How to build a mastering chain

What is a mastering chain?

Mastering is the last stage before a song is released. You take the finished mix, a single stereo file, and make small changes to the whole thing at once, mostly to its tone and its level, so it sits at a sensible loudness next to other records. The mastering chain is the set of plug-ins that does it, in the order the audio passes through them.

Because everything goes through every stage, each move is bigger than it looks. Lift 10 kHz by 2 dB on a vocal and only the vocal changes; do it on a master and the hi-hats, the vocal, the reverb and the synths all get brighter together. That's why mastering settings are so much smaller than mixing settings. The difference between the two jobs is covered in mixing vs mastering; this page is about building the chain itself.

A typical chain has four stages: EQ for tone, a compressor to hold the mix together, saturation or a soft clipper to take the edge off the tallest peaks, and a limiter at the end to set the final loudness and the ceiling. The lab above has those four, in that order, running on the full mix loop.

Reading the lab

The four buttons under the header are the chain. Click one to switch that stage on or off, and the line under its name says what it's doing to this loop right now. The knobs for each stage sit in their own group below the screens.

The big screen shows the whole loop. The pale shape is the mix going in and the pink shape is the master coming out, both at their real levels, so as you turn the limiter up the pink shape gets fatter and flatter against the dashed ceiling. The amber lane underneath adds up the gain reduction from the compressor and the limiter.

The loudness screen is the one to watch. Each column is a version of the loop: the amber block rises to its integrated loudness in LUFS and the pink line marks its true peak. The left pair is the original and your master as you'd export them. The right pair is the same two files after Spotify's normalisation, which is how most people will hear them.

Match loudness is on by default. It plays the master at the same loudness as the original, so the A/B button tells you what the chain did to the sound with the extra loudness taken out. The panel on the right explains the stage you last touched, using numbers measured from the loop, and every reading on the page comes from the same processing you're hearing.

The full mix is built from Sounds Like You loops, but a mastering chain makes most sense on a song of your own. Load a bounce of your mix with the + Your sound button and it runs through the chain in place of ours. The lab sums it to mono and loops the first 30 seconds, and the file stays on your device.

Mastering chain order

Order changes the result because each stage reacts to whatever the one before it sends. A compressor after an EQ hears the EQ'd mix; a limiter at the end hears everything. The order below is the one most masters use, and the reasons hold even when you swap individual plug-ins.

The limiter goes last because its job is to stop anything coming out above the ceiling. An EQ boost after it raises the peaks the limiter just held down, so they can pass the ceiling again. The same goes for saturation, a stereo widener or anything else that changes the level.

EQ goes before the compressor so the compressor isn't reacting to problems you're about to remove. Leave a 30 Hz rumble in and the compressor turns the whole mix down every time it swells. The EQ before or after compression lesson covers the cases where the other order wins, which mostly come up on single tracks.

The soft clipper sits just before the limiter because the two split one job. The clipper shaves the fastest peaks, which are mostly drum hits, and the limiter catches what's left without having to turn the whole mix down as hard.

Getting the mix ready

Mastering can't fix a mix. If the vocal is too quiet, turning up 3 kHz on the master brightens everything else along with it, so go back and change the vocal level. The best thing you can do for a master is listen to the mix on a second system, like headphones or a phone speaker, and fix what you hear there first.

Take any limiter off the mix bus before you start, so the mastering chain gets the real peaks. Then check the level: the loudest peaks of the mix should sit somewhere below 0 dBFS, and a few decibels of headroom gives every stage a sensible level to work with. Plug-ins modelled on analogue gear, like tape and valve saturators, react to how hard you hit them, so the input level changes their sound.

Pick a commercial track in a similar style to compare with as you go. Turn it down until it matches the loudness of your master before you compare, or you'll end up chasing its level. The reference tracks lesson shows how to set that up.

Building the chain, stage by stage

EQ

Mastering EQ works on the balance of the whole record, like a touch more weight under the kick or a little more air on top, and the moves are small. A 1 dB shelf on a master is a clear change, and if you find yourself reaching for 4 or 5 dB, the problem belongs back in the mix.

