Abstract
Auditory brainstem responses (ABRs) to monaural 70 dB nHL clicks in normal adults were recorded with wide bandwidth and stored digitally. Using digital to analogue conversion with expanded time, the ABRs were input to a variable band-pass filter. The order of averaging and filtering was unimportant. Low-pass effects were unremarkable. High-pass filtering in the commonly used range 20 Hz to 500 Hz produced severe waveform distortion, causing emergence of artifactual peaks, amplitude increase and decrease, absolute latency decrease and interpeak interval changes. Digital high-pass filtering gave identical results. The distortion increased with the filter order. Using a zero phase characteristic drastically reduced ABR distortion and abolished latency changes except for poorly resolved peaks. Butterworth and boxcar filter moduli gave indentical results. At common high-pass frequencies, filter phase distortion dominated the mean square error in ABR estimation. Off-line digital filtering of ABR is both practicable and desirable, especially for facilitating inter-laboratory pooling of data.