Multiband compression and dynamic EQ basics: when to split the spectrum, and when not to

A normal compressor hears the whole signal and turns the whole signal down. That is a problem when only one part of the sound misbehaves. A bass note that blooms at 80 Hz pulls the entire bass down, including the pluck you wanted to keep. A singer who gets harsh at 3 kHz on the loud lines gets squashed everywhere, not just in the harsh bit.

Multiband compression and dynamic EQ are the two tools built for that problem. Both react to level, and both only touch part of the spectrum. This guide explains how each works, how to set one up step by step, and how to choose between them. If threshold, ratio, attack and release are new to you, start with our compression basics and EQ basics guides first.

Multiband compression: split, then compress

A multiband compressor uses crossover filters to cut the signal into frequency bands, runs a separate compressor on each band, then adds the bands back together.

Ableton's Multiband Dynamics is a clear example to learn on, because its manual spells the design out. It splits audio into up to three bands, and Ableton says it was "designed primarily as a mastering processor". Each band has both an upper and a lower threshold, so the device can do downward compression (turn loud signals down), upward compression (bring quiet signals up), downward expansion (push quiet signals further down) and upward expansion. Ableton points out that with three bands and two thresholds each, one instance can provide six types of dynamics processing at the same time.

That flexibility is also the trap: six processors means six chances to make things worse.

The crossover matters more than people think

Splitting a signal with filters and putting it back together is not free. Ableton is candid about this in its notes on the crossover filters of EQ Three, its three-band EQ: the steeper 48 dB setting gives "more drastic filtering" but costs more CPU, and the filters are tuned to sound like an analog filter cascade, so the 48 dB mode "does not provide a perfect linear transfer". In plain terms, even with every band's compressor doing nothing, the device can colour the sound a little.

Third-party designs handle this in different ways. FabFilter's Pro-MB, for instance, offers Dynamic Phase, Linear Phase and Minimum Phase modes, and its help notes that oversampling, at extra CPU cost, improves the high-end response in the Minimum Phase and Dynamic Phase modes. Either way, bypass-test the plug-in with no compression applied before you trust it on a master.

Setting up a multiband compressor, step by step

1. Start from flat. Load the plug-in with no preset, or zero out the preset. Presets compress every band by default, which hides what you are actually fixing.
2. Name the one problem. "The low end is uneven on the sustained notes" is a problem. "Make it sound more pro" is not.
3. Set the crossover around the problem. Solo the band and sweep the crossover until the band contains the misbehaving range and as little else as possible. Ableton's EQ Three example of FreqLo at 500 Hz and FreqHi at 2,000 Hz gives a low band of 0 to 500 Hz, a mid band of 500 to 2,000 Hz and a high band above that.
4. Compress only that band. Leave the others at a 1:1 ratio, or bypassed. Lower the threshold until gain reduction appears only on the problem moments.
5. Set attack and release by ear. Low bands usually want slower times so the compressor does not track individual bass cycles and distort; high bands can move faster.
6. Match the level and compare. Use the output gain so bypassed and processed are equally loud, then switch between them. If you cannot hear a clear improvement at equal level, take it off.

Dynamic EQ: one band that moves on its own

A dynamic EQ is an ordinary EQ band with a level detector attached. FabFilter puts it simply: dynamic EQ "changes the gain of an EQ band dynamically, depending on the level of the input signal". In Pro-Q 4, any Bell or Shelf band can be made dynamic, with a dynamic range of -30 to +30 dB; a negative value works like compression and a positive one like expansion. By default the band listens to a band-limited version of the input matching its own frequency range, so it reacts to what happens in that range rather than to the whole mix.

The key difference from multiband compression is shape. A multiband compressor works on wide blocks set by crossovers. A dynamic EQ works on a bell or shelf you can make as narrow as one resonance, and there is no crossover to sum back together. FabFilter describes it as allowing edits "similar to a multi-band compressor, but in a way that's often more intuitive".

Setting up a dynamic EQ band, step by step

1. Find the frequency first, statically. Boost a narrow bell and sweep until the problem jumps out: a harsh vowel, a boomy note, a ringing snare overtone.
2. Turn that boost into a dynamic cut. Return the static gain to zero and set a negative dynamic range, say -3 to -6 dB to begin with.
3. Set the threshold so the band only moves on the offending notes. Watch the band's gain meter; it should sit still most of the time.
4. Narrow or widen the Q until the cut catches the problem without dulling the notes around it.
5. Level-match and compare, exactly as above.

Which one, when

Our rule of thumb, which is analysis rather than anything a manual says:

• One specific resonance or harsh frequency that comes and goes: dynamic EQ. A narrow band is more precise than any crossover block.
• A whole region that is uneven, such as a bass whose low end swings in level from note to note: multiband compression on one band, or a dynamic low shelf.
• Controlling one region of a sound when another sound plays: either tool with an external sidechain. Our sidechain compression guide covers multiband ducking in detail.
• A finished mix that "needs glue": probably neither. Try a gentle single-band bus compressor first, as in our bus processing guide.

What we think

Multiband compression has a reputation as the secret mastering weapon, and that reputation does real harm to home producers. A three-band preset on a master bus makes dozens of choices about your tonal balance, decided by somebody who never heard your song. When it works, it is because it is quietly re-EQing the mix in a way you could have done on purpose, in the mix, on the instrument that caused the problem.

So our artist-first position is this: fix things at the source, and reach for multiband tools only for problems that are genuinely dynamic, meaning they change over time. A bass that is always too boomy needs a static cut. A bass that is boomy only on one note needs a dynamic EQ band on that note. A vocal that is only harsh when the singer pushes is a perfect dynamic EQ job, and doing it with a narrow band leaves the softer, more intimate lines untouched, which is often where the emotion of the performance lives.

There is a taste argument too. Heavy multiband processing evens out the very variations that make a performance feel played. A kick that is slightly fatter on the downbeat, a guitar that gets brighter as the player digs in: those are not mistakes. Every band you compress is a decision to make the record a little more uniform, and uniform is rarely what people remember about a song.

If you remember one thing, make it step six: level-match and switch. If the processed version is not clearly better at the same loudness, the honest move is to delete the plug-in.

Key takeaways

• A multiband compressor splits the signal with crossovers and compresses each band separately; a dynamic EQ moves a single EQ band in response to level.
• Crossovers can colour the sound even with no compression applied, so bypass-test before trusting one on a master.
• Use dynamic EQ for narrow problems that come and go, multiband for broader regions that swing in level, and static EQ for anything that is always wrong.