Model-fitting approach to array calibration

The authors propose an approach to array calibration which can be used in a general correlated signal environment and requires little knowledge of calibration system geometry. A set of measured phases of signals which are radiated or reflected by point targets and received at the various antenna elements can be modeled as a Gaussian random vector. The positions of the angles and antenna elements can be linked to the measured phases. Optimum estimates of these positions can be obtained by applying maximum-likelihood (ML) estimation to the measured phases. A Newton ascent algorithm, terminated by a statistical stopping criterion, is developed for implementing the ML estimation of the array geometry. A tolerance theory based on this criterion is used to evaluate the accuracy of the calibrated array geometry and provide array design guidelines. An experiment is included to verify the proposed approach.<>