True Peak vs Inter-Sample Peaks

Your mix looks clean. Every sample sits below 0 dBFS. The meters are happy. You export, upload to your distributor, and three days later someone tells you the master sounds distorted on Apple Music.

Nothing went wrong with your file. Something went wrong between your samples.

What a Sample Peak Actually Is

Digital audio is a series of discrete measurements. At 44.1kHz, your DAW captures 44,100 amplitude snapshots per second. Each one is a sample. The peak meter in your DAW shows you the highest of those snapshots, the loudest individual measurement in the file.

That number is real and useful. It is also incomplete.

The problem is what happens between those snapshots. Audio is not actually a series of disconnected points. It is a continuous waveform. When a digital file gets converted back to analog (which happens every time someone listens to it) the converter reconstructs the waveform from those discrete samples using a process called interpolation. That reconstructed waveform does not always peak at the same point your sample meter showed. Sometimes it peaks higher.

That higher point is the inter-sample peak.

What a True Peak Actually Is

A true peak meter estimates the level of that reconstructed waveform. It oversamples the audio (typically by a factor of 4x or higher) and measures what the signal actually does between samples. The result is a more accurate picture of the maximum level your audio will reach in the real world.

True peak is measured in dBTP, decibels true peak, to distinguish it from dBFS, which refers to sample peaks. A file that reads -0.1 dBFS on a sample peak meter might read +0.3 dBTP on a true peak meter. Those 0.4 decibels are real. They are just hiding between your samples.

Most professional mastering tools measure true peak. Most DAW channel meters do not. This is one of several reasons why checking your rendered file outside the DAW before delivery matters. What you see on the channel strip and what is actually in the file are not always the same thing.

Why Lossy Encoding Makes This Worse

Inter-sample peaks are a fact of digital audio even before encoding. Lossy encoding (the conversion from WAV to AAC, Ogg Vorbis, or MP3) makes them significantly worse.

Lossy codecs work by discarding audio information the encoder determines is perceptually irrelevant, then reconstructing the signal at playback using mathematical models. That reconstruction process introduces its own inter-sample behavior. A file that measured +0.3 dBTP before encoding might measure +1.0 dBTP or higher after. The encode did not add distortion intentionally. It just did not have enough headroom to reconstruct the signal cleanly.

This is why every major streaming platform specifies a true peak ceiling for delivered masters. Spotify asks for -1 dBTP on quieter masters and -2 dBTP on louder ones. Apple Music asks for -1 dBTP. Apple Digital Masters specifically requires -1 dBTP as part of the certification standard, because the AAC encode that follows needs room to work. The platforms are not being precious about headroom. They are accounting for what their encoders do to a file.

A master delivered at 0 dBFS sample peak with inter-sample peaks at +0.5 dBTP will come out of the AAC encoder with audible distortion. Not clipping in the obvious sense. Something subtler: a hardness, a brittleness on loud transients that was not in the original. Most listeners will not identify it as distortion. They will just feel like something is slightly off. That feeling is the encode running out of headroom.

What True Peak Limiting Does

A true peak limiter operates on the oversampled signal, not the sample-domain signal. It catches and controls inter-sample peaks that a standard sample peak limiter would miss entirely.

The practical difference: a standard limiter set to 0 dBFS will allow inter-sample peaks to pass through above 0 dBFS, because it is only measuring and acting on samples. A true peak limiter set to -1 dBTP will catch those peaks before they exceed the ceiling, regardless of where they fall between samples.

Most modern mastering limiters have true peak limiting modes. FabFilter Pro-L 2, iZotope Ozone's Maximizer, and Sonnox Limiter all offer true peak measurement and limiting. If your limiter does not have a true peak mode, your ceiling is not where you think it is. That is not a hypothetical. It is a gap in your signal chain.

True peak limiting does introduce a small amount of additional processing compared to sample peak limiting, because the oversampled signal has to be brought back down to the original sample rate after limiting. Done well, this is inaudible. Done with a limiter that handles it poorly, it can introduce subtle smearing on transients. This is one reason why limiter choice matters at mastering. Not all true peak implementations are equal.

