What Is PCM Audio?

PCM turns continuous sound into a stream of numbers. A microphone converts sound waves into a fluctuating electrical voltage, and an analog-to-digital converter (ADC) measures that voltage at fixed, evenly-spaced instants — each measurement is a sample, a number representing the amplitude of the sound at that moment. Store those numbers in sequence and you have PCM audio: a literal, ordered list of amplitude readings. To play it back, a digital-to-analog converter (DAC) reads the numbers at the same rate and reconstructs a smooth analog voltage, which drives speakers or headphones to recreate the original pressure wave. Nothing is modeled, predicted, or perceptually discarded — PCM stores the waveform directly, which is why it is both the simplest scheme and the highest-fidelity one. Its faithfulness is governed entirely by two settings: how often the ADC measures (sample rate) and how precisely it records each measurement (bit depth).

PCM encoding is exactly two operations. First, sampling: the ADC takes amplitude readings at a fixed rate — the sample rate — which sets the time resolution and, by the Nyquist theorem, the highest frequency the recording can contain (half the sample rate). Second, quantization: each sampled amplitude, which is a continuous value, is rounded to the nearest level on a fixed scale defined by the bit depth. Sampling handles the time axis; quantization handles the amplitude axis. The rounding in quantization introduces a small error — quantization noise — that sets the noise floor, and finer bit depth means smaller rounding steps and lower noise. 'Linear PCM' (LPCM), the form used in WAV, AIFF, and CD, means the quantization levels are evenly spaced. Understanding these two independent steps is the key to all digital audio: sample rate and bit depth are not vague quality knobs but the precise definitions of temporal and amplitude resolution.

Three numbers fully specify a PCM stream. Sample rate — how many samples per second — with 44,100 Hz the CD standard and 48,000 Hz the video and broadcast standard; higher rates like 96 and 192 kHz are used in production. Bit depth — how many bits encode each sample — where 16-bit gives 65,536 amplitude values (~96 dB dynamic range), 24-bit gives 16.7 million (~144 dB), and 32-bit float extends the range enormously for processing. Channels — how many independent streams — with mono carrying one, stereo two, and surround five, seven, or more. Together, sample rate, bit depth, and channel count determine both the audio quality and the exact data rate: multiply them (rate × bytes-per-sample × channels) and you get the bytes per second. Because nothing is compressed, PCM size and quality are perfectly predictable from these three numbers alone — unlike lossy formats, where a bitrate setting stands in for quality.

The way the numbers are written to a file matters for compatibility. Sample format: 8-bit PCM is conventionally stored as unsigned integers (0-255, with 128 representing silence), while 16-, 24-, and 32-bit integer PCM use signed two's-complement values centered on zero, and 32-bit float uses IEEE floating point. Endianness: multi-byte samples can be stored little-endian or big-endian, and this is a defining difference between the two main uncompressed containers — WAV (and its RIFF structure) is little-endian, while AIFF (Apple's IFF-based format) is big-endian, though the underlying samples are otherwise the same. Channel interleaving: in a stereo or surround file the samples are usually interleaved — left sample, right sample, left, right — so playback hardware can stream them in order, rather than stored as separate blocks per channel. These details are invisible in normal use but explain why a raw PCM stream needs a container (WAV, AIFF) or an explicit format description to be interpreted correctly — the same bytes mean different things at different bit depths, sign conventions, and byte orders.

Not all PCM is the plain linear kind. Linear PCM (LPCM) uses uniform quantization steps and is what WAV, AIFF, CD, Blu-ray, and HDMI use. But several PCM relatives trade exactness for efficiency. A-law and μ-law (mu-law) are companded PCM used in telephony: they apply logarithmic quantization so that 8 bits carry roughly the perceptual range of 13-14 linear bits, which is why phone systems get intelligible voice at 8-bit, 8 kHz. ADPCM (Adaptive Differential PCM) stores the difference between consecutive samples and adapts its step size, achieving modest compression while remaining simple to decode — used in older games, some voice recorders, and formats like IMA/DVI ADPCM. DPCM stores raw differences without adaptation. These are technically lossy or semi-compressed schemes built on the PCM idea, distinct from true LPCM. When people say 'PCM' without qualification in a music context, they almost always mean linear PCM — the uncompressed, uniformly-quantized form that serves as the quality reference.

