True peak vs sample peak: why a file below 0 dBFS can still clip

Your sample-peak meter can stay below zero while a true-peak meter shows an overload. The meters are checking different things: the stored samples and the waveform reconstructed from them.

Sample peak is the highest stored sample value. True peak estimates the highest level between those samples as well. When the reconstructed waveform rises above the stored values, you have an intersample peak.

Why meters disagree

A sample-peak meter only reads stored values, so it cannot show every peak in the reconstructed waveform. ITU-R BS.1770-5 describes a true-peak estimate using oversampling and filtering, including four-times oversampling at 48 kHz.

Encoding adds another step to check. AAC, MP3 and Opus alter the waveform, and the decoded file can have higher peaks than the PCM master. Leave headroom, then check the encoded result when possible.

What happened in our AAC and Opus tests

Our five-source export test showed why a fixed margin needs checking. After a tonal change, we could no longer assume that our −2 dBTP PCM ceiling covered every lossy export. At a separate −1 dBTP comparison ceiling, two AAC files decoded to +0.27 and +0.70 dBTP.

On the same EDM source, lowering the PCM ceiling did not produce a matching drop in the decoded peak. A −2 dBTP PCM master decoded to −0.69 dBTP; −3 decoded to +0.28; −4 decoded to −2.02. These are results from that file and encoder, rather than a rule about all AAC exports.

The resulting export policy uses −2 dBTP for WAV/FLAC and the main PCM master. AAC, Opus and MP3 receive a constant trim to −4 dBTP immediately before encoding. In the five-source test, decoded AAC measured −3.88 to −2.02 dBTP and Opus −3.65 to −3.30 dBTP, with no detected clipped runs.

What to look for in a mastering report

• Sample peak tells you the tallest stored sample.
• True peak estimates reconstruction between samples.
• Clipped runs describe repeated flat sample sequences, a separate kind of evidence.
• Codec round-trip checks the delivery format rather than assuming the PCM result survives unchanged.

Read the final file’s true peak alongside its loudness. A limiter setting describes what you asked for; the measurement tells you what the rendered audio reached. An encoded copy needs its own check.