← All Sonaudit guides

Loudness and peaks

What Is True Peak in Audio? dBTP vs. LUFS

See why a reconstructed waveform can peak above its stored samples, what dBTP measures, and why true peak cannot replace a loudness reading.

Published Updated

True peak is an estimate of the highest level reached by the continuous waveform represented by a digital audio signal. It is written in dBTP. Unlike a sample-peak meter, it looks between the stored samples by reconstructing the waveform at a higher effective sample rate.

That distinction matters because the reconstructed curve can rise above every individual sample. A file can remain below 0 dBFS on a sample-peak meter yet produce a true-peak result near or above 0 dBTP during conversion, encoding, or playback.

Why sample peak can miss the highest point

A PCM file stores measurements at discrete moments. Sample peak simply finds the largest absolute stored value and expresses it relative to digital full scale in dBFS. It is exact for those sample values, but the samples are not the complete shape that a converter reconstructs.

When adjacent high-level samples describe a steep high-frequency waveform, the interpolation between them can form a higher crest. This is often called an inter-sample peak. The International Telecommunication Union defines true peak in the continuous-time domain and describes oversampling as the way to estimate it.

dBTP, dBFS, and LUFS answer different questions

dBFS describes stored digital sample level, while dBTP describes an estimated reconstructed peak. Both are level measurements near the top of the signal path. LUFS instead estimates programme loudness over time using frequency weighting and, for integrated loudness, gating.

A track can therefore meet an integrated LUFS target and still exceed its true-peak ceiling. Lowering one short peak may barely change integrated loudness, while changing compression across the whole programme can move LUFS substantially without producing a new highest peak.

  • Sample peak in dBFS: the largest value stored in the PCM samples.
  • True peak in dBTP: an estimate of the largest reconstructed waveform value.
  • Integrated loudness in LUFS or LKFS: a gated programme-level loudness estimate.
  • Loudness Range in LU: a description of loudness variation, not a peak ceiling.

How a true-peak meter estimates the waveform

ITU-R BS.1770-5 describes a four-times oversampling path for 48 kHz audio, producing a 192 kHz intermediate representation before the maximum is converted to dBTP. Higher input sample rates require proportionally less oversampling for the same intermediate rate. The interpolation filter and numerical implementation affect how closely a meter follows the reference tolerance.

Sonaudit uses a four-phase FIR analyser. Its recorded official-test coverage applies to the 48 kHz true-peak files in the EBU Loudness Test Set v5, and its inter-sample clipping decision was cross-checked with Apple's afclip. It can measure other sample rates, but the project's published official-tolerance record is limited to the named 48 kHz material.

There is no universal true-peak target

A true-peak number becomes a pass or fail only when a delivery rule supplies a ceiling. The appropriate margin depends on the destination, codec workflow, and current receiver specification. A broadcast profile, a podcast recommendation, and a music-mastering workflow should not borrow one another's threshold without checking the source requirement.

Also distinguish an original PCM preflight from an encoded-file test. Lossy encoding can create peaks that were not present in the source measurement. If the destination provides an encoder, round-trip tool, or ingest validator, use it on the actual deliverable or audition encode rather than assuming the PCM result predicts every later stage.

How to investigate a high true-peak result

First confirm the selected profile and the unit. Compare true peak with sample peak; a wide difference can indicate that the reconstructed waveform, rather than a stored full-scale sample, is driving the result. Return to the source session, adjust gain or limiting with the delivery goal in mind, render again, and rescan.

Do not normalize a finished file automatically just to make one number green. Peak control can change the sound, and the correct remedy may be in the limiter, sample-rate conversion, or earlier mix. After the PCM master passes its intended preflight, complete any destination-specific encoding and listening checks.