Waveform Audio

RIFF Container Structure

WAV uses the RIFF (Resource Interchange File Format) container, shared with WebP and AVI. The file begins with a 12-byte master header: the ASCII string RIFF, a 32-bit little-endian size value covering the rest of the file, and the WAVE FourCC identifier. After this header, data is organized into chunks, each with a 4-byte ID, a 4-byte size, and its payload.

Two chunks are mandatory. The fmt chunk (with trailing space) describes the audio format: a 16-bit format tag (0x0001 for PCM, 0x0003 for IEEE float), channel count, sample rate, byte rate, block alignment, and bits per sample. The data chunk contains the raw audio samples, interleaved by channel in little-endian byte order.

Optional chunks can appear in any order after fmt . The fact chunk stores the total sample count for non-PCM formats. LIST chunks with INFO sub-chunks carry metadata — IART for artist, INAM for title, ICRD for creation date. These fields are limited to ASCII text and a few hundred bytes each.

PCM Encoding: The Math

The byte rate of uncompressed PCM audio is deterministic: sample rate x bit depth x channels / 8. For CD-quality audio: 44,100 x 16 x 2 / 8 = 176,400 bytes per second. Multiply by 60 for the per-minute figure: 10,584,000 bytes, or 10.09 MB. A three-minute song occupies 30.27 MB. A 90-minute album fills 908 MB.

This predictability is both a strength and a constraint. File sizes are large but never surprising. There is no variable bitrate, no quality parameter, no encoder choice. The audio data is a direct numerical record of the waveform.

Bit Depth Comparison

16-bit integer PCM provides 96 dB of dynamic range with 65,536 discrete amplitude levels. This is sufficient for final delivery formats — CD, broadcast, and most streaming services.

24-bit integer PCM extends dynamic range to 144 dB with 16.7 million amplitude levels. Recording studios use 24-bit to preserve headroom during capture. The extra 48 dB of range means quiet signals stay well above the noise floor even when levels are conservative.

32-bit float PCM uses IEEE 754 single-precision floating point. Dynamic range exceeds 1,500 dB in theory, but the practical advantage is different: floating point samples cannot clip. Values above 1.0 or below -1.0 are preserved exactly, allowing gain adjustments after recording without permanent distortion. Most DAWs process audio internally in 32-bit float regardless of the source format.

Dithering

Reducing bit depth — from 24-bit to 16-bit for CD mastering, for example — requires truncating or rounding the lower bits. Truncation introduces quantization distortion correlated with the signal. Dithering adds a small amount of noise before truncation, randomizing the quantization error and converting distortion into a faint, constant noise floor. Triangular probability density function (TPDF) dither is the standard for most conversions. Noise-shaped dither pushes the added noise into frequency ranges where human hearing is least sensitive, typically above 15 kHz.

Skipping dither when downsampling bit depth is a common mistake. The resulting artifacts are subtle on loud material but clearly audible in fade-outs and quiet passages as a gritty, metallic distortion.

Broadcast Wave Format (BWF)

BWF extends WAV with a bext chunk containing broadcast-specific metadata: description (256 ASCII characters), originator, originator reference, origination date and time, time reference (sample count from midnight, enabling frame-accurate synchronization), and a UMID (Unique Material Identifier). The coding history field records the signal chain — sample rate conversions, codec passes, and equipment used.

BWF is mandatory in many broadcast environments. The EBU (European Broadcasting Union) R68 and R128 standards reference it explicitly. BWF files are structurally identical to WAV files with the added bext chunk — any WAV player can read the audio, though non-BWF-aware software ignores the metadata.

iXML Metadata

Professional field recorders (Sound Devices, Zoom F-series) embed iXML chunks in WAV files. This XML-structured metadata includes take number, scene, tape ID, track names, circled take flags, GPS coordinates, and pre-roll timecode. Post-production workflows use iXML to automate media management. The iXML specification is maintained by the iXML consortium and has been through multiple revisions, though no formal standards body governs it.

RF64: Breaking the 4 GB Barrier

Standard WAV files use 32-bit size fields, capping the maximum file size at 4,294,967,295 bytes (just under 4 GB). At 24-bit, 96 kHz, stereo, this limit is reached after approximately 124 minutes. For long-form recording — concerts, location sound, scientific capture — this is insufficient.

RF64 replaces the RIFF header with RF64 and adds a ds64 chunk containing 64-bit size values. This extends the maximum to 16 exabytes. The audio data and all other chunks remain identical. Most professional recording software and hardware support RF64 transparently, switching from WAV to RF64 automatically when the 4 GB boundary approaches.

Limitations

Beyond the 4 GB size cap without RF64, WAV has several constraints. There is no built-in chapter or cue point standard widely supported across players — the cue chunk exists but implementation is inconsistent. Metadata support is primitive compared to ID3v2 in MP3 or Vorbis comments in FLAC. No native compression exists within the WAV container for PCM data. File integrity depends entirely on the RIFF size fields being correct; truncated files often fail to open rather than playing what is available.

When to Use WAV vs Alternatives

Use WAV for recording, editing, and mixing where maximum compatibility with audio hardware and DAW software is required. Use FLAC instead of WAV for archival storage — FLAC compresses PCM audio by 40-60% with bit-perfect reconstruction, saving significant disk space with no quality loss. Use AIFF when working in Apple-centric production environments, though AIFF and WAV are functionally equivalent for PCM audio with only byte-order and metadata differences. Use WAV over MP3 or AAC whenever the workflow involves further processing — lossy formats should only appear at the final delivery stage. For files exceeding 4 GB, verify that your toolchain supports RF64 before recording, or use FLAC which has no practical size limit.