In this paper we consider the problem of direction of arrival estimation from a set of beamformed measurements corrupted by correlated Gaussian noise. We investigate the performance of sparse signal processing (SSP) on correlated and pre-whitened data, focusing on accuracy and false alarm rate. We considered both a fixed-grid and a gridless method, and performed several simulations. From our analysis we conclude that SSP applied to tapered and beamformed data without prewhitening is robust, but results in a lower accuracy than in the pre-whitened case. Furthermore, by comparing the two methods, we also observe that, for the scenario considered, the fixed-grid method performs as good as the gridless one.