Digital Audio: How Sound is Transformed and Transmitted Digitally

Every time you stream a song, record a voice note, or join an online class, your device is performing a quiet bit of magic. It takes sound, which exists in the real world as continuous waves of pressure, and turns it into numbers that a computer can store, copy, and send anywhere. This conversion sits at the heart of modern communication. Understanding how sound becomes digital, gets squeezed into manageable file sizes, and travels across networks helps explain why an MP3 fits on your phone while a studio recording occupies far more space. Let us walk through the journey from a sound wave to a digital file.

From sound waves to numbers 🔗

Sound is a physical event. When you speak or play an instrument, you create vibrations that travel through the air as waves of changing pressure. A microphone picks up these vibrations and converts them into a continuously varying electrical voltage. This smooth, unbroken signal is called analog because the voltage is directly analogous to the air pressure that created it.

Computers, however, do not understand smooth waves. They only work with discrete numbers, ones and zeros. To bridge this gap, we use a device called an analog-to-digital converter (ADC), which measures the electrical signal at regular intervals and records each measurement as a number. When you play the file back, a digital-to-analog converter (DAC) reverses the process, rebuilding the voltage and driving a speaker so you hear sound again.

The audio spectrum and human hearing 🔗

Before digitising sound, it helps to know what range of sound matters. The human ear can detect vibrations roughly between 20 Hz and 20,000 Hz, measured in hertz, which counts how many times a wave completes a full cycle per second. Low frequencies produce deep sounds like a bass drum, while high frequencies produce sharp sounds like a whistle. Sounds below 20 Hz are called infrasound and those above 20 kHz are ultrasound. Both exist, but our ears simply cannot perceive them. This audible band, from 20 Hz to 20 kHz, is what engineers call the audio spectrum, and it defines how detailed our digital capture needs to be.

Sampling and the sampling rate 🔗

The process of measuring the analog signal at fixed time intervals is called sampling. Each measurement is a sample, and the number of samples taken every second is the sampling rate or sampling frequency. The most common sampling rate in consumer audio is 44.1 kHz, which means 44,100 measurements are taken every second. Video and professional work often use 48 kHz, while studio recordings may go up to 96 kHz or higher.

The higher the sampling rate, the more closely the digital version traces the original wave. A rate that is too low cannot capture fast-changing sounds and produces a distorted result. A rate that is far higher than needed only wastes storage space without adding sound that anyone can hear.

Why 44.1 kHz? The Nyquist theorem 🔗

The choice of sampling rate is not arbitrary. It follows a rule known as the Nyquist theorem, which states that to accurately capture a frequency, you must sample at least twice as fast as that frequency. Since the upper limit of human hearing is about 20 kHz, a system needs to sample at more than 40,000 times per second to reproduce everything we can hear. The 44.1 kHz standard sits just above this threshold, comfortably covering the audible range while leaving a small buffer for the filters that prevent unwanted distortion called aliasing. This is why 44.1 kHz became the benchmark for compact discs and remains common for everyday audio today.

Quantization and bit depth 🔗

Sampling decides how often we measure the wave, but we also need to decide how precisely to record each measurement. This is called quantization, and it depends on the bit depth, the number of bits used to store each sample. A 16-bit recording can represent 65,536 possible amplitude levels, which is enough for music distribution. Professional recording often uses 24-bit depth for greater precision during editing.

Think of bit depth as the markings on a ruler. The more closely spaced the markings, the more accurately you can read a measurement that falls between two of them. Two factors, therefore, decide how faithfully a digital file represents the original sound: the sampling rate and the bit depth. Together they also determine the raw data rate, which is simply the sampling rate multiplied by the bits per sample.

Compression and storage 🔗

Uncompressed audio is accurate but bulky. A standard CD-quality file carries a data rate of around 1,411 kilobits per second, so a three-minute song can easily occupy 30 to 50 megabytes. That was fine for a physical disc, but it becomes a problem when you want to store thousands of songs on a phone or stream them over a mobile network. Compression solves this by reducing file size, and it comes in two distinct flavours.

