A simple procedure is described for finding the direction to the source of an acoustic signal—primarily the human voice. The method, in effect, constructs the cross correlation of severely distorted signals from two microphones. The distortion, a combination of time‐domain processing and digital logic, is such to give strong preference to direct‐path early arrivals over reflections that follow, and to produce streams of pulses whose cross correlations can be computed very simply. The method has been implemented as a combination of custom hardware, and a simple microprocessor. The implementation operates in real time (with ∼0.2‐s delay) and produces outputs to indicate either the azimuth angle, or inter‐microphone delay. Possible uses include self‐aiming directional microphones, automatic aiming or switching of video cameras, and remote identification of talkers in nonvideo teleconferencing.