Sensitivity analysis of high-resolution array processing algorithms

Summary form only given. The sensitivity of the maximum-likelihood and MUSIC (multiple signal classification) algorithms, which are used to estimate the directions of arrival of signals from multiple sources, has been quantified. A first-order sensitivity analysis of these algorithms was performed to determine the effect of modeling errors (i.e. system errors, which cause differences between the true and assumed array manifolds), assuming a perfectly known covariance matrix (the infinite-data case). The effect of phase and gain errors in the array elements or in the receivers and the effect of errors in the element locations were studied for two closely spaced sources. Both uncorrelated and correlated sources were considered. Formulas were developed for the error introduced into the direction-of-arrival (DOA) estimates produced by these algorithms, when the modeling errors are small, as a function of the source locations and other system parameters. These formulas have been validated by computer simulations.<<ETX>>