Audio Loudness
& Delivery

01 — The loudness problem

For decades, broadcast audio was controlled only by peak level — the maximum instantaneous amplitude of the signal. The rule was simple: keep peaks below 0 dBFS. But a signal can have its peaks just touching 0 dBFS and still sound very quiet, if those peaks are brief and the programme content is sparse. Conversely, a heavily compressed signal can keep its peaks just below the limit while sounding dramatically louder than anything around it.

The result was loudness inflation: advertisers compressed their spots to sound louder than programmes; programmes were mixed louder to compensate; networks applied broadcast limiters; everything got louder together. By the early 2000s the difference in perceived loudness between the quietest and loudest content on air had become an audience complaint significant enough to prompt regulatory action in the United States and coordinated standards work across Europe and Australia.

The solution was to agree on a perceptual loudness measurement — a meter that approximates how the human ear experiences loudness over time, rather than just catching the highest peak. The result was the ITU-R BS.1770 algorithm, which underpins every major loudness standard in use today.

02 — How loudness is measured

ITU-R BS.1770 and frequency weighting

The ITU-R BS.1770 algorithm applies a K-weighting filter to the audio signal before measuring its level. The filter has two stages: a high-shelf pre-filter that boosts high frequencies (reflecting the increased sensitivity of the ear at higher frequencies) and a high-pass filter that rolls off sub-bass content. After filtering, the signal’s energy is summed across channels, with the surround channels weighted slightly lower than front channels.

The result is a measure of loudness in LUFS — Loudness Units relative to Full Scale. The same measurement is sometimes called LKFS (Loudness, K-weighted, relative to Full Scale) in the North American ATSC A/85 standard. The two terms describe exactly the same algorithm and produce exactly the same reading; the different names reflect which standards body published which document first.

Momentary, short-term, and integrated

A loudness meter typically displays three related readings simultaneously.

The integrated measurement uses gating to exclude silence and very quiet passages from the calculation. Two gates are applied: an absolute gate that discards any block below −70 LUFS, and a relative gate that discards any block more than 10 LU below the ungated mean. This prevents long periods of silence — credits, fade-outs — from dragging the integrated reading below target.

True peak vs sample peak

Digital audio is made up of discrete samples. A sample-peak meter reads the level at each sample point. But when a digital signal is converted to analogue — or is resampled to a different sample rate — the reconstruction process can produce peaks between the original samples that are higher than any individual sample value. These are called inter-sample peaks.

A true peak meter uses oversampling (typically 4× the original sample rate) to detect inter-sample peaks before they become a problem. Broadcast and streaming delivery standards universally specify a true peak limit — commonly −2 dBTP — rather than a sample peak limit. If your DAW only shows sample peaks, you may be delivering audio that clips on the platform’s decoders without knowing it.

03 — The standards

Several standards govern loudness measurement and delivery targets around the world. All are built on ITU-R BS.1770, but they apply slightly different targets and tolerances.

EBU R128

Published by the European Broadcasting Union, EBU R128 is the standard adopted by most European broadcasters and by ABC, SBS, and the Australian free-to-air networks. It targets:

ATSC A/85 (North America)

The Advanced Television Systems Committee standard, widely used in the United States. It targets −24 LKFS integrated loudness and −2 dBTP true peak — 1 LU louder than EBU R128’s target but with a slightly more conservative true peak. Free TV Australia aligns with the ATSC target of −24 LUFS rather than the EBU −23 LUFS.

Free TV Australia Operational Practices

Australian free-to-air broadcasters follow a loudness standard derived from ATSC A/85, specifying −24 LUFS integrated, −2 dBTP, and a maximum LRA of 15 LU. All five commercial networks and the national broadcasters are bound by this specification for linear broadcast delivery.

04 — Platform loudness targets

Different platforms apply loudness normalisation differently. Broadcast platforms require you to deliver audio already at the target level. Streaming video platforms typically accept audio louder than target and turn it down on ingest. Music streaming platforms normalise during playback and may or may not touch the delivered file at all. Understanding which applies determines how to approach your final mix.

A note on music vs programme loudness

The gap between broadcast/streaming video (−24 LUFS) and music streaming (−14 LUFS) is intentional. Television and film audio includes wide dynamic range — quiet dialogue, ambient room tone, loud effects — and the low target preserves that range. Music is typically more compressed and listeners expect higher sustained energy. If you are delivering the same audio to both broadcast television and a music platform, you will need separate mixes at different targets.

05 — Audio channel configurations

Broadcast and streaming delivery requires audio in specific channel configurations. The required tracks and their channel order are defined precisely — a mislabelled or reordered channel can result in dialogue in the wrong speaker or a phase-cancelled mix on the platform.

