In this experiment to determine critical bandwidth, the masked threshold of tone by two narrow-band noises on either side of the tone was measured. The noises were of equal spectrum level and were symmetrically separated on a linear frequency scale with respect to tone. When the separation of the two noises was incresed, masked threshold of the tone first remained constant, but dropped after a critical frequency separation was reached. In our experiment, we used the narrow-band noises without bandwidth on the peak and with steep slopes (200dB/oct). Each masker noise had the spectrum level from-10dB to 50dB SPL. Tones at 250, 500, 800, 2000, 4000 and 6000Hz were masked by two noises and each tonal duration varied from 3ms to 350ms.
The results were as follows:
In this method, the critical bandwidths showed a clear continuing decrease as the signal frequency was decreased below 6, 000Hz than those estimated by Zwicker et al. (1957), Scharf (1970) and, Zwicker and Terhardt (1980). Although the bandwdth values were generally somewhat smaller than those estimated by Moore and Glasberg (1983), the way with unequalized TDH39 earphone that they varied with signal frequency was similar to that described by Moore and Glasberg. If the critical bandwidth is related to the equivalent rectangular bandwidth (ERB) function of the auditory filter proposed by Moore and Glasberg, this may indicate a need to revise the classical critical band function.
The width of the critical band in the same signal frequency was independent of sound pressure from 20dB to 80dB SPL. We also found that the masked threshold of short duration signal was independent of critical bandwidth. Moreover, masked threshold at short delays indicated no threshold elevation or overshoot into critical bandwidth of a 15ms, 2000Hz signal by two narrow-band noises.
View full abstract