Pulse-code modulation, or PCM, represents a sampled analog signal as a sequence of binary code words.

The process has three conceptual stages:

1. sample the continuous-time signal;
2. quantize each sample to one of $L$ amplitude levels;
3. encode the selected level as bits.

If

$$L=2^b,$$

then each sample requires $b$ bits. With sampling frequency $f_s$, the uncoded PCM bit rate is

$$R_b=b f_s.$$

Sampling error and quantization error are different. A sufficiently high sampling rate can prevent aliasing for a band-limited input, while increasing the number of quantization bits reduces amplitude resolution error.

PCM therefore converts a waveform problem into a digital-data problem. Once samples are represented as bits, source coding, error-control coding and digital modulation can be applied without needing to preserve the original analog amplitude directly.