Gain staging

Two questions that drive producers mad:

1. Every track in my session peaks safely below zero. Why is the master clipping?
2. If I turn the signal up going into an effect, and down by the same amount coming out, shouldn't it sound exactly the same?

The answer to the second one is right here. This compressor is set up on a drum loop. Before you touch it, a prediction: if you lower the Threshold by 6 dB and then use Match to bring the loudness back, will it sound the same as before?

Same loudness. Different drums. The hits are flatter, the room sound between them pumps up, the snare has lost some crack. Lowering the threshold by 6 dB is exactly the same, to the compressor, as feeding it 6 dB more signal: either way, the drums now reach 6 dB further past the point where it starts working. Turning the result back down restores the level, but not the shape.

Gain staging means knowing which parts of your chain care how loud the signal is, and choosing that loudness on purpose.

Some stages don't care how loud you are

A plain volume fader is linear: double what goes in and you get double out, nothing else changes. Turn up before it and down after it by the same amount, and you're back exactly where you started. A normal EQ behaves the same way. It doesn't matter whether you hit it hard or soft; the curve is the curve.

So if your chain is only faders and EQ, you can push levels around freely and the sound won't change, only its size. (As long as you don't hit a hard ceiling somewhere, which we'll get to.)

Some stages care a lot

A compressor, a saturator, a tape emulation, a guitar amp: these are nonlinear. Doubling what goes in does not double what comes out. They treat loud moments differently from quiet ones. That's their whole job.

That's why engineers have talked for decades about "hitting the tape harder" or "driving the console". On analog gear, the input level was one of the most important tone controls in the studio. Feed a tape machine a little hotter and the drums get thicker and the transients soften. Same idea, same reason.

The compressor works the same way through a different door. Its threshold is fixed, so the louder the signal arrives, the more of it crosses the line and gets turned down. It won't necessarily work "twice as hard" for 6 dB more input; ratio, knee and timing decide the result. But it will work differently, every time.

Here's the classic drive version of the experiment, as a chain:

The practical rule: find the first stage in your chain whose sound changes with level, and set the level going into it on purpose. Everything before it is just plumbing. Everything after it gets whatever that stage decided.

Why the master clips when no track does

Now the first question. Imagine two tracks, each peaking at −6 dBFS, which is half the full-scale ceiling. If they happen to be identical and perfectly in step, their peaks land at the same instant and add up: half plus half is a whole. The sum hits 0 dBFS.

Real tracks aren't identical copies, so their peaks usually don't line up perfectly, and the sum lands somewhere below that. But a kick and a bass that both hit hard on the downbeat, or a stack of layered synths playing the same chord, can line up well enough to push the master over. Ten tracks that each look modest can add up to far more than any of them alone.

The fix is boring and reliable: turn the tracks down, not the master. Leave room at the sum for the loudest moment of the song, which is usually the densest chorus, not the verse you've been looping.

Headroom inside the computer, and outside it

Most modern music software calculates in floating-point numbers, which have an enormous range. Inside the mix, a channel can often go "over zero" without anything actually clipping, and turning it down later recovers it cleanly. Check how your own software behaves; it isn't true of every plugin or every engine.

The file you export is a different story. A WAV or MP3 has a hard ceiling at 0 dBFS, and so does your audio interface's output. Anything over that gets chopped flat: clipping, the harsh crackle you hear on overcooked masters. And some plugins clip internally, on purpose or not, however clean the master meter looks.

So if you hear distortion that stays even after you pull the master down, look upstream. Turning down the end of the chain can't un-clip something that already clipped earlier. Digital audio explains where those hard ceilings come from.

Where the rules bend

• A hot input is sometimes the sound. Plenty of great records were made by pushing gear into its sweet spot. Gain staging doesn't mean keeping everything quiet; it means knowing which stage you're driving and why.
• Your meters may be in different places. A track meter might read before the plugins, after them, or after the fader. Know where each one is listening before you trust it.
• Compressors with auto-makeup will hide the level change from you. Turn auto-makeup off when you want to hear what level is doing.

Make it yours

Pick a chain of your own, for example a bass through a saturator and a compressor. Play its loudest phrase, then for each stage ask: if I push 6 dB more into this, does the sound change, or just the size?

Find the first stage that answers "the sound". Try it a few dB hotter and a few dB cooler, matching the final level each time. Choose the version you like, then check the loudest moment of the song still has headroom at the master.

Save the chain, and write a one-line gain map next to it: "bass drives the saturator about 4 dB hot, compressor clean, final level set at the channel fader". When you reopen the project in six months, you'll know which knob is actually a tone control.

Quick check

1. Your vocal chain sounds harsher since you raised the mic preamp, even though you turned the channel fader down to match. Why? Something in the chain is nonlinear (a compressor, a saturator, the preamp itself) and is now working harder. The fader restored the level, not the tone. Lower the input to that stage instead.
2. Two synth layers each peak at −6 dBFS, playing the same notes in the same octave. The master clips. A pad and a hi-hat each peak at −6 dBFS and the master is fine. Why the difference? The synth layers are nearly the same wave in step, so their peaks add almost fully. The pad and hats are different sounds whose peaks rarely coincide.
3. The master meter never goes red, but you still hear crackle on the loudest chorus. Where do you look? Earlier in the chain: a plugin or bus that's clipping internally. A quiet final meter can't remove distortion that happened before it.

Next steps

• Go deeper: Making room with EQ. Levels are healthy now, but the keys and the bass still blur into one smear. That isn't a level problem.
• Sideways: Distortion and saturation. Choose the nonlinear stage for its sound, not by accident.