Loudness normalization for DJs: a LUFS target or a true peak ceiling?

Integrated loudness is the loudness of a whole track as one number, in LUFS (loudness units relative to full scale). ITU-R BS.1770 defines how it is measured: the signal is weighted to follow the ear's sensitivity and averaged over the whole track, with quiet passages gated out. EBU R128 uses the same measurement for broadcast with a target of -23 LUFS, and streaming services turn tracks up or down at playback to targets of their own. For DJ use the target is set far louder than broadcast's, nearer the level club tracks are mastered at.

Every track then gets the gain that brings its integrated loudness to the target. Loud masters come down and quiet ones come up. The exception is a dynamic master, whose peaks sit far above its average level: the gain it needs would push its peaks over full scale. Such a track takes one of two routes. Either a limiter, or another kind of peak control, takes its peaks down by the amount they would go over, so it can reach the target. Or it stops where its peaks allow and stays below the target.

The result: perceived loudness is matched closely across the library, and the trim knob mostly stays where it is from one track to the next. The cost is peak control on the dynamic tracks, whose highest peaks change by the amount it takes off. The louder the target, the more tracks need it; a quieter target needs it on fewer tracks and leaves the whole library quieter.

The weak point, once the gain is written into the file: a file has nothing above full scale. With a ceiling of -0.5 dBTP, a track whose peak-to-loudness ratio is larger than the gap between the target and the ceiling cannot reach the target by gain alone: above 9.5 dB at -10 LUFS, above 11.5 dB at -12 LUFS. The gain it needs would push its peaks over full scale, and written into a 16- or 24-bit file those peaks would clip. So every such track is either peak-controlled, which takes its loudest transients down (and an MP3 or AAC has to be decoded and encoded again for that), or left below the target, which gives up on that track the matching the approach is for. These are mostly the dynamic and the older masters, the tracks a gain change should help most. How many there are depends on the library and the target. Overs are common to begin with: five of the six commercial files measured below had true peaks above 0 dBTP as sold.

True peak is the highest level a track's waveform actually reaches, including peaks that fall between two samples, which a sample-peak meter misses. BS.1770 Annex 2 measures it by oversampling, in dBTP. This approach works out, for every track, the single gain that puts its true peak on one ceiling, and applies that gain and nothing else.

Bake'n Deck's Headroom works this way. The ceiling is -0.5 dBTP by default, the figure AES TD1008 §7B gives for high-rate codecs, and it can be changed. Quiet tracks are raised, and tracks whose peaks sit above the ceiling are lowered. No limiter, compression or EQ is ever applied. FLAC, AIFF, WAV and ALAC get the exact gain. MP3 and AAC are adjusted by rewriting their global_gain field in 1.5 dB steps, with no re-encoding, so they land at most one step below the ceiling.

The result: every track uses all the headroom it has. Older masters, CD rips and vinyl recordings that were left with spare headroom come up to full level, and the dynamics of every track stay exactly as mastered. What remains are the loudness differences that come from how dense the masters are. Two tracks with the same true peak can be several dB apart in loudness, because a dense master stays close to its peak all the way through and a dynamic one reaches it only on transients. A DJ hears that difference and handles it with the trim or the fader, as part of mixing.

Bake'n Deck takes this approach on purpose, and its line is simple: leave the trim at 12 o'clock and no track goes over 0 dB. Every file's true peak sits at or under the ceiling, by construction and without a limiter, so nothing on the stick drives a channel into clipping on its own. A gain change does the most for files that never used their headroom, often older releases and self-made rips, because it can bring them all the way up, and those are where most of the trim-knob work on a CDJ setup comes from. Differences in density stay, because only peak control could remove them, and they are left to your ears. The measurement itself is described in How Headroom measures loudness and true peak.

Illustrative numbers for three lossless files: a dense club master, a dynamic master and a quiet older master. The LUFS target is -10 LUFS, and both approaches keep true peaks at or under -0.5 dBTP.

Both approaches bring the older master C up by 6 dB, and C is the track that would otherwise need the most trim. They differ on A and B. The LUFS target brings A down and B up, with peak control on B. The true peak ceiling leaves A and B 4.9 dB apart, which is exactly the difference between their peak-to-loudness ratios, and changes nothing on either apart from 0.3 dB of gain.

Neither one is right for every library. If you want a stick that sounds even without touching the trim, the LUFS target is built for exactly that. If you want every master left as it was released and every track at its full level, with the remaining differences left to your ears, the true peak ceiling is.

rekordbox has its own Auto Gain. The rekordbox 7.2.18 Instruction Manual describes it in three places:

Preferences, Controller category, Mixer tab: “[Auto Gain] Select [Enable] to automatically adjust the audio volume of the track loaded to a deck. You can also adjust it manually using the [Auto Gain] knob on the [GRID EDIT].” (p. 242)

The GRID EDIT panel, PERFORMANCE mode: “Display and change the Auto Gain value detected by track analysis.” (p. 140)

EXPORT mode, DUAL PLAYER layout: “[AUTO GAIN] loads the attenuation amount set when analyzing the track, and can be adjusted manually. The settings for this function are saved for the track.” (p. 122)

So Auto Gain is one volume value per track. It is set when the track is analysed, can be changed by hand, is saved with the track in rekordbox's library and is applied on rekordbox's own decks. rekordbox does not document what the analysis measures or what level it aims for. The manual mentions neither LUFS, loudness nor true peak, lists no reference level setting for Auto Gain, and does not say whether the value can raise a track or only lower it; it calls the value an attenuation amount. On Pioneer DJ's forum, staff described it in 2017 as trying to “guesstimate the best setting”, and in 2019 answered a question about setting a library-wide normalization level with “Unlikely - that's what the gain knob is for.”

