This work deals with the problem of detecting deterministic sinusoids in white noise. The procedure, due to kumaresan and Tufts (T-K), is briefly described as a minimum norm exact-linear-prediction design. This method has proved a better performance than currently reported methods in the same field, mainly for very short data records or small number of sensors for angle of arrival estimation in sonar o radar arrays. The algorithm is tested against maximum entropy and phased arrays in order to stablish its performance as concerns with statistical stability, tracking and ROC characteristics.