In this paper, the presence detection of long-code direct sequence spread spectrum signals (DS-SS) in low signal to noise ratio (SNR) using the sensor array are considered. A sparse method, which is fast and computationally efficient, is proposed for the presence detection of these signals. The proposed method is based on the phase linearity of the cross channel terms of the wideband spatial covariance (R) matrices. In addition, a new sample R matrix estimation technique, which is sparse and called as block of consecutive frequency bins (BCFB), is proposed. The proposed BCFB technique provides a fast signal detection which is the desirable for the real time applications. It is shown in various simulations that the proposed detection method can detect long-code DS-SS signals in very low SNR.