Analysis and Ranking of the Acoustic Disturbance Potential of ...

Report No'. 6945
BBN Systems and Technologies Corporation
where CR represents the critical ratio in terms of signal level relative to
spectrum level noise. This gives the bandwidth in Hz of the band of masking
noise that contains power equal to that of the signal tone (see Scharf 1970).
Johnson ( 1968b) , Terhune ( 198 1 ) and others have used equation (2) to calculate
masking bandwidths in Hz for marine mammals based on the assumption that signal
power equals masking power. Figure 2.29 shows the results of such calculations,
expressed as a percentage of the center frequency of the masking band.
Based on the available critical ratio data and the equal power assumption,
masking bands often appear to be on the order of 1/6th to 1/3rd of an octave in
width, i.e., bandwidth equa1.s 11.6 to 23.2% of the center frequency (Fig.
2.29). If one of these "rules of the thumbt1 were spictly true, the critical
ratios at several frequencies would be as follows:
100 Hz 1 10 kHz 100 kHz
1/3 octave 13.6 dB 23.6 d~ 33.6 dB 43.6 dB
1/6 octave 10.6 20.6 30.6 $0.6
As evident from Figure 2.29, the critical ratios at low frequencies (human, cat)
exceed those expected if the masking bandwidth is 1/3 octave. In contrast,
critical ratios for marine mammals listening at most higher frequencies are
somewhat lower than those expected if the masking bandwidth were 1/3 octave, or
even 1/6 octave, particularly if one ignores the harp and ringed seal data that
have been questioned by Moore and Schusterman ( 1987).
When attempting to calculate the radius of audibility of marine mammal calls
or industrial noise in the presence of background noise, several workers have
assumed that masking bands are 1/3 octave wide (e.g., Payne and Webb 1971; Gales
1982; Miles et al. 1987). Gales (1982) also considered the possibility that, at
frequencies below 450 Hz, the masking bandwidth exceeds 11'3 octave. As evident
from Fig. 2.29, masking bandwidth may indeed exceed 1/3 octave at low frequencies
if marine mammals listening in water are similar to terrestrial mammals listening
in air. If so, noise power in the masking band will be higher than calculated
from the 1/3 octave assumption, and the radius of audibility of low frequency
sound would be less than that calculated. Conversely, for higher frequencies
where the masking bandwidth seems to be less than 1/3 octave based on critical
ratio data for marine mammals, the radius of audibility could be somewhat greater
than calculated assuming a masking bandwidth of 1/3 octave. All of these
estimates depend on the validity of the equal power assumption, i.e., that a
narrowband sound signal is masked when total noise power in the masking band
equals or exceeds the power of the signal.
The equal-power assumption may not accurately represent the width of the
masking band (Scharf 1970; Kryter 1985). Other methods, measure the masking band
directly by manipulating the bandwidth of sounds masking a signal. The term
"critical bandf1 is used for direct empirical measures of the masking band (Scharf
1970). In humans, the critical band in Hz is about 2.5 times wider than the
critical ratio equal-power band at the same center frequency. This means that
humans can detect a signal whose level is somewhat less than the band level of