The eleven stages, and why they are in that order

What each stage in the chain is for, and why moving one of them changes everything downstream.

Open any mastering tool and you get a row of boxes: EQ, compressor, saturation, limiter. It looks like a menu. It is not a menu — it is a sentence, and the words are in a particular order for reasons that are mostly invisible until you move one.

Here is the chain, in the order audio actually travels through it.

① cleaning EQ        ⑤ de-esser        ⑨  limiter
② glue compressor    ⑥ multiband       ⑩  safety clipper
③ tonal EQ           ⑦ saturation      ⑪  loudness normalisation
④ tonal compressor   ⑧ stereo width

Eleven stages, and they fall into three groups that behave completely differently.

Three kinds of stage

Corrective stages fix a problem. The cleaning EQ, the de-esser, the safety clipper. They exist because something is wrong, and when nothing is wrong they should do nothing at all. A corrective stage that is always working is a misconfigured stage.

Character stages make a choice. The compressors, the saturation, the stereo width, the tonal EQ. There is no correct setting for these — there is only the setting that produces the record you want. This is where a master acquires a personality, and it is the part that cannot be automated by measurement alone.

Destination stages hit a number. The limiter's ceiling and the loudness normalisation. These are not decisions in the same sense: -14 LUFS is a specification, not a taste. You are told the number by the platform and you arrive at it.

Most confusion about mastering comes from treating one group as another — arguing about the loudness target as if it were a matter of taste, or expecting a compressor to have a correct answer.

Why the order is what it is

Four decisions in the sequence do most of the work.

Cleaning EQ comes before the glue compressor. A compressor reacts to whatever energy it is given, and low-frequency energy has the most of it. If your mix has rumble at 30 Hz that nobody can hear on any speaker, a compressor placed before the cleaning EQ will still duck the entire track every time that rumble peaks. You would hear the whole master breathe in sympathy with something inaudible. Removing it first means the compressor only responds to music.

Tonal EQ comes after the glue compressor, not before. The glue stage changes the balance slightly — it always does, because compression is level-dependent and the loudest parts of a mix are rarely spread evenly across the spectrum. If you set your tonal balance first, the compressor then moves it, and you are correcting a moving target. Doing the broad compression first and the tonal decisions second means what you hear when you set the EQ is what comes out.

The de-esser comes before the multiband compressor. Both react to high frequencies. A harsh sibilant hitting the multiband's high band pulls that whole band down — so an "s" on a vocal quietly ducks every cymbal in the arrangement for a fraction of a second. Catching the sibilant first with a narrow tool means the broad tool never sees it.

Saturation comes before the stereo width stage, and both come before the limiter. Saturation generates new harmonic content, and the width stage acts on the difference between left and right — so saturating afterwards would generate harmonics that were never placed in the stereo field deliberately. And both raise peak level, which is exactly what the limiter is there to handle. Put either after the limiter and you have undone the limiter's work.

What "combined effect" actually means

The stages are not independent, and three interactions catch people out.

The multiband can undo your EQ. You add 3 dB at 60 Hz because the mix is thin. The multiband's low band now receives 3 dB more energy, crosses its threshold more often, and compresses harder — handing back a good part of what you just added, but unevenly, only on the loud parts. The mix now sounds thin in the choruses and full in the verses. The EQ move was real; the multiband answered it.

Saturation makes the limiter's job easier, which is the whole point. Saturation rounds peaks and adds harmonics, which raises the average level without raising the peak. A track that has been gently saturated needs less limiting to reach the same loudness — and limiting is where masters die. Almost every master that sounds crushed is one where the limiter was asked to do work that an earlier stage should have done.

Width changes tone. Widening the stereo image raises the level of the difference between the channels, and in most music that difference is concentrated in reverb, cymbals and room tone — high frequencies. Push the width up and the master gets brighter, whether or not you touched an EQ.

Every stage in a chain is downstream of every earlier one. There is no such thing as an isolated adjustment.

Why nothing here makes it louder

The last stage normalises to a loudness target, which means every gain decision in the ten stages before it is given back. This surprises people constantly, and it is the subject of its own note.

The consequence worth carrying into everything that follows: the chain does not control how loud your master is. It controls what your master sounds like at a loudness that was decided elsewhere. Once you accept that, the settings stop being a competition and start being a description.

Where to go next

The three notes that follow take the groups in turn — the two compressors, the three ways to make a peak smaller, and the stages that decide what it sounds like. None of them will tell you a correct setting, because there isn't one. They will tell you what each control is actually doing, which is the part that transfers.