Start with a low cut, a high-pass filter set somewhere between 20 and 30 Hz. Very little music lives down there, and on a lot of mixes that range holds rumble and stray energy you can't hear on most speakers. On the lab's full mix, a 30 Hz cut takes that region down by about 1.7 dB and barely touches the kick. Then use wide shelves or wide bells for tone. The lab has a low shelf at 120 Hz for the weight of the kick and bass, and an air shelf at 10 kHz for cymbals and breath.

Some mastering EQs have extra modes. A linear-phase EQ doesn't shift some frequencies in time against others, which some people prefer on a whole mix (see linear-phase EQ), and a mid-side EQ lets you treat the centre and the sides of the stereo image separately, covered in mid-side EQ. Neither is needed to make a good master.

Compression

A compressor on a master moves the whole mix up and down together, which is what engineers mean by glue. Because it reacts to everything at once, a loud kick turns the vocal and the pads down with it, so keep it gentle: a ratio of 1.5:1 to 2:1 and a threshold set for 1 to 2 dB of gain reduction on the loudest hits.

The numbers are easy to work out. At 2:1, a hit that goes 4 dB over the threshold comes out 2 dB over, so the compressor takes 2 dB off it. In the lab, a −14 dB threshold at 2:1 gives up to about 1.9 dB of gain reduction on the full mix, and the compressor is working by more than 1 dB for about a quarter of the loop.

Use a slow attack, around 10 to 30 ms, so the front of each drum hit gets through before the compressor reacts; the lab's is fixed at 30 ms. Set the release so the gain recovers between beats, or use an auto release if the compressor has one. Watch the meter: if it never gets back to zero, the compressor is turning the whole song down and you'll start to hear pumping. The VCA bus compressor from SSL consoles is a common choice for this job, and so are slow valve designs like the Manley Variable Mu. The bus compression lesson goes further into the settings.

Saturation and soft clipping

This stage is optional, and it's the easiest one to overdo. Saturation from a tape or valve plug-in rounds off the peaks and adds harmonics, which makes a mix feel denser and a little louder for the same peak level. A soft clipper does a similar thing with less colour: it bends the very tops of the tallest peaks and leaves everything under them alone.

On a master the target is a decibel or two off the peaks, no more. In the lab, 6 dB of drive into the soft clipper shaves about 0.8 dB off the tallest drum hits, and the distortion it adds sits around 35 dB below the mix, where it's hard to hear. Turn the drive up and watch the coach: once the distortion climbs past about −30 dB, it warns that you may start to hear grit on the drums.

The reason to bother is the limiter. Every decibel the clipper takes off the drum transients is a decibel the limiter doesn't have to turn the whole mix down for. The comparison in soft clipping vs hard clipping vs limiting shows how the two fail differently when you push them.

Limiting

The limiter sets the loudness of the master. Its gain, which some limiters call input or threshold, turns the mix up into it, and its ceiling is the highest level it lets out. Peaks stay at the ceiling and everything underneath gets louder, so the gap between the peaks and the average level shrinks; that gap is the crest factor.

Set the ceiling to −1 dBTP and switch on true peak detection if your limiter has it. A true peak is the level the wave reaches between the samples, and it can be higher than any single sample. Turning the file into MP3 or AAC for streaming can push it higher again. Spotify recommends keeping true peaks under −1 dB, and under −2 dB for masters louder than −14 LUFS. The lab's limiter reads sample peaks, so you'll see its true peak land a few tenths of a decibel above the ceiling, and the coach tells you when it does. That overshoot is what a limiter's true peak option is there to stop.

In the lab, 4 dB of gain into a −1 dB ceiling takes the full mix from −15.1 to about −11.5 LUFS, and the limiter's deepest gain reduction is about 5 dB on the loudest hit. That's a moderate master. Most of the time the limiter is barely moving, and it only digs in on the biggest peaks. For more on how it works, including lookahead and release, see limiting.

How loud should a master be?

Streaming services measure the integrated loudness of every track and play them all at roughly the same level, so a master only needs to be as loud as the music calls for. Spotify and YouTube aim for about −14 LUFS and Apple Music for about −16. Spotify also has a Loud setting at −11 LUFS and a Quiet one at −19 for listeners who change it.

