Dolby NR is based on reducing the dynamic range of the sound during recording and expanding it during playback developed by Dolby Laboratories. In this it is similar to other analogue noise reduction systems, however it is the most widely used. The process is known as companding.
Several types of Dolby NR were developed, including A, B, C and S. Most widely used is the B type, which allows for acceptable playback on devices without noise reduction; most pre-recorded cassettess available on the market use Dolby B. There is also a system called HX Pro, which operates on a different principle.
Dolby B (and C which is similar) are a form of dynamic preemphasis. The background hiss of a tape white noise is unnoticeable if it is masked by a stronger audio signal, especially at higher frequencies. This is called psychoacoustic masking. When the tape is recorded, the amplitude of the signal in the higher frequency registers is used to determine how much pre-emphasis to apply - a lower level signal is boosted by about 10dB (Dolby B) or 20dB (Dolby C). As the signal rises in amplitude, less and less pre-emphasis is applied until at the "Dolby level" (+3 VU), no signal modification is performed. On playback, the opposite process is applied (deemphasis), based on the signal level. Thus as the signal level drops, the higher frequencies are progressively more strongly filtered, which also filters the constant background noise level. The two processes cancel out as far as the signal is concerned, so it is reproduced faithfully, but only one process (the de-emphasis) is applied to the noise, which is thereby reduced.
The calibration of the recording and playback circuitry is important for faithful cancellation of the complementary processes, and is easily upset by poor quality tapes, dirty playback heads or using incorrect bias levels. This usually manifests itself as muffled-sounding playback, or "breathing" of the noise level as the signal varies.
HX or "Headroom eXtension" is a method for further increasing the dynamic range of a cassette tape. Because tape is magnetic, it is inherently non-linear in nature, due to the hysteresis of the magnetic particles. If an analogue signal were recorded directly onto magnetic tape, it would be reproduced extremely distorted, due to this non-linearity. To overcome this, a high frequency signal is mixed in with the recorded signal, which "pushes" the envelope of the signal into the linear portion of the transfer characteristic of the tape. The high frequency signal, called bias, is inaudible on playback, as its wavelength is too short to be captured by the magnetic medium. An audio signal that contains a large amount of high frequency components has a tendency to partially act as a bias signal in itself, so the addition of a fixed bias can push the envelope a little too much for these signals, into the second non-linear region where the tape becomes saturated. Dolby HX modulates the bias level to allow for the self-biasing effect of high frequencies, thus preventing the signal from entering the saturation region. The effect is to reduce distortion of loud high-frequency signals, which can sound very harsh otherwise.
How Dolby B works
Dolby HXPro
External link