Pulse Statistics, Analysis of Room Acoustics

Abstract

Many sounds of speech and music more nearly resemble pulsed wave trains than continuous sounds abruptly terminated as used in classical reverberation measurement. It is therefore not surprising to find that two rooms can differ markedly in acoustical quality even if they appear identical under reverberation analysis which ignores details of short transients. This paper introduces a pulse statistics point of view which takes immediate account of the pulse-like nature of common sounds. Fundamentally, the method consists in examining the response of the room to a short pulse. The walls are replaced by an array of image sources (simple images if the walls are hard, or appropriately modified if there is absorption). These image arrays are then considered statistically. From this approach one can derive such classical quantities as reverberation time and mean free path. One can also analyze the detailed nature of discrete reflections including interference effects and thus obtain an average correlation between room geometry and the character of its pulse response. Idealized experiments in a hard-walled rectangular room are employed to illustrate the essential features of this approach. A point source emits an exponentially damped 3000-c.p.s. wave train of about 2-milliseconds duration. The received signals are recorded logarithmically on an oscillograph and the system is calibrated for quantitative results. Several dozen discrete reflections can be measured and correlated with calculation. The pulses merge into a more or less continuous background after a time which is calculated and confirmed experimentally. Detailed differences arise according to the positions of the source and microphone in the room. This approach appears useful for studying auditoriums by scale models and for investigating speech intelligibility in rooms.