Appearance
For audio preservation masters, capture lossless audio at a minimum of 48 kHz / 24-bit and store it as Broadcast WAV (BWF) or FLAC — both are bit-for-bit faithful, so the real choice is BWF's universal support versus FLAC's smaller size and built-in integrity checks. Never use MP3 or any lossy codec for masters; lossy formats permanently discard audio data and belong only to access copies. Generate lossy derivatives from the lossless master, never the other way round.
Which lossless format should the master use?
The contest is between Broadcast WAV and FLAC. Both reconstruct the original samples exactly, so quality is identical; the differences are operational.
| Factor | Broadcast WAV (BWF) | FLAC |
|---|---|---|
| Lossless | Yes | Yes |
| Typical size | 100% | 40-60% smaller |
| Embedded metadata | bext chunk | Vorbis comments |
| Built-in checksums | No | Per-frame + MD5 of audio |
| 4 GB file-size limit | Yes (classic WAV) | No |
| Tool ubiquity | Very high | High |
BWF is the long-standing default. FLAC has gained acceptance as a preservation master because it halves storage and carries internal integrity data; the classic WAV 4 GB limit also makes FLAC attractive for very long recordings (RF64/BW64 lifts the WAV limit but support is narrower).
What sample rate and bit depth do I capture at?
Follow IASA TC-04: a minimum of 48 kHz / 24-bit for analogue sources, with 96 kHz / 24-bit common for music, fragile media, or material you may analyse closely. Bit depth governs dynamic range and noise floor; 24-bit gives headroom that 16-bit does not. You can always downsample later for access copies, but you can never recover resolution you failed to capture, so err toward the higher justified setting.
How do I encode and verify a FLAC master?
Encode with maximum compression and built-in verification so the encoder proves losslessness as it works:
bash
flac --best --verify --keep-foreign-metadata \
-o master.flac source.wav
# --verify re-decodes during encoding and aborts on any mismatchTo independently confirm later, decode back to WAV and compare audio data:
bash
flac -d master.flac -o roundtrip.wav
# compare audio MD5 stored in the FLAC against the source
metaflac --show-md5sum master.flacIf the stored MD5 matches the source audio's MD5, the FLAC is a perfect copy.
What metadata and fixity should I attach?
Embed structured provenance and keep an external checksum. For BWF, the bext chunk records originator, origination date and coding history (the chain of equipment and conversions). For FLAC, use Vorbis comments. In both cases store a separate SHA-256 for fixity:
bash
sha256sum master.flac >> fixity.txtFLAC's internal per-frame checksums catch corruption during playback, but an external fixity record is what lets your preservation system detect silent bit-rot over years.
What pitfalls trip people up?
A few mistakes recur:
- Encoding the master as MP3 — irreversible loss; masters must be lossless.
- Capturing at 16-bit when the source warranted 24-bit headroom.
- Hitting the 4 GB WAV limit mid-recording — use FLAC or BW64 for long content.
- Forgetting coding history — without it, future curators cannot judge provenance.
- Normalising or applying noise reduction to the master — keep the master raw; process derivatives instead.
Key Takeaways
- Masters must be lossless: choose Broadcast WAV (BWF) or FLAC; never MP3.
- Capture at least 48 kHz / 24-bit (often 96 kHz / 24-bit for music and fragile media).
- FLAC saves 40-60% storage, avoids the 4 GB WAV limit, and carries internal checksums.
- Encode FLAC with
--verifyand confirm the stored audio MD5 matches the source. - Embed provenance (BWF bext / FLAC Vorbis comments) and keep an external SHA-256.
- Keep the master unprocessed; apply noise reduction and downsampling only to access derivatives.
Frequently Asked Questions
Is WAV or FLAC better for audio preservation?
Both are lossless and equally faithful. Broadcast WAV (BWF) is the traditional preservation master because of universal support; FLAC is lossless too and roughly 40-60% smaller, with built-in checksums. Many archives now accept FLAC as a preservation master.
What sample rate and bit depth should preservation masters use?
The IASA TC-04 guidance for analogue sources is a minimum of 48 kHz / 24-bit, with 96 kHz / 24-bit common for music and fragile originals. Capture at the highest justified setting; you cannot recover detail you did not digitise.
Is MP3 ever acceptable for preservation?
No. MP3 is lossy and discards audio data permanently, so it is unsuitable as a preservation master. MP3 or AAC are fine only for access/streaming derivatives generated from the lossless master.
Why use Broadcast WAV (BWF) instead of plain WAV?
BWF extends WAV with a bext chunk that carries structured metadata such as originator, date and coding history, which plain WAV cannot hold. This embedded provenance is valuable for preservation and is the reason BWF is preferred.
How do I verify a lossless audio conversion?
Decode the FLAC back to WAV and compare, or use FLAC's built-in verify. Running flac with the verify flag re-decodes during encoding and aborts on any mismatch, guaranteeing the FLAC is a perfect copy of the source.
What about embedded metadata and checksums?
Embed structured metadata (BWF bext or FLAC Vorbis comments) and store a separate fixity checksum such as MD5 or SHA-256. FLAC additionally stores per-frame and whole-stream checksums internally, helping detect corruption early.