Abstract
When performing subjective tests of an audio system, it is necessary to use appropriately selected programme material to excite that system. Programme material is often required to be wide-ranging and representative of commonly consumed audio, whilst having minimal selection bias. A random radio sampling procedure was investigated for its ability to produce such a stimulus set. Nine popular stations were sampled at six di↵erent times of day over a number of days to produce a 200-item pool. Musical and signal-based characteristics were examined; the items were found to span a wide range of genres and years, and physical similarities were found between items in the same genre. The proposed method is beneficial for collecting a wide and representative stimulus set.