Stereo (2.0)

Two channels: Left (Ch 1) and Right (Ch 2). In broadcast MXF delivery this is typically the first stereo pair in the file. For streaming delivery without a surround mix, stereo is the minimum accepted configuration. Always deliver a proper stereo mix — not a downmix of a 5.1 sum — where stereo is the primary deliverable.

5.1 surround (6 channels)

Six channels in the ITU standard order: Left, Right, Centre, LFE, Left Surround, Right Surround. In broadcast MXF the channel order is typically L, R, C, LFE, Ls, Rs on channels 1–6. In a ProRes MOV file the channel order may vary between applications — verify in your DAW’s export settings. The LFE channel carries bass management information only; it is not the subwoofer output. LFE content should be appropriate for bass-enhanced reproduction, not simply a low-shelf of the full mix.

Dolby Atmos

Dolby Atmos is an object-based audio format that adds height channels and the concept of audio objects — sounds that can be positioned and moved in three-dimensional space around the listener, rather than being fixed to a specific speaker channel. A typical Atmos delivery includes a 7.1.2 bed (seven ground-plane channels, one LFE, two height channels) plus object metadata.

For broadcast and streaming delivery, Atmos is typically delivered as an ADM Broadcast WAV (Audio Definition Model) — a single polyphonic WAV file carrying both the audio beds and the object metadata. Within an IMF package, Atmos is delivered as an IAB track (Immersive Audio Bitstream). A 5.1 or stereo fallback must always be delivered alongside any Atmos deliverable.

06 — Technical delivery specs

Sample rate

48 kHz is the universal sample rate for broadcast and post-production audio. It is the only sample rate accepted for broadcast MXF delivery. 44.1 kHz is the compact disc standard and is used by music distribution platforms, but is not appropriate for picture delivery. Do not deliver 44.1 kHz audio in a broadcast or streaming video deliverable.

96 kHz and 192 kHz are high-resolution sample rates used during music recording for headroom in the recording process. For delivery, even to music streaming platforms, 44.1 kHz or 48 kHz at 24-bit is the accepted format. Delivering at a higher sample rate does not improve perceived quality for the listener and wastes file size.

Bit depth

24-bit is the delivery standard for all professional broadcast and streaming audio. It provides 144 dB of theoretical dynamic range — far more than is needed for any programme, but the additional headroom ensures that the dithering artefacts that appear at very low levels are inaudible. Never deliver 16-bit audio to a broadcaster or streaming platform; it is not accepted by any current broadcast specification.

32-bit float is used internally within DAWs for processing, as it provides virtually unlimited headroom during a session. It is not a delivery format — convert to 24-bit integer on export.

Audio codec for delivery

Broadcast and streaming mastering delivery requires uncompressed PCM audio. This is the audio format inside MXF OP1a files, ProRes MOV files, and Broadcast WAV files. The platform encodes the delivered PCM to its own delivery codec (AAC, Dolby Digital, Dolby Atmos, etc.) for consumer playback.

07 — M&E stems and deliverables

Long-form productions — feature films, episodic drama, documentary — typically require several audio deliverables beyond the final mix. These elements support international versioning, re-editing, and archiving.

M&E (Music and Effects)

The M&E track is the complete final mix minus dialogue — music, sound effects, and ambience are present, but all recorded and post-produced speech is removed. M&E is used by international distributors to replace the original-language dialogue with a dubbed version. A fully filled M&E (sometimes called a fully filled M&E or M&E with fills) must include replacement ambience and effects wherever dialogue was removed, so the dubbed version does not have audio holes under the replacement speech.

Stems

Stems are pre-mixed submixes of discrete elements of the final mix: typically Dialogue, Music, and Effects (DME). Each stem is mixed and balanced as it appears in the final programme, but delivered as a separate file. When all stems are summed together they must reproduce the final mix exactly. Stems allow a re-editor to adjust the balance of individual elements — raising effects under an action sequence, reducing music under a dialogue retake — without access to the original session.

Music cue sheets and session files

Most long-form streaming and broadcast deliveries require a music cue sheet alongside the audio files — a log of every piece of music used in the programme, including title, composer, publisher, timing in and out, and usage type (background, visual vocal, theme, etc.). This is a rights clearance requirement, not a technical one, but it is typically collected by the post facility and submitted with the delivery. Platforms also sometimes request the original mix session files (Pro Tools, Nuendo, or DaVinci Fairlight sessions) as an archive deliverable, particularly for Originals where the platform retains long-term rights. Confirm all non-audio deliverables with the commissioning platform before completing post-production.