The Secondary Surveillance Radar (SSR) is a transponder system used in air traffic control. Due to growing traffic densities, it is increasingly likely that a ground station receives a mixture of responses of various aircraft, partly overlapping in frequency and time. Currently such "collisions" are disregarded, at a loss of system performance and reliability. In this article, we propose to equip the ground station with an antenna array, and investigate techniques to blindly separate such a mixture based on source waveform properties. At base-band, a received SSR signal consists of a binary sequence with alphabet ¼ ½ , modulated by a complex exponential due to the residual carrier frequency. We present three algebraic algorithms to compute the separating beamformers by taking into account the particular modulation format of the received signal. The Cramer-Rao Bound is derived, extensive simulations are presented, and an experimental platform has been built to collect measurement data and demonstrate the algorithms.