Ariane
13
t
hE
t
EChnICaL
and
thE
P
hILOSOPhICaL
B
aCkGrOUnd
OF
t
hE
a
rIanE
d
IGItaL
a
UdIO
L
EvELLEr
(Or, Why Automatic Level Control Doesn’t Have to be Processing’s Weak Link!)
What is Levelling?
Levelling is the automatic process of reducing the long-term dynamic variations in level of an audio source. Also
called Automatic Gain Control or, AGC.
Why Use Levelling?
Levelling is required for multiple purposes, but the most important is to force a given program’s level variations
to best conform to the requirements of human hearing of the consumer in their expected listening environment.
Average levelling control is done on the basis of the audio signal’s average level over time, while peak levelling
control is based upon the highest instantaneous value. While peak amplitude value can be determined by simple
measurement of the audio’s absolute value at any given moment in time, average control can only be derived by
analysis over a specified period of time.
Levelling for effect
Levelling can be for artistic purposes, or simply as a control utility. These two separate purposes are distinction
are not usually made as most processors’ effect is taken as a inseparable. The Ariane was created specifically
to allow control without any effect. Automatic level control can be implemented in ways that are very audible.
For our purposes, however, levelling is being considered purely for its merit as a utility for the maintenance or
improvement of audible perception. The Ariane has been designed to have little or no sound effect of its own
The Arianes level-managed output will allow the following dynamics processing to work with an extremely con-
sistent signal, so that the expected effect will itself be more consistent and predictable.
Levels and Human Hearing
It is accepted among many learned students of human perception that human hearing, complex as it is, is more
sensitive to the long-term average of sound level than it is to instantaneous peaks. The human ear/brain per-
ception system integrates sound levels over time. That is, a sound of a given level that is sustained for a length
of time will sound louder than a shorter time burst of the same level. Depending upon duration and intensity, a
particular sound can actually be lower in level than another, yet be perceived to sound louder. The sensitivity of
human hearing to time as well as amplitude is critical when considering how to control listening levels over time.
RMS Average
One very useful way to determine average level is called Root Mean Squared, or RMS for short. This is a math-
ematical process where the audio’s absolute value is squared, then an average of this signal is taken over a
given length of time, then the square root is taken of that average. The result is a control signal that corresponds
to the power level of the audio over time, rather than its peak level.
RMS measurement, with its requisite integration, corresponds nicely with human hearing. By using an RMS sys-
tem of measurement and control, a level controller can work remarkably well in conjunction with human hearing.
Thus, by choosing to analyse our signal with RMS detection, we can build a simple system whose control is
much less distracting to the human ear than a simple peak-based control system.
The Nature of Program Audio
The creation of programs and presentations of audio generally requires the mixing and level adjustment of multi-
ple sources. It is the nature of ‘raw’ unprocessed audio sources (such as live microphones) to be wide ranging in
their levels, consistency and other qualities. These sources are the ones that most obviously will require a level-
ler’s automatic assistance. Yet it is also the nature of audio that some sources, such as produced music, will be
very consistent in level and quality, even while their initial absolute level may be unknown. These sources will
not require as much control, if any, as the unprocessed material. What is most typical of all, is that these multiple
types of sources may need to be conformed to each other before their final combined presentation to the end
user. Taking advantage of coincidental properties of human hearing and program audio, it doesn’t take a genius
to note that the audio for programming should be presented in a way that humans find most easy to accept.
While extreme loudness variations are disturbing, corrective control systems themselves can add effects that are
equally if not more disturbing. In other words, in audio levelling, the cure can be worse than the illness. By seek-
ing out and utilizing only those aspects of audio control that are utilitarian AND pleasing to the ear, while keeping
at arm’s length those methods of audio control that are generally found to be distracting, and then matching this
knowledge base with an understanding of the nature of typical program audio, an intelligent control system can
be designed that best meets the needs of successful, listenable programming over the long term.
Windowing Release
While we humans respond to average level in determining loudness, we also crave detail. That is, minute varia-
Содержание DSPXtra-FM
Страница 63: ...www bwbroadcast com...
