Thresholds for perception of ripple depth in the passband of a low-pass filter

Filters are often used in research related to speech and tone perception. Idealized filters with flat pass-bands, flat stop-bands, and infinitely steep rolloff are impossible to implement, thus practical filters often have a frequency response curve that has a certain degree of ripple, and always have finite rolloff. Despite this, there is a dearth of literature concerning the sensitivity of the ear to such artifacts. In the current study we examined the threshold of sensitivity to ripple depth in the pass-band of a digital low pass finite impulse response (FIR) filter. FIR filters are often used due to their lack of phase distortion, and due to the fact that their ripple depth can be closely controlled. Using an adaptive threshold detection paradigm, we found the threshold of noticeable ripple depth to be 2.7 dB (SD = 1.45) for filtered noise, and 4.89 dB (SD = 2.09) for a filtered chirp signal. This is significantly larger than the difference limen in intensity (DLI) for noise or pure tones. Notably, the threshold for filtered noise is significantly lower and more uniform across subjects, compared to the threshold for a filtered chirp signal. We conclude that these issues must be addressed when the necessity for filtering arises, though they cannot be considered independently of the signals themselves.