← All Sonaudit guides

Loudness and peaks

LUFS, True Peak, and Sample Peak: What Each Audio QC Metric Tells You

Separate programme loudness from sampled and reconstructed peaks, and avoid using one number as a substitute for the others.

Published Updated

An audio file can meet a loudness target and still exceed a peak ceiling. It can also keep every stored sample below 0 dBFS while reconstruction produces a higher inter-sample peak. LUFS, sample peak, and true peak describe different properties; no single one replaces the other two.

Sonaudit shows the measurements together because delivery specifications often combine them. The chosen profile decides whether a value is merely reported or compared with a threshold.

Integrated LUFS describes programme loudness

Integrated loudness estimates the perceived loudness of the measured programme over time. The ITU-R BS.1770 algorithm applies frequency weighting and gating so very quiet sections do not dominate the final value. EBU Mode also defines momentary and short-term views, while Loudness Range describes variation rather than average level.

Integrated LUFS is useful for comparing a complete programme with a delivery target. Momentary and short-term maxima help find brief or sustained loud passages that the integrated number can hide. Loudness Range helps describe how widely the programme moves, but it is not a peak measurement and it does not decide whether an inter-sample overload exists.

  • Integrated: the gated measurement from the beginning to the end of the programme.
  • Momentary: a short 400 ms loudness view in EBU Mode.
  • Short-term: a 3 second loudness view in EBU Mode.
  • LRA: a descriptor for the distribution of short-term loudness over the programme.

Sample peak sees the stored sample values

Sample peak is the highest absolute value among the discrete samples in the file, normally expressed in dBFS. It is direct and useful for finding samples at or near digital full scale. A run of consecutive full-scale samples can support a clipping finding because the waveform has been held against the numerical limit.

The limitation is in the name: sample peak looks at sample points. A digital-to-analogue converter or a later sample-rate conversion reconstructs a continuous waveform between them, and that reconstructed curve can rise higher than either neighbouring sample.

True peak estimates what happens between samples

True-peak analysis oversamples the signal to estimate the maximum of the reconstructed waveform. The result is expressed in dBTP. This is why a file can show a sample peak below 0 dBFS and a true peak closer to, or above, 0 dBTP.

Sonaudit uses a four-phase FIR true-peak analyser. Its 48 kHz results were checked against the EBU Loudness Test Set v5 true-peak material, and the inter-sample clipping decision was cross-checked with Apple's afclip in the project's validation suite. Other sample rates remain measurable, but the project's published official-tolerance record is specifically for 48 kHz material.

Use the metric named by the delivery rule

If a receiver specifies integrated loudness and maximum true peak, check both in the same scan. Do not substitute RMS for LUFS, sample peak for true peak, or a brief loudness maximum for the integrated programme value. Also verify whether the rule applies to a complete mix, a chapter, a short-form item, or a dialogue-gated measurement; the measurement scope changes the conclusion.

These numbers describe level, not overall sound quality. A file can pass every loudness and peak threshold while containing an edit click, a swapped channel, unwanted silence, or creative problems that only listening reveals.