Multiband compression explained for producers who keep stacking EQs

You've notched a resonant bump at 110 Hz on the bass. It fixes the two notes that boomed and thins out everything else. What you wanted was a notch that only exists while those notes play - that's dynamic EQ, and we'll get to it.

But the more common problem is messier: the low E swallows the mix, the A blooms somewhere else, the whole bottom octave swells depending on what's played. The loud frequency moves with the pitch, so no notch - static or dynamic - can chase it. That's the job for a multiband compressor on the low band: it controls the level of a zone, not a spot.

What multiband compression does

A broadband compressor watches the full-range signal and reduces the gain of everything when the level crosses the threshold. Put one on a drum bus and the kick triggers it - the snare, hats, and overheads all duck on every kick hit. That's the source of bus-compressor pumping.

A multiband compressor splits the incoming signal into frequency bands using crossover filters, runs each band through its own independent compressor, then sums the bands back together at the output. A peak in the low band only pulls the low band down. The mid and high bands keep running at full level unless their own thresholds get crossed. You get frequency-dependent dynamics control: gain reduction targeted at the part of the spectrum that has the dynamic problem, with the rest of the signal untouched.

This is different from EQ, which is a static gain change per frequency, and different from dynamic EQ, which uses EQ-shaped curves on narrower bands. Multiband sits in the middle: wider bands than dynamic EQ, with full compressor controls (threshold, ratio, knee, attack, release) on each one.

The three jobs it does better than anything else

Taming a bass that booms on some notes and not others. Bass instruments rarely have even level across their range. Resonances in the instrument, the room, or the pickup, or just the way certain notes sit in the mix, mean different pitches bloom by different amounts in the sub region. If the unevenness traces back to one fixed resonance - same Hz every time - skip ahead, that's a dynamic EQ job. But when the loud frequency follows the notes, a low-band multiband compressor with the crossover around 100–120 Hz doesn't care where in the band the energy lands: it catches whichever note gets loud and leaves the quiet ones alone. The mid and high content of the bass - the pick attack, the growl, the harmonics that carry the part on a phone speaker - stays completely untouched because it lives above the crossover.

Controlling vocal sibilance with more flexibility than a fixed de-esser. A standard de-esser sets a detection band and a reduction band and locks them. That works fine when the sibilance sits in one place. Vocals that change register, or vocals where you've already applied a presence boost that brought up the harshness, often have sibilance that wanders across a wider range, or that varies in severity with performance dynamics. A multiband compressor with a single band from 5-10 kHz, fast attack, short release, and a 4:1 to 8:1 ratio gives you full control over how hard each syllable gets caught and how quickly the band recovers.

Glueing a mix bus without the kick pumping the highs. This is the classic case. Throw an SSL-style compressor on the mix bus and the kick and bass dominate the gain-reduction circuit. Every kick hit ducks the cymbals and vocal. A two- or three-band multiband with most of the work happening on the low band (up to roughly 150 Hz, 2:1 to 3:1, a few dB of gain reduction) leaves the upper bands sitting at light or zero compression. You get cohesion across the bass region without the breathing on the top end.

When to reach for something else

If the problem lives at one fixed frequency - a room mode, an instrument body resonance, the 110 Hz bump from the intro - a dynamic EQ with a narrow Q is more natural than a multiband band. The test: does the boom sit at the same frequency every time, or does it follow the notes? Same frequency every time means dynamic EQ. Follows the notes means multiband. Dynamic EQ works in surgical strips; multiband works in wider zones.

If the problem is one resonant pitch on a guitar or a single vowel ringing on a vocal, a dynamic EQ with a narrow Q is usually more natural than a multiband band. Multiband works in wider zones; dynamic EQ works in surgical strips.

Dialling it in without smearing transients

Start with every band bypassed or with all thresholds at maximum. The most common way to ruin a mix with a multiband is to leave the factory preset's four active bands running. Sound on Sound's Mike Senior puts it bluntly: "use the minimum number of bands required to solve it, you're much more likely to get transparent results without unflattering sonic side-effects."

Place the crossover at a frequency where there isn't much spectral content, so the bands don't fight each other. For bass control that usually means 80-120 Hz. For vocal mud, 200-300 Hz. For sibilance, 5-7 kHz. Solo the band while you sweep to make sure you're catching the problem material and not the surrounding harmonics.

Set the threshold by looping the worst instance of the problem and lowering until the gain-reduction meter shows 3-6 dB on those moments, then back off until only the offenders trigger reduction. Ratio depends on the job: 2:1 to 3:1 for bus glue, 3:1 to 4:1 for bass and proximity control, 4:1 to 8:1 for sibilance because the events are short enough that high ratios stay invisible.

Attack is where transients get killed. On low bands, 5-15 ms lets the front edge of each note through before the compressor clamps; faster than that and the bass loses punch. On high bands chasing sibilance, 1-3 ms or instant is what you want because the sibilant peak is over in milliseconds. If your compressor has lookahead - Pro-MB offers up to 20 ms - a small amount (1-5 ms) cleans up fast-attack artifacts on any band.

Release should be tuned to the tempo. Start with the quarter-note value in milliseconds (60,000 divided by BPM) and halve from there until the band recovers between notes without pumping. For long sustained 808s, releases of 250 ms or more are normal. For sibilance, 20-50 ms keeps the band ready for the next word.

Then gain-match the bypass A/B. If the compressed version sounds smaller, thinner, or duller at matched level, the compression is making the problem louder rather than solving it.

The starting point worth memorising

One band, low crossover at 100 Hz, 3:1 ratio, 10 ms attack, 100 ms release, threshold set for 3 dB of gain reduction on the loudest notes. That single setup, dropped on a bass track or a mix bus, fixes more low-end problems than another EQ in the chain.