So a master at −9 LUFS gets turned down 5 dB on Spotify, while a master at −14 plays as it is. Once both are at −14, the louder master has nothing left to show for its extra limiting except flatter drums. The lab shows this on the right-hand pair of columns: push the limiter from 4 dB to 12 dB of gain and the exported master gets under 3 dB louder, then Spotify turns it down by about 5.3 dB to the same −14 LUFS.

Quiet masters work differently. Spotify only turns a track up as far as its peaks allow, so a master at −15 LUFS with peaks already at −0.9 dBTP stays at −15. That's the original loop in the lab: its peaks already sit just above Spotify's −1 dBTP limit, so any gain would push them further over and Spotify leaves it where it is.

None of that means every master should be −14 LUFS. Some styles, like a lot of club music, are mastered louder because the density is part of the sound, and that's a fair choice as long as you hear what the limiter is costing you. The streaming loudness and normalising vs limiting lessons go through the services in detail.

Comparing at matched loudness

A master is louder than the mix it came from, and people hear a louder version as better, usually as fuller and brighter. That makes a plain bypass switch useless for judging a master. Switch the chain off and the mix drops by 3 or 4 dB, and the master wins every time for that reason alone.

The lab handles this with Match loudness, which turns the master down to about the original's level while you compare. In your DAW, do the same thing by hand. Read the integrated loudness of the mix and of the master on a LUFS meter, put a gain plug-in after the limiter, and turn it down by the difference whenever you A/B. Take it out before you export.

How to build a mastering chain, step by step

This is the order the lesson in the lab follows, and it works on most mixes.

1. Bounce or open the finished mixTake any limiter off the mix bus and leave the peaks a few decibels below 0 dBFS. Load a reference track and turn it down to a similar loudness.
2. Start with EQSet a low cut between 20 and 30 Hz. Make any tonal changes with wide shelves or bells, keeping each one under about 2 dB.
3. Compress gentlyStart at 2:1 with a 30 ms attack and an auto or medium release. Lower the threshold until the loudest hits show 1 to 2 dB of gain reduction.
4. Shave the peaks, if it helpsUse a soft clipper or saturation set to take a decibel or so off the tallest peaks, and back it off if you can hear crunch.
5. Finish with a limiterSet the ceiling to −1 dBTP with true peak detection on. Raise the gain until the master reaches the loudness you want, watching the gain reduction on the loudest hits.
6. Check it at matched loudnessTurn the master down to the mix's loudness and A/B. If you can't say what improved, ease off the stage that's doing the most.
7. ExportBounce at 24-bit, or at 16-bit with dither switched on as the last step, and check the exported file on a LUFS meter.

Starting settings

Places to begin on a typical pop or electronic mix. Every mix needs its own adjustments, and the A/B at matched loudness decides where you end up.

Common mastering chain mistakes

Fixing the mix on the master

If you need 4 dB of EQ to make the vocal sit, or heavy compression to tame one loud snare, the fix belongs in the mix. Mastering moves change everything at once, so a fix for one part costs every other part something.

Too much limiting

Past about 6 dB of gain reduction on the loudest hits, most mixes start to lose the front edge of the drums, and the quiet parts come up around them. On streaming, that loudness gets turned back down. Back the gain off until the drums come back, and check the right-hand columns in the lab to see what a listener would actually get.

Putting something after the limiter

An EQ boost, a saturator or a stereo widener after the limiter can push the peaks back above the ceiling. Keep the limiter last, with dither after it only when you export at 16-bit.

Ignoring true peak

A limiter that reads sample peaks can let the wave pass the ceiling between samples. Use a limiter's true peak mode, or leave a little more room, and read the exported file on a meter that shows dBTP.

Judging by loudness

The master will always sound better than the mix on a plain bypass, because it's louder. Compare at matched loudness every time you change something.

Training your ears to hear limiting

The quiz under the lab runs the full chain at matched loudness with the limiter turned up by a hidden amount: 2, 6 or 11 dB of gain. Your job is to say how hard it's working. Listen to the very front of the kick and snare, because that's where heavy limiting shows first, then to the gaps between hits, which fill up as the limiter works harder. Eight rounds take a few minutes, and your best score is saved in your browser.

Questions people ask

Keep going