PCM, or pulse-code modulation, is the method of representing an analog signal digitally by measuring its amplitude at regular intervals and storing each measurement as a binary number. In digital audio, PCM is the foundation of almost everything: CDs, WAV and AIFF files, studio recording, Blu-ray soundtracks and the internal processing of audio interfaces and plugins.
A PCM stream is fully described by three parameters: the sample rate, the bit depth and the number of channels. Together they determine the frequency range, the dynamic range and the data rate of the signal, and lossless and lossy codecs ultimately decode back to PCM for playback.
How PCM encodes an audio waveform
PCM conversion happens in three steps. An anti-aliasing filter first removes frequencies above half the sample rate; the converter then samples the filtered signal at fixed intervals; finally, each sample is quantized to the nearest of a fixed set of levels and stored as an integer or floating-point number.
Playback reverses the process. A digital-to-analog converter turns the numbers back into voltages, and a reconstruction filter smooths the result into a continuous waveform. When the signal is properly band-limited, the sampling theorem guarantees exact reconstruction of everything below the Nyquist frequency; the only inherent error comes from quantization.
| Stage | What happens | Main source of error |
|---|---|---|
| Anti-aliasing filter | Removes content above the Nyquist frequency | Filter ripple or phase effects |
| Sampling | Measures amplitude at fixed intervals | Clock jitter |
| Quantization | Rounds each sample to a discrete level | Quantization noise |
| Storage or transmission | Writes samples as binary words | None if lossless |
| Reconstruction | Converts numbers back to a smooth waveform | Filter and analog stage quality |
Sample rate, bit depth and data rate
The sample rate sets the highest frequency that can be represented, which is half the sample rate, known as the Nyquist frequency. The bit depth sets the number of quantization levels and therefore the theoretical dynamic range, which grows by about 6.02 dB per bit.
| Sample rate | Nyquist frequency | Typical use |
|---|---|---|
| 8 kHz | 4 kHz | Traditional telephony |
| 44.1 kHz | 22.05 kHz | Audio CD and music distribution |
| 48 kHz | 24 kHz | Video, broadcast and most production |
| 96 kHz | 48 kHz | High-resolution recording and processing |
| 192 kHz | 96 kHz | Archival and specialist work |
The theoretical dynamic range of a format is roughly 6.02 × bits + 1.76 dB for a full-scale sine wave. Real converters fall short of these figures because of analog noise; the best 24-bit converters reach roughly 120 to 130 dB, not the theoretical 146 dB.
| Bit depth | Theoretical dynamic range | Typical use |
|---|---|---|
| 16-bit integer | About 98 dB | CD, final distribution |
| 24-bit integer | About 146 dB | Recording and mastering |
| 32-bit float | Very large, effectively no internal clipping | Mixing engines and field recorders |
| 64-bit float | Even larger | Internal plugin and DAW processing |
The uncompressed data rate equals sample rate × bit depth × channels. Stereo CD audio therefore runs at 1,411.2 kbps, and a stereo 48 kHz, 24-bit stream needs 2,304 kbps, about 17 megabytes per minute.
PCM variants and containers
Plain PCM as used in audio files is properly called linear PCM (LPCM), because its quantization levels are evenly spaced. Other variants change the spacing or encode differences between samples to save bandwidth.
| Variant | Principle | Typical use |
|---|---|---|
| Linear PCM | Uniform quantization levels | WAV, AIFF, CD, Blu-ray, studio work |
| A-law and μ-law | Logarithmic companding to 8 bits | Telephony, G.711 voice |
| DPCM and ADPCM | Encode differences or predictions | Low-bandwidth voice, older games |
| Floating-point PCM | Mantissa and exponent per sample | DAW processing and 32-bit float recording |
PCM is a representation, not a file format. WAV, Broadcast WAV and AIFF are containers that store PCM samples with a header describing rate, depth and channel layout. The same PCM data also travels over AES3, S/PDIF, HDMI and USB audio interfaces.
PCM compared with compressed and alternative formats
Every playback chain ends in PCM, so the useful comparison is how the audio is stored or transmitted before decoding. Lossless codecs reproduce the original PCM exactly, while lossy codecs reconstruct a different PCM signal that is designed to sound similar.
| Representation | Relation to PCM | Data rate | Exactness |
|---|---|---|---|
| Uncompressed PCM | Stored directly | Highest | Bit-exact |
| Lossless codecs | Compressed, decoded to identical PCM | About 40 to 70 percent of PCM | Bit-exact |
| Lossy codecs | Perceptually coded, decoded to similar PCM | About 5 to 20 percent of PCM | Not bit-exact |
| DSD | 1-bit delta-sigma stream, not PCM | About 2.8 Mbps per channel for DSD64 | Usually converted to PCM for editing |
Practical choices and common mistakes
Choosing PCM settings is a balance between headroom, compatibility and storage. Recording at a higher resolution than the final format is common practice because it leaves room for gain changes and processing without accumulating noise.
- Record at 24-bit or 32-bit float, then reduce to 16-bit only at the final export with dither.
- Match the project rate to the delivery, using 48 kHz for video and 44.1 kHz or 48 kHz for music.
- Avoid repeated sample rate conversions, since each one adds filtering and potential artifacts.
- Higher sample rates do not mean better sound by default; above about 48 kHz the benefits are mainly for processing headroom.
- Truncating bit depth without dither produces correlated distortion on quiet passages and fades.
Explained without jargon
PCM is the way a computer writes down sound as numbers. Many thousands of times every second, it measures how strong the sound wave is and stores that measurement.
Played back in order, those numbers rebuild the sound, much like the frames of a film create smooth movement. More measurements per second capture higher pitches, and more precise measurements capture quieter details without hiss.
A typical use case
A video production company records interviews for a documentary on a portable recorder set to 48 kHz, 32-bit float PCM in Broadcast WAV files. The 32-bit float format protects against clipping when a speaker suddenly laughs or raises their voice, so the sound recordist can set conservative levels.
In the edit, the editor works at the same 48 kHz rate to avoid conversion. After mixing, the final soundtrack is exported as 48 kHz, 24-bit PCM for the broadcast master, while a separate lossy stream is encoded from that PCM master for online platforms. The interviews remain free of clipping and audible noise, and no quality is lost between recording and the master.