Abstract
The dominant mode rejection (DMR) adaptive beamformer attenuates loud interferers by directing beampattern nulls toward signals contained in the dominant subspace [Abraham/Owsley, Proc. Oceans, 1990]. The dominant subspace is defined by the eigenvectors associated with the largest eigenvalues of the sample covariance matrix. DMR performance is primarily determined by how closely the eigenvectors of the sample covariance matrix match the true interferer directions. Random matrix theory (RMT) describes how the accuracy of the sample eigenvectors varies with the interferer-to-noise ratio (INR), array size, and the number of snapshots used to estimate the sample covariance. A simplified analytical model based on RMT predicts the mean DMR notch depth as a function of INR, array size, interferer location and the number of snapshots. This talk compares the RMT predictions with experimental results obtained by cancelling array strum in data from a vertical array deployed in the Philippine Sea. The array strum interference predominantly falls within the subspace spanned by the first two eigenvectors of the covariance matrix. Notch depth statistics obtained using a large set of receptions show good agreement between theory and experiment. [Work supported by ONR.]