A joint transmitter/receiver optimization problem for multiuser communications in a symmetric crosstalk environment based on the minimum mean squared error (MMSE) criterion is addressed. It is assumed that each user employs an identical transmit pulse shaping filter and sees identical direct and crosstalk channel responses. All the channels are assumed to be bandlimited with identical support that, in general, extends beyond the basic Nyquist zone [-(1/2T), 1/2T], corresponding to the common symbol period T. Necessary conditions for the optimum transmitter and receiver are derived and several useful insights generated. In particular, the performance advantage obtained by exploitation of the cyclostationary nature of the interference is demonstrated, corroborating earlier results. An algorithmic procedure to solve for the jointly optimal transmit-receive pair is outlined and used for numerical evaluation of the system performance as a function of several parameters of interest such as system bandwidth, the number of users and the number of points at which the transmitter is nonzero in the set (f: f/sub 0/+k/T) where T is the symbol period, and f/sub 0/ /spl isin/[-(1/2T),1/2T].

Future wireless systems will have to support multimedia services such as voice, data and fax. Issues in providing multimedia services on wireless include multiple access, bandwidth rationing and power control. We address the latter two issues. CDMA is assumed to be the access method. Different services have different quality (QoS) requirements, maximum power and/or minimum rate constraints. In order to achieve the required QoS they can alter their power and/or rate of transmission. Different rates are represented by varying the processing gain. Since users interfere with each other, achieving each user's QoS requirement is completely coupled with the powers. We formulate this as a constrained optimization problem. The objective functions used are minimum power and maximum rates. The former reduces interference seen by other cells due to the current cell, the latter attempts to achieve the best possible throughput for users in the current cell. We develop bounds on the total number of users of each class that can be supported simultaneously while maintaining adequate QoS and meeting resource constraints.

Two dimensional, separable signal models are proposed for modeling the effects of intersymbol and inter-track interference in high-density magnetic storage media. The optimum (minimum mean-squared error) linear and decision feedback equalizers based on these models are developed. The use of a two dimensional LMS algorithm for adaptive updating of the equalizer coefficients is discussed and analytical results presented. Finally, simulation evidence is provided to predict the performance improvements achievable based on these models vis-a-vis one dimensional equalization.>