PCM is nearly everywhere uncompressed digital audio appears. Audio CDs are 16-bit, 44.1 kHz stereo LPCM. WAV files are almost always PCM in a RIFF container; AIFF files are PCM in an IFF container. Blu-ray supports LPCM up to 24-bit, 192 kHz and up to 7.1 channels. Every professional and consumer audio interface captures PCM from its ADC and streams it to the computer over USB, Thunderbolt, or PCIe. HDMI and S/PDIF carry LPCM as their uncompressed audio path between devices. DAWs work in PCM (often 32-bit float) internally for all editing and mixing. Telephony uses companded 8-bit, 8 kHz PCM (μ-law or A-law). Even when you play an MP3, the decoder's final output is PCM sent to the DAC. In short, PCM is both the capture format at the microphone end and the playback format at the speaker end — compression only exists in between, for storage and transmission.

PCM is uncompressed: every sample is stored exactly as measured, so the file is large but perfect. All codecs are defined relative to it. Lossy codecs (MP3, AAC, Opus, Vorbis) take PCM as input and discard perceptually-masked detail to shrink the file dramatically; their decoded output is an approximation of the original PCM, close but not identical. Lossless codecs (FLAC, ALAC, WavPack) also start from PCM but compress it reversibly, so their decoded output is bit-for-bit identical to the original PCM at roughly half the size. This is the crucial mental model: PCM is the source and the reference. A lossless codec is judged by whether it reconstructs the PCM exactly (it does); a lossy codec is judged by how close its reconstruction sounds to the PCM (very close at high bitrates). Quality comparisons, null tests, and mastering checks all measure the codec output against the original PCM, which is why keeping a PCM (or lossless) master matters.

PCM is not the only way to digitize audio. DSD (Direct Stream Digital), the format behind the SACD, takes a fundamentally different approach: instead of many-bit samples at a moderate rate, it uses 1-bit samples at an extremely high rate — 2.8224 MHz (64 times CD's 44.1 kHz) for standard DSD64, and higher for DSD128 and DSD256. Each 1-bit sample only records whether the signal is going up or down relative to the last, using a sigma-delta modulation scheme; the density of ones and zeros encodes the waveform. DSD advocates argue it captures a smoother, more analog-like signal; critics note it is harder to edit (most DSD workflows convert to high-rate PCM for processing anyway) and pushes quantization noise into ultrasonic frequencies that must be filtered. In practice, PCM dominates because it is easy to edit, mix, and convert, and modern high-rate PCM matches or exceeds DSD on measurable performance. DSD remains a niche audiophile format; PCM is the universal working standard.

Because PCM is uncompressed, its size is exact arithmetic: bytes per second = sample rate × (bit depth ÷ 8) × channels. CD-quality 16-bit / 44.1 kHz stereo is 44,100 × 2 × 2 = 176,400 bytes per second, which is 1,411 kbps, about 10.1 MB per minute, or 605 MB per hour. Stepping up to 24-bit / 48 kHz stereo gives 2,304 kbps, about 16.5 MB per minute. High-resolution 24-bit / 96 kHz stereo is 4,608 kbps, about 33 MB per minute — so a standard 4 GB WAV holds only around two hours. Mono halves every figure; surround multiplies it by the channel count. These fixed costs are the entire reason lossy codecs exist: a 128 kbps MP3 is roughly one-eleventh the size of CD PCM. For planning storage and bandwidth, PCM size is the predictable baseline, and any codec's savings are measured against it. When file size matters, you compress the PCM; when quality and editability matter, you keep it.

Use PCM whenever audio is being captured, edited, or held at full quality. Recording: your interface captures PCM, so tracking is always PCM. Editing and mixing: DAWs work in PCM (commonly 32-bit float) internally, and every cut, fade, and process operates on raw samples. Mastering: final masters are PCM right up until the moment they are encoded to a distribution format. Quality assurance: you compare a compressed file against the original PCM to judge how much was lost. Archival: keep masters as PCM (WAV or AIFF) or losslessly-compressed PCM (FLAC, ALAC) so no detail is ever discarded. For storage and distribution, compress that PCM into FLAC for lossless or MP3/AAC/Opus for lossy delivery — but always keep the PCM master, because you can generate any compressed format from it, while you can never fully rebuild PCM from a lossy file.