Compressive sampling provides a new and interesting tool to optimise measurements of physical phenomena with a small number of sensors. The essential idea is that close to perfect reconstruction of the observed phenomenon may be possible when it can be described by a sparse set of basis functions. In this paper, we show how to apply compressive sampling techniques to the recording, analysis and synthesis of spatially extended sound fields. Numerical simulations demonstrate that our proposed method can dramatically improve the playback of spatialised sound fields as compared, for example, with High Order Ambisonics.