True Peak

True peak (measured in dBTP — decibels True Peak) estimates the maximum level of the analog waveform that will be reconstructed by a digital-to-analog converter (DAC). Sample peak — what a standard dBFS meter shows — only measures the amplitude of individual digital samples. The critical difference: when a DAC reconstructs a continuous analog waveform from discrete samples, the resulting curve can exceed the amplitude of any individual sample. These are called inter-sample peaks (ISP), and they occur when the digital signal has been limited or clipped close to 0 dBFS. The math works like this: imagine two adjacent samples at 0 dBFS. The DAC draws a smooth curve between them. If the phase relationship of the frequencies in the signal causes the curve to overshoot, the reconstructed analog level can be 1-3 dB above the digital sample peak. On a consumer playback system, this means the DAC clips — distortion that was not visible on any sample-based meter. The ITU-R BS.1770 recommendation defines the algorithm for true peak measurement, which uses 4x oversampling of the digital signal to estimate the analog waveform between samples. This is why broadcast standards (EBU R128, ATSC A/85) specify true peak limits, not sample peak limits. The practical numbers: EBU R128 mandates -1 dBTP maximum for broadcast. Netflix Sound Mixing Specifications specify -2 dBTP for original content. Amazon Prime targets -2 dBTP. Apple Music and Spotify normalize to approximately -1 dBTP. The pattern is consistent: no professional delivery spec allows 0 dBTP. If your limiter ceiling is set to 0 dBFS and your sample peaks show 0.0 dBFS, your true peak is likely +0.5 to +1.5 dBTP. You are clipping and do not know it.

A common mistake: setting the limiter ceiling to -0.1 dBFS thinking it provides a safety margin. It does not. -0.1 dBFS is a sample peak reading. The true peak at that level can be +0.4 to +1.0 dBTP depending on the spectral content of the signal. High-frequency content (sibilance in vocal recordings, cymbal crashes, synth leads) produces larger inter-sample peaks because the DAC interpolation has less room to fit a smooth curve between samples that are already near maximum amplitude. The correct approach: set your true peak limiter ceiling to -1 dBTP for broadcast deliverables, -1.5 to -2 dBTP for streaming. Use a limiter that measures true peak, not sample peak — FabFilter Pro-L2, Waves L2, or iZotope Ozone Maximizer all have true peak modes. In DaVinci Resolve Fairlight, the built-in limiter has a true peak ceiling setting. Enable it. The limiter applies inter-sample peak detection and reduces gain only when a true peak would exceed the ceiling, preserving perceived loudness. Another gotcha: codec encoding raises peaks. When your master file gets encoded to AAC (Apple), Ogg Vorbis (Spotify), or Opus (YouTube), the lossy compression can push the peak level up by 0.5-1 dB. This is why Netflix specifies -2 dBTP: the extra headroom accounts for codec-induced peak increases. If you deliver at -1 dBTP, the AAC encode for Apple devices may clip. Test your master after encoding, not just in your DAW. A roundtrip through ffmpeg encoding to AAC and back will reveal whether your true peak ceiling was sufficient.