What Is Audio Clipping? How to Prevent Distorted Audio

Peak, dBFS and gain describe different things

Sample peak is the largest stored sample magnitude. In a normalized digital representation, 1 is full scale and is referenced as 0 dBFS. Half that amplitude is about -6.02 dBFS: 20 times log10(0.5). Gain multiplies amplitudes; adding about 6 dB nearly doubles them.

A floating-point editor may temporarily hold values above full scale, but a fixed-range output stage can clip them. Analog microphones, preamps and other stages can also distort before the digital meter reaches 0 dBFS. A low peak reading does not prove a recording was never clipped earlier.

A generated signal measured by the actual analysis engine

Our reference is six seconds of a 1 kHz sine wave, mono, 48 kHz, 16-bit PCM. The clean version has amplitude 0.5. A second version multiplies it by three, then hard-clamps values to the range -1 to +1 before writing PCM. This intentionally destroys the peak shape.

The generator clamps 156,000 of 288,000 samples in that second file. Tubelexity's existing LUFS analysis helper, using FFmpeg core 0.12.10, measures both files. The existing peak-normalization filter then attenuates the clipped file to a -3 dBFS sample-peak target; it does not attempt reconstruction.

The readings below are from those actual files, not estimates from the gain slider. Their small true-peak/sample-peak difference demonstrates that the measurements are not interchangeable. A steady synthetic tone is not representative program audio and these results are not universal target recommendations.

Normalization changes level, not the missing waveform

Peak normalization applies gain based on the measured peak and a target. Loudness normalization targets a perceptual loudness measurement and may need dynamic processing when a peak ceiling would otherwise be exceeded. Neither operation knows which original values were removed by clipping.

The normalized test file has a lower maximum, but consecutive equal-height samples remain at the damaged peaks. A clipping indicator based only on near-full-scale samples may stop warning after attenuation. Listen as well as measure: a clean-looking meter is not proof that distortion has disappeared.

• Adjust measured level without claiming to repair clipping
• Understand LUFS, dBFS and normalization targets

Compression, limiting and headroom

A compressor reduces gain above a threshold according to its ratio, attack and release. It can reduce the difference between louder and quieter passages. Makeup gain raises the processed signal again and can consume the headroom you just created. A limiter is a more restrictive form of peak control, but the exact algorithm determines what it catches.

A sample-peak limiter is not automatically a true-peak limiter. True peak estimates the reconstructed waveform between samples; it can exceed stored sample peaks. Lossy encoding can also change peaks. Measure after export, not only before compression or MP3 encoding, and do not describe a sample ceiling as a verified true-peak guarantee.

For an unclipped recording, a modest ceiling below full scale is a practical starting point, not a universal mastering specification. Different delivery requirements and program material call for different choices. For the controlled peak-normalization test we disable extra limiting so attenuation is the only change.

• Control dynamic range and check makeup gain
• Check peak changes after EQ boosts
• Apply gain cautiously and recheck the export

Prevent overload before trying to hide it

During recording, reduce input gain if the loudest expected passage overloads. Moving a microphone or choosing a different recording level can address an upstream problem that later software cannot undo. During editing, inspect levels after boosts, summed tracks and effects, not only the original track.

If distortion persists at low playback volume, compare an earlier source or another playback device to separate recorded damage from speaker overload. These browser tools analyze and adjust files; they are not an automatic restoration or mastering service. Long audio and large files can exceed browser memory, particularly on mobile, so begin with a short representative section.