What the values show, measured. We read the Auto Gain values rekordbox 7 set for six commercial tracks after analysis and measured the same files with Headroom (BS.1770-4) on 2026-10-02. Every value equals about -10 LUFS minus the track's integrated loudness, to within 0.1 dB:

So Auto Gain is loudness normalization to about -10 LUFS, the first approach on this page, applied at playback. It raises tracks as well as lowering them, whatever the manual's "attenuation amount" suggests, and it does not hold back for peaks: track 3, whose true peak was -0.3 dBTP, comes up 3.5 dB to about +3.2 dBTP, and three of the six end above 0 dBTP.

rekordbox can do that with gain alone because it never writes the value into a file. The overs are caught at its output instead. [Output Level], in the same Mixer tab, lowers the output of every deck (-3 to -21 dB, or NONE, with the internal mixer), and the manual adds: "If the level is too high, clipping is likely to occur when applying FX or mixing. Clipping will enable the limiter to prevent distortion, but at the same time it will spoil the attack of a sound." (p. 242). A file on a USB stick has no such stage after it, which is the weak point of a LUFS target written into the file, and the reason Bake'n Deck writes a true peak ceiling instead.

Where the value lives decides what a CDJ gets. It is saved in rekordbox's library, not written into the audio file, and Pioneer DJ staff wrote in 2019: “Auto gain data is not stored on the export as there are no devices which support it.” A CDJ playing a USB stick plays each file at the level it has. Both approaches on this page differ from Auto Gain in that one respect: they write the gain into the file, so it reaches the deck. rekordbox Auto Gain: why it does not reach your CDJ has the details.

What is the difference between LUFS normalization and true peak normalization?

LUFS normalization sets every track to the same integrated loudness, so tracks sound about equally loud. True peak normalization sets every track's highest peak to the same level, so every track uses all of its headroom. Both use ITU-R BS.1770 measurements. Dense and dynamic masters have different gaps between peak and loudness, so the two give different results on the same library.

Does normalizing to a LUFS target need a limiter?

Written into the file, yes, for every track whose peak-to-loudness ratio is larger than the gap between the target and the ceiling: with a -0.5 dBTP ceiling, above 9.5 dB at -10 LUFS and above 11.5 dB at -12 LUFS. Without peak control those tracks either clip or stay below the target, and the louder the target, the more tracks that applies to. rekordbox's Auto Gain is applied at playback, so its output level setting and a limiter can catch the overs.

With the trim at 12 o'clock, can a track go over 0 dB?

Not with files levelled to a true peak ceiling: every true peak sits at or under the ceiling, -0.5 dBTP by default. With a LUFS target written into the file, a track whose peaks would go over is either peak-controlled or left below the target; with neither, it clips. And many files have true peaks above 0 dBTP as sold.

Will a true peak ceiling make every track equally loud?

No. Every track ends up with the same true peak, but at the same peak a dense master sounds louder than a dynamic one, sometimes by several dB. Those differences stay, and you handle them with the trim or the fader. Quiet files that had spare headroom do come up to full level.

Which approach keeps the trim knob still?

A LUFS target, most of all. A true peak ceiling removes the trim moves that come from quiet files with unused headroom and leaves the ones that come from differences in density.

Does rekordbox Auto Gain measure LUFS?

rekordbox does not document it, but it behaves that way. Measured on six tracks, each Auto Gain value equals about -10 LUFS minus the track's integrated loudness, to within 0.1 dB, and it raises tracks as well as lowering them. The manual calls the value an attenuation amount set when the track is analysed and mentions neither LUFS nor true peak.

Does rekordbox Auto Gain reach a CDJ?

No. The value is saved with the track in rekordbox's library and is not written into the audio file, and Pioneer DJ staff have written that Auto Gain data is not stored on the export. A CDJ reading a USB stick plays the level that is in the file.

I play from rekordbox too. What happens to Auto Gain after the files change?

Its stored value still describes the old level, because it was set when the track was analysed. Re-analyse the tracks, or turn Auto Gain off in Preferences under Controller, Mixer, to hear the level in the files. That is also the level a CDJ plays.

Are MP3 and AAC files re-encoded?

A gain change alone can be written into the global_gain field of an MP3 or AAC file in 1.5 dB steps, with no re-encoding, and that is all Headroom ever does to them. Peak control changes the samples themselves, so applying it to an MP3 or AAC file means decoding it and encoding it again.

1. ITU-R BS.1770: how integrated loudness and true peak are measured; Headroom measures to BS.1770-4
2. EBU R128: the broadcast loudness target, -23 LUFS
3. AES TD1008: the -0.5 dBTP figure in §7B
4. rekordbox 7.2.18 Instruction Manual: pp. 122, 140 and 242
5. Pioneer DJ Forums, “Auto Gain - confused”: official staff comments of 2017-10-16 and 2019-04-14
6. Pioneer DJ Forums, “Autogain / Normalisation”: official staff comment of 2019-12-17
7. How Headroom measures loudness and true peak: method and parity numbers