Set recording levels slightly low rather than high because digital clipping is irrecoverable

Digital distortion from an over-driven input is a hard ceiling: the waveform clips and the data is permanently destroyed. A quiet recording can always be amplified or normalised in post, regaining the detail and dynamic range that was captured. Setting input gain conservatively — aiming for peaks well below maximum — is therefore the safer default, especially in unpredictable acoustic environments where a sudden shout or passing vehicle can spike the level. This is the same signal-to-noise reasoning as gain staging in mixing, with the added constraint that a real-world sound event has no second take.

Examples

Recording a busy market: set levels so talking voices peak around -12 dBFS, leaving headroom for a sudden shout or nearby truck without clipping.

Assessment

A student’s field recordings peak at -2 dBFS and contain sporadic clip artefacts. What should they have done differently, and why can’t the clipping be fixed in post?