Intersample peaks are rogue transients that fall between samples. Your DAW's standard meters cannot see them - but they can cause real distortion when your master is converted, encoded or played back.
Why they happen
Digital audio stores a waveform as a finite number of samples per second - 44,100 at 44.1 kHz. Standard peak meters read the values of those samples. But when the samples are converted back to a continuous waveform, the reconstructed curve can rise higher between two samples than either sample itself. That hidden peak is an intersample peak.
So a master can read -0.1 dBFS on a standard meter while the reconstructed signal actually exceeds 0 dBFS. It clips in the converter, or when encoded to a lossy format, even though your meters showed no clipping at all. In loud masters, this is common.
Why they matter
Intersample peaks can produce audible distortion - harshness, crackle, a brittle top end - especially after conversion to AAC or MP3 for streaming. Sometimes a little is inaudible, or even suits the style. But when it is not, the meters will still say "all is well".
Measuring them
- Intersample peak meters oversample internally - running at two, four or more times the session rate - to catch peaks between the samples.
- True peak meters estimate the actual peak of the reconstructed signal and report it in dBTP. They are the standard specified by streaming services and broadcasters.
Both use more CPU. Put one at the end of your chain, after the maximizer.
When clipping is acceptable
Clipping is usually described as a fault, and often it is. But fixing every last trace can sometimes make a master sound over-processed or weaker. In moderation, subtle clipping can bring content forward and make it feel more aggressive - and some genres rely on it. Clipping is situational: listen first, then look at the meters. Seeing red can bias your ears; decide whether it sounds wrong before deciding whether it looks wrong.
Reducing intersample peaks at the maximizer
- Ceiling - lowering it leaves room for intersample peaks. A ceiling around -1 dBTP is a common safe choice for streaming. Once set for a format, leave it.
- True peak mode - many maximizers have a true peak setting that limits intersample peaks directly.
- Oversampling - runs the limiter internally at a higher sample rate, so its detection is more precise, then returns to the session rate. The session sample rate does not change. It can sound smoother - or, sometimes, over-processed. Toggle it and listen.
- Lookahead - lets the limiter round off peaks before they arrive. Short lookahead times (around 1-2 ms) preserve transients.
- Attack and release - set them for the best sound, then leave them.
- Stereo linking - linked or independent left and right limiting can change how peaks are handled.
Reducing them before the maximizer
- Trim the premaster's level.
- Use reductive EQ to remove energy that is not needed - excessive sub or harsh highs.
- Control peaky elements with multiband compression or de-essing.
- Use mid/side processing to tame excessive side energy.
The less work the maximizer has to do, the fewer intersample peaks it creates.
Do this now
- Put a true peak meter after your maximizer. Push the maximizer with its ceiling at -0.1 dBFS and note how far true peak exceeds it.
- Turn on true peak mode or oversampling and compare the sound and the readings.
- Export to AAC or MP3 and listen for any new distortion.
Adapted, with thanks, from the lesson notes of J.M. Brooklyn Audio Mastering.