Reduced-Dimension Decision Feedback Equalizers for High-Speed Wireless Digital Communications

Purdue University, West Lafayette IN

Investigators

Abstract

ABSTRACT The primary problem addressed is the deleterious effect of dynamic multipath propagation on high-speed wireless digital communications supporting real-time video and multimedia applications. Digital TV (DTV) is a target application along with Fourth Generation Cellular Communication Systems. Reflections off buildings and other structures cause ghosts and other artifacts in the reconstructed video stream. These multipath effects can wreak havoc on all functions of the receiver causing loss of synchronization and loss of carrier/timing recovery as well as inter-symbol interference. Discrete ghosts are often associated with multipath conditions of large, flat surfaces, such as tall steel buildings. However, multipath conditions can occur in areas with hills and dense foliage where a signal can travel through many paths to the receiver. When the signal is bounced back from a moving object, complex ghosts can arise with a time-varying nature that is very challenging for the equalizer to overcome. To accommodate delay spreads of up to 40 microseconds, current generation Digital TV receiver chips employ a Decision Feedback Equalizer (DFE) with more than 100 Feed-Forward taps and more than 400 Feed-Back taps. This constitutes a large number of taps to adapt implying slow convergence and high computational complexity. Future high-speed wireless multimedia applications will place even more stressful demands on equalizer performance. The investigator will develop fundamental advances in the area of reduced-dimension Decision Feedback Equalizers (DFE's). These include methods for dimensionality reduction based on a powerful reduced-rank adaptive filtering algorithm referred to as the Multi-Stage Nested Wiener Filter (MSNWF). The MSNWF provides convergence speeds better than the Recursive Least Squares (RLS) algorithm, but at a much reduced computational complexity due to the fact that that the MSNWF constrains the overall weight vector to lie in a data-adaptive, low-dimensional subspace. The MSNWF has been successfully employed to effect a reduced-rank, linear equalizer for applications in Third Generation Cellular Communications based on both CDMA Technology and the European GSM EDGE System. The investigator will also develop other means for dimensionality reduction including the use of the channel shortening concept employed in OFDM systems to reduce the length of the cyclic prefix, and symbol waveform compression achieved through simple linear filtering at the receiver. In addition, the investigator will study the use of reduced-dimension DFE's for equalization when block space-time coding is employed under frequency selective multipath conditions.