Digital audio is the representation of sound as numerical data that can be stored, processed, transmitted and reproduced by digital electronic systems. Acoustic sound is a continuous physical phenomenon involving variations in air pressure, whereas digital audio represents measurements of those variations using discrete numerical samples. An analogue-to-digital converter, or ADC, samples an electrical representation of sound at regular intervals and quantises each measurement into a finite numerical value.
The resulting sequence of numbers can be stored in files, transmitted through networks, edited using software and processed by digital signal processors. During playback, a digital-to-analogue converter, or DAC, reconstructs an electrical waveform from the numerical samples, which can then drive headphones, loudspeakers or other audio equipment.
Sampling rate and bit depth are two fundamental properties of digital audio. The sampling rate specifies how many measurements are taken per second and is expressed in hertz or samples per second. Common rates include 44.1 kHz and 48 kHz, with higher rates used in some production and specialist applications. According to the Nyquist-Shannon sampling theorem, a band-limited signal can be reconstructed from samples when the sampling rate is greater than twice the highest frequency component being represented.
Human hearing is often described as extending approximately to 20 kHz for young listeners under ideal conditions, which helps explain the historical use of sampling rates around 44.1 kHz for consumer audio. Practical systems also employ anti-aliasing filters to prevent frequencies above the usable bandwidth from being incorrectly represented as lower frequencies.
Bit depth determines how precisely each sample's amplitude can be represented. With a fixed number of bits, there are a finite number of possible amplitude values. Greater bit depth permits finer quantisation and increases the theoretical dynamic range while reducing quantisation error. A commonly used 16-bit representation provides 65,536 possible quantisation levels per sample, while 24-bit audio provides many more. In professional recording, greater bit depth provides additional headroom and makes low-level processing less susceptible to quantisation limitations.
Bit depth should not be confused with bitrate: bit depth describes the precision of individual samples, while bitrate describes the number of bits used per unit of time for a particular encoded stream. The uncompressed bitrate of PCM audio depends on sample rate, bit depth and number of channels.
Digital audio is often stored as pulse-code modulation, or PCM. In straightforward PCM, each channel is represented by a sequence of numerical amplitude samples. Stereo audio has separate left and right channels, while surround formats can contain substantially more channels. PCM itself is generally an uncompressed representation, although the underlying samples can subsequently be placed inside different containers or processed using lossless or lossy compression.
File formats such as WAV and AIFF can contain PCM audio, but the container format and the audio encoding are conceptually different. This distinction is important because a file extension does not always uniquely determine the codec or internal representation. Digital audio can therefore involve several layers: sampled audio data, an encoding or compression method, a file container and metadata.
Digital audio can be edited mathematically in ways that are difficult or impossible with purely analogue systems. Software can adjust volume, equalisation, dynamics, timing, pitch, stereo balance and other properties by manipulating numerical samples. Digital effects can include reverberation, delay, filtering, compression, limiting and noise reduction. Multiple tracks can be mixed into a final waveform, and the same source can be duplicated without the generational degradation associated with repeated analogue copying. However, digital processing is not automatically perfect: poorly designed algorithms, clipping, excessive processing, inadequate conversion or inappropriate sample-rate conversion can introduce audible artefacts. Good digital audio engineering therefore depends on both adequate technical resolution and appropriate signal-processing practices.
Digital audio is fundamental to modern telecommunications, broadcasting, music production, streaming, film, games, accessibility systems and computing. Voice calls convert speech into digital representations so that network infrastructure can transport it efficiently. Music services use compressed digital audio to reduce storage and bandwidth requirements. Recording studios often work with high-resolution PCM because it provides flexibility for editing and mixing before final distribution. Digital audio can also be encoded using perceptual codecs such as MP3 or AAC, or lossless codecs such as FLAC. The appropriate representation depends on the application's priorities, including fidelity, file size, latency, processing requirements and compatibility.
Digital audio has several advantages over analogue representation. Numerical data can be copied and processed consistently, stored compactly using compression, transmitted with error-detection mechanisms and manipulated by computers. Digital systems can also incorporate metadata such as track titles, artist names and timestamps. Nevertheless, digital audio requires conversion between the physical analogue world and numerical representations. The quality of the final result depends on microphones, analogue circuitry, converters, clocks, sampling and quantisation, signal processing and playback equipment. The phrase “digital” therefore does not itself guarantee high quality; it describes the representation method, while quality depends on the complete capture, processing, encoding and reproduction chain.
Digital audio should be distinguished from compressed audio. Compression is optional: PCM can be uncompressed, while digital audio can also be compressed using lossless or lossy techniques. It should also be distinguished from analogue audio, where the signal varies continuously rather than being represented as discrete numerical samples.
Digital audio can be lossless in the sense that a lossless codec can reproduce the original digital samples exactly, but this does not mean that the original acoustic event can be reconstructed perfectly because microphones, analogue conversion and sampling already establish the original digital representation. Digital audio is therefore best understood as a numerical model of an audio signal whose usefulness comes from its ability to be stored, processed, transmitted and reproduced predictably by digital systems.
