Digital audio is a numerical representation of sound in which an analogue audio signal is converted into discrete numerical samples that can be stored, processed and reproduced by digital systems. During analogue to digital conversion, the continuously varying electrical signal is measured at regular intervals determined by the sampling rate, and each measurement is represented by a numerical value with a resolution determined by bit depth.
Common sampling rates include 44.1 kHz and 48 kHz, while higher rates such as 88.2 kHz, 96 kHz or 192 kHz are also used in some production contexts. Common bit depths include 16 bit and 24 bit, with 32 bit floating point widely used internally in many digital audio workstations. Digital audio allows sound to be copied, edited, transmitted and processed as data. The stored numbers do not constitute sound by themselves. During playback, a digital to analogue converter reconstructs an electrical waveform that can be amplified and reproduced through speakers or headphones. Digital audio is therefore a complete chain involving sampling, numerical representation, storage, processing, conversion and playback rather than merely a file format.
Sampling rate describes how frequently the analogue waveform is measured. A sampling rate of 44.1 kHz means that 44,100 samples are taken per second for each channel. The relationship between sampling rate and frequency representation is described by the Nyquist sampling theorem. To represent frequencies accurately without aliasing, the sampling rate must be greater than twice the highest frequency component being represented, and practical systems use anti aliasing filters because real signals and filters have finite characteristics.
Human hearing is commonly described as extending to approximately 20 kHz for young healthy listeners, although individual hearing varies and the upper limit generally decreases with age. This helps explain why rates such as 44.1 kHz and 48 kHz are sufficient for conventional audible frequency ranges when properly implemented. Higher sampling rates can provide wider transition bands for analogue filters and may be useful in particular production or processing contexts, but they do not automatically make a recording audibly superior. The appropriate rate depends on the application, equipment, processing requirements and delivery specification.
Bit depth determines how precisely each sample value can represent signal amplitude. Greater bit depth provides more possible numerical levels and therefore allows lower quantisation error and a greater theoretical dynamic range. Sixteen bit audio provides 65,536 possible integer levels per sample, while 24 bit audio provides 16,777,216 possible integer levels.
In ideal conditions, each additional bit increases theoretical signal to quantisation noise ratio by approximately 6 dB, so 24 bit systems provide substantially more theoretical dynamic range than 16 bit systems. Real equipment has additional limitations such as analogue noise, converter performance and clocking, so theoretical values should not be confused with practical system performance. Digital audio systems also use floating point arithmetic internally, which can provide very large numerical headroom during processing.
This does not mean that an analogue converter can record unlimited level. If the signal clips at the input converter, information may already be lost. Bit depth is therefore particularly important for recording and low level resolution, while internal floating point processing provides flexibility during editing and mixing.
Digital audio can be stored in many different file formats and encoding systems. WAV and AIFF are common containers for uncompressed PCM audio, while formats such as FLAC use lossless compression that reduces file size without discarding audio information. Lossy codecs such as MP3 and AAC reduce file size by discarding information according to perceptual models. A container describes how data and metadata are organised, while a codec describes how audio is encoded and decoded. These concepts are often confused.
A WAV file can contain different forms of audio data, although PCM is common, and an MP3 file uses a particular lossy compression system. Lossless compression can reconstruct the original digital audio exactly, whereas lossy compression produces a representation that differs from the source. Streaming platforms commonly use their own delivery and encoding pipelines, meaning that the audio supplied by a producer may undergo additional processing before reaching listeners. Professional workflows therefore usually retain high quality lossless masters and create distribution formats as required rather than treating a compressed consumer file as the definitive source.
Digital audio provides extensive opportunities for editing and processing. Once sound has been converted into numerical data, software can manipulate levels, timing, pitch, frequency content, dynamics, spatial placement and many other properties. Digital audio workstations can cut and move regions, create crossfades, correct timing, adjust pitch, automate parameters and apply effects. Processing can be destructive or non destructive depending on the workflow.
Digital copies can be exact duplicates of the source data, unlike repeated analogue copying where generation loss can occur. This makes digital systems particularly powerful for complex production. However, digital flexibility can also encourage excessive editing and processing. A recording can be technically corrected to the point where natural performance variation disappears. Digital technology therefore changes what is possible but does not determine what is musically appropriate. The quality of the final result still depends on recording technique, musical decisions, monitoring, processing choices and the judgement of the people using the system.
Digital audio involves several technical issues beyond sampling and bit depth. Quantisation error occurs when an analogue amplitude cannot be represented exactly by the available numerical levels, although properly designed systems can reduce its audible significance substantially. Dithering adds controlled low level noise when reducing bit depth and can improve the statistical behaviour of quantisation error. Aliasing occurs when frequency components above the system's representable range are incorrectly mapped into lower frequencies, which is why anti aliasing filtering is important.
Digital clipping occurs when a signal exceeds the maximum representable value in a fixed point system, producing severe distortion. In floating point environments, internal processing can provide considerable headroom, but conversion to fixed point or analogue output still requires appropriate level management. Clocking determines the timing of sample conversion, and poorly controlled timing can introduce errors, although modern professional systems can achieve very low levels of such problems. These concepts demonstrate that digital audio is an engineering discipline as well as a convenient method of storing music.
Digital audio is used throughout contemporary music, film, broadcasting, games, communications, archiving and scientific applications. It allows audio to be stored on computers, transmitted through networks, copied rapidly and processed using algorithms. A digital recording can be synchronised with video, edited frame accurately, distributed internationally and incorporated into complex multimedia systems. Digital audio also supports multichannel and immersive formats in addition to conventional mono and stereo.
Metadata can store information such as artist names, titles, dates and technical specifications, although support varies between formats and platforms. Digital preservation is nevertheless not automatically permanent. Files require readable formats, functioning storage media, data integrity checks and migration strategies. A perfectly copied file can still become inaccessible if the software, hardware or format required to decode it disappears. Long term preservation therefore requires management of both the data and the technological environment surrounding it.
Digital audio is sometimes described as either inferior to or inherently better than analogue audio, but such a simple comparison is technically inadequate. Digital and analogue systems have different mechanisms, limitations and characteristic behaviours. A well designed digital system can provide extremely low noise, precise editing, stable copying and wide dynamic range, while analogue systems can provide different forms of saturation and physical behaviour. Audible results depend on implementation, recording levels, converters, microphones, monitoring, processing and production choices.
Digital audio also does not mean that music is generated without physical components. Digital playback still requires transducers, amplifiers, analogue circuits and speakers or headphones. Understanding digital audio therefore requires viewing it as a representation and processing system connecting acoustic sound with numerical information. Its importance comes from the ability to manipulate, store and distribute that representation with great precision, while recognising that numerical accuracy alone does not guarantee musical quality.
