Stabilisation of adaptive array patterns by signal space projection

Adaptive array (and sidelobe canceller) patterns are found to fluctuate considerably from one set of data to another, when there are degrees of freedom to spare, unless a very considerable amount of data integration is performed. A highly effective cure for this effect is to determine a basis for the signal space for the system and project the normally obtained optimum weight vector into the signal space. The resultant vector gives patterns which are virtually identical on successive data blocks. The basis need not be of eigenvectors and can be obtained by quite simple processing of the data.<
>