Sparse Linear Array Synthesis Using Exponential Analysis

This paper presents an exponential analysis technique for synthesizing a sparse non-uniform linear array using equidistant samples of the array factor of a dense uniform linear array. The problem statement is explored in a realistic production noise setting (i.e., we model uncertainties/tolerances during production as Gaussian noise), which motivates some slight oversampling. The collected samples are organized in the form of a Hankel matrix. A Cadzow iteration provides an accurate lower-rank approximation of the Hankel matrix. This lower-rank Hankel matrix is then used to obtain an equivalent sparse reduced array that accurately approximates the performance of the original dense array. Numerical experiments demonstrate that the newly developed method is more robust and accurate compared to the methods previously reported.