Pulse Code Modulation (PCM): Complete Beginner's Guide

Pulse Code Modulation (PCM) is the recipe that turned the telephone network digital: sample the voice, quantize each sample, and encode every quantized value as a binary number. Those 1s and 0s can then travel over any digital link, be regenerated perfectly at each repeater, and arrive at the far end as clean as they left.

PCM matters because it is the grandfather of all digital media. The same three steps — sample, quantize, encode — sit inside CDs, WAV files, digital TV, and every VoIP call you make. Understand PCM and you understand the skeleton of the entire digital world.

Key Concepts

• Sampling: voice is band-limited to ~4 kHz and sampled at 8 kHz (the Nyquist rate).
• Quantization: each sample is rounded to one of 256 levels (8 bits).
• Encoding: each level becomes an 8-bit binary word, sent serially.
• Bit rate: n bits × fs samples/s — the speed of the digital stream.
• Companding (A-law/μ-law): a clever trick that quantizes quiet sounds finely and loud sounds coarsely, matching how ears hear — used in every phone network.
• Regenerative repeater: re-reads the bits and re-transmits fresh pulses, so noise never accumulates — PCM's superpower over analog.
• Frame: one complete set of samples from all channels sharing a TDM link.

Formulas

• Bit rate: Rb = n·fs bits/s — n bits per sample, fs samples per second.
• Levels: L = 2ⁿ; telephone PCM uses n = 8, so 256 levels.
• Bandwidth need: roughly Rb/2 Hz minimum (Nyquist) for the baseband channel.
• SQNR with companding: telephone-quality ≈ 38 dB — as good as 12-bit linear quantization, from only 8 bits.
• Assumptions: uniform sampling at the Nyquist rate, and the 64 kbps figure assumes 8 kHz × 8 bits with no extra framing overhead.

Worked Example

Build one telephone voice channel. Step 1: sample the 4 kHz voice band at fs = 8 kHz. Step 2: quantize with n = 8 bits (μ-law companded). Step 3: bit rate Rb = 8 × 8000 = 64,000 bits/s = 64 kbps — the famous DS0, the atom of the phone network. Step 4: bundle 24 such channels with TDM framing (one extra bit per frame): 24 × 8 + 1 = 193 bits per frame, 8000 frames/s → 193 × 8000 = 1.544 Mbps — the T1 line that carried business calls for decades. Step 5: at the far end, a DAC turns the bits back into the original voice. Sixty-four kilobits per second, and your grandmother's voice crosses continents unchanged.

Common Misconceptions

• "Digital voice is always worse than analog." Wrong: PCM with proper sampling is transparent to the ear, and unlike analog it does not collect hiss at every repeater.
• "More bits per sample is the only way to improve quality." Not quite: companding squeezes 12-bit quality out of 8 bits by spending precision where ears are sensitive.
• "PCM is outdated." Wrong: every modern codec (MP3, Opus, AAC) starts with PCM samples — compression is just PCM with the redundancy squeezed out afterwards.

Real-World Applications

• The global telephone network was rebuilt on 64 kbps PCM channels (T1/E1 hierarchies).
• Audio CDs store 44.1 kHz × 16-bit PCM — "PCM" is literally what a WAV file contains.
• VoIP systems (WhatsApp calls, Zoom) packetize PCM-like samples, usually after compression.
• Broadcast and studio chains move audio as PCM over AES/EBU and SDI links.

Sample it, round it, write it in binary — PCM's three steps are almost insultingly simple, yet they digitised the world's voice, music, and video. Every fancier codec is just PCM wearing a compression costume.