If my DAW works in 32-bit float, does level matter?

For internal noise and distortion, barely: the headroom is so large you won't clip the summing bus. But it matters for everything else. Plugins that model hardware are calibrated to a specific level and behave differently depending on what you feed them, meters stop being useful, and any conversion to fixed point on export does clip. Healthy levels aren't superstition, they're the condition for every other decision being comparable.

What level should I record at?

Peaks around -12 to -6 dBFS leave you plenty. That reference keeps headroom for unexpected peaks and puts the signal in the zone plugins are designed for. Recording up against 0 dBFS is a habit inherited from 16-bit, where the noise floor forced you to push; with 24-bit that reason disappeared.

What is 0 VU and why do people mention -18 dBFS?

VU meters on analogue hardware read 0 at the optimum working level, and moving to digital meant deciding which dBFS value that corresponded to. The most widespread convention puts it around -18 dBFS. That's why many compressor and preamp emulations «come alive» with a signal around there: it's the level they were calibrated for.

Can't I just fix the level with the fader?

The fader sits at the end of the chain, so it lowers what leaves the channel but not what went into the plugins. If you pushed a signal into a compressor 12 dB too hot, the fader doesn't undo the compression that caused. That's what input gain is for, because it acts first.

Is internal clipping audible?

On a floating-point bus, no: the number can pass 0 without breaking and come back down. What is audible is clipping inside a fixed-point plugin, at your interface's physical output and in the exported file. There the ceiling is real, and crossing it distorts.

Fundamentals9 min read

Gain staging: why your mix sounds wrong before you start

Healthy levels from one end of the chain to the other, how much headroom to leave, why the fader isn't the place to fix it, and what actually changes in the digital domain.

The problem nobody sees coming

There's a kind of mix that no EQ or compressor will fix, because the fault happened earlier: the tracks came in at wildly different levels, each plugin got more or less signal than it expected, and from then on every decision was made on a crooked foundation. It sounds dull or harsh for no obvious reason, and the more you work on it, the worse it gets.

Gain staging is simply keeping healthy levels at every point in the chain: input, plugins, channel, bus and master. It doesn't improve the sound by itself — it isn't processing — but it puts everything else in a position to work. It's the part of the craft that never shows off and that you only notice when it's missing.

Why level changes the sound, even in the digital domain

There's a widespread idea that internal level doesn't matter in a modern DAW because floating point has headroom to spare. That's true as far as clipping the bus goes: you won't break the sum by working hot. But it doesn't follow that level is irrelevant, and conflating the two is what leads to mixes where nothing behaves as it should.

Any plugin modelling hardware — preamps, compressors, tape, saturators — is calibrated to a reference level and reacts differently to what you feed it. That's not a quirk: it's the model working.
All non-linear processing (saturation, distortion, some limiters) depends on input level by definition.
Meters are only useful within the range they were designed for. With everything pinned to the ceiling they all read the same and stop informing you.
Exporting to a fixed-point format does have a real ceiling: what was a large number inside gets clipped outside.

The chain, point by point

Recording: peaks between -12 and -6 dBFS. It leaves headroom and stops anyone re-taking over an unexpected peak.
Each plugin's input: if the plugin has its own meter, aim for the signal to sit in its middle range, not up against the top.
Channel: with the fader at 0 dB (unity), the track should be near its place in the mix. If it isn't, adjust the gain before the fader, not the fader.
Group buses: summing several tracks raises the level; check the bus isn't arriving hot at the master.
Master: keep at least 3 dB free throughout the mix. The master isn't where you go to gain volume.

The channel point pays off the most and is the one most often skipped. If you get used to leaving faders near unity and resolving differences with input gain, you keep the fader's whole travel for actual mixing, instead of spending it compensating for tracks that arrived wrong.

Starting a mix on a straight foundation

Before touching a single plugin, balance. Set every fader to zero, pull the master down if you need to, and adjust each track's gain until the whole thing sounds roughly balanced: the vocal is audible, the drums hold, nothing buries anything. It's dull and it takes ten minutes, and from there everything you do starts from somewhere reasonable.

That initial balance is also the best diagnostic tool there is: if the mix works on faders alone, everything after is refinement. If it doesn't, no plugin will save it and the problem is the arrangement. For the next step — controlling dynamics without flattening them — there's the compression guide, and to avoid losing your master headroom on the way, the one on LUFS and loudness.

Frequently asked questions

If my DAW works in 32-bit float, does level matter?: For internal noise and distortion, barely: the headroom is so large you won't clip the summing bus. But it matters for everything else. Plugins that model hardware are calibrated to a specific level and behave differently depending on what you feed them, meters stop being useful, and any conversion to fixed point on export does clip. Healthy levels aren't superstition, they're the condition for every other decision being comparable.
What level should I record at?: Peaks around -12 to -6 dBFS leave you plenty. That reference keeps headroom for unexpected peaks and puts the signal in the zone plugins are designed for. Recording up against 0 dBFS is a habit inherited from 16-bit, where the noise floor forced you to push; with 24-bit that reason disappeared.
What is 0 VU and why do people mention -18 dBFS?: VU meters on analogue hardware read 0 at the optimum working level, and moving to digital meant deciding which dBFS value that corresponded to. The most widespread convention puts it around -18 dBFS. That's why many compressor and preamp emulations «come alive» with a signal around there: it's the level they were calibrated for.
Can't I just fix the level with the fader?: The fader sits at the end of the chain, so it lowers what leaves the channel but not what went into the plugins. If you pushed a signal into a compressor 12 dB too hot, the fader doesn't undo the compression that caused. That's what input gain is for, because it acts first.
Is internal clipping audible?: On a floating-point bus, no: the number can pass 0 without breaking and come back down. What is audible is clipping inside a fixed-point plugin, at your interface's physical output and in the exported file. There the ceiling is real, and crossing it distorts.