Digital Audio
A light controlled by a dimmer switch changes continuously from one brightness
level to another ranging from total darkness to fully on. This is fundamentally
analog: theoretically, a dimmer has an infinite number of states. A light controlled
by a regular switch, on the other hand, has two states—on and off—and nothing
in between. This is fundamentally digital. To understand what happens when an
analog audio signal is converted into a digital signal, we must look at what the
ADC in a recording system does.
SAMPLING AND QUANTIZING
In order to convert a continuous analog signal into a series of numbers, the ADC
has to sample the incoming analog waveform. To do this, the ADC measures the amplitude of
the incoming waveform some number of times every second and assigns a numerical value to
the amplitude. The number of samples taken per second is a frequency and is given in hertz
(Hz), just as the number of cycles per second of a waveform is given in Hz. This frequency is
called the sampling rate. The act of assigning an amplitude value to the sample is called
quantizing and the number of amplitude values available to the ADC is called the sample
resolution.
It is worth noting that the term “sampling” has at least three distinct meanings in music
technology, each at a different timescale. The first meaning, given above, is the periodic
measuring of an analog waveform’s amplitude. In this sense, an individual sample represents an
imperceptibly small amount of time. The second meaning is the act of recording single notes of
musical instruments to be triggered later by MIDI messages. These are the samples forming
sample libraries that are commonly used by composers to mimic the timbres of acoustic
instruments (see Chapter 13, “Sampling Techniques”). The third meaning is drawn from hip-hop
and other pop genres, in which a recognizable portion of an existing recording, such as several
measures of music or a characteristic vocal sound, is used in the creation of a new song.
Throughout this discussion, the term “CD-quality” will be used as shorthand for the digital audio
specifications associated with the audio CD medium. “CD-quality” is really a misnomer. The
actual quality of the audio varies greatly depending on the microphones, preamps, and ADCs
used to record it and the DACs, amps, and speakers used to listen to it.
The Sampling Rate
The sampling rate of CD-quality digital audio is 44,100 Hz (44.1 kHz), meaning
that the ADC measures the amplitude of the incoming analog waveform 44,100
times per second for each channel of audio (two channels in the case of stereo
audio). Another way of saying it is that the ADC measures the amplitude, waits a
short period of time, takes another measurement, waits a short period of time,
takes another measurement, and so on (see Figure 6.2).
You may remember from geometry that between any two points on a continuous
line there are an infinite number of other points. This means that, while the ADC
is waiting between samples, an infinite number of points are ignored. If the ADC
actually tried to sample every point, storing the data would use up all of the hard drives,
flash drives, and CDs in the world before digitizing even a millisecond of analog audio. The fact
that an infinite number of points are thrown away when the ADC digitizes an analog
signal has consequences in the resultant digital signal.
The Nyquist Frequency
The consequence of capturing only a finite (rather than infinite) number of
samples is that the resultant digital audio can contain no frequency that is higher
than one-half of the sampling rate. This maximum frequency is called the Nyquist
frequency after the physicist and engineer, Harry Nyquist, who contributed to what is known as
the sampling theorem. For CD-quality digital audio:
Nyquist frequency = 1⁄2 × Sampling rate
Nyquist frequency = 1⁄2 × 44,100 Hz = 22,050 Hz = 22.05 kHz
What this means is that the maximum frequency that can be present on a CD,
including fundamentals and all partials, is 22,050 Hz. As a result, every recording system
first gets rid of any frequencies above the Nyquist frequency, using a filter, before digitizing the
signal (see Figure 6.3).
The infinite number of samples that the ADC threw away during digitization were necessary to
represent all of the frequencies above the Nyquist frequency. Is the loss of all of the frequencies
above 22,050 Hz, the Nyquist frequency, an acceptable compromise?
If you remember the discussion of the range of human hearing, we can hear frequencies
(fundamentals and partials) up to about 20,000 Hz when we’re young. Since 20,000 Hz <
22,050 Hz, digital audio sampled at the CD-quality sampling rate contains all of the frequencies
that we can hear, and thus the compromise works.
Higher Sampling Rates
Despite the fact that the CD-quality sampling rate properly represents all of the frequencies that
we can hear, higher sampling rates are frequently used during recording, even when the result
is eventually reduced to the CD-quality sampling rate or to a compressed format for distribution.
One common sampling rate is 48 kHz, which is only marginally higher than the CD-quality
sampling rate. This sampling rate is the standard for audio that accompanies video; a video file
ripped from a DVD will most often have audio at 48 kHz. Fortunately, even the built-in audio
outputs of a computer can usually handle 48 kHz, so you don’t have to do anything special to
play it back.
There are two higher than CD-quality sampling rates often used in recording: 96 kHz and 192
kHz. The Nyquist frequencies for these two sampling rates are 48 kHz and 96 kHz respectively,
both of which are far above human hearing. In addition, some audio interfaces support 88.2
kHz.
There are various arguments used to justify high sampling rates. There are the sound- quality
justifications, including being able to perform the initial filtering far out of the range of human
hearing, being able to capture ultrasonic frequencies that may combine in performance to
produce audible components within the range of human hearing, or that the sound has more
“sparkle.” These arguments have been hotly debated with no clear resolution. There are also
cultural justifications, including “everyone else is doing it” and “I can use these numbers to
convince my clients that my studio is state- of-the-art.”
Regardless of the reason, recording practice routinely includes 96 kHz and sometimes 192 kHz,
even when the final product is distributed at 44.1 kHz or as a compressed audio file (more on
that later).
