This paper presents a new phase-alignment algorithm for distributed beamforming [1].¹ This scheme is more energy-efficient than existing solutions, since it requires each node to transmit only once, but it converges at least as fast as all other previously proposed algorithms. In contrast with all other studies of distributed beamforming, where noise had been neglected, here it is explicitly included in the signal model and shown to have a considerable impact on synchronization that cannot be ignored. With and without noise, analytical expressions that characterize the performance of the new algorithm are provided, with emphasis on various limiting regimes of interest.