This paper reviews and compares several techniques for multichannel sound panning over loudspeakers, as well as their application to computationally efficient 3-D sound spatialization over headphones. Ambisonic and pairwise amplitude (or intensity) panning approaches are reviewed within a general encoding/decoding framework. The performance of the different techniques is evaluated by use of objective criteria of auditory localization, and in terms of reconstruction of interaural time differences and HRTF spectral cues at the listener's ears. Improved approaches to efficient binaural synthesis are discussed, whereby interaural time differences are explicitly controlled. Specifically, the performance of the ``binaural B format" encoding/decoding scheme is evaluated, and practical applications of this technique are reviewed.