If the problem covers a wide range, like a low end that swells on every kick or a top end that gets bright in the choruses, use a multiband compressor. Its bands are built to cover a range evenly.
If the problem is one narrow spot that only shows up sometimes, like a single bass note that booms or a vocal that turns harsh at 3 kHz on its loudest words, use a dynamic EQ. And if the problem is there all the time, use neither: a plain EQ cut does the job.
At a glance
| Property | Multiband compressor | Dynamic EQ |
|---|---|---|
| How it isolates a range | Crossover filters split the sound into bands | A filter feeds the detector; the sound itself isn't split |
| What moves | The level of a whole band | The gain of a bell or shelf |
| Shape of the cut | Flat across the band, steep at the edges | Deepest at the centre, easing off either side |
| Narrowest useful band | About two octaves | As narrow as the EQ's Q allows |
| With nothing compressing | Flat level, but most crossovers shift the phase | Exactly the original |
| Usual number of bands | Three or four, covering the whole spectrum | One or two, only where needed |
| Typical job | Evening out the low end of a mix, taming a bright top | A boomy note, a harsh resonance, esses |
What do they have in common?
Both are compressors that only care about part of the frequency range. A normal compressor listens to the whole sound and turns the whole sound down, so a loud kick drags the vocal down with it. Multiband compression and dynamic EQ both get around that by listening to one band and turning down only that band.
The controls you'll use most are the same on both: a threshold the band has to cross, a ratio that sets how hard it comes down, and attack and release times for how fast. In the lab the two modes share exactly the same detector maths, so when you switch between them the only things that change are how the band is cut out and how its gain is applied.
Both are also a moving EQ. When a band comes down by 4 dB, that range of frequencies is 4 dB quieter at that moment and the tone changes with it, which is the curve the tone screen in the lab draws as the loop plays.
How a multiband compressor works
A multiband compressor starts with crossover filters. A three-band compressor with crossovers at 150 Hz and 2.5 kHz sends everything under 150 Hz to one compressor, everything from 150 Hz to 2.5 kHz to a second, and everything above 2.5 kHz to a third. The outputs are added back together at the end, and when no band is compressing they add up to a flat level at every frequency.
Each band is compressed as one block. If the low band comes down 6 dB, everything in it comes down 6 dB, from the lowest sub to just under the crossover, and the slopes of the crossover filters decide how sharply that stops at the edges. The multiband compression lesson covers setting one up from scratch.
In this lab the multiband mode is a three-band compressor with only the middle band working. The frequency knob sets the centre of that band and the width knob sets how far apart its crossovers are, so at 100 Hz and three octaves wide they sit at 35 and 283 Hz. They're Linkwitz-Riley filters with slopes of 24 dB per octave, the usual crossover design because the bands add back up flat.
How a dynamic EQ works
A dynamic EQ starts with an ordinary EQ band, usually a bell or a shelf, and adds a threshold to it. A filter matching the band feeds the detector, which hears only that part of the sound. When the band gets louder than the threshold, the EQ band's gain moves down, and when it falls back the gain returns to where it was.
The sound itself is never split. With the threshold above everything in the band the bell sits at 0 dB, and a bell at 0 dB changes nothing, so a dynamic EQ at rest passes the sound through exactly as it went in.
The width of a dynamic EQ band is set like any EQ band's, with Q. In the lab it's shown in octaves so it lines up with the multiband mode: one octave is a Q of about 1.4 and half an octave is about 2.9. Most dynamic EQ plug-ins let you go far narrower than that.
The shape of the cut
Set both to the same band and the same gain reduction, and the tone screen shows two different curves. The multiband cut has a flat bottom across the band and steep sides at the crossovers. The dynamic EQ cut is deepest at the centre frequency and eases off gradually either side, which is the shape of any bell.
In step 2 of the lab, a one-octave dynamic EQ band at 100 Hz comes down by up to 4.3 dB on the loudest kicks and bass notes. At that moment the bell takes the full 4.3 dB off at 100 Hz but less than a decibel at 50 Hz and 200 Hz. Give the multiband the same 4.3 dB of gain reduction and its deepest cut is only about 2.5 dB.
