LUFS, true peak & dynamic range explained

Quick reference

Metric	Measures	Target or diagnostic?
Integrated LUFS	Average perceived loudness, whole track	Target (~−14 LUFS)
Short-term LUFS	Loudness over a 3 s window	Diagnostic
Momentary LUFS	Loudness over a 400 ms window	Diagnostic
Sample peak (dBFS)	Highest individual digital sample	Incomplete on its own
True peak (dBTP)	Real peak after reconstruction / codec	Spec ceiling (−1 dBTP; −2 Amazon)
LRA (LU)	Loudness variation across the track	Descriptive
DR	Crest-style peak-to-average gap	Diagnostic
PLR	True peak − integrated LUFS	Diagnostic
PSR	True peak − loudest short-term LUFS	Diagnostic

LUFS: perceived loudness

LUFS (Loudness Units relative to Full Scale) is the modern standard for measuring how loud something sounds, defined in ITU-R BS.1770 (current revision BS.1770-4). It applies K-weighting, a curve (a high-pass filter plus a high-frequency shelf) that approximates how we perceive loudness, so it tracks perception far better than raw peak or RMS. You'll see it three ways:

• Integrated LUFS. One loudness figure for the whole track. This is what streaming platforms read to decide how much to turn you up or down.
• Short-term LUFS. Loudness over a 3-second window, useful for spotting the loudest section.
• Momentary LUFS. A 400 ms window, for catching transients.

Lower (more negative) is quieter: −16 LUFS is quieter than −9 LUFS. For the specific numbers each platform targets, see streaming loudness targets.

Sample peak (dBFS) vs true peak (dBTP)

Sample peak (measured in dBFS) is simply the highest individual digital sample in the file. True peak (dBTP) reconstructs the continuous waveform between samples, the way a digital-to-analog converter or a lossy codec does, and reports the real maximum. Those inter-sample peaks routinely exceed the highest sample.

This is why a master can read exactly 0.0 dBFS yet clip once Spotify encodes it to Ogg Vorbis: the true peak was above 0 dBTP all along, and the codec exposes it. Spotify asks for −1 dBTP, or −2 dBTP if the master is hotter than −14 LUFS, and Amazon asks for −2 either way. Loud records often sit above those figures on purpose, so read yours and decide rather than discover it later. Soneam measures true peak with 4× oversampling, the minimum specified in ITU-R BS.1770.

LRA: loudness range

Loudness Range describes how much the loudness varies across a track. Broadly, it's the spread between its quieter and louder passages, expressed in LU. (Technically it's derived from the statistical distribution of gated short-term loudness, not a simple max-minus-min, which is why brief silences don't skew it.) A ballad with big dynamic swells has a high LRA; a wall-of-sound master has a low one.

LRA is descriptive, not a target: use it to confirm the master matches the music's intent.

Dynamic range: DR, PLR and PSR

Three related figures describe how much life is left in the master.

• DR (dynamic range). Most often the TT / Pleasurize Music Foundation "DR" value (the figure behind the Dynamic Range Database), a crest-style measure of the gap between peaks and average level. Higher means more punch and space; aggressively limited masters trend low.
• PLR (peak-to-loudness ratio). True peak minus integrated LUFS. A quick read on how much the whole master was pushed: a PLR around 8+ usually breathes, while 3–5 signals heavy limiting. The "right" range is genre-dependent.
• PSR (peak-to-short-term ratio). True peak minus the loudest short-term LUFS, so it reflects dynamics in the densest section rather than averaged across the song. A loud master can show a healthy PLR overall yet a crushed PSR in the chorus.

None of these has a single "correct" value. They're diagnostic. They tell you whether the master's dynamics fit the genre, and whether you've limited harder than the music actually needs.

Which numbers actually matter when you deliver

When you hand off a master, only two are hard targets:

1. Integrated LUFS. Set it to the genre-appropriate level (≈ −14 LUFS for most modern releases; quieter for dynamic genres).
2. True peak (dBTP). The specs ask for −1 dBTP (−2 for Amazon). Above that you are accepting whatever the encode does with it, which loud records often do deliberately.

The rest of them (LRA, DR, PLR, PSR) are there to verify the first two were achieved without crushing the music. If your PLR or PSR has collapsed and the master is hotter than the song needs, it's worth a second look, since you may be trading dynamics for level the platform will simply turn back down. Loudness is a creative call, and so, in a smaller way, is true peak: the specs ask for −1 dBTP (−2 for Amazon), and competitive masters exceed it deliberately.

A worked example: reading a real master

Numbers mean more together than apart. Here's a single pop master, straight off the meter, and what each figure is actually telling you:

Metric	Reading	What it says
Sample peak	0.0 dBFS	Looks maxed but safe, and it's lying to you (see the next row)
True peak	+0.6 dBTP	Already over full scale between samples; whether that is audible depends on the codec and the material
Integrated LUFS	−8.9 LUFS	Loud; Spotify Normal will turn it down ~5 dB to sit at −14
Loudest short-term	−6.1 LUFS	The chorus is the dense part driving that average
LRA	4 LU	Narrow range; consistent, intentionally so for the genre
PLR	9.5	Whole-track headroom is fine on paper…
PSR	6.7	…but the chorus is noticeably more crushed than the song average

Read top to bottom, the story is clear: the master looks clean on a sample meter, the true peak is already over full scale between samples, and the chorus is crushed harder than the rest of the track (low PSR). Whether −8.9 LUFS suits this record is a creative call, and so is the +0.6 dBTP in a smaller way: Spotify asks for −2 dBTP at this loudness, and loud records routinely ship above that without anyone noticing. What you want to avoid is getting there without having looked.

See your own numbers

The free Soneam loudness tool reports all of these (integrated LUFS, true peak, sample peak, LRA, DR, PLR and PSR) from a WAV or FLAC, and plays the master back the way each streaming platform will after normalization. Analysis runs in your browser; nothing is uploaded or stored.

And when the numbers are where you want them, the next step is getting the client to hear the master properly and sign off on it: see how to send a master to a client for approval.

Frequently asked questions

What is LUFS, in plain terms?

LUFS measures how loud audio sounds to a listener over time using a K-weighting curve that approximates human hearing. It's defined in ITU-R BS.1770 and is what streaming platforms use to normalize playback.

What's the difference between dBFS and dBTP?

dBFS (sample peak) is the highest individual digital sample; dBTP (true peak) estimates the real peak after reconstruction or lossy encoding using oversampling, and is often higher. A file at 0.0 dBFS can exceed 0 dBTP and clip after encoding.

What is a good PLR or DR value?

There's no universal number; both are diagnostic and genre-dependent. Roughly, a PLR around 8 or higher tends to breathe while 3 to 5 indicates heavy limiting. Dynamic genres sit higher, dense genres lower by design.

What's the difference between PLR and PSR?

PLR compares true peak to the whole track's integrated loudness; PSR compares true peak to the loudest short-term loudness, exposing how much dynamic range survives in the densest passage.

Is LRA a target I should aim for?

No. LRA is descriptive; it tells you how much loudness varies across the track. Use it to confirm the master's dynamics match the music's intent, not as a number to hit.

Which metrics actually matter when delivering to streaming?

Integrated LUFS and true peak (dBTP) are the two the platforms act on: one decides how far playback turns you down, the other decides whether the encode distorts. LRA, DR, PLR and PSR are diagnostics that tell you what the loudness cost the music.