What is headroom in audio?

Headroom in plain English

The word comes from the space above your head in a room. Audio has a ceiling too. In a digital file it's 0 dBFS, the highest level a 16-bit or 24-bit file can store, and anything that tries to go above it gets cut off flat, which is called clipping. Headroom is how far below that ceiling your loudest moment sits.

You measure it from the peak, the single highest point of the waveform, because that's the part that would hit the ceiling first. If the loudest snare hit in your mix peaks at −8 dBFS, the mix has 8 dB of headroom. Turn it up 8 dB and the headroom is gone; turn it up 9 and that snare clips.

There's a floor as well. Every recording system has some noise, from the microphone preamp, the converter or the limits of the file format, and that's called the noise floor. Your music lives in the space between the two. Headroom is the part you keep free at the top, so that a louder note than you expected, or a boost you make later, still fits.

Headroom says nothing about how loud something sounds. Two tracks can both peak at −1 dBFS and differ in loudness by 10 dB, because loudness depends on the average level under the peaks. That's measured in LUFS, which the LUFS lesson covers in full. This page is about the peaks and the room above them.

Reading the lab

The lab plays a loop through a gain control and a limiter and measures what comes out. For this lesson, the reading to watch is TP, the true peak meter next to the big screen. The number under it is the highest point the loop reaches anywhere, in dBTP, and the headroom is the distance from that number up to 0. The Headroom figure in the Right now panel does that subtraction for you, so a true peak of −6.2 dBTP reads as 6.2.

The crest factor screen shows the same thing as a picture. For each version of the loop, the pink line is its highest sample peak and the amber block is its average level. The dashed pink line near the top is 0 dBFS, so the gap between the pink line and the dashed line is your headroom, and the gap between the pink line and the amber block is the crest factor, which decides how much headroom a sound needs.

Gain is plain volume, so every loudness and peak reading moves by exactly the amount you turn it. The Limiter knob turns the loop up into a limiter with a ceiling at −1 dBTP, which is how a master is made loud. Streaming is switched off here. The Right now panel describes what your settings are doing, in numbers measured from the loop you're hearing.

The drum loop comes from Wonky Percussion and the bass from Flow, two Sounds Like You packs, and the full mix is built from both. The vocal is a public domain recording from Freesound and the sine wave is made in your browser. You can also load one of your own sounds with the Your sound button. The lab turns every loop up or down so its peaks land close to 0 dBFS, so the readings won't match your file's own peak level.

How much headroom do you need?

It depends on the stage, because the risk changes. While you record, the danger is a performance that gets louder than you planned; during the mix, it's the plug-ins and buses that add level as you work. On the final master, the only thing left to protect against is what happens when the file is played or encoded.

Treat these as starting points. In a floating-point DAW, a channel peaking at −4 dBFS is fine; it's the recording and the final file that have hard limits, and they're the two places where clipping can't be taken back.

Headroom when recording

Recording is where you need headroom most, because the converter in your audio interface turns the signal into a fixed-point file with a hard top. If a vocal hits that top, the peak is gone. You can turn the take down afterwards, but the flat top is part of the recording now, and you'll hear it as a crackle or a harsh edge on the loudest words.

The old advice to record as hot as possible came from 16-bit recording and tape, where a quiet signal sat close to the noise. A 16-bit file has a theoretical range of about 96 dB between its noise floor and 0 dBFS. A 24-bit file has about 144 dB, far more than any converter can deliver; the best ones manage around 120 dB. That means a 24-bit recording peaking at −10 dBFS still has over 100 dB of range below it, which is more than you can hear in any normal room.

So set the preamp gain with the performer singing or playing as loud as they will in the take, and aim for peaks around −10 to −6 dBFS. Suppose a singer's loudest line peaks at −14 dBFS in the soundcheck, and in the take they push harder and the same line comes in 8 dB higher. It peaks at −6 dBFS, with room to spare. Set the soundcheck peak at −4 dBFS instead and that take clips by 4 dB.

The sample rate and bit depth lesson goes further into why 24-bit gives you this extra room.