The multiband cuts less because its band is made from two crossovers, and each crossover filter has a slope. When the band is narrow, the slopes of its two edges overlap, so the band never reaches full level and its neighbours fill in part of the gap when it's turned down. With the lab's 24 dB per octave crossovers, a band can't take more than these amounts off the sound even when you mute it completely:
| Band width | Band muted | Band down 6 dB | Bell down 6 dB |
|---|---|---|---|
| Half an octave | −5.1 dB | −2.2 dB | −6.0 dB |
| One octave | −8.9 dB | −3.3 dB | −6.0 dB |
| Two octaves | −18.8 dB | −5.1 dB | −6.0 dB |
| Three octaves | −30.3 dB | −5.7 dB | −6.0 dB |
So a multiband band needs to be two or three octaves wide before it cuts the way its gain reduction meter suggests. Plug-ins with steeper crossovers can cut a thinner slice, but the overlap never goes away completely. A bell has no such limit, because its gain is the cut at its centre frequency at any width.
Steps 6 to 8 show it on a vocal. A half-octave dynamic EQ band at 3 kHz takes up to 4.1 dB off the loudest words and leaves the rest of the voice alone. The same band as a multiband barely reaches the threshold, and even with the ratio all the way up and the meter showing 18 dB of gain reduction, the sound only gets 4.4 dB quieter at 3 kHz.
What each one listens to
A multiband compressor's detector hears its own band, the slice the crossovers cut out. A dynamic EQ's detector hears whatever its filter lets through, which is normally a band-pass filter at the same frequency and width as the bell. Press Band only in the lab to hear either.
That changes how a threshold behaves. In step 3 the dynamic EQ's band-pass filter around 100 Hz peaks at −6.9 dBFS. Switch to multiband in step 4 and the band between the 71 and 141 Hz crossovers peaks at −10.5 dBFS, because the crossover slopes take some of the level off at the centre, so with the threshold left at −18 dB it works for 6% of the loop instead of 26%. Settings don't carry over from one to the other; set the threshold by watching the gain reduction.
Many dynamic EQs can also take their trigger from a different track. FabFilter Pro-Q, from version 3 on, lets each dynamic band listen to an external sidechain, so a bell at 60 Hz on the bass can dip only while the kick plays. The kick and bass lesson covers that trick. Many multiband compressors take an external sidechain too.
Does a multiband compressor change the sound when it's doing nothing?
Its level stays flat, but its phase doesn't. A crossover filter delays the frequencies below and around its crossover point more than the ones above it, so when the bands are added back together the level comes out flat while the waveform has changed shape. Take step 4's settings, turn the threshold back up to 0 dB, and the coach runs a null test, subtracting the result from the original. An identical copy would cancel to silence. The multiband leaves a residue only about 4 dB down.
You rarely hear that on its own. It becomes a problem when the multiband sound is blended with an unprocessed copy, in parallel processing or on a drum bus with a dry send, because the phase-shifted frequencies partly cancel against the dry ones. Some multiband compressors offer linear-phase crossovers to avoid it; FabFilter Pro-MB has a linear-phase mode, at the cost of extra latency and some pre-ringing. The linear-phase vs minimum-phase EQ lesson explains that trade.
A dynamic EQ at 0 dB has no phase shift at all, since there's nothing in the audio path doing anything. Once the bell moves, it shifts the phase around its frequency the way any EQ bell does, and only by as much as it's cutting at that moment.
When to use each
Reach for a multiband compressor when the problem is broad and the range has clear edges. The low end of a mix that swells on every kick is the standard case: a band from the bottom up to about 120 or 150 Hz, compressed gently, keeps the weight even without touching the chords and vocal above it. A top band above 5 kHz that comes down a decibel or two in the loud sections is another. Multiband compressors are also behind heavier effects like OTT, which pushes every band up and down at once.
Reach for a dynamic EQ when the problem is narrow or only shows up sometimes. A bass line where one note booms louder than the others is the clearest example, because the boom sits at that note's frequency and nowhere else. A vocal that turns harsh around 2 to 4 kHz only when the singer pushes is another, and so is a guitar resonance that rings on certain chords. De-essing with a dynamic EQ band is common too.
| Job | Multiband | Dynamic EQ |
|---|---|---|
| Even out the whole low end of a mix | Good | Works, as a low shelf |
| One bass note that booms | Too broad | Good |
| Vocal harshness on loud notes | Takes body with it | Good |
| Esses | Works, as a high band | Good |
| Duck the bass under the kick | Possible, with a sidechain | Good, with a sidechain |
| Heavy sound design, like OTT | Good | No |
| A problem that's there all the time | Use a static EQ | Use a static EQ |
Plenty of plug-ins sit between the two. A dynamic EQ band set to a low shelf behaves a lot like the low band of a multiband compressor, and iZotope Ozone has both a Dynamic EQ module and a multiband Dynamics module so you can pick per job. Ableton Live's Multiband Dynamics and Logic Pro's Multipressor are multiband compressors, and Tokyo Dawn Labs' TDR Nova is a free dynamic EQ.
