Limiting vs clipping

At a glance

What's the difference between limiting and clipping?

Both tools have a ceiling, a level the sound may not pass, and both are there to stop peaks. They differ in what they do to a peak that reaches it.

A clipper works on one sample at a time. If a sample is under the ceiling it passes untouched, and if it's over, the clipper replaces it with the ceiling value (a hard clipper) or bends it down on a curve (a soft clipper). Nothing else is touched, and because no time is involved a clipper has no attack or release controls and needs no lookahead. It does change the shape of the waveform: loud peaks get flat tops. This page sticks mostly to hard clipping; the gentler, rounded kind is covered in soft clipping vs hard clipping vs limiting.

A limiter works on the level over time. It measures how far a peak is going to go over, turns the whole signal down by that much, then lets the level rise again over its release time. Most limiters use lookahead, which means delaying the audio slightly so the turning down can start before the peak arrives; the one in this lab looks 5 ms ahead and recovers with an 80 ms release. The wave keeps its shape, and the price is that quieter sounds near each peak get turned down with it. A limiter is a compressor taken to the extreme, with a ratio so high that nothing passes, and the limiting lesson goes through its controls one by one.

Reading the lab

The lab is set to 9 dB of drive for every step. Drive turns the sound up before it reaches the ceiling, so with 9 dB the loudest hits of the drum loop would land at about +8 dBFS if nothing stopped them. Button 3 is the hard clipper and button 4 the limiter, and you can press those keys while it plays.

On the big screen the pale shape is the loop after drive and the pink shading is everything over the ceiling. The coloured shape in front is what comes out. With the clipper you'll see it stop dead at the dashed line; with the limiter the whole shape shrinks around the loud hits. The close-up shows one of those hits sample by sample, which is where the two look least alike.

Match loudness is on unless a step turns it off. It plays each version at the same loudness as the original, so you hear the change in shape without a change in level fooling you. The written readout in the lab measures what's happening on the sound you've picked, including how much louder it is at the same peak.

Hear the limiter on the drums Hear the clipper on the drums

Is clipping louder than limiting?

At the same peak level, yes, and on drums by a lot. Run the drum loop through both with 9 dB of drive and the hard clipper comes out 7.0 dB louder than the original while the limiter manages 3.9 dB. Measured in LUFS, the loudness unit streaming services use, the original sits at −19.9 LUFS, the limited version at −14.9 and the clipped version at −11.9. All three peak at or just under 0 dBFS.

The reason is in what each one touches. Only 2.4% of the samples go over the ceiling, all of them at the tops of the loudest hits. The clipper changes that 2.4% and leaves the other 97.6% exactly as it was. The limiter starts turning down before each of those hits, because of the lookahead, and takes around 80 ms to come back up, so the tail of the kick and whatever plays just after it get turned down too. That costs average level, and average level is most of what makes something sound loud.

The lab measures this with the crest factor, the gap between the peak level and the RMS (average) level. The drum loop starts with a crest factor of 18.6 dB. Anything that shrinks it lets you raise the average while the peak stays put, and the clipper shrinks it fastest. The same test on the other loops in the lab gives these figures, each measured on the loop as the lab plays it:

The clipper wins every row, but its lead is biggest on drums and smallest on the vocal. Drum peaks are short and separated by gaps, which is the worst case for a limiter: it ducks everything for tens of milliseconds to catch a peak a few milliseconds long. A vocal's level rises and falls over whole syllables, so the limiter's gain moves with the words and wastes less.

Compare them on the full mix

Which sounds better, clipping or limiting?

Neither, in general. They go wrong in different ways, and the one that sounds better is the one whose failure you can't hear on that sound.

A clipper fails by distorting. Cutting the top off a wave adds harmonics, new frequencies at whole-number multiples of the note. Put the lab's 110 Hz sine through the hard clipper with 9 dB of drive and it gains a harmonic at 330 Hz only 12 dB below the note, another at 550 Hz 21 dB down, and more all the way up. On a drum hit those harmonics last as long as the clipped part, a millisecond or two, and mostly come across as a harder, slightly crunchier attack. On a held note they last as long as the note, and you hear them as fizz or buzz.

A limiter fails by moving the level. Push it and you hear pumping, the quiet parts of the mix dipping after every loud hit and swelling back, and the hits themselves lose their snap because the limiter is at its lowest gain exactly when they land. On the steady sine the limiter adds almost nothing, because it settles on one gain and stays there, which is why it's the better choice for anything sustained.

The limiter can distort too. If its release is faster than one cycle of a bass note (a 40 Hz note takes 25 ms per cycle), the gain wobbles along with the wave and bends its shape. Most limiters keep the release long enough by default, and the lab's 80 ms is safely above that.

Hear a clipped vocal Hear it limited See the harmonics on a sine

Clipping, limiting and true peaks

A digital file stores samples, and a speaker plays the smooth wave drawn through them. That wave can rise above the highest sample, so a file whose samples never pass 0 dBFS can still have a true peak above it. Those overs can distort when the file is turned into an MP3 or AAC stream, or in a cheap converter.

Clipping makes this worse. A flat top with sharp corners is exactly the shape that overshoots when it's redrawn. The hard-clipped drums from the lab, measured with a true peak meter, read 0.0 dBFS on their samples and +0.9 dBTP between them. The lab's limiter, which only watches samples, read +0.3 dBTP.

