 |
Audio settingsIn order to properly audition the examples and the content of interactive exercises in lessons 11 and 12, it is necessary that the audio system of your computer is of sufficiently high quality and is properly calibrated. Neglecting this will most certainly result in misleading results and false conclusions. Please, make sure that following requirements are met: Audio hardware requirementsThe computer system should be equipped with a high quality audio card or external audio interface capable of reproducing a stereo signal in the frequency range of 20Hz-20kHz, with flat (+/- 1dB) frequency response, at least 16-bit resolution, true signal-to-noise ratio of at least 90dB, nonlinear distortion below 0.5%. The audio interface should be able to operate with sampling frequencies of 44.1kHz (preferably without internal resampling to 48kHz), 32kHz, 22.05kHz and 16kHz. Gamers' and cheap multimedia sound cards are not recommended. Also sound adapters built into PC mainboards and general purpose audio outputs in laptop microcomputers are usually inappropriate. You may verify the quality of your sound card using a freeware Rightmark Audio Analyzer. The audio interface should be connected to a pair of high quality headphones or monitor speakers with a reasonably flat frequency response. Small plastic multimedia desktop speakers or those built into laptop computers are pretty bad due to high nonlinear distortions and very limited frequency range. Walkman type headphones are also usually inappropriate. The following test audio files are recommended to verify the proper operation of your sound card and monitors. The sine wave with different sampling frequencies should be reproduced with consistent and high sound level and no signs of audible distortion. The stereo white noise file consists of four short samples that should appear on the left, right, middle and outside (due to phase inversion) of the stereo field, with equal color and loudness. The resolution test contains a speech sample buried in white noise. You should be able to hear and understand the spoken sentence despite the noise. The linearity test contains two very high frequency tones with slightly different and varying frequencies. These should be barely audible. In the case of nonlinearities in your audio system a weak difference "siren" tone will be heard as a product of demodulation (listen carefully). Software requirementsIn order to save on download time, most audio examples have been compressed using the popular MP3 technique. A great care has been paid to appropriate compression settings to ensure no audible artifacts would be added by this compression. Examples of coding-related impairments have been also re-compressed to MP3 with high bit rate, and thoroughly tested against possible error accumulation. In order to playback these files properly, the operating system should be configured for flawless support of *.wav and *.mp3 files, both offline and online (streaming, embedded into web pages). We do not recommend using third-party audio players. The native Windows Media Player offers good compatibility with all formats used in this course. It is a good idea to setup the network buffering option to at least 5sec (to access this setting in WMP, select Tools, Options, Performance, and change the Network Buffering from "default" to a chosen value). The following examples should play back without stutter, drop-outs and other artifacts. All examples should also sound identical. Audio calibrationBefore attempting to perform any masking experiments, it is very important to calibrate the output level of the audio interface and monitors/headphones. The calibration file linked below is a sequence of 2.2kHz tone bursts with alternating level. The sequence is repeated 4 times. Listen carefully to these soft tones and adjust the output level of your sound card (or the amplification system) so that you hear only the louder tone from the sequence. The softer tone should fall below your hearing threshold, while the stronger tone should be barely audible. After the proper adjustment, the level of the louder tone will reflect approximately 0dB. From now on, the sound level should not be changed.
|