| Direction of arrival (DOA) estimation for transmitted sources is useful in many applications including radar and sonar systems, law enforcement, and traffic management. A number of highly accurate methods exist for solving this problem, unfortunately they all require precise knowledge of the characteristics of the receiving antenna array. These characteristics, which include sensor locations, gain and phase response, and mutual coupling, are rarely perfectly known in real situations. Calibration methods are required to estimate the true array characteristics before accurate DOA estimation can be performed.; Self-calibration algorithms estimate both source DOAs and perturbed array response vector parameters simultaneously. Calibration errors are usually assumed to be small and a first order approximation to the perturbed array response vector is often used to simplify the estimation procedure. This dissertation investigates the performance of these algorithms and shows how they break down in the presence of moderate to severe errors. A novel algorithm is proposed which eliminates the small error assumption. Simulation results demonstrate the improved performance of the new technique. The performance of the new technique is tested as a function of source signal DOA separation, array size, number of snapshots, signal to noise ratio, and sensor location perturbation. The technique is extended to the cases of gain/phase errors, simultaneous gain/phase and sensor location errors, disjoint sources, and a mixture of simultaneous and disjoint sources. Additionally, a new Cramer-Rao bound on self-calibration with unknown source covariance is developed. |