True Peak Targets by Context

The right true peak ceiling depends on where the master is going.

Streaming platforms. -1 dBTP is the standard ceiling for most platforms. Spotify, Apple Music, Tidal, and Deezer all specify -1 dBTP for standard delivery. Amazon Music asks for -2 dBTP. For masters that are already heavily limited and running hot, -2 dBTP is a reasonable conservative choice across the board. For more detail on how each platform handles this, see Loudness Targets and Mastering for Streaming Platforms.

Apple Digital Masters. -1 dBTP is required as part of the certification specification. The ADM pipeline runs your file through Apple's AAC encoder as part of the verification process, and the headroom is not optional. See What Is Apple Digital Masters and Why Does It Matter?

CD and lossless digital. True peak management still matters even without lossy encoding. CD Red Book audio is 16-bit/44.1kHz, and the dither process applied during bit depth conversion can push peaks slightly. -0.3 to -0.5 dBTP is a reasonable ceiling for CD masters. For lossless streaming delivery, -0.5 to -1 dBTP gives the platform room to apply normalization gain without introducing clipping.

Vinyl pre-masters. True peak is less critical here than in digital delivery. The cutting lathe is not an encoder. Inter-sample peaks in a vinyl pre-master are not going to distort on playback the way they would through a lossy codec, but a clean master with well-managed peaks is easier to cut than one already pushing hard.

Broadcast. EBU R128 and ATSC A/85 both specify -1 dBTP as part of their loudness standards. If your work ever crosses into post production or broadcast delivery, -1 dBTP is non-negotiable.

What to Do If You Find Inter-Sample Peaks After the Fact

A true peak limiter operates on the oversampled signal, not the sample-domain signal. It catches and controls inter-sample peaks that a standard sample peak limiter would miss entirely.

The practical difference: a standard limiter set to 0 dBFS will allow inter-sample peaks to pass through above 0 dBFS, because it is only measuring and acting on samples. A true peak limiter set to -1 dBTP will catch those peaks before they exceed the ceiling, regardless of where they fall between samples.

Most modern mastering limiters have true peak limiting modes. FabFilter Pro-L 2, iZotope Ozone's Maximizer, and Sonnox Limiter all offer true peak measurement and limiting. If your limiter does not have a true peak mode, your ceiling is not where you think it is. That is not a hypothetical. It is a gap in your signal chain.

True peak limiting does introduce a small amount of additional processing compared to sample peak limiting, because the oversampled signal has to be brought back down to the original sample rate after limiting. Done well, this is inaudible. Done with a limiter that handles it poorly, it can introduce subtle smearing on transients. This is one reason why limiter choice matters at mastering. Not all true peak implementations are equal.

One thing worth knowing about workflow: engineers who switch on true peak mode after the fact, at the end of a session, often find they dislike the sound. The limiter is suddenly doing more work than it was designed into the chain to do, catching peaks it was never accounting for, and the result can feel constrained or slightly processed. The better approach, particularly for loud, dense program material like EDM, metal, or heavily produced hip-hop, is to engage true peak mode from the start of the mastering session. When the limiter is working with true peak awareness throughout, it shapes the transients and dynamics with that ceiling in mind rather than fighting against peaks it was not anticipating. The master tends to breathe more naturally, and the true peak limiting becomes part of the sound rather than a correction applied on top of it.

For quieter, more dynamic material where inter-sample peaks are less of a concern, this matters less. But if you know going in that the master is going to be loud, start with true peak mode on. It is much easier to build toward a ceiling than to retrofit one.

The Short Version

Sample peaks are what your DAW shows you. True peaks are what your listeners actually hear. Inter-sample peaks are the gap between those two things: real audio energy hiding between measurements that reveals itself the moment a converter or encoder reconstructs the waveform.

A standard peak meter set to 0 dBFS is not a true peak ceiling. It is a sample peak ceiling. Those are different numbers, and the difference matters most exactly when it is hardest to catch: after the render, inside the encode, on the platform, in the listener's ear.

Measure true peak. Limit to the appropriate ceiling for your delivery format. Check your rendered files before you send them.

The samples you cannot see are the ones that will cause you trouble.