Array calibration and modeling of steering vectors

The use of structure in estimating covariance matrices has shown great promise for adaptive arrays when the data for estimating a covariance matrix is limited. When applied to adaptive beamforming or detection the result is a significant gain in effective signal-to-noise ratio when compared to unstructured estimates. In practice, the array steering vectors are not as predicted in the ideal models and the covariance is mis-modeled by the assumed structure. This mis-modeling can severely limit the performance in the presence of jamming or strong background clutter. The impact of the mis-modeling is examined and array manifold calibration is shown to reduce the impact while retaining the benefits of structured covariance estimation.