Headroom inside a DAW

Inside your DAW the rules are looser. Ableton Live, Logic, Pro Tools and FL Studio all mix in 32-bit or 64-bit floating point, a number format that can hold values far above 0 dBFS. A channel meter showing red inside the mixer isn't clipping by itself; pull the master fader down by the same amount and the peaks come back intact.

The ceiling comes back wherever the audio leaves floating point. Your master output becomes fixed-point at the interface's converter, so anything over 0 dBFS clips on its way to your speakers. A 16-bit or 24-bit WAV export clips it too, while a 32-bit float WAV keeps it. And some plug-ins clip on purpose or by design, like a limiter, a clipper, or a plug-in modelled on analogue hardware that saturates when you drive it.

Those analogue-modelled plug-ins are the reason to keep channel levels moderate even in floating point. Many of them are set up so that 0 VU on their meter sits at around −18 dBFS, which matches how studio gear is lined up. Feed one a vocal averaging −6 dBFS and it behaves like a real compressor or console driven 12 dB hotter than normal, which may add distortion you didn't ask for.

Keeping each stage at a sensible level, from the preamp to the master bus, is called gain staging, and it has its own lesson: gain staging.

Sounds that need more headroom

Two sounds peaking at the same level can differ in loudness by more than 10 dB. What decides it is the crest factor: how far a sound's peaks sit above its average, or RMS, level. A sound with a high crest factor has short spikes that reach the top while the rest of it is much quieter, so it uses up headroom without sounding loud.

In the lab, the drum loop peaks at −0.7 dBTP and reads −15.8 LUFS, with its peaks 17.6 dB above its average level. A sine wave peaking at −0.9 dBTP reads −2.5 LUFS, because its peaks sit only 3.0 dB above its average, which makes it more than 13 LU louder though both have under 1 dB of headroom left.

That's why drums and percussion fill a mix's headroom first, and why the mix bus meter often jumps on the kick and snare while the pads and vocals barely move it. Each drum hit starts with a transient a few milliseconds long that reaches far above the body of the sound. Sustained sounds like pads and held synth notes have low crest factors, so they get loud without using much headroom.

The full mix in the lab has a crest factor of 16.2 dB. If you mixed it so that its average level sat at −18 dBFS, its peaks would land at about −2 dBFS, which is why a sensible average level and a sensible amount of headroom usually go together. The dynamic range and crest factor lesson goes further into the measurement.

True peak headroom

A digital file only stores samples, and the waveform your speakers play is the smooth curve drawn through them. That curve can rise higher between two samples than either sample does, so a file whose highest sample is at −0.3 dBFS can produce a wave that reaches above 0 dB once it's turned back into sound. A true peak meter works out those in-between peaks, usually by calculating four points for every sample, and reports them in dBTP.

The bass loop in the lab is a real example. Its highest sample is at −0.5 dBFS, so a sample peak meter says there's half a decibel of headroom. The true peak meter reads 0.0 dBTP, and that headroom isn't there.

This matters most for the final master. Streaming services encode your file to AAC, Ogg Vorbis or MP3, and encoding can push peaks higher still. Spotify's advice to artists is to keep the true peak under −1 dBTP, and under −2 dBTP for masters louder than −14 LUFS, because loud masters are the ones most likely to distort when they're encoded. Set your final limiter's ceiling to −1 dBTP with its true peak option switched on and you've left that headroom. The true peak lesson covers the measurement in detail.

Headroom for mastering

A mix leaves headroom so mastering has something to work with. Say your mix has a true peak of −6 dBTP. The mastering engineer can turn it up by 5 dB before those peaks reach a −1 dBTP ceiling, and that part costs the sound nothing. To go louder still, the last processor in the chain, usually a limiter, holds the peaks at the ceiling while everything under them keeps rising.

The lab's last step shows that second part. The full mix already peaks at −0.2 dBTP, and the Limiter knob pushes it 6 dB into a −1 dBTP ceiling. The limiter turns the loudest moments down by up to 6.9 dB, the true peak settles at −1.1 dBTP and the mix comes out 3.2 LU louder. That loudness comes out of the crest factor, which shrinks from 16.2 to 12.2 dB. The limiting lesson shows how far you can push that before it hurts the sound.

