Abstract
As the internet gets more and more popular, the demands for high speed transmission network become more and more urgent. Asymmetric digital subscriber lines (ADSL) technology is a new transmission technique designed to provide several megabits per second downstream and hundreds of kilobits per second upstream through telephone lines (copper wires). As the highly dispersive property of telephone lines, one form of multicarrier modulation techniques, called discrete multitone (DMT) modulation, has been selected by the American National Standards Institute (ANSI) and the European Telecommunication Standards Institute (ETSI) for ADSL to combat the severely distorting channel. In a DMT system, every symbol consisting of N samples is cyclic prefixed by its last v samples to eliminate inter-symbol interference (ISI), where v is the length of channel impulse response. The addition of cyclic prefix would cause a data rate loss of the ratio v/(N+v) because of the long extension of channel impulse response, which is often the case in telephone lines. In order to reduce the data rate loss, a time domain equalizer is placed in the receiver front end to shorten the channel impulse response. Many optimal solutions to find the equalizer coefficients based on different approaches are published. Those optimal solutions often need complicated matrix operations and rely on accurate channel estimation. J. Chow, J. Cioffi, and J. Bingham proposed a solution that employs frequency domain least mean squares (FLMS) algorithm to train the equalizer. Although it has a large performance loss compared with those optimal solutions, it is still famous for its feasibility on real-time implementation. In this thesis, we propose a low-complexity algorithm that uses time domain least mean squares (TLMS) to train the equalizer. Simulations show that our algorithm outperforms that developed by J. Chow et al. for a few dB in high SNR case and performs fairly in low SNR case. On the other hand, our algorithm remains the advantages of the frequency domain algorithm such as processing in the absence of channel estimation and low hardware complexity, which make it suitable for real-time processing.