Differences in loudness levels between programs and channels have long annoyed viewers of analog TV services. Viewers might expect that digital broadcasts would not have such a problem. But, the more ample dynamic range available in digital broadcast is making the problem worse.
The Dolby LM100 broadcast loudness meter uses an electronic method called “Dialogue Intelligence” to measure the subjective loudness of broadcast programming.
Studies of a European satellite platform have suggested that loudness differences between channels can be 15dB or more. Another study of programming being ingested for playout found that a single genre of content had average loudness values that varied by up to 17dB.
One contributing factor to these differences is the fact that broadcasters usually use peak-program meters (PPM) or volume-unit (VU) meters to measure and normalize material. But neither of these account for the listener’s perception of how loud he or she perceives the material at home. These measurements indicate absolute signal levels. The listener’s perception is based on the frequency response of the ear and the average level of the signal, not the peaks and periods of near silence.
Currently, broadcasters use PPM or VU meters to align the program peaks to the same level. For program material that has a narrow dynamic range (small average-to-peak ratio), this does lead to a rough form of loudness normalization. Aligning the peak level also roughly aligns the average level. But the larger the dynamic range of the material, the worse the results of a peak-alignment technique. Broadcasters have tried many different methods over the years to calculate how loud a human will assess a signal to be. Early analog solutions used filters to mimic the frequency response of human ears. But broadcasters did not adopt these methods because they were expensive, and PPMs and VUs still do not use any frequency response curve.
More recent techniques of varying complexity have been proposed, based on models of the human ear. But these techniques tend to be computationally expensive and unrealistic for real-time use in a broadcast environment.
In studies, Dolby found that the variation in measurement accuracy, when compared with the variations between individual listeners — between the simpler and the more complex measurement systems — is actually quite minimal. Therefore, the internationally standardized LeqA measurement, which produces accurate results in real-time with acceptable processing costs, is the ideal choice for the basis of a loudness algorithm.
The importance of dialogue
Mathematically, the LeqA measurement is a long-term average assessment of a signal, using the A-weighting frequency response curve based on measurements of the response characteristics of the human ear at conversational levels. The human ear is most sensitive to the frequencies used in dialogue. But simply filtering the frequency content of a signal to the response curve of the human ear is an over-generalization of how people assess loudness.
For example, when watching television, a viewer will adjust the volume so that dialogue is clearly audible at a conversational level. But he or she is unlikely to adjust the volume upwards during a quiet section of a program where there is no dialogue, and unlikely to adjust it downwards following a short impulsive noise. In other words, the listener understands the program dynamics and is able to separate this from his or her assessment of loudness.
Dolby has developed an electronic method called “Dialogue Intelligence” that locates the same dialogue portions of a signal as a human listener and uses them to control when the loudness measure is applied. The company employs this method in its LM100 broadcast loudness meter. The unit measures the subjective loudness of broadcast programming and provides a numerical measurement that broadcasters can use to adjust different content or channels to similar levels, on either analog or digital TV services.
John Couling is the professional product manager for Dolby Laboratories.