Skip to content

Clipping in Audio: Digital vs Analogue Explained

Send feedback

Last updated 6 min readTags:MixingMasteringDistortion

Learn what clipping does to audio, how digital and analogue clipping differ, and how to find and fix unwanted distortion in your recordings and exports.

Clipping happens when a signal exceeds a system’s level limit and its peaks are cut off or rounded. It changes the waveform and adds distortion. You may hear crackling, a fizzy edge or less impact on drum hits, though brief clipping is not always obvious.

Digital clipping is often described as harsh and analogue clipping as warm. Either can sound unpleasant. The result depends on the clipping characteristic, how far you drive it and the source material. Digital processing has an additional concern: aliasing, which can introduce unwanted frequencies.

Hard clipping and soft clipping

Hard clipping cuts off the signal at a fixed threshold. Below that threshold, the waveform passes unchanged; above it, the output cannot rise any further. A sine wave develops flat tops and bottoms as it is driven harder.

Soft clipping makes the transition more gradual. The peaks become rounded as the signal approaches the limit. It usually produces less high-frequency distortion than an abrupt hard clip at a comparable amount of peak reduction.[1]

Both types add harmonics to a sine wave. The difference is not simply “hard clipping makes odd harmonics; soft clipping makes even ones”. A symmetrical soft clipper also produces odd harmonics. Asymmetry, where positive and negative parts of the waveform are treated differently, can introduce even harmonics.[2]

Hard and soft describe the clipping characteristic. Both can occur in software and analogue circuits.

Where digital clipping happens

In conventional 16-bit or 24-bit integer audio, 0 dBFS is the maximum sample level. A signal that exceeds the available range must be reduced before conversion to that format; otherwise, its peaks can be clipped.

Inside a DAW, the situation is different. Many audio engines use floating-point processing, which can carry levels above 0 dBFS without cutting off the peaks. Ableton Live, for example, has substantial internal floating-point headroom.[3]

The important question is where the signal meets a limit:

Stage What to check
Recording input An overloaded preamp or analogue-to-digital converter may distort before the audio reaches the DAW. Reduce the recording level at the input.
Plug-in A plug-in may deliberately saturate or clip, even when the DAW itself has spare headroom. Check its input level and processing settings.
Hardware output A signal above full scale needs attenuation before conversion to analogue. Internal floating-point headroom does not extend to the physical output.
Integer export Check the level before export to a 16-bit or 24-bit file. Peaks above the format’s range cannot be preserved unchanged.

A 32-bit float export can preserve samples above 0 dBFS. You can subsequently reduce their level without losing the original peaks, provided they were not already clipped earlier in the chain. Playback still requires a suitable level.[4]

A red meter therefore needs context. It may indicate an internal level above 0 dBFS, or it may identify an actual overload. It does not mean the same thing at every point in the session.

How analogue clipping differs

Analogue equipment has electrical limits. An amplifier cannot produce an indefinitely large output voltage; beyond its available headroom, the waveform becomes distorted.

Some circuits approach this limit gradually. Others clip abruptly. A valve stage is not guaranteed to sound gentle, and a transistor stage is not guaranteed to sound harsh. The circuit, operating level and material all matter.

Analogue clipping itself has no digital sample rate, so it does not create aliasing. It can still produce strong distortion. A digital recording of an overloaded analogue signal preserves that distortion; turning the recording down afterwards does not remove it.

Why aliasing matters in a digital clipper

Clipping generates higher-frequency content. In digital processing, content above half the sample rate, the Nyquist frequency, can appear as lower frequencies through aliasing. Those frequencies need not belong to the original harmonic series.[5]

For example, a 48 kHz system has a Nyquist frequency of 24 kHz. A 30 kHz component generated during clipping aliases to 18 kHz. On a 10 kHz test tone, that 18 kHz component is not a harmonic of the input.

Oversampling runs the clipping process at a higher internal sample rate. A filter removes content above the final bandwidth before the signal returns to the session’s sample rate. This reduces aliasing, though it does not guarantee its complete removal.[5]

Oversampling also uses more CPU, and its filters can affect the result. Compare the available settings on the actual audio. It cannot undo clipping already present in a recording or remove the distortion caused by excessive drive.[6]

How to find and fix unwanted clipping

Start with a passage where the problem is clearly audible, such as a loud snare hit or vocal phrase. Keep the listening level comfortable and compare changes at roughly equal loudness.

  1. Check the source. Bypass processing and listen to the original recording. If it is already distorted, a new recording at a lower input level is the most reliable remedy.
  2. Find where the distortion appears. Check the track, group buses, master processing and output in signal order. Bypass one suspect processor at a time.
  3. Reduce the level before the overloaded stage. For a plug-in, lower its input or place a gain control before it. A fader after the processor only makes its distorted output quieter.
  4. Check the export. Import the rendered file into a clean session and listen to the same passage. If the problem appears only in the export, check its format, output level and any processing applied during rendering.

For example, if a drum bus becomes brittle when you increase the clipper’s drive, reduce that drive first. Then compare hard and soft settings, followed by oversampling. Match the output level after each change. A cleaner result with oversampling suggests aliasing contributed, but the setting can alter other aspects of the sound too.

If the peaks have already been cut off, a simple gain reduction cannot restore them. De-clipping software may estimate a repair; it cannot guarantee recovery of the original waveform.

When deliberate clipping is useful

A clipper can reduce brief drum peaks before a limiter. The limiter then receives a signal with smaller peaks, though the clipper adds its own distortion.

Increase the drive gradually and listen for changes to the attack, cymbals and vocal consonants. Stop when the added distortion or loss of impact outweighs the benefit. There is no clipping amount that suits every source.

Check the final export with a true-peak meter. A sample ceiling below 0 dBFS does not guarantee that reconstructed peaks stay below full scale. Oversampling and true-peak control serve related but different purposes; one does not automatically provide the other.[6]

The dynamic range control guide explains how clipping fits alongside compression and limiting.

What Sonalix’s peak readings tell you

A true-peak reading identifies the highest estimated waveform level and whether there is a headroom concern. It does not establish that audible clipping has occurred.[7]

Likewise, a file below full scale may contain clipping from an earlier stage. Use Sonalix’s measurements to guide your checks, then inspect the session and listen to the affected passage. The per-metric severity guide explains how to interpret the ratings.

Sources

  1. Julius O. Smith III, Physical Audio Signal Processing: Nonlinear distortion, including hard and soft clipping.
  2. Julius O. Smith III, Physical Audio Signal Processing: Enhancing even harmonics.
  3. Ableton Live manual: Mixing.
  4. Audacity manual: Digital audio fundamentals.
  5. Jatin Chowdhury: Practical considerations for antiderivative anti-aliasing.
  6. FabFilter Pro-L 2 help: Oversampling.
  7. Sonalix: Per-metric severity.

Was this page helpful?

    Type to search titles, headings and summaries.