User Manual

9. Important notes on digital clipping
The digital words representing an audio signal at each moment have a
maximum possible positive value and a minimum possible negative value.
These are defined by the bit depth of the file format. Any attempt to force an
audio signal beyond these maximum permitted values (for example, by
applying excessive gain), will result in the audio signal being clipped. There
are other ways in which a signal can become clipped, and some of these are
far from obvious. We all know that clipping distortion sounds quite unpleasant
and is to be avoided.
PEAK-NORMALIZED SIGNALS
A 'Normalize’ process allows a file to be processed in such a way that the
maximum peak level within the file just reaches (but does not exceed) the
digital zero or clipping point. This is desirable because it means that the file is
as ‘loud’ as possible without clipping. This maintains the best signal-to-noise
ratio, especially at low bit word lengths.
the L3’s peak limiter is useful in situations where a higher average sound level
is required. It allows the typical level of signals to be even further increased by
gently pulling down the Gain of waveform peaks, with no audible nonlinear
distortion. The L3 can simultaneously re-scale the audio data so that the
limited peak signals approach or just reach digital zero.
However, by storing soundfiles at the maximum possible level, there is a risk
that any subsequent processing may take these peak levels too high. This will
result in clipping distortion. Peak limiting to 0dB, by whatever means, leaves
no margin for any subsequent increase in peak level.
Intuitively, you might think that simple Gain reduction could be applied without
incurring the risk of clipping, while any increase in Gain would surely cause
clipping. This is true. You might also think that applying an EQ boost at any
frequency might result in clipping, depending on the peak energy level within
the band being equalized. This is also true.
What is far less obvious is that applying an EQ cut also runs the risk of
causing clipping. To prove this would take a lot of math, but the following
description should help get the point across.
At any instant, the peak level of a signal may be the result of several
components at different frequencies and at different phases relative to each
other. Some components will add while others will subtract. But what happens
if you ‘EQ out’ a frequency that would otherwise be subtracting from the peak
level by virtue of its phase? The peak is now higher than it was. For most
audio material, this effect will be relatively small. Typically, peak levels will
increase by about 0.3 dB. However, it is possible that under unfavorable
Waves L3 software guide page 24 of 26