Answers are grounded in this paper and cite their sources.
Sparse arrays and low-resolution analog-to-digital converters (ADCs) reduce the cost of large arrays, but coarse quantization degrades direction-of-arrival (DOA) estimation. A 1.5-bit ADC quantizes each real and imaginary part into three levels, and the existing convex formulation for 1.5-bit data uses only the signs of the unquantized samples. In this paper, we require each unquantized value to lie in the quantization interval of its ternary output. Combined with atomic norm minimization on the uniform linear array that contains the sparse array, these constraints give the 1.5-bit atomic norm minimization (1.5B-ANM) estimator, which also recovers the signals at the missing sensor positions without normalization. Then, we propose 1.5B-ANM-D, which further penalizes the deviation of each unquantized value from its conditional mean, computed from the fraction of zero outputs. Both estimators are solved by semidefinite programming. In simulations, 1.5B-ANM-D outperforms covariance fitting on a nested array with few snapshots or a large threshold, whereas covariance fitting is more accurate on a coprime array. With sources at least one Rayleigh limit apart and 100 snapshots, both estimators locate 11 sources with six sensors on the nested array and 9 on the coprime array, with errors below half that limit in almost every run.
No takes yet. Share an insight, caveat, or question.
Cheng et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: