A Dolby Digital Primer (part 3)
In modern film sound technique, dialogue is normally mixed at a level between -25 to -31dBFS Leq. In the dubbing theatre and cinema the gain will be set so this gives an acoustic level of 85dB Leq. This allows peaks to achieve a level of up to 116dB SPL, which most people would regard as being sufficiently loud. This works fine in the cinema, but not so well at home. Audio levels in signals delivered to the home via whatever medium are set according to the peaks in the signal not the average level, or the level of the dialogue. This results in a situation where a viewer watches a movie with the sound comfortably loud and the dialogue clearly audible. The viewer then switches to a rock music program and suddenly he is blasted out of his chair into the hallway - the level is now excessively loud. The reason for this is that if the dialogue in the movie is mixed at -31dBFS Leq, the volume control will have to be turned up quite high to hear it clearly. Peaks in the soundtrack will be loud, but only occasionally. The rock music programme however will be mixed at a level more like -8dBFS and is therefore considerably louder. The result, as you know, is that you can’t watch TV without constantly adjusting the volume control.
Dolby Digital has an answer to this. Whatever the signal, the full dynamic range is encoded. But since AC-3 has a 24-bit range, then the signal can be normalized by whoever oversees the encoding process so that whatever the nature of the audio, it comes out at an acceptable level for the viewer. A Dolby Digital encoder, such as the DP569, includes a control word, dialnorm, in the bitstream to set the signal, whatever its nature, to an appropriate listening level. A film could have a dialnorm value of -31, which corresponds to no reduction; a news report could have a dialnorm of -14, giving a 17dB reduction so that the newsreader’s voice is the same level as the dialogue in the movie.
Dynamic Range Compression is controlled by two words in the Dolby Digital bitstream - dynrng and compr. Remember that this does not change the audio data, it changes the way it is replayed in the decoder. Dolby Laboratories are of course masters at manipulating dynamic range, and there is more to Dynamic Range Compression than a simple logarithmic compressor. The essence of Dynamic Range Compression is that there is a null zone around the average signal level where there is no compression and the audio passes through at unity gain. Signals that exceed the highest level of the null zone are reduced in level, signals below the null zone are raised. The result is that a lot of the time nothing happens, but when the signal would be too loud or too quiet for its intended application, it will be reduced in level or boosted as appropriate. If the source material already has a reduced dynamic range, then it will live mostly in the null zone so it will not be compressed any further. Dolby Digital encoders such as the DP569 offer several compression 'profiles’ for specific applications. For film soundtracks, for example, dialogue is often the quietest component and if levels that were quieter than the dialogue were brought up, then camera and other types of noise might be revealed. In film mode therefore, most compression is done above the null zone. For speech however, a 5:1 compression ratio with up to 15dB of boost is used below the null zone and a 20:1 ratio, almost limiting, is used above it. Selection of mode is surprisingly easy, but subject to interpretation by the person doing the encoding depending on the source material. A good domestic decoder will allow the sound track to be played back without compression, or dynrng (the terminology in a domestic decoder can vary) can be applied for 'easier’ listening.
A special condition exists when Dolby Digital material is replayed through the RF (aerial) input of a TV set. Dolby Digital has the ability to reproduce a wide dynamic range, but in this situation it must reduce that range to match analogue sources such as live transmissions or VHS soundtracks. Accordingly, an 11dB boost is given to the signal. Obviously, boosting the signal by such an amount would in itself create an over-level situation frequently, and the reduction of the 5.1 channels to two or one increases that probability. Therefore compr can be enabled in the encoding process to protect the signal against RF overmodulation.