This dissertation was presented to the Faculty of the Graduate School of The University of Texas at Austin in partial fulfillment of the requirements for the degree of

Ph.D. in Electrical Engineering

Multicarrier modulation (MCM) techniques such as orthogonal frequency division multiplexing (OFDM) and discrete multi-tone (DMT) modulation are attractive for high-speed data communications due to the ease with which MCM can combat channel dispersion. With all the benefits MCM could give, DMT modulation has an extra ability to perform dynamic bit loading, which has the potential to exploit fully the available bandwidth in a slowly time-varying channel. In broadband wireline communications, DMT modulation is standardized for asymmetric digital subscribe line (ADSL) and very-high-bit-rate digital subscriber line (VDSL) modems. ADSL and VDSL standards are used by telephone companies to provide high speed data service to residences and offices.

In an ADSL receiver, an equalizer is required to compensate for the channel's dispersion in the time domain and the channel's distortion in the frequency domain of the transmitted waveform. This dissertation proposes design methods for linear equalizers to increase the bit rate of the connection. The methods are amenable to implementation on programmable fixed-point digital signal processors, which are employed in ADSL/VDSL transceivers.

A conventional ADSL equalizer consists of a time-domain equalizer, a fast Fourier transform, and a frequency domain equalizer. The time domain equalizer (TEQ) is a finite impulse response filter that when coupled with a discretized channel produces an equivalent channel whose impulse response is shorter than that of the discretized channel. This channel shortening is required by the ADSL standards. In this dissertation, I first propose a linear phase TEQ design that exploits symme- try in existing eigen-filter approaches such as minimum mean square error(MMSE), maximum shortening signal to noise ratio (MSSNR) and minimum intersymbol interference (Min-ISI) equalizers. TEQs with symmetric coefficients can reach the same performance as non-symmetric ones with much lower training complexity.

Second, I improve Min-ISI design. I reformulate the cost function to make long TEQs design feasible. I remove the dependency of transmission delay in order to reduce the complexity associated with delay optimization. The quantized weighting is introduced to further lower the complexity. I also propose an iterative optimization procedure of Min-ISI that completely avoids Cholesky decomposition hence is better suited for a fixed-point implementation.

Finally I propose a dual-path TEQ structure, which designs a standard single-FIR TEQ to achieve good bit rate over the entire transmission bandwidth, and designs another FIR TEQ to improve the bit rate over a subset of subcarriers. Dual-path TEQ can be viewed as a special case of a complex valued filter bank structure that delivers the best bit rate of existing DMT equalizers. However, dual-path TEQ provides a very good tradeoff between achievable bit rate vs. implementation complexity on a programmable digital signal processor.