A compressor on the mix bus watches one signal and applies one gain reduction to everything. Whatever is loudest sets the compression for the whole record, usually the kick, sometimes the snare.
That’s fine when the goal is glue. It stops being fine when the problem is confined to a frequency range. A bass note that ducks the entire mix every bar is a low-frequency problem being solved with a full-band tool, and no attack or release setting fixes it, because the detector can’t tell a bass note from a vocal.
Split the signal into bands and give each one its own compressor and the bass note only controls the bass band.
Where the crossovers go
Four bands, with crossovers around 120 Hz, 800 Hz and 5 kHz, covers most material. What each band is for:
Below 120 Hz. Kick and bass. This band’s job is consistency: the low end should sit at the same level from verse to chorus. Slow attack so the kick transient survives, slow release so the band doesn’t move audibly. Two to four decibels.
120 Hz to 800 Hz. Body. Snare fundamental, guitar and piano weight, the lower half of the vocal. This is where a mix gets congested and where most of the energy lives, so it’s the band to be most careful with. Gentle ratios.
800 Hz to 5 kHz. Presence. Vocal intelligibility, snare crack, guitar attack. Compress this band and the mix gets more consistent in the range the ear is most sensitive to, which is either exactly what you want or immediately obvious.
Above 5 kHz. Air and cymbals. Fast attack costs nothing here, and a little reduction evens out a bright hi-hat without a de-esser’s side effects.
Move the crossovers to the material. If the bass sits high, the first crossover goes up. The defaults are a starting point, not a fixed grid.
What mid/side adds on top
Bands separate by frequency. Mid/side separates by position. The problems that need both are common.
A wide synth pad and a centred bass occupying the same octave is the standard example. A band-only compressor can hold that octave down, but it holds down both the bass and the pad. Split the band into mid and side and the centre can be squeezed hard while the edges stay open: the bass gets controlled, the pad keeps its size.
Running four bands across mid and side means eight compressors, each with its own threshold, ratio and timing. That’s a lot of parameters, and the honest answer is that most mixes don’t need all of them. The way to use it is to find the one band and one channel that has the problem, work there, and leave the other seven alone.
If you can’t say which band and which channel you’re fixing, you don’t need a multiband yet.
Attack, release and the mistake with numbers
Set attack and release in real milliseconds and read them as milliseconds. Eight bands snapping to a sparse switch grid make the same labelled position mean different things in different bands, and a “medium” release on a 60 Hz band and on a 10 kHz band are not the same setting.
Low bands want slow numbers because the waveform itself is slow. A 40 Hz cycle is 25 milliseconds long, so a 5 ms release on that band is chasing individual cycles and will distort. Start around 100 ms at the bottom and work down as you go up.
High bands can take fast numbers because the content is fast. Ten milliseconds attack and 50 release on the top band is unremarkable.
Solo the band before you set it
Every multiband mistake comes from setting a band by looking at its gain reduction meter. Solo it, listen to what’s in it, then set the threshold by ear against that. A band that reads 6 dB of reduction on a meter can be doing nothing you can hear, or destroying the snare, and the meter reads the same either way.
Then unsolo and check against bypass at matched level. Multiband compression almost always comes out louder, and louder always sounds better for the first few seconds.
When not to reach for it
A mix that needs four bands of correction usually needs an EQ move instead. Multiband compression is for problems that come and go: the chorus that gets boomy, the verse where the cymbals jump forward. If the problem is there for the whole song at the same level, it’s static, and a static problem is an EQ problem. Fixing it with a compressor means you’ve built a filter that changes shape depending on how loud the section is, which is harder to predict and harder to undo.
A starting point
- Bypass everything. Set the crossovers to the material with all bands bypassed, using solo to hear where they land.
- Find the one band with the problem. Set it alone, with the others still bypassed.
- Aim for 2–4 dB in that band. More than that on a mix bus is a mix problem being covered up.
- Only then decide whether it needs to be split into mid and side. Most of the time it doesn’t, and knowing that saves you four sets of parameters.
- Match levels and A/B against the full bypass, not against the last thing you changed.