For a mastering engineer, the precise headroom figure on your mix doesn't change much. In floating point they can turn a mix peaking at −10 dBFS up by 5 dB without losing anything. What causes trouble is a mix that has already clipped somewhere, or one with a limiter left on the master bus, because the limiter has already flattened the peaks and there's nothing left to shape. Send a mix peaking around −6 to −3 dBFS with nothing limiting it, exported as a 24-bit or 32-bit float WAV, and it will be easy to work with.

Headroom in analogue gear

Analogue equipment has headroom too, measured differently. Professional gear works at a nominal level of +4 dBu, the level its meters show as 0 VU, and most of it can go a good deal higher before it distorts, often to somewhere above +20 dBu. The difference between the two, often around 20 dB, is that device's headroom. Analogue distortion also tends to come on gradually as the level rises, where a digital file clips at one exact level.

When a studio connects analogue gear to digital converters, it picks a digital level to match 0 VU. The European broadcast standard EBU R68 lines 0 dBu up with −18 dBFS, and the American SMPTE RP155 lines +4 dBu up with −20 dBFS. Either way, a signal sitting at its normal analogue level has 18 to 20 dB of headroom in the digital file, which is why so many plug-ins treat −18 dBFS as their normal level. The dBFS vs dBu vs dBV comparison explains those units.

How to leave headroom in a mix, step by step

1. Record with room to spareSet each input so the loudest part of the performance peaks around −10 to −6 dBFS on a 24-bit recording.
2. Trim before you mixIf a sample or a recording comes in peaking near 0 dBFS, turn it down with the clip gain or a utility plug-in at the top of the channel, before any other plug-ins, so they all receive a moderate level.
3. Start the mix with the loudest partSet the drums or whatever drives the track so the mix bus peaks around −10 dBFS, then bring everything else in around them.
4. Watch the mix bus as you add partsEvery new part adds level. If the bus starts peaking above −3 dBFS, select the channel faders and pull them all down by the same amount, which keeps the balance you've built.
5. Turn your monitors up to compensateA mix with 6 dB of headroom sounds 6 dB quieter than a mastered track. Raise the speaker or headphone level so you judge it at a comfortable volume.
6. Check the exportBounce at 24-bit or 32-bit float, look at the peak and true peak readings, and make sure nothing on the master bus is limiting or clipping.

Common headroom mistakes

Recording as hot as possible

A big waveform on screen looks reassuring, but on a 24-bit recording it buys you nothing except a greater chance of clipping. If a take has clipped, the only real fix is another take; declipping plug-ins can rebuild some of the peaks, and they work best when the clipping is light.

Trusting the sample peak meter on a master

A master that reads −0.1 dBFS on an ordinary peak meter can have true peaks above 0 dB, and those are the ones that distort after encoding. Check with a true peak meter, or set your limiter to true peak mode.

Leaving a limiter on the mix bus for mastering

It's tempting to send the version that sounds as loud as a release. That version has already lost its headroom and its peaks, and the mastering engineer can't get them back. Send the mix without the limiter, and send the limited one as a reference if you want them to hear what you liked.

Mistaking a quieter mix for a weaker one

When you turn a mix down to leave headroom, it sounds smaller, simply because quieter music sounds less exciting. Match the level with your monitor control before you decide anything about the balance.

Turning the master fader down to hide a clip

If a plug-in on a channel or bus is clipping, pulling the master fader down makes the meter go green, but the clipping happened earlier and it's still in the sound. Turn down the input to whatever is clipping.

Train your ears

Peak meters are good at protecting headroom and poor at telling you which version is louder. The quiz under the lab plays two versions of a loop with the screens covered and asks which has the higher integrated loudness. In some rounds the louder one has been through the limiter and has lower peaks, so a peak meter would point you the wrong way. Eight rounds take a few minutes, and your best score is saved in your browser.

Questions people ask

Keep going