EBU R 128 — loudness normalisation and permitted maximum level
The European Broadcasting Union's recommendation, first published February 2010, read here in its November 2023 (V5) revision. Five pages, and the document that ended peak-normalised levelling in broadcast. It is the first standards text this spoke holds, and it closes the measurement half of coverage edge 8.
What it recommends
- Programme Loudness normalised to a Target Level of −23.0 LUFS, with ±1.0 LU tolerance only where the target is not practically attainable (live programmes), and an explicit warning that drifting to the edge of the tolerance must not become normal practice. Quality-control workflows get ±0.2 LU for measurement error.
- True Peak not above −1 dBTP during production, measured to ITU-R BS.1770 and EBU Tech 3341, tolerance ±0.3 dB for signals band-limited to 20 kHz. Distribution paths with data reduction may require lower.
- Measure the whole signal, without weighting toward speech, music or effects.
- Loudness Range (Tech 3342), Maximum Momentary and Maximum Short-term Loudness (Tech 3341) as descriptive measures, not as normalisation targets.
The measurement itself belongs to ITU-R BS.1770, which defines LU and LUFS; R 128's own contribution is the level gate (BS.1770 equation 7), so that Programme Loudness measures the foreground rather than being dragged down by silence. The revision history is short and worth reading as engineering: the relative gate moved from −8 to −10 LU in 2011, and the target tolerance tightened to ±0.5 LU in 2014 before the current wording.
Why the EBU wrote it
The reasoning clauses state the problem plainly. Peak normalisation "has led to considerable loudness differences between programmes and between broadcast channels," and those differences are "the cause of the most viewer/listener complaints." The quasi-peak meter of EBU Tech 3205 does not reflect loudness and was never meant to indicate a long-term average, so the industry had been levelling with an instrument that measured the wrong thing.
That is the analog-vs-digital argument's neighbour, arrived at from the broadcast side: the loudness war was not a taste dispute but a metering failure, and it stopped being winnable when the meter changed.
Where this wiki's subject sits
R 128 governs broadcast. Streaming is handled by a supplement (R 128 s2) and the platforms set their own targets, which is why spotify-loudness-normalization normalises to −14 LUFS rather than −23. Both numbers come from the same measurement definition, which is the point: once BS.1770 exists, a target is a policy choice and no longer an argument about meters.
See loudness-normalization for what this means for a record that leaves a DAW.