On definitions of signal duration, evaluated on close-range airgun signals

In impact assessments for underwater noise, the duration of a transient signal is often expressed by the 90%-energy signal duration τ90 %. Consequently, the rms sound pressure is computed over this duration. Using a large set of measurements on marine-seismic airgun signals, it is shown that τ90 % is often very close to the interval between the primary and secondary pulse (the bubble period) or a small integer multiple thereof. In this situation τ90 % is a measure of the duration of the relative silence between primary and secondary peaks, which is not the intended measure. Rarely, τ90 % quantifies the duration of the main peak, leading to a much lower value of τ90 %. Since the number of peaks included in τ90 % is sensitive to the nature of the signal, relatively small differences in the signal lead to large differences in τ90 %, causing instability in any metric based on τ90 %, e.g., the rms sound pressure. Alternative metrics are proposed that do not exhibit these weaknesses. The consequences for the interpretation of sound pressure level of a transient signal, and the benefits of using a more stable metric than τ90 % are demonstrated.