How to choose between them, step by step
- Find the problem with a narrow boostPut a static EQ bell on the track, boost it 8 to 10 dB with a high Q and sweep it until the problem jumps out. Note the frequency, then take the boost off.
- Decide how wide it isWiden the bell until it covers the whole problem. Under about an octave calls for a dynamic EQ. Two octaves or more, or a range running all the way to the bottom or top, is a job for a multiband band.
- Check whether it comes and goesWatch the problem across the song. If it's there all the time, a static EQ cut is simpler and you can stop here.
- Set the threshold by the gain reductionLower it until the band comes down 2 to 4 dB on the loud moments and returns to zero between them. If it never gets back to zero, raise the threshold.
- Solo the bandUse the plug-in's listen or solo button to check the band is on the sound you're after and nothing else.
- Compare at matched loudnessBypass it at the same level and listen. If you can't tell what improved, back the threshold off or take it out.
Common mistakes
Using either one as a static EQ
If the band is turned down for most of the song, it's working as a slow, wobbly EQ cut. The lab's coach warns you past 60% of the loop. Swap it for a static cut and keep the dynamic band for problems that come and go, like the loud chorus notes or the odd boomy kick.
Making multiband bands narrow
Pulling two crossovers close together to target one resonance gives you a band that can only cut a few decibels and that shifts the phase on either side of it. A dynamic EQ is built for that job.
Turning on every band because it's there
A four-band compressor doesn't need four bands compressing, and each extra one is another set of controls to get right and another thing moving in the mix. Start with only the band that has a problem and leave the rest alone.
Blending a multiband sound with a dry copy
Parallel processing and multiband crossovers don't mix well, because the crossover's phase shift partly cancels against the dry signal. If you want a parallel blend, switch to linear-phase crossovers, or use a multiband that leaves the phase alone while a band is at rest, like Pro-MB in Dynamic Phase mode. A dynamic EQ mostly avoids the problem, because nothing in its audio path moves until the band does.
Judging by loudness
Both make the sound quieter while they work, so once makeup gain is added the processed version can end up louder and win the comparison for that reason alone. The A/B in the lab matches the loudness for you. In your DAW, match it by ear before you decide.
Training your ears to tell them apart
The quiz under the lab plays each round untouched, through a dynamic EQ, or through a multiband compressor. Both are set to the same two-octave band and pushed hard enough to hear, and the loudness is matched, so the clue is the shape of the cut: the dynamic EQ dips most at the centre, while the multiband takes the whole slice down at once and sounds wider. Start on the bass loop, where the band sits on the notes, then try the vocal.
Questions people ask
Is dynamic EQ better than multiband compression?
Neither is better across the board. A dynamic EQ is better for one narrow problem, like a single boomy bass note or a harsh resonance, because a bell can be as thin as you like and does nothing at rest. A multiband compressor is better for controlling a whole range, like everything under 150 Hz, because its bands cover a range evenly and add back up flat.
Can a dynamic EQ replace a multiband compressor?
Mostly, yes. Set a dynamic EQ band to a low or high shelf with a ratio and it behaves a lot like the bottom or top band of a multiband compressor. What you lose is a band with a flat middle and clean edges, which is what you want when a whole range like 300 Hz to 2 kHz has to come down together.
Is a de-esser a dynamic EQ?
A split-band de-esser is very close to one: it listens to the high end and turns only that part down when an s goes over the threshold. A dynamic EQ band set to a high shelf, or a bell somewhere between 5 and 9 kHz, does the same job. The difference is mostly in the controls, since most de-essers show little more than frequency, threshold and range.
Does multiband compression cause phase problems?
The crossover filters shift the phase of the frequencies around each crossover, even when nothing is compressing. On its own you rarely hear it, because the bands still add back up to a flat level. It shows up when you blend the processed sound with a dry copy, as in parallel processing, where the shifted frequencies partly cancel. Linear-phase crossovers avoid it at the cost of latency, and FabFilter Pro-MB's Dynamic Phase mode leaves the phase alone until a band's gain moves.
How many bands should I use on a dynamic EQ?
One for each problem you can hear, and usually no more than two or three. Each band should be doing something you can point to when you solo it. If a band is turned down for most of the song, replace it with a static EQ cut, which does the same job with less to go wrong.
Should multiband compression or dynamic EQ go on the master?
Either can, lightly. On a master, a multiband compressor is the usual choice for keeping the low end even from section to section, and a dynamic EQ is the usual choice for one resonance that only appears in the loud parts. Keep either to 1 to 3 dB of gain reduction, and fix the problem in the mix instead if you can.