Loudness Range (LRA)

Definition

The Loudness Range (LRA) measures the variation in loudness within a program, that is, its macrodynamics. The method is defined by EBU Tech 3342:

1. short-term loudness values are calculated (3 s rectangular window, K-weighting from ITU-R BS.1770-4) with an overlap of at least 2.9 s between consecutive windows, that is, one value every 100 ms or less;
2. absolute gate: values below −70 LUFS are discarded;
3. the average loudness of the remaining values is calculated (average of the energies) and the relative gate is set at −20 LU below this average; lower values are discarded;
4. the remaining values are sorted and the 10th percentile (L10) and the 95th percentile (L95) are calculated;
5. LRA = L95 − L10, in LU.

The lower percentile at 10% prevents a long fade-out from dominating the measurement; the upper one at 95% prevents a single very loud event (a hit, an explosion) from widening the result. The relative gate at −20 LU (wider than the −10 LU of integrated loudness) keeps the quiet but real passages, such as a verse at pianissimo. The measurement is independent of sample rate and format.

Parameters and units

• Unit: LU.
• Input: 3 s short-term values, step ≤ 100 ms.
• Gate: absolute −70 LUFS; relative −20 LU.
• Percentiles: 10th and 95th.
• Typical values by genre:

Musical function

LRA describes how much the song moves between quiet and loud sections: verse and chorus, break and drop, pianissimo and fortissimo. It is a measure distinct from microdynamics (crest factor, PLR, PSR): a limiter working on transients greatly reduces PLR and barely changes LRA, while the arrangement, level automation and slow bus compression act mainly on LRA. An LRA that is too low makes the song flat and uniform; too high, in a genre listened to in the car or on headphones on the street, it forces the listener to adjust the volume.

Practical example

A 3:00 pop ballad, about 1800 short-term values:

The compression raised the verses by more than 4 LU while leaving the chorus almost unchanged: the LRA dropped from 9.2 to 5.1 LU and the ballad lost its crescendo. By reducing the multiband range and using +1.5 dB of automation on the chorus, it returns to 7.5 LU, a value consistent with the genre.

Common mistakes

• Confusing LRA with crest factor or PLR: they measure different time scales.
• Reducing LRA only with the limiter: the result is a master squashed on the transients with sections that remain unbalanced.
• Measuring LRA on 20–30 s excerpts: the statistics don't contain enough values and the result doesn't represent the song.
• Including long silences or reverb tails in the file: the gate excludes them only below the relative threshold.
• Using the LRA of a different genre as a target.

How to check it

• App analysis: the Loudness Range (EBU 3342) row uses the 3 s short-term values calculated every 100 ms, the absolute gate at −70 LUFS and the relative gate at −20 LU, and returns the difference between the 95th and 10th percentiles.
• Dedicated meters: Youlean Loudness Meter, iZotope Insight, FabFilter Pro-L 2 report LRA in real time and over the whole program.
• Diagnosis: compare the LRA of the mix and the master: a drop of more than 2–3 LU indicates that mastering has altered the shape of the song.
• Comparison with the reference: analyze the reference and the master in full in the app; a similar LRA, within the same genre, indicates consistent macrodynamics.