EETimes - How audio codecs work about:reader?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttps%2Fwww.eetimes.co...

eetimes.com

EETimes - How audio

codecs work
EETimes
6-8 minutes

Editor’s note: this article begins an ongoing

series of technical reviews of audio codecs,
how they work, and common applications.
Introduction
Audio codecs are among the most important
innovations affecting the communcations and
entertainment industries. Digital sound
techniques have enabled whole new
approaches to traditional analog technologies
for more than fifty years. Originally confined to
telephone company infrastructure
applications, digital respresentations of audio
is ubiquitous today. Many of the most popular

1 of 8 4/30/22, 18:45
EETimes - How audio codecs work about:reader?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttps%2Fwww.eetimes.co...

consumer products rely on digital audio, and

more are being introduced almost daily.
Unlike many other computer-based
information technologies, digital sound has an
absolute “gold standard” for comparison.
Everybody has personal, first-hand
exprerience with live music and face-to-face
conversation. Ordinary telephone sound
quality is the expectation for audio
telecommunications. This expectation has
unexpected consequences. For example,
when GSM phones were field tested users
objected to the absolute silence present
during periods of no speech. “Comfort noise”
had to be injected into the audio signal to
conform to users telephone-based
expectations. Ultimately, audio codec sound is
about consumers' quality expecttions.
There are many factors that affect sound
quality. The similarity between the sound
originated by the source and the data stored
by the computer, usually known as fidelity, is

2 of 8 4/30/22, 18:45
EETimes - How audio codecs work about:reader?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttps%2Fwww.eetimes.co...

determined by the sound format and the

recording method. Lossy audio compression
formats like mp3 degrade the audio quality in
an almost imperceptible way because of
removed information in the audio data. In
spite of this degradation, the most important
task is the acquisition of the sound data.
An introduction to digital sound recording is
necessary prior to studying digital sound
formats. Many of the concepts dealing with
sound compression formats have their basis
in digital sound recording.
Recording equipment
Analog recording acquires the original signal
with the highest fidelity possible. Due to noise
and nonlinearities originated from electronic
devices, analog signals may suffer several
quality losses and distortions. Digital signals
are immune to these kinds of effects, however
sound sources are in most cases continuous
and real, making analog recording
unavoidable.

3 of 8 4/30/22, 18:45
EETimes - How audio codecs work about:reader?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttps%2Fwww.eetimes.co...

The basic steps for digital sound storage and

processing are:
Sample and Hold
The sampling and hold stage takes a time
variant voltage signal and breaks it into
sections where the voltage remains constant
for a given period of time, known as a
sampling period. This operation is extremely
important because it prepares the signal for
digitization. Theoretically, the input signal
must be frequency band limited with its
highest frequency as half of the inverse
sampling period. This boundary is known as
the Nyquist Frequency. If a signal contains
frequencies beyond the Nyquist Frequency it
needs to be pre-filtered to avoid a distortion
called “aliasing”. A commonly utilized
Sampling Frequency is 44100 Hertz; this
means that the recorded sound cannot have
any component beyond 22050 Hertz.

Figure 1 Basic sample and hold circuitry

4 of 8 4/30/22, 18:45
EETimes - How audio codecs work about:reader?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttps%2Fwww.eetimes.co...

captures an analog signal for conversion.

Analog to Digital Conversion (ADC)
After signal preparation, performed by the
sample and hold, every single constant signal
step is digitalized by the Analog to Digital
Converter. The voltage level is transformed
into a binary representation, better known as
Pulse Code Modulation. The quality of the
analog to digital conversion depends on the
number of possible voltage levels, which is
calculated by raising 2 to the number of bits
(also called quantization bits). The most
common conversion utilizes 16 bits, but some
equipment goes up to 24 bits.

Figure 2 The complete conceptual A/D

converter.
Digital Signal Processing
The digitalized signal consists now of a
sequence of numbers, the sampling
frequency and the number of quantization
bits. Many sound effects can be added to the

5 of 8 4/30/22, 18:45
EETimes - How audio codecs work about:reader?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttps%2Fwww.eetimes.co...

raw audio data using digital signal processing

(DSP). Instead of analog audio recording, a
sound can be generated mathematically.
Digital synthesizers simply generate a
sequence of numbers; representing every
voltage level, assign a number of quantization
bits and a sampling frequency.

Figure 3 Conversion from continuous time to

discrete time results in quantization of the
input signal.
Digital to Analog Conversion
In order to hear the sound from a digital
source it needs to first be transformed into an
analog signal. This task is performed by the
Digital to Analog Converter (DAC) taking a
digital sequence of numbers as input and
delivering a stepwise voltage signal. The
number of bits and the speed of conversion
limit the DAC.
Filtering (Sample and Hold Inverse Operation)
To recover the signal it is necessary to take

6 of 8 4/30/22, 18:45
EETimes - How audio codecs work about:reader?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttps%2Fwww.eetimes.co...

the stepwise voltages coming out of the

D.A.C. and smooth them. A low band pass
filter that eliminates every component over the
Nyquist frequency performs this smoothing.
The final result is an analog signal.
Summary
Besides the digital algorithm utilized to
process sound, the quality is mostly
determined by the acquisition method.
Compression algorithms alter the digital
signal properties to achieve smaller file sizes
with reasonable quality degradation. For
example, if voice is recorded using a sampling
frequency of 44100 Hertz it would be
reasonable for compression to decrease the
sampling frequency down to 8000 Hertz. The
sound is not greatly affected since voice does
not have significant frequencies over 4000
Hertz. The signal can be re-sampled digitally
to 44100 Hertz again and sent to the audio
device. All compression techniques
compromise the quality, However each of

7 of 8 4/30/22, 18:45
EETimes - How audio codecs work about:reader?url=https%3A%2F%2Ffanyv88.com%3A443%2Fhttps%2Fwww.eetimes.co...

them takes their own approach to make the

degradation as unperceivable as possible.
The situation can be compared to image
compression (similar concepts are applied),
but in that case for a 2D grid of data.
About the authors
Christopher Davis received his B.S. in
Computer Science from the University of New
Mexico in 2001 and his M.S. in Computer
Science from the University of New Mexico in
2005. He is currently employed by Respec
Information Technologies as a contractor for
Sandia National Laboratories. He can be
reached by email at
Victor M. Vergara is a Doctoral candidate in
Eletrical Engineering at th University of New
Mexico. He received his B.S.E.E. degree from
the University of Panama, and M.S.E.E. from
UNM. He can be reached by email at

8 of 8 4/30/22, 18:45