Loudness
How loud is a song, a podcast or a commercial? Peak meters can't tell you. Loudness measurement — LUFS — can, and it's what broadcasters and streaming services use to make everything play back at a similar level.
Peak level is not loudness
Digital level is measured in dBFS — decibels below full scale, the largest value a file can store (0 dBFS). A peak meter shows how close the signal gets to that ceiling.
But two files that both peak at −1 dBFS can sound very different in loudness. A dynamic acoustic recording reaches −1 dBFS only on a few drum hits; a heavily limited pop master sits just below it all the time. Our ears judge loudness by the average energy over time, weighted by how sensitive we are to different frequencies — not by the peaks.
LUFS and LU
LUFS (Loudness Units relative to Full Scale) is a measure of perceived loudness, standardized in ITU-R BS.1770. It's the number you'll see in delivery specs. LKFS is the same unit under its American name.
- Values are negative: −23 LUFS is quieter than −14 LUFS.
- LU (Loudness Unit) is a difference in loudness: going from −23 to −20 LUFS is +3 LU. One LU equals one decibel.
- The measurement applies a K-weighting filter first: it reduces low bass and slightly boosts the upper mids, roughly matching how the ear perceives loudness. Sub-bass adds a lot of energy but sounds less loud than its level suggests.
Momentary, short-term and integrated
Loudness changes over time, so it is measured over three windows:
- Momentary (M) — 400 ms
- Follows the music almost instantly. Useful while mixing, too jumpy to judge a whole piece.
- Short-term (S) — 3 seconds
- A steadier reading of the current passage — a verse, a chorus, a line of dialogue.
- Integrated (I) — the whole programme
- One number for the entire file, from start to end. This is the value delivery specs and streaming services use.
Gating: ignoring silence
A podcast with long pauses shouldn't measure quieter just because of the pauses. Integrated loudness therefore uses gating: passages quieter than −70 LUFS are ignored (absolute gate), and so are passages more than 10 LU below the average of the rest (relative gate). The result reflects the parts you actually hear.
Loudness range (LRA)
The loudness range, in LU, describes how much the loudness varies over the programme — the difference between the quieter and louder passages, ignoring the extreme 10 % at the bottom and 5 % at the top (EBU Tech 3342).
- Low LRA (around 3–6 LU): consistently loud — heavily compressed pop, commercials, radio promos.
- Medium LRA (around 6–12 LU): typical for many mixes and speech programmes.
- High LRA (15 LU or more): very dynamic — classical music, film soundtracks.
LRA is not good or bad in itself; it tells you whether the dynamics fit the listening situation. A podcast for commuters with LRA 18 LU means constant volume adjustments.
True peak
A sample peak meter shows the highest sample value in the file. But when the samples are turned back into a continuous waveform — by a D/A converter, or after encoding to MP3 or AAC — the signal between two samples can rise higher than either sample. These inter-sample peaks can clip even though no sample reaches 0 dBFS.
True peak, in dBTP, estimates the real peak of the reconstructed waveform by oversampling the signal (BS.1770 uses 4× oversampling). Delivery specs set a true-peak limit — typically −1 or −2 dBTP — to leave room for this, and for the extra peaks lossy encoding adds.
EBU R128 and other standards
EBU R128 (2010) is the European Broadcasting Union's recommendation for loudness normalization in broadcasting. It replaced peak-based levelling, which encouraged ever-louder programmes and commercials, with a single loudness target. Similar standards exist elsewhere — ATSC A/85 in the US, ARIB TR-B32 in Japan — and the measurement behind all of them is ITU-R BS.1770.
Common targets
Integrated loudness targets, from quiet to loud:
| Destination | Integrated | True peak | Notes |
|---|---|---|---|
| EBU R128 (European broadcast) | −23 LUFS ±0.5 LU | ≤ −1 dBTP | ±1 LU allowed for live programmes |
| ATSC A/85 (US broadcast) | −24 LKFS ±2 dB | ≤ −2 dBTP | |
| Netflix | −27 LKFS ±2 | ≤ −2 dBTP | Measured on dialogue only |
| Podcasts (Apple's recommendation) | −16 LUFS ±1 | ≤ −1 dBTP | |
| Spotify | −14 LUFS (normalization) | ≤ −1 dBTP | −2 dBTP recommended for masters louder than −14 LUFS |
| Apple Music | −16 LUFS (Sound Check) | ≤ −1 dBTP | |
| YouTube | −14 LUFS (normalization) | Only turns loud content down |
What normalization means for your music
Streaming services measure every track and adjust playback so tracks play at a similar loudness. A master at −8 LUFS on a −14 LUFS service is simply turned down by 6 dB. It doesn't sound louder than other songs — but it has lost the dynamics that heavy limiting took away.
That doesn't mean every master must hit −14 LUFS. Many genres are mastered louder for good reasons. It means loudness is a creative choice, not a competition: measure, compare with the reference tracks you are aiming for, and keep the true peak in check.
Services that turn quiet tracks up usually do so only as far as the true-peak headroom allows, so very dynamic tracks may still play somewhat quieter.
Measuring loudness with StudioTools
- Measure Loudness analyses whole files — integrated loudness, loudness range, true peak, max momentary and short-term — and can store the result in Broadcast Wave files.
- The preview meters show momentary, short-term and integrated loudness and true peak live while you listen.
Quick reference
| Term | Meaning |
|---|---|
| dBFS | Level relative to digital full scale (0 dBFS = maximum). |
| LUFS / LKFS | Perceived loudness, K-weighted, relative to full scale. |
| LU | Difference in loudness; 1 LU = 1 dB. |
| M / S / I | Momentary (400 ms) / short-term (3 s) / integrated (whole programme) loudness. |
| LRA | Loudness range: how much loudness varies, in LU. |
| dBTP | True peak: estimated peak between samples, relative to full scale. |