Abstract
Almost all communication systems need to deal with the distortion problems due to the inter-symbol interference (ISI). This phenomenon occurs when a sequence of signals is transmitted through a bandlimited channel. Adjacent signal pulses may disperse and overlap with one another in time. To address this issue, proper filters are necessary at the transmitter end and/or the receiver end, requiring special techniques based on so-called Nyquist’s pulse-shaping criterion. Among them, the raised-cosine filter is one of the most widely used. In addition, equalization is another technique to combat ISI. Various algorithms for equalization have been proposed in the literature. In this thesis, Minimum Mean Square Error (MMSE) equalizer is adopted. We propose a cost-effective architecture to realize such an equalizer. The most difficult part of realizing an MMSE type of equalizer lies in the computation of the inverse of an auto-correlation matrix. Direct computation often consumes too much hardware. The contribution of this thesis is mainly a new architecture to implement the iterative process of Gauss-Seidel algorithm for inverse matrix computation. In this architecture, resource sharing is done in a way that the hardware utilization is dramatically improved, the target perform is achieved, while the design size is significantly reduced by 63% as compared to a naïve version. In summary, we use only five complex multipliers in this design to perform various tasks required in an equalizer such as the channel estimation, and the computation of auto-correlation matrix and the equalizer’s adaptive coefficients.