The loudness war: how it changed mastering
Last updated 6 min readTags:MasteringLoudnessDynamics
What the loudness war changed, why normalisation reduces the advantage of louder masters, and how to judge loudness without losing impact.
The loudness war is the competition to make recordings play louder than other releases at the same volume setting. Engineers reduce peaks with compression, limiting or clipping, then raise the overall level. Taken too far, that process can weaken drum attacks and introduce distortion.
Loudness normalisation reduces the advantage of this approach: a louder master is turned down during playback. It does not undo the processing. For your own master, the useful question is whether extra limiting improves the sound once both versions play at a comparable loudness.
Why records became louder
At a fixed playback volume, a louder recording can seem more impressive in a quick comparison. That gave artists, labels and engineers a reason to ask for louder releases. Digital mastering made increasingly high average levels possible, even though the maximum sample level remained fixed.
Simply turning up a recording eventually causes clipping. A limiter reduces the highest peaks so the rest of the signal can be raised without exceeding its ceiling. Repeated across competing releases, this practice increased average levels and reduced the difference between peaks and the body of the music.
Emmanuel Deruty's 2011 analysis of 4,500 tracks found that crest factor fell by about 3 dB from the early 1980s. This is evidence about a historical sample, rather than a figure that describes every genre or today's releases.[1]
What excessive loudness processing can change
Peak reduction changes more than a meter reading. Depending on the material and settings, you may hear:
- Weaker attacks: the initial hit of a kick or snare becomes less distinct from the sound that follows it.
- Pumping: the rest of the mix dips and recovers as loud hits trigger gain reduction.
- Distortion: bass notes become gritty, or vocals and cymbals acquire a rough edge.
- Less contrast within a passage: sustained sounds stay close in level to drum hits, so the rhythm feels less forceful.
Earl Vickers provides a documented example with two releases of Chris Isaak's “Wicked Game”. His comparison identifies bass clipping distortion in the chorus of the 2006 compilation version.[2]
Compression and clipping can also be deliberate parts of a production's sound. A dense master may suit the music. The problem arises when you prefer a version only because it plays louder, without noticing what the processing has changed.
Peak contrast and loudness range are different
The phrase “lost dynamics” can refer to several things. These measurements describe different aspects of a recording:
| Measurement | What it describes | What it cannot establish alone |
|---|---|---|
| Integrated LUFS | The track's overall measured loudness | Whether the processing sounds good |
| Crest factor | The difference between peak and RMS level | How much a chorus rises above a verse |
| Loudness range (LRA) | The spread of short-term loudness values across the recording | Whether individual drum attacks remain intact |
A song can retain a quiet verse and a loud chorus while its drum peaks are heavily limited. Conversely, a steady arrangement can have a low LRA without excessive compression. LRA describes variation over longer periods than individual transients; it is not a test for over-limiting.[3]
A low crest factor also needs context. The arrangement, sound selection and frequency balance affect it. Compare versions of the same track before treating a reading as evidence that the master needs changing.
What loudness normalisation changes
Loudness normalisation adjusts playback gain according to measured loudness. For example, Spotify's Normal setting uses −14 LUFS. It applies a common gain adjustment during album playback and adjusts tracks individually in playlists or shuffle playback.[4]
For two individually normalised tracks, the arithmetic might look like this:
| Master | Integrated loudness | Gain adjustment to −14 LUFS | Result |
|---|---|---|---|
| A | −8 LUFS | −6 dB | −14 LUFS |
| B | −14 LUFS | 0 dB | −14 LUFS |
Both reach the same measured loudness. They need not sound identical in loudness or impact: their tonal balance, arrangement and moment-to-moment levels still differ. Any distortion or altered attacks in Master A remain after it is turned down.
The advantage also depends on playback conditions. Spotify states that its web player and third-party devices do not use normalisation. Listeners can also choose different playback levels. Where normalisation is absent or disabled, a louder file still plays louder at the same volume setting.[4]
Should you master to −14 LUFS?
There is no single LUFS value that suits every track. Spotify recommends −14 LUFS integrated, but a service's playback reference does not establish the best sound for your music. Mastering engineer Ian Stewart's iZotope guide likewise advises against treating −14 LUFS as a universal target.[4][5]
A louder master can be a valid choice if you want its density and processing character. A quieter one can be a valid choice if further limiting weakens the sound. Judge that choice at matched loudness, with suitable reference tracks for the genre.
How to judge the loudness of your master
- Make two versions. Keep your current master, then make a second with less final limiting. Keep the other processing unchanged so the comparison has a clear purpose.
- Match their playback loudness. Measure the same passage in both versions and reduce the louder version's monitoring level. Check the match by ear. A small level advantage can influence your preference.
- Listen to the busiest passage. Check whether the kick and snare still have distinct attacks, whether bass notes distort, and whether cymbals or vocals become rough. Listen for the whole mix dipping after strong hits.
- Check a transition. Play through a verse into a chorus. Decide whether the change still has the impact you intend; a whole-track LUFS figure cannot answer that.
- Measure the exported file. Check integrated LUFS and true peak after the final render. Spotify advises true peaks below −1 dBTP, or below −2 dBTP for masters louder than −14 LUFS, to reduce the risk of encoding distortion.[4]
If another engineer will master the track, agree on delivery requirements. Keep a version without final loudness maximisation, and send your limited version as a reference if it represents the sound you want.
You can analyse your mix with Sonalix to check its loudness and true peak before release. Use the readings alongside a level-matched comparison to decide whether the final processing has improved the track.
Sources
- Emmanuel Deruty: “'Dynamic Range' & The Loudness War”, Sound On Sound, September 2011
- Earl Vickers: “The Loudness War: Background, Speculation and Recommendations”, supporting material for the 2010 AES paper
- EBU Tech 3342: Loudness Range, 2011 edition
- Spotify for Artists: Loudness normalization
- iZotope: How to master for streaming platforms