Loudness

Loudness is how loud a sound seems to a listener, as opposed to how high its electrical or digital level is. Two tracks that peak at exactly the same level can sound very different in loudness, because the ear is more sensitive to some frequencies than others and responds to average energy over time, not to individual peaks.

In production, loudness is measured in LUFS (Loudness Units relative to Full Scale), the unit used by broadcast standards and by streaming services to normalize playback.

What is loudness?

Loudness is a perception. The American National Standard on acoustical terminology (ANSI/ASA S1.1) defines it as the attribute of auditory sensation by which sounds can be ordered on a scale from quiet to loud. That makes it different from the physical quantities engineers measure directly: voltage, sound pressure level in decibels, or digital sample values in dBFS.

In plain terms: a meter shows how much signal there is; loudness is how big that signal feels to your ears. A bass-heavy track can hit the top of a peak meter and still sound quieter than a bright vocal track at a lower level, because human hearing is far less sensitive to deep bass than to the midrange where speech lives.

Loudness vs level vs volume

• Level is a measured quantity: volts, dB SPL or dBFS.
• Volume usually means the setting of a playback control, or informally the level.
• Loudness is the perceived result, which depends on level, frequency content, duration and the listener.

The different meanings of “loudness”

• Psychoacoustics: the subjective magnitude of a sound, measured in phons and sones.
• Broadcast and streaming: a standardized measurement of program loudness in LUFS (or LKFS), used to match levels between programs and songs.
• Consumer hi-fi: the “loudness” button on older amplifiers, which boosts bass and treble at low volume to compensate for the ear’s reduced sensitivity to those frequencies when listening quietly.
• Music: dynamics in performance, written as piano, forte and so on.

The history of loudness measurement

Equal-loudness contours

The scientific study of loudness began with Harvey Fletcher and Wilden A. Munson at Bell Telephone Laboratories, who in 1933 published the first equal-loudness contours. By asking listeners to match tones of different frequencies to a reference, they showed how strongly hearing depends on frequency: we are most sensitive in the region of roughly 2–5 kHz, and much less sensitive to low bass, especially at low listening levels. The work of D.W. Robinson and R.S. Dadson in 1956 refined these Fletcher-Munson curves and led to the ISO 226 standard, most recently revised as ISO 226:2023.

Eberhard Zwicker developed a model, later refined with Hugo Fastl, that calculates loudness from the ear’s critical bands (the Bark scale). This approach, and a second model by Brian Moore and Brian Glasberg, are standardized in ISO 532.

The loudness war

Digital audio has a hard ceiling at 0 dBFS. From the 1990s onwards, the only way to make a CD sound louder than the competition was to reduce the gap between peaks and average level, using heavier and heavier limiting and clipping. This competition became known as the loudness war. Its cost was reduced punch and dynamic contrast, audible distortion on some records, and listener fatigue.

From peak meters to LUFS

Broadcasters had a parallel problem: programs and especially commercials jumped in loudness even when their peaks were controlled. The ITU developed an objective measurement, building on research by Gilbert Soulodre at Canada’s Communications Research Centre, and published it as ITU-R BS.1770 in 2006. It introduced K-weighting and the unit LKFS (loudness, K-weighted, relative to full scale). The current revision is BS.1770-5.

In 2010 the European Broadcasting Union published EBU R128, developed by its PLOUD group chaired by Florian Camerer. It adopted the BS.1770 measurement, popularized the name LUFS (identical to LKFS), and set a broadcast target of −23 LUFS. In the United States, the CALM Act passed the same year made the ATSC A/85 practice, with a −24 LKFS target, the basis for stopping commercials from being louder than programs.

Streaming normalization

Streaming services then applied the same idea to music: measure every track and play it back at a common loudness. Once a very loud master is simply turned down to match everything else, the competitive advantage of extreme loudness largely disappears. Spotify, Apple Music, YouTube, Tidal and Amazon Music all normalize playback today, although their targets and methods differ.

How loudness is measured

Phons and sones

The traditional psychoacoustic units are the phon and the sone. A sound’s loudness level in phons equals the level in dB SPL of a 1 kHz tone judged equally loud. The sone is a linear scale: 1 sone is the loudness of a 1 kHz tone at 40 dB SPL, and each increase of about 10 phons roughly doubles perceived loudness. That is the origin of the rule of thumb that +10 dB sounds about twice as loud.

