Co-array Music under Angle-Independent Nonidealities

Abstract

The difference co-array is crucial in determining the number of resolvable sources in direction-of-arrival (DoA) estimation. This virtual array of pairwise sensor position differences enables sparse arrays to identify vastly more sources than sensors. However, the idealized assumptions giving rise to the co-array, such as isolated omnidirectional gain patterns, may not hold in practice. Consequently, the applicability of the co-array model to real-world arrays needs to be investigated thoroughly. In this work, we consider a general class of angle-independent departures from the ideal model caused by nonideal sensors or compression of the array measurements. We study the impact of these nonidealities on DoA estimation using co-array MUSIC, assuming that the array is calibrated and that an infinite number of snapshots is available. We establish that proper use of the calibration data enables unbiased DoA estimation of more sources than sensors. Nonidealities may nevertheless cause subspace swap at low SNR.