Adaptive array nulling algorithms for pulsed communication systems
D.J. Farina · 1993
The application of adaptive arrays to communications systems is addressed for the specific case of a pulsed desired signal with known on and off transition timing. The application allows one to estimate a covariance matrix when the desired signal is on that is distinct from a second matrix estimated when this signal is off. Adaptive antenna weights that attempt to maximize antenna gain on the desired signal and null out the interferers are determined from these two covariance matrices. Accurate knowledge of the antenna array manifold (response to incoming signals) is not required. The thesis extends upon the Burst Acquisition (BA) Algorithm proposed by B. Agee for this problem. A Normalized Covariance Subtraction Thermal Noise Correction (NCS-TNC) Algorithm is proposed that has enhanced performance against a single non-stationary interferer. The normalization and thermal noise correction could also be applied directly to the BA Algorithm. The thesis also provides a new Eigen Vector Correlation (EVC) Algorithm which is designed to handle multiple interferers with independently time-varying power levels. The BA Algorithm is shown not to be able to handle scenarios with this type of interference. The EVC Algorithm performance is characterized as related to the scenario characteristics (number of interferers, positions, powers, pulsing) and the estimation of the covariance matrices. In addition, a method to determine the desired signal transitions is suggested. The EVC Algorithm is shown to reject both continuous and most pulsed interferers. The exception is a pulsed interferer that has the same period as the desired signal, has a equal or smaller duty cycle, and is on only when the desired signal is on. Like many adaptive algorithms, the EVC Algorithm is less successful when applied to scenarios with signals closely spaced in terms of beamwidths, especially when many signals exist. A drawback of the EVC Algorithm is that it is computationally intensive, compared to the BA Algorithm. Computer simulations are used to illustrate the performance of the algorithms. A discussion of computational aspects of all of the algorithms and options is also presented.