Inter-sample peaks: why audio can clip below 0 dBFS
Last updated 5 min readTags:MasteringTrue PeakLoudness
Inter-sample peaks can exceed 0 dBFS even when your DAW meter stays below it. Learn how to measure true peak, choose a ceiling and check your export.
Your DAW's peak meter reads −0.1 dBFS, but a true-peak meter on the exported file reads +0.8 dBTP. Neither reading is necessarily wrong. The first measures stored samples; the second estimates the highest level of the waveform those samples represent.
That waveform can peak between samples, above the level shown by a sample-peak meter. These inter-sample peaks can cause clipping during playback or later processing, even when no stored sample exceeds 0 dBFS.
For a release-ready file, check true peak as well as sample peak. A ceiling of −1 dBTP is a sensible starting point, though some delivery guidelines recommend more headroom.
What are inter-sample peaks?
A digital audio file stores samples: measurements of the waveform's amplitude at regular intervals. During playback, a digital-to-analogue converter reconstructs a continuous waveform from those samples.
The reconstructed waveform is not a series of straight lines between points. Its curves can reach their highest levels between the stored samples.
The EBU's true-peak meter tests include a tone with samples just below 0 dBFS but an expected true peak of +3 dBTP. This is a deliberately chosen test signal, rather than a typical music example, but it demonstrates why a sample-peak reading cannot establish the waveform's maximum level. [1]
An inter-sample peak does not have to exceed 0 dBTP. It becomes a potential overload when the reconstructed level exceeds what a later stage can handle.
Sample peak and true peak
| Measurement | What it tells you | Unit |
|---|---|---|
| Sample peak | The largest absolute value among the stored samples | dBFS |
| True peak | An estimate of the reconstructed waveform's highest absolute level, including between samples | dBTP |
A true-peak meter uses oversampling to estimate the waveform between the original samples. ITU-R BS.1770 describes the measurement method. [2] For more detail, see true peak and dBTP.
True peak is still an estimate. It does not predict the exact result of every converter or codec, so a reading just below 0 dBTP leaves little margin.
When can inter-sample peaks cause clipping?
Clipping can occur when a stage cannot accommodate the signal's peak level. The main risks are:
- Playback conversion. Some digital-to-analogue converters can overload as they reconstruct a waveform whose peaks exceed full scale.
- Sample-rate conversion. The new samples can fall closer to the waveform's peaks. Their values may exceed the range of a fixed-point output unless the level is reduced. [2]
- Lossy encoding and decoding. AAC, MP3 and similar formats change the waveform. The decoded result can have higher peaks than the source, with a risk of clipping later in the signal path. Apple's mastering guidance recommends checking encoded versions as well as the original file. [3]
A positive true-peak reading indicates a risk; it does not prove that audible clipping has occurred. Playback equipment and processing paths differ.
What true-peak ceiling should you use?
For a finished file intended for release, −1 dBTP is a useful starting ceiling. Treat it as a practical margin, rather than a guarantee against every later conversion.
Follow any specific delivery instructions. Spotify currently recommends peaks below −1 dBTP, or below −2 dBTP when the master is louder than −14 LUFS integrated. These are its published mastering recommendations, rather than a universal rule for all releases. [4]
If you are sending a mix to a mastering engineer, use their requested export settings. There is no need to add a limiter solely to meet a streaming ceiling at that stage.
How to reduce inter-sample peaks
1. Measure the complete export
Analyse the file you intend to deliver with a true-peak meter or file analyser. Check the whole track; its highest peak may occur outside the loudest-sounding section.
2. Reduce output gain if that is sufficient
If the mix sounds right and only needs more headroom, turn it down after the final processor. Additional compression or limiting is not required merely to lower its peak level.
For example, a file at +0.8 dBTP needs a 1.8 dB reduction to reach −1.0 dBTP. A gain reduction at this point lowers the whole signal without changing its dynamics. Re-export and confirm the result.
This does not repair distortion already introduced earlier in the chain. If the audio is already clipped, correct the overload at its source.
3. Use true-peak limiting when you need to control peaks
If you need to retain more average loudness, use a limiter with a true-peak mode and set the required ceiling. Listen to the result as well as checking the meter.
Without a true-peak mode, leave more output headroom and verify the export with a separate true-peak meter. Oversampling alone does not establish a particular true-peak ceiling.
4. Check after conversion
Measure again after any sample-rate conversion. If possible, audition an AAC or MP3 version too. A codec preview or a decode into floating-point audio can reveal overs without first clipping them to fit a fixed-point file. [3]
If the converted version exceeds your intended ceiling, reduce the source level further and repeat the check.
Will a small overshoot be audible?
Not necessarily. The audible result depends on the signal, the amount and duration of any clipping, and the playback path. A meter warning alone cannot tell you whether a listener will hear distortion.
Loudness normalisation is no substitute for checking the file. It changes playback gain; it cannot repair clipping already introduced before that adjustment. Spotify applies normalisation during playback, and its availability depends on the device and settings. [4]
For a file you intend to release, measure the final export, leave the required headroom and listen to any converted versions you can check.