A clipper takes everything above a ceiling and flattens it. That’s the whole operation. The peaks stop where you tell them to, the rest of the waveform is untouched, and the level you gain is the difference between the old peak and the new one.
Because peaks are short and rare, you can usually take several decibels off them before anything sounds wrong. That headroom is what a clipper buys you: a track that measures the same peak level and reads several LUFS louder, without a compressor’s pumping.
What it costs is harmonic distortion at the corner. The sharper the corner, the more of it, and the brighter the result.
Why mid/side clipping isn’t the same as stereo clipping
Set a stereo clipper and a mid/side clipper to identical settings and they do different things to the same file. This isn’t a subtlety in the implementation. It’s arithmetic.
Clipping is nonlinear, and a nonlinear operation doesn’t commute with the mid/side matrix. Flattening left and right and then computing mid and side gives a different signal from computing mid and side, flattening those, and summing back. The order changes the answer.
In practice: clipping left and right protects each speaker’s peak. Clipping mid and side protects the centre’s peak and the difference signal’s peak independently, which means a loud centred kick no longer forces the wide reverb through the same corner it does. The stereo image survives harder settings.
That’s the reason the mid/side switch is a mode rather than a convenience.
Which shape
Shapes differ in how the curve approaches the ceiling, and the approach is audible.
Limit reaches the ceiling by pulling gain down instead of bending the waveform, so it colours least. It’s the everyday setting and the one to start from when you want level without character.
Soft meets the line at unity slope and arrives flat. Forgiving, with a bit of harmonic colour as it catches transients. Good on a mix bus.
Hard is a corner. It’s the brightest of the shapes because a corner generates the most harmonics, and it’s the best at raising perceived loudness. It’s also the easiest to overdo on a full mix.
Sine takes the longest, smoothest route to the ceiling of the shapes that clip. Between Soft and Limit in character.
Fold doesn’t clamp at all. Past the quarter turn the waveform folds back on itself. This is a distortion effect rather than a peak control, and it’s the reason the side lane wants its own shape selector. Fold on the sides with Limit on the mid gives cymbals and room an edge while the vocal and the kick keep their shape.
The setting that catches people out
The five shapes are level-matched by construction: each one passes signal untouched below the knee and arrives there at unity slope. Switching shape changes what happens to the peaks and not how loud the track is.
That’s deliberate, and it means the shape comparison is honest. It also means you can’t tell the shapes apart by whether one sounds louder, which is how most people compare saturation. Listen to the transients instead: the snare edge, the pick attack, the top of a hi-hat.
If you A/B two shapes and one sounds better, it’s better. If one sounds louder, something is level-matching them badly and the comparison is worthless.
Oversampling, and why the ceiling isn’t at zero
A corner generates harmonics above the audible band. At 44.1 kHz there’s nowhere for them to go, so they fold back down as inharmonic content in the top octaves, heard as a harsh fizz. Oversampling gives them room. Two times handles most material; four is worth it on bright, dense mixes.
The ceiling wants to sit a little below zero — around -0.3 dB — because a converter reconstructs the waveform between the samples you clipped, and the reconstructed peak can exceed the sample peak. That headroom is what stops an encoder or a DAC from clipping again on the way out, after you’ve already decided where the ceiling is.
How much is too much
There’s no number. The signs are:
- The snare loses its front edge and starts to sound like it’s behind the mix.
- The bass note stops moving. Sustained low content is where flattening shows up as a change in pitch content rather than a change in level.
- The mix gets brighter without you touching an EQ. That’s harmonic distortion, and past a point it reads as fatigue rather than presence.
Blend it back with Mix if you’ve gone past. Parallel clipping keeps the original transients underneath as an anchor, which is a better safety net than backing the drive off, because it lets you keep the level you gained.
A starting point
- Drive at zero. A clipper that arrives already clipping can’t be compared against not having it.
- Ceiling at -0.3 dB, oversampling at 2x, Limit on both lanes.
- Raise the drive until you can hear it, then back off until you can’t.
- Switch to mid/side and try a more aggressive shape on the side lane alone. This is where the mode pays for itself.
- Check mono. Independent clipping across mid and side can move the balance in ways that only show up when the sides cancel.