Multiband compression is a dynamics process that splits an audio signal into two or more frequency bands with crossover filters, applies an independent compressor to each band and then sums the bands back together. In mastering it lets the engineer control the level of one frequency region, such as a booming low end or harsh upper midrange, without the rest of the mix reacting to it.
A conventional full-band compressor reduces the gain of the whole signal whenever any part of it crosses the threshold, so a loud kick drum can make vocals and cymbals duck as well. Multiband compression removes that cross-band interaction, which is why it sits between equalization and dynamic range compression in the mastering chain.
How multiband compression processes the signal
The process has three stages: band splitting, per-band dynamics and recombination. Each stage affects the sound, so a multiband compressor can change the tone even when no compression is happening.
- Crossover filters divide the signal into bands, typically three to five in mastering.
- Each band feeds its own detector and gain stage, with its own threshold, ratio, attack, release and makeup gain.
- The processed bands are summed back into a full-range signal, ideally with a flat response when no gain reduction occurs.
The crossover design determines how cleanly the bands recombine. Linkwitz-Riley filters of 24 dB per octave are common because their outputs sum to a flat magnitude response, but they introduce phase shift around each crossover. Linear-phase crossovers avoid that phase shift at the cost of latency and possible pre-ringing on sharp transients.
| Crossover type | Strength | Weakness |
|---|---|---|
| Minimum-phase, gentle slope | Smooth, natural band transitions | Bands overlap and interact more |
| Minimum-phase, steep slope | Clear band separation | More phase shift near crossovers |
| Linear-phase | No phase distortion when bands sum | Latency and pre-ringing |
| Dynamic EQ style bands | Processing only around a center frequency | Narrower scope than a full split |
Key parameters and typical mastering settings
In mastering, multiband compression is used with gentle settings. The goal is to stabilize a frequency region, not to reshape the mix, so gain reduction is usually measured in one to three decibels rather than in the larger amounts used on individual tracks.
| Parameter | Typical range | What it controls |
|---|---|---|
| Number of bands | 3 to 5 | Resolution versus complexity |
| Crossover points | Around 100–250 Hz, 1–3 kHz and 5–10 kHz | Where bands divide |
| Ratio | 1.5:1 to 3:1 | Strength of compression above threshold |
| Attack | About 5–50 ms, slower in low bands | How quickly transients are caught |
| Release | About 50–300 ms, longer in low bands | How quickly gain recovers |
| Gain reduction | Usually 1–3 dB per band | Audible density and control |
Low bands need slower time constants because a fast release can follow the individual cycles of a bass waveform and create distortion. A 50 Hz wave has a period of 20 ms, so a release shorter than that modulates the wave itself. High bands tolerate faster settings because their cycles are much shorter.
Multiband compression compared with related tools
Several mastering tools address similar problems, and choosing the right one avoids unnecessary side effects. The decisive question is whether the problem is static, present all the time, or dynamic, appearing only at certain moments.
| Tool | Problem it solves | Best when |
|---|---|---|
| Static EQ | Constant tonal imbalance | The issue is always present |
| Full-band compression | Overall level variation and glue | The mix should move as one unit |
| Multiband compression | Frequency regions that vary in level | One area jumps out only at times |
| Dynamic EQ | Narrow resonances that appear intermittently | A specific frequency needs control |
| De-esser | Sibilance between about 5 and 10 kHz | Only the vocal consonants are harsh |
| Mid/side processing | Imbalance between center and sides | Width or center elements need control |
Many engineers now prefer dynamic EQ for narrow problems because it touches less of the spectrum. Multiband compression remains the better choice for broad regions, such as controlling the whole low end of a dense rock master.
Practical uses in mastering
Multiband compression is a corrective tool first and a creative tool second. Typical mastering tasks include the following.
- Taming an uneven low end, where some bass notes are much louder than others.
- Controlling harshness in the 2–5 kHz region when distorted guitars or bright synthesizers peak.
- Reducing sibilance across the whole mix when only a stereo master is available.
- Adding density to a band with low ratios and slow settings, a form of subtle upward stability.
- Preparing for limiting by removing frequency-specific peaks that would otherwise trigger heavy limiter gain reduction.
Some multiband processors also offer upward compression or expansion per band. Upward compression raises quiet details in a band, for example low-level ambience in the high frequencies, and is used carefully because it also raises noise and room tone.
Common mistakes and how to avoid them
The flexibility of multiband compression makes it easy to damage a mix that was already balanced. Experienced engineers treat it as a last resort and always compare the result at matched loudness.
| Mistake | Audible result | Correction |
|---|---|---|
| Too many bands | Unnatural, disjointed tone | Use only the bands that solve a problem |
| Heavy gain reduction | Flat, lifeless master | Keep reduction to a few decibels |
| Fast release in the low band | Bass distortion and pumping | Lengthen release beyond the waveform period |
| Makeup gain used as EQ | Tonal balance drifts with level | Use EQ for static tone changes |
| No loudness-matched comparison | Louder version sounds better by default | Match LUFS before judging |
A frequent root cause is fixing in mastering what belongs in the mix. When a single instrument causes the problem, stem mastering or a mix revision usually gives a cleaner result than multiband processing of the stereo file.
In everyday language
A normal compressor turns the whole song down a little whenever it gets too loud. Multiband compression is smarter: it splits the sound into bass, middle and treble and only turns down the part that is causing trouble.
Imagine a choir where only the basses sometimes sing too loudly. Instead of asking the whole choir to sing softer, the conductor quietly signals just the bass section. Everyone else keeps their normal volume, and the overall sound stays balanced.
Seen in the real world
A mastering engineer receives a funk track in which some bass notes around 80 Hz are 6 dB louder than others because of a resonant room during recording. A full-band compressor would make the vocals and guitars duck on every loud note.
The engineer sets a low band below 180 Hz with a 2:1 ratio, 30 ms attack and 200 ms release, adjusted so that only the loud notes trigger about 2 dB of gain reduction. The other bands are left unprocessed.
The bass line now sounds even, the limiter at the end of the chain works less hard, and the master reaches its loudness target without audible pumping in the vocals.