RESOLUTION IMPROVEMENT AND LEAST SQUARE NULL STEERING IN NARROWBAND AND BROADBAND ARRAYS

Abstract

Adaptive array processing systems which respond to an unknown interference environment are currently of considerable interest. The development of adaptive arrays had taken off with the advancement in digital computing techniques. As a result, adaptive arrays can be more easily implemented, and are more versatile and reliable. Adaptive arrays have also acquired entirely new processing possibilities, extending its applicability into many fields including radar, sonar, communications, astronomy, seismology and ultrasonics. In this dissertation, two methods for improving the performance of adaptive arrays are presented and investigated. In Part I of the dissertation, a new technique to reduce the variances of the source-tracking response zeros of narrowband adaptive arrays in the presence of random amplitude and phase errors is investigated. The zero variances are reduced by forming several spatially shifted power inversion arrays from the original system, averaging the weight vectors of these arrays, and determining the zeros of the spatially averaged weight vector. Equations for the optimal averaging weights and the ensuing variance performance are derived and verified by simulation results. It is found that the new technique may reduce the zero variance substantially when compared with taking the entire system as a single array. In Part II of the dissertation, a new algorithm to track and reject an unknown broadband source by approximating the inter-element delay in a 2- element adaptive array is investigated. Essentially, by constraining the response of the L-tap processing filter to correspond to that of a delay in the least squares sense over the frequency band of interest, the algorithm reduces the number of adaptive parameters in the broadband system to merely an unknown gain and delay which can then be adjusted to track the unknown source by using the LMS algorithm. Compared with using the conventional LMS algorithm, the new algorithm has the advantage that, at about the same complexity, estimate for the source location is available and the convergence speed is significantly improved.