoders for long codewords

Error control codes are used in virtually every digital communication system. Traditionally, decoders have been implemented digitally. Analog decoders have been recently shown to have the potential to outperform digital decoders in terms of area and power/speed ratio. Analog designers have attempted to fully understand and exploit this potential for large decoders. However, large codes are generally still implemented with digital circuits. Nevertheless, in this thesis a number of aspects of analog decoder implementation are investigated with the hope of enabling the design of large analog decoders. In this thesis, we study and modify analog circuits used in a decoding algorithm known as the sum-product algorithm for implementation in a CMOS 90 nm technology. We apply a current-mode approach at the input nodes of these circuits and show through simulations that the power/speed ratio will be improved. Interested in studying the dynamics of decoders, we model an LDPC code in MATLAB's Simulink. We then apply the linearization technique on the modeled LDPC code in order to linearize the decoder about an initial state as its solution point. Challenges associated with decoder linearization are discussed. We also design and implement a chip comprised of the sum-product circuits with different configurations and sizes in order to study the effect of mismatch on the accuracy of the outputs. Unfortunately, testing of the chip fails as a result of errors in either the packaging process or fabrication