LUFS: the BS.1770 algorithm

The LUFS measurement used in every modern loudness meter follows a fixed recipe:

1. K-weighting: the signal passes through a high-shelf filter (roughly +4 dB in the upper range, modeling the acoustic effect of the head) and a high-pass filter around 38 Hz that strongly reduces the lowest frequencies, reflecting how little they contribute to perceived loudness.
2. Mean square: the filtered signal is squared and averaged, giving a measure of power rather than peaks.
3. Channel weighting: in surround formats, left, right and center count at full weight, the surround channels are weighted about +1.5 dB, and the LFE channel is excluded.
4. Gating: for integrated loudness, the audio is measured in 400 ms blocks overlapping by 75%. An absolute gate ignores blocks below −70 LUFS (silence), and a relative gate ignores blocks more than 10 LU below the average of the rest, so quiet passages do not drag the result down.

Momentary, short-term and integrated

• Momentary loudness uses a 400 ms window and moves quickly.
• Short-term loudness uses a 3-second window, useful while mixing.
• Integrated loudness is the gated measurement over the whole track or program, the number delivery specs refer to.

True peak

A normal sample-peak meter reads the highest sample value, which can never exceed 0 dBFS. But when the samples are converted back to a continuous waveform, or encoded to a lossy format such as AAC or Ogg Vorbis, the reconstructed signal can overshoot between samples. True-peak meters estimate these inter-sample peaks by oversampling (BS.1770 describes 4× oversampling for 48 kHz audio) and read in dBTP. This is why delivery specs ask for a true-peak ceiling such as −1 dBTP rather than a sample-peak ceiling.

Loudness range (LRA)

Integrated loudness is a single number and says nothing about how much the loudness moves. Loudness Range, defined in EBU Tech 3342, measures that movement in LU: it takes the short-term loudness values, gates them (−70 LUFS absolute, −20 LU relative), and reports the spread between the 10th and 95th percentiles. A dense pop master may have an LRA of only a few LU; a film soundtrack or a classical recording can have much more. See dynamic range for how LRA relates to other dynamics measurements.

Loudness targets: streaming and broadcast

Targets are the levels at which platforms or broadcasters play content back, not rules for how loud your master must be. Platform behavior changes over time, so check the platform’s own documentation before an important release.

Platform or standard	Playback / delivery target	Notes
Spotify	−14 LUFS (Normal); −11 LUFS (Loud); −19 LUFS (Quiet)	Whole albums are normalized together when played as an album. Quieter masters are raised only as far as their peaks allow, keeping 1 dB of headroom; a limiter is used only in the Loud setting. Spotify recommends masters below −1 dBTP, or −2 dBTP for loud masters (Spotify for Artists).
Apple Music	−16 LUFS	Sound Check switched to LUFS-based measurement, following AES TD1008 (MeterPlugs).
YouTube, Tidal, Amazon Music	Around −14 LUFS (widely measured and reported)	Mostly turn loud content down rather than quiet content up; details are not always officially documented.
AES TD1008 (2021)	−16 LUFS for music, −18 LUFS for speech and mixed content	In album mode the loudest track may play somewhat louder than −16 LUFS. Recommendations for internet streaming and on-demand audio, including album normalization (summary).
EBU R128 (Europe)	−23 LUFS, maximum −1 dBTP	Broadcast television and radio.
ATSC A/85 (US)	−24 LKFS	US television, supporting the CALM Act.

Loudness in practice

Don’t master “to −14” just because Spotify plays at −14

The most common mistake in modern mastering is treating a playback target as an upload rule. If a track mastered at −8 LUFS is played on Spotify’s Normal setting, it is simply turned down by about 6 dB; nothing else happens to it. A track mastered to exactly −14 LUFS is left as it is. What matters is how each version sounds after normalization: the −8 LUFS master will have more density and less punch, the −14 LUFS master more dynamics and punch. Choose the balance that suits the music, not a number.

Typical integrated loudness by genre

These ranges describe what many commercial masters measure before platform normalization. They are observations, not targets.

Genre	Typical integrated loudness
EDM, dubstep, hardstyle	about −6 to −9 LUFS, sometimes louder (varies widely)
Pop, hip-hop, trap	about −7 to −9 LUFS
Rock and metal	about −8 to −11 LUFS
Jazz, classical, acoustic	about −12 to −18 LUFS or lower
Podcasts and speech (distribution)	about −16 to −19 LUFS; broadcast −23/−24

How engineers make things sound louder without just limiting

• Control the low end: sub-bass uses a lot of headroom but, because of K-weighting and human hearing, adds little perceived loudness. Tight, well-balanced bass leaves room for everything else.
• Build density in stages: gentle compression, saturation or clipping on individual tracks and buses, rather than one limiter doing 8 dB of gain reduction on the master.
• Use the midrange: the ear is most sensitive between roughly 2 and 5 kHz, so balance there has a big effect on perceived loudness.
• Leave true-peak headroom: a final ceiling around −1 dBTP (lower for very loud masters) protects against distortion when platforms encode to lossy formats.

