Audio Mastering Guide: LUFS, True Peak, Limiting, Dither and Export.

Audio mastering is the final stage of technical refinement and perceptual optimisation before a mix is delivered to streaming services, physical formats or large PA systems in clubs and festivals.

It cannot rescue a fundamentally broken mix. Its job is to make controlled, final decisions about tonal balance, dynamics, loudness, peak margin, sequencing and delivery quality.

---

1. Mastering processing chain

A mastering chain usually uses small, deliberate moves rather than dramatic sound-design changes:

[Final mix] -> [Linear-phase / M-S EQ] -> [Bus compressor] -> [Dynamic EQ / multiband]
                                                                        |
[WAV deliverable] <- [Dither] <- [True Peak limiter] <- [Clipper] <-----+

The exact order is not fixed. A mastering engineer may move processors, use parallel paths or omit entire stages depending on what the mix needs.

1.1 Master EQ — linear phase and Mid/Side

• Linear-phase EQ: Can preserve relative phase through an EQ move, but it may introduce pre-ringing and latency. It is a tool, not an automatic upgrade over minimum-phase EQ.
• Mid/Side processing: Separates the sum and difference components of a stereo signal.
  • Mid: Often contains kick, bass, lead vocal and other centre information.
  • Side: Contains stereo difference information and can be shaped carefully to manage width and brightness.

A high-shelf lift above 10 kHz on the Side channel is a common example, but it must not weaken mono compatibility or exaggerate noise and harshness.

1.2 Master compression and multiband compression

• Subtle bus compression: Low ratios such as 1.2:1 to 1.5:1, slower attacks and programme-dependent release can reduce small dynamic differences while preserving transients.
• Multiband compression: Splits the signal into frequency bands so low-frequency peaks do not trigger the same gain reduction in the mid and high bands.

Multiband processing can solve a narrow problem, but it can also change the mix balance quickly. Use it only when the crossover and gain-reduction behaviour are understood.

1.3 Dynamic EQ, saturation and clipping

• Dynamic EQ: Reduces or boosts a specific frequency only when its detector crosses a threshold. It is useful for intermittent harshness or resonances.
• Saturation: Adds harmonics and may increase perceived density.
• Clipping: Removes the fastest peaks before the final limiter. A controlled clipper can reduce limiter workload, but excessive clipping produces audible distortion and weakens transients.

---

2. Loudness measurement — LUFS and True Peak ceiling

Modern loudness workflows use ITU-R BS.1770 and related delivery recommendations to measure perceived programme loudness alongside peak level.

  0 dBFS  ----------------------------- maximum digital sample level
 -1 dBTP  ----------------------------- example True Peak ceiling
 -6 LUFS  ----------------------------- very dense example master
-14 LUFS  ----------------------------- common normalisation reference example

These values describe different things. LUFS measures loudness over time; dBTP estimates reconstructed peak level. A loudness target is not a peak ceiling.

2.1 Loudness units — LUFS / LKFS

• Integrated LUFS — LUFS I: Average loudness over the measured programme.
• Short-Term LUFS — LUFS ST: Loudness over a moving three-second window, useful for comparing sections such as a verse and drop.
• Momentary LUFS — LUFS M: A faster 400 ms view of loudness changes.

2.2 True Peak and inter-sample peaks

When a digital signal is reconstructed by a DAC or encoded to a lossy format, the continuous waveform between stored samples can exceed the highest sample value.

• True Peak limiters: Use oversampling to estimate these reconstructed peaks and keep them below a chosen ceiling.
• Delivery margin: A ceiling around -1 to -2 dBTP is commonly used for lossy streaming delivery because it leaves room for transcoding and playback variation. The required value still depends on the platform specification.

---

3. Platform normalisation versus club masters and DJ edits

Platform / environment	Example integrated loudness	Example True Peak ceiling	Delivery approach
Streaming services	Often around -14 to -10 LUFS, depending on programme and strategy	Commonly -1 to -2 dBTP	Preserve dynamics and leave transcoding margin.
YouTube Music	Normalisation often discussed around -14 LUFS	Around -1 dBTP is a common delivery choice	Louder uploads may be turned down.
Club / DJ playback	Frequently denser, sometimes around -8 to -5 LUFS	Often closer to 0 dBTP, depending on the master	Prioritise impact, but check distortion and translation on a PA.
DJ edits / bootlegs	Often matched to the surrounding DJ library	Depends on source quality and intended use	Aim for consistent perceived level without crushing every transient.

These are working references, not universal standards. Platforms change behaviour, and club music is not governed by a single loudness target.

---

4. Dither, export, quality control and stem mastering

4.1 Applying dither

Dither is low-level random noise added before fixed-point bit-depth reduction. It decorrelates quantisation error so very low-level signals degrade into noise rather than structured distortion.

• Core rule: Apply dither once, as the final processing step, when reducing to the target fixed-point bit depth.
• Do not add it automatically: If the deliverable remains 24-bit and no bit-depth reduction occurs, another dither stage may be unnecessary.

4.2 Quality control and referencing

Before final delivery:

1. Listen on contrasting systems: studio headphones, nearfields, a mono phone speaker and a car system.
2. Check for unwanted distortion, clicks, phase problems and encoding artefacts.
3. Compare against suitable references at matched loudness. A louder reference will otherwise appear better even when its tonal balance is worse.
4. Confirm metadata, file length, channel format, sample rate, bit depth and start/end silence.

4.3 Stem mastering

In stem mastering, the engineer receives several subgroups — for example drums, bass, synths and vocals — instead of only the stereo mix. This allows limited balance and processing changes without reopening the full session.

The extra control comes with extra responsibility: if stems do not sum exactly to the approved mix, the mastering stage can turn into a remix.

4.4 Introduction to Dolby Atmos and spatial audio

Immersive mastering manages beds and objects in a three-dimensional delivery format such as 7.1.4. It requires dedicated monitoring, loudness and True Peak compliance, metadata validation and a binaural check for headphone playback.

A value such as -18 LUFS may appear in specific delivery workflows, but the final requirement must come from the current platform or distributor specification rather than from a generic stereo-mastering rule.