Phased array processing: direction of arrival estimation on reconfigurable hardware
Jasper D. Vrielink · 2009
A beamforming system consists of three different parts, the beamformer, the beamsteering, and the parameter estimation. In this thesis the parameter estimation is described, the Direction Of Arrival (DOA) estimation in particular. Two popular DOA estimation algorithms are described. The first algorithm is MUltiple SIgnal Classification (MUSIC), and the second algorithm is Estimation of Signal Parameters by Rotational Invariance Techniques (ESPRIT). Both algorithm can be used to estimate the DOAs of multiple signals. Covariance Matrix Differencing (CMD) is an extension to MUSIC to improve the performance of the MUSIC algorithm. This CMD extension is also described in this thesis. A model of MUSIC and a model of ESPRIT are made in Matlab to analyse the performance, and the effects of different test scenarios on the DOA estimation. Both algorithms are compared by means of these test scenarios. The performance of the CMD extension is also analyzed by means of a set of test scenarios. Based on the superior performance of the MUSIC algorithm when the Signal to Noise Ratio (SNR) is low, the MUSIC algorithm is chosen to be implemented on the reconfigurable architecture. The MUSIC algorithm is implemented on the Montium2 architecture. The implementation is described by means of pseudo code. Implementation aspects such as, accuracy, computational load, and scalability are analyzed. The complete implementation requires 1.5 million clock cycles on the Montium2. This number of clock cycles results in an execution time of 7.5ms. A practical example of a beamforming system used as a satellite television receiver, mounted on the roof of a car, shows that this is an acceptable execution time in this particular situation.