Compare at matched loudness

Louder almost always sounds better in a quick comparison, which is why so many processing decisions are fooled by level. When you compare a plugin on and off, two mixes, or your master against a reference, match their loudness first. Most modern meters and many plugins include an automatic gain-match feature for exactly this reason.

Albums need a different approach

A quiet ballad on an album should not be pushed to the same integrated loudness as the loudest single. Album normalization, used by Spotify and recommended by AES TD1008, keeps the level relationships between songs, so set the song-to-song balance by ear while mastering the album.

Loudness meters in your DAW

DAW	Built-in loudness metering
Logic Pro	Loudness Meter plugin
Cubase / Nuendo	SuperVision, with loudness modules
Pro Tools	Loudness metering in the Pro Limiter plugin
Fender Studio Pro (formerly Studio One)	Level Meter with LUFS and EBU R128 modes
Reaper	JS: Loudness Meter Peak/RMS/LUFS
Ableton Live	No dedicated LUFS meter; use a third-party or Max for Live meter
FL Studio	Wave Candy has a BS.1770 loudness mode; use a third-party meter for true peak and LRA

Popular third-party meters include Youlean Loudness Meter (free and paid versions), iZotope Insight, NUGEN Audio VisLM and Waves WLM Plus. Limiters such as FabFilter Pro-L 2 include loudness and true-peak metering, and tools such as MeterPlugs’ Loudness Penalty preview how much each platform will turn a master down.

The loudness war: documented examples

• Metallica, Death Magnetic (2008): widely criticized for its extremely compressed and clipped CD master. Fans noticed that the versions released for the video game Guitar Hero III sounded more dynamic, and the album became a symbol of the loudness war.
• Red Hot Chili Peppers, Californication (1999): frequently cited for audible clipping and distortion caused by aggressive loudness processing.
• Iggy and the Stooges, Raw Power (1997 remix): often named as one of the loudest CDs ever released.
• Daft Punk, Random Access Memories (2013): mastered by Bob Ludwig and widely praised for keeping its dynamics at a time when most pop releases did not; it went on to win the Grammy for Album of the Year.

Common mistakes and myths

• Myth: you must master to exactly −14 LUFS. −14 LUFS is a playback reference on some platforms, not an upload requirement. Many excellent masters are much louder or quieter.
• Myth: streaming services compress your music. Normalization is a fixed gain change for the whole track or album, exactly like turning the volume knob. Only when a platform raises a quiet track beyond its available headroom may limiting be involved, and Spotify, for example, uses a limiter only in its Loud setting.
• Myth: a peak meter tells you how loud a track is. Peaks and loudness are different things. Two tracks peaking at −1 dBFS can differ by 10 LU or more in loudness.
• Myth: true-peak limiting is completely transparent. It prevents inter-sample overs, but heavy true-peak limiting on dense material can still dull transients. Many engineers shape peaks earlier with a clipper or bus processing and let the final limiter do less.
• Mistake: making every album track equally loud. It flattens the emotional shape of the record.
• Mistake: comparing processing at different levels. The louder option almost always wins; match loudness before judging.

Loudness vs related terms

Comparison	The difference
Loudness vs volume or level	Level is measured (dBFS, dB SPL, volts); loudness is how that level is perceived.
LUFS vs RMS	RMS is the average power of the raw waveform. LUFS adds K-weighting and gating so the number matches perception much better.
LUFS vs LKFS	The same measurement with two names: LKFS in ITU and ATSC documents, LUFS in EBU documents.
LU vs dB	A loudness unit (LU) is a relative step on the LUFS scale; 1 LU corresponds to 1 dB.
Sample peak vs true peak	Sample peak reads the highest sample value; true peak estimates the waveform between samples after reconstruction.
Loudness vs dynamic range	Loudness is how loud something is on average; dynamic range and LRA describe how much its level varies.

Standards and research

Standard	What it covers
ISO 226:2023	Normal equal-loudness-level contours for pure tones.
ISO 532-1 and 532-2 (2017)	Methods for calculating loudness: Zwicker and Moore-Glasberg models.
ITU-R BS.1770 (2006; current revision -5)	The loudness and true-peak measurement algorithm behind LUFS/LKFS.
EBU R128 (2010), Tech 3341, Tech 3342	European broadcast loudness target, meter behavior (EBU Mode) and Loudness Range.
ATSC A/85	US television loudness practice (−24 LKFS).
AES TD1008 (2021)	Loudness recommendations for internet streaming and on-demand audio.

Related terms

• Fletcher-Munson curves: why loudness depends on frequency.
• RMS: average level, the ancestor of LUFS metering.
• Headroom: the space between your peaks and 0 dBFS.
• Limiter: the final tool for controlling peaks and raising loudness.
• Normalization: how loudness is matched by platforms and in your DAW.
• Dynamic range: how much the level varies.
• Mastering engineering: where final loudness decisions are made.

FAQ