LUFS Short-Term (short-term loudness)

Definition

Short-term loudness is loudness calculated over a sliding rectangular window of 3 s, without gating. The definition is contained in EBU Tech 3341, which is based on the method of ITU-R BS.1770-4:

L_S(t) = −0.691 + 10 · log10( Σ_i G_i · (1/T) · ∫ y_i(τ)² dτ ), with the integral extending from t − 3 s to t and y_i the signal of channel i after K-weighting.

The measurement must be updated at a rate of at least 10 Hz (every 100 ms or less). The highest value reached in the song is the maximum short-term (Max S). The series of short-term values is also the basis for calculating the Loudness Range (EBU Tech 3342) and the PSR. Tech 3341 provides two display scales: the EBU +9 scale (from −18 to +9 LU around the reference value) and the EBU +18 scale (from −36 to +18 LU).

Parameters and units

• Unit: LUFS; differences in LU.
• Window: 3 s, rectangular, no gating.
• Update rate: ≥ 10 Hz.
• Max short-term: highest value of the song; in compressed pop it exceeds the integrated loudness by 1–3 LU, in orchestral music by 4–8 LU.
• Typical section values (modern pop): verse −12/−10 LUFS, chorus −9/−7 LUFS, EDM drop −7/−5 LUFS.

Musical function

Short-term follows loudness phrase by phrase and section by section. It is used to balance verse, pre-chorus and chorus, to verify that the drop really hits harder, to compare the same section of the song with the reference and to decide the master bus automation. It is also the most useful figure for mixing toward a target: if the reference has its chorus at −8 LUFS short-term, your own chorus should be compared at that level before judging EQ and compression.

Practical example

A 3:30 pop song analyzed by section:

Integrated loudness −10.1 LUFS, maximum short-term −8.3 LUFS: the final chorus is the loudest part, as expected. If the verse were at −9.5 LUFS, the contrast with the chorus would shrink to 0.7 LU and the song would lose its sense of opening up: in that case you lower the verse or lighten the elements that fill it.

Common mistakes

• Using short-term as a compliance value: delivery specifications refer to integrated loudness.
• Reading short-term in the first 3 s of playback, when the window still contains the initial silence.
• Comparing the short-term of your own chorus with the verse of the reference.
• Pushing the limiter until every section has the same short-term: the shape of the song is erased.
• Confusing the maximum short-term with the peak: they are different quantities, one is energy over 3 s, the other instantaneous.

How to check it

• App analysis: the Max short-term (3 s) row is the highest value of loudness calculated over 3 s windows (one 400 ms block plus 26 steps of 100 ms) updated every 100 ms, with K-weighting.
• By section: export verse and chorus as separate files and analyze them in the app; the integrated loudness of a single-section file coincides with its average short-term.
• Dedicated meters: Youlean Loudness Meter (short-term history graph), FabFilter Pro-L 2, iZotope Insight.
• Comparison with the reference: align the two tracks on the same section and compare the short-term values before judging timbre.