Two compressors in a row, and why that isn't a mistake

Attack time is the single biggest character control in mastering, and almost nobody adjusts it.

There are two compressors in the chain, one after the other, separated only by an EQ. People see this and assume redundancy — surely one compressor, set correctly, could do the work of two?

It could not, and the reason is worth understanding, because it is the same reason a compressor has an attack knob at all.

Here are the two, at their defaults:

Same threshold. Everything else inverted. That is not two attempts at the same job.

What attack time actually does

A drum hit is not one event. It is a transient — a spike lasting a few milliseconds — followed by a body that lasts much longer. Your ear uses the transient to decide what the instrument is and how hard it was struck, and the body to decide how loud it is.

A compressor with a fast attack catches the transient. It clamps down within a millisecond, so the spike never gets through at full height. The result is a sound that is even, controlled, and slightly dull — you have removed the part your ear was using to judge impact.

A compressor with a slow attack lets the transient pass. By the time the gain reduction arrives, twenty milliseconds later, the spike is already out the door; what gets compressed is the body behind it. The transient now stands further above its own body than it did before.

This is the whole trick, and it is genuinely counter-intuitive: a compressor with a slow attack makes drums hit harder. It does not add anything. It lowers everything except the part you notice.

The glue compressor: fast, gentle, and always on

1 ms attack, 1.5:1 ratio. Fast enough to catch everything, gentle enough that catching everything does no harm.

That combination is deliberate. A fast attack at a high ratio is destructive — it is exactly how you flatten a mix. A fast attack at 1.5:1 barely moves anything: at a ratio that low, a signal 6 dB over the threshold comes out 4 dB over instead of 6. Two decibels, applied consistently to everything.

What that buys is cohesion. Because the compressor responds to the whole mix and applies the same small gain movement to all of it, elements that were recorded separately start moving together. The bass ducks fractionally when the kick lands. The vocal breathes with the snare. None of it is audible as compression; what you hear is a record that sounds like it happened in one place.

The 200 ms release matters as much as the attack. Long enough that the gain reduction does not recover between one beat and the next, so the compressor settles into a steady state rather than pumping in time with the music. A glue compressor should be doing something you can measure and not quite hear.

The tonal compressor: slow, firm, and selective

20 ms attack, 3:1 ratio. Now the settings make sense in the other direction.

The slow attack means transients pass through untouched. The 3:1 ratio means that once gain reduction does arrive, it arrives seriously — a signal 6 dB over threshold comes out 2 dB over. So the body of every sound gets pushed down firmly while its leading edge does not.

The 100 ms release — shorter than the glue stage — lets the compressor recover between hits, so each new transient meets a compressor that has already let go. Longer release here and successive drum hits would each be quieter than the last.

Then the +1 dB make-up gain brings the level back. The peaks are where they were; everything between them is 2 or 3 dB higher. The master is denser, and the drums hit harder than before you compressed. Both, at once.

The glue compressor decides whether the record sounds like one thing. The tonal compressor decides whether it hits you.

Why one stage cannot do both

Set a single compressor to 1 ms and you get cohesion and lose punch. Set it to 20 ms and you get punch, but transients now pass into the rest of the chain unhandled — and the limiter at the end has to deal with every one of them, which is where crushing comes from.

There is no attack time that is both fast and slow. Splitting the job into two stages is not extra processing; it is the only way to get both behaviours out of a mechanism that only has one attack setting.

The order matters too. Gentle-and-fast first means the tonal compressor receives a signal that is already even, so its threshold behaves predictably instead of being triggered by whichever section happened to be loudest.

What goes wrong

Both compressors set fast. The most common self-inflicted injury. You get the flattening twice, and the master arrives at the limiter with nothing left to give. It will be loud and it will be lifeless, and no amount of EQ afterwards brings the impact back — the information is gone.

Thresholds set from a number rather than a meter. -20 dB means nothing without knowing where the mix sits. A quiet mix will never reach it and both stages do nothing; a hot mix sits permanently below it and both stages run flat out. The threshold is a relationship, not a value.

Compressing a mix that was already compressed. The single most useful thing to know about your own mix is how much dynamic range it has left. Wavely reads it — the LRA figure in the analysis — and adapts the compression accordingly, which is why the chain does less to an already-squashed mix than to a dynamic one.

The thing measurement can't give you

Elsewhere in these notes there is a distinction between things you can copy from a reference record and things you cannot. Tonal balance you can copy: measure the difference between two spectra and apply it.

Compression you cannot. You can measure that a reference record has 4 LU of dynamic range, but that number does not tell you whether it got there with a fast attack and a low ratio or a slow attack and a high one — and those two routes to the same number sound nothing alike. One is a record that sits still. The other is a record that punches.

Which is why the compression settings in every preset here were chosen, and only the tonal balance and the width were measured. It is also, honestly, the part of mastering that stays a craft.