Aiming at the requirement of real-time signal detection in the passive surveillance system, a wideband array signal detection algorithm is proposed based on the concept of power focusing. By making use of the phase difference of the signal received by a uniform linear array, the algorithm makes the power of the received signal focused in the Direction Of Arrival (DOA) with improved cascade FFT. Subsequently, the probability density function of the output noise at each angle is derived. Furthermore, a Constant False Alarm Rate (CFAR) test statistic and the corresponding detection threshold are constructed. The theoretical probability of detection is also derived for different false alarm rate and Signal-to-Noise Ratio (SNR). The proposed algorithm is computationally efficient, and the detection process is independent of the prior information. Meanwhile, the results can act as the initial value for other algorithms with higher precision. Simulation results show that the proposed algorithm achieves good performance for weak signal detection.