Phased array calibration and beamforming using signal processing

Doctoral thesis, 2007

Array antennas have a wide range of applications which includes ground and airborne radar, weather radar and anti-collision radar. In addition, equally important applications exist for communications, both on earth and to satellites in space, as well as for medical applications such as the detection and treatment of tumors. The modern trend in array design is to use digital beamforming and to integrate the antennas with the rest of the system more closely. This adds new possibilities, but also contributes to make the system expensive. To make full use of the new technology, this thesis presents a direction dependent signal processing array calibration method. The new method, called local calibration, also allows for larger array imperfections and, possibly, a lower overall array cost. In contrast to other non-parametric calibration methods, the proposed method can handle position errors and shows good performance also for sparse calibration grids, which is demonstrated by high resolution direction of arrival estimation. The local calibration has also been included in a convex beampattern optimization, which makes it possible to achieve a uniform sidelobe level also with unknown sensor position errors in the array. The embedded element pattern is included in the optimization by extracting the direction dependence of the received power from the calibration. Another application for the local calibration method is to reduce the number of necessary calibration measurements by calculating pseudo calibration data from a sparse grid of measured calibration data. Array signal processing is often performed assuming Uniform Linear Arrays (ULAs), although this model is far too simple to reveal the true behavior of array antennas. Throughout this thesis, more realistic models including mutual coupling and array imperfections are used. Specifically, the impact of mutual coupling and position errors on the adaptive beamforming performance in linear and circular arrays is studied. In the studied cases, the circular array is found to be more sensitive to neglecting the mutual coupling. Robust adaptive methods are commonly used to handle model uncertainties, such as unknown mutual coupling, but even better performance can be achieved if these methods are combined with a calibration.