A mastering limiter fixes this with a true peak mode, which oversamples the signal so it can see and catch the peaks between samples; FabFilter Pro-L 2 has one, and so do most current mastering limiters. A clipper can only catch them by cutting the shape again, which brings back the corners that caused them. So on a master the clipper goes somewhere before the end, and the limiter, with true peak detection on, goes last.

Clipping also adds harmonics above the highest frequency the sample rate can hold, and those fold back down as aliasing, extra tones that aren't in tune with the music. Clipper plug-ins with an oversampling setting avoid most of it, so switch it on when you push one hard.

When to use a clipper

Reach for a clipper when the peaks are short. A drum transient lasts a few milliseconds, so cutting 2 or 3 dB off it adds harmonics for a few milliseconds, and most listeners hear a slightly harder drum with no obvious distortion. It also costs no loudness between the hits, which a limiter can't manage.

• On a drum bus, to take the top off the loudest hits before they reach the master.
• Straight on a kick or snare channel, when one or two hits poke out far above the rest.
• Before the limiter on a loud master, so the limiter has fewer fast peaks to duck for.
• On anything you want crunchier, where the distortion is the point. That's closer to saturation than peak control.

Dedicated clipper plug-ins include KClip from Kazrog and StandardCLIP from SIR Audio Tools, and Ableton Live's Saturator has a Soft Clip switch.

When to use a limiter

A limiter is the right tool when the peaks are wide, the sound is sustained, or the result has to stay clean whatever arrives. That covers the very last plug-in on almost every master, because it's the only one of the two that can hold a true peak ceiling. It also covers vocals, pads, strings and held bass notes, where a clipper's harmonics last long enough to hear.

A limiter is also the safer choice when you can't predict the input, such as a live stream, a podcast bus or a sample you'll pitch around. A clipper that sounds fine on the material you set it up with can turn nasty on a louder or longer note, whereas a limiter's worst case is a moment of pumping.

Keep the gain reduction modest. On a full mix, a limiter catching 2 to 3 dB of peaks usually goes unnoticed; past about 6 dB you'll usually hear it working on a dense mix, and the gain reduction lesson shows how to read the meter while you set it.

How to use a clipper and a limiter together

The clipper is better at drum peaks and the limiter is better at everything else, so the usual master chain gives each the part it's good at. Split the work and neither has to do enough to be heard.

1. Find out how much you needPut a limiter on the master on its own and push it until the track is as loud as you want. Note the gain reduction on the loudest hits; say it's 6 dB.
2. Put a clipper in front of itInsert a clipper before the limiter with its ceiling at 0 dBFS. Raise its input gain until it's taking 2 to 3 dB off the loudest drum hits, then listen to the snare and back off as soon as you hear crunch.
3. Bring the limiter back downThe clipper has removed the fastest peaks, so the limiter needs less. Lower its input until it's catching 2 to 3 dB at the same final loudness.
4. Set the ceilingSet the limiter's ceiling to −1 dBTP with true peak detection on. Spotify recommends staying under −1 dBTP, and under −2 dBTP for masters louder than −14 LUFS.
5. Check at matched loudnessBounce a version with the limiter only and one with both, match their loudness, and pick by ear. On drum-heavy music the version with the clipper usually keeps more punch; on a ballad it may make no difference at all.

The same split works one level down. Clip the drum bus by a couple of decibels and the master limiter has far less to catch, because the drums are where most of a mix's sharpest peaks come from. The mastering chain lesson shows where both sit among the other processors.

Is a clipper the same as clipping on export?

The maths of a hard clipper at 0 dBFS is the same as a file that clips on export: everything over the line is cut flat. The difference is control. With a plug-in you choose the ceiling, how hard you drive it, whether it rounds or cuts and whether it oversamples, and you can see how much it's taking off.

Most DAWs mix in floating point, which can hold levels far above 0 dBFS, so a red light on the master doesn't distort anything inside the session. It does when the mix is written to a 16-bit or 24-bit WAV, or reaches your audio interface, because those stop at 0 dBFS. A limiter last on the master, set to a ceiling under zero, stops that from happening by accident.

Common mistakes

Asking one of them to do everything

A limiter taking 8 dB off a drum-heavy mix pumps, and a clipper taking 8 dB off it crunches. Split the work so each is under 3 dB, as in the steps above.

Clipping sustained sounds

A clipper treats a 40 Hz sub note the same as a 2 ms snare hit, so a long bass note gets flattened for its whole length and turns into buzz. Clip the drums on their own bus before they meet the bass, or limit the bass instead.

Putting the clipper after the limiter

The limiter's true peak ceiling only holds if nothing changes the signal after it. Drive a clipper after it and the sharp corners come straight back, overs and all, so the limiter goes last.

Deciding by loudness

At the same peak level the clipper is louder, and louder wins a quick comparison almost every time. With Match loudness off in the lab, the clipped drums beat the limited ones for no reason except their extra 3 dB. Compare at matched loudness.

Hear it without Match loudness

Training your ears to tell them apart

At matched loudness, a clipper changes the front edge of each hit and a limiter changes what happens just after it. The quiz under the lab plays the selected sound, drums to start with, through either the hard clipper or the limiter at matched loudness, with the screens covered, and asks which one you heard. For the limiter, listen to the gaps between the hits, where the quieter sounds dip after each loud one. For the hard clipper, listen to the very start of the snare for a papery crackle.

Questions people ask

Keep going