Lossless compression 🔗

Lossless compression shrinks a file without throwing away any sound. It works by finding and removing redundancy in the data, then repackaging it more efficiently, much like reorganising items to fit a smaller box without leaving anything behind. When the file is played, the compression is fully reversed, and the audio is restored bit for bit, identical to the original. The popular FLAC format works this way, typically reducing files to roughly half their original size. The trade-off is that lossless compression cannot shrink files nearly as much as the alternative.

Lossy compression and psychoacoustics 🔗

Lossy compression achieves much smaller files by permanently discarding audio data, but it does so cleverly. It relies on psychoacoustics, the study of how humans actually perceive sound. The encoder identifies parts of the audio that our ears are unlikely to notice and removes them. For example, if a loud sound and a much quieter sound occur close together in frequency, our brain naturally masks the quieter one, so the encoder simply deletes it.

This approach, also called perceptual coding, was one of the most influential audio technologies of the twentieth century. According to analysis from Sound on Sound, lossy formats let you control the balance between size and quality through the bit rate, often using variable bit rate encoding that lowers the data rate during simpler passages of music. The result is a file a fraction of the original size that still sounds convincing to most listeners. The catch is that once the data is gone, it cannot be recovered.

Common audio formats 🔗

A format is essentially a container that defines how audio data is stored and labelled. The three most commonly studied formats illustrate the spectrum from raw quality to convenient portability.

WAV 🔗

The WAV format, short for Waveform Audio File Format, was developed by Microsoft and IBM for Windows computers. It stores uncompressed audio, usually as raw PCM data, which means it keeps every sample exactly as it was captured. WAV files support sample rates up to 192 kHz and bit depths up to 32-bit, which is why they remain the default working format in professional audio software. The downside is size, since nothing is discarded a WAV file is essentially a long list of every sample value. This makes WAV ideal for recording, editing, and archiving, but inconvenient for casual music libraries.

MP3 🔗

The MP3 format, formally MPEG-1 Audio Layer III, is the lossy format that transformed how the world listens to music. Developed largely at the Fraunhofer Institute and standardised in the early 1990s, its goal was to shrink CD-quality audio into a much smaller file with little noticeable difference. It does this using psychoacoustic modelling combined with mathematical techniques to discard inaudible information. An MP3 can compress audio to roughly a tenth of the original size. At its highest bit rate of 320 kbps, it sounds excellent for ordinary listening, which is why it became the standard for sharing music and distributing podcasts across nearly every device.

AIFF 🔗

The AIFF format, or Audio Interchange File Format, was created by Apple in the late 1980s and is essentially the Macintosh counterpart to WAV. Like WAV, it stores uncompressed PCM audio and offers the same high sample rates and bit depths, delivering the highest quality with large file sizes. There is no audible difference in sound quality between a WAV and an AIFF of the same specifications. One area where AIFF stands out is its strong support for metadata, allowing files to carry details such as artist names, album artwork, and track notes. Both formats now play on Windows and Mac systems alike, so the choice often comes down to ecosystem and metadata needs rather than sound.

Choosing the right format 🔗

The decision usually depends on purpose. For recording, mixing, and long-term archiving where quality must be preserved, an uncompressed format like WAV or AIFF is the right choice. For everyday listening and sharing where file size and compatibility matter most, a high bit rate MP3 works well. For those who want smaller files with no quality loss, a lossless compressed format like FLAC offers a middle path. Each format reflects a different answer to the same underlying question: how much quality are you willing to trade for convenience?

What do you think? If lossy formats like MP3 discard sound that most people cannot hear, is there any real value in storing music in uncompressed formats for the average listener? And as storage and bandwidth keep getting cheaper, do you think lossy compression will eventually become unnecessary?

References

1. https://www.techtarget.com/whatis/definition/analog-to-digital-conversion-ADC
2. https://en.wikipedia.org/wiki/Audio_frequency
3. https://www.sciencedirect.com/topics/engineering/nyquist-theorem
4. https://en.wikipedia.org/wiki/44,100_Hz
5. https://www.soundonsound.com/techniques/what-data-compression-does-your-music
6. https://www.izotope.com/en/learn/whats-the-difference-between-file-formats