What is subtractive EQ?

Subtractive EQ in plain English

Every EQ move is either a boost or a cut. Subtractive EQ is the habit of reaching for the cut first. When a sound has a problem, you look for the range of frequencies causing it and turn that range down, instead of turning up everything around it to cover the problem.

Say an acoustic guitar sounds boomy. You could boost the top end until the boom feels smaller in comparison, but the boom would still be there, and the hiss of the recording would come up with the treble. A cut of a few decibels on the boom fixes what was wrong and leaves the rest of the guitar as it was.

Any EQ can do this, including the stock one in your DAW. The EQ lesson explains the controls themselves. This page is about using them to take things away: which problems to look for, how to find them by ear, how much to cut and how to tell when you've gone too far.

Reading the lab

The lab starts on an acoustic guitar recording because it has two real problems in it. On the big screen is the guitar's spectrum, low frequencies on the left and high ones on the right, with the original drawn as a pale line and the output of the EQ as a teal shape over it. The pink curve is the EQ itself, and wherever it dips below the middle line, that range is being cut.

Band 1 is a high-pass filter, switched off to begin with. Band 2 is a narrow bell for hunting down problems, and bands 3 and 4 are spare bells at 600 Hz and 3 kHz for your own experiments. Drag a numbered point to move it, or use the controls under the screen. The Right now panel measures what each change does to this recording, including how much of the guitar's energy sits in the range the selected band covers.

Listen to this band plays only the slice of the sound around the selected band, which is how you check what you're about to remove. The range strip under the screen does the same for six fixed ranges. A/B switches the EQ off and sets the original to the same loudness as the EQ'd sound, because cuts make a sound quieter, and quieter can sound worse even when the tone is better.

Why cut before you boost?

Cuts take level away. A 6 dB cut halves the height of the wave in that range, so the overall level of the sound goes down and never up. Boosts use up headroom, and across a mix of thirty tracks those small boosts add up to a louder, harder-to-manage mix.

Cuts also hide better. A narrow boost picks out one frequency and makes it ring, which is easy to hear as a flaw. A narrow cut removes a thin slice of the sound, and a dip like that is much harder to hear than a peak of the same size. That's why a deep cut at a high Q can take out one ringing note with almost no other change.

In a mix, when two sounds share a range, the louder one covers the quieter one, which is called masking. Cutting a little from the part that matters less often does more for the important part than boosting it would, and it keeps the overall level down while you do it.

Boosting still has its place. If a sound is missing something, a small, wide boost is the right tool, and the additive vs subtractive EQ lesson compares the two approaches directly.

Common problems to cut

Most problems that subtractive EQ fixes live in the same few places. The frequencies below are where to start listening, and they shift with every instrument and every recording. Use them to know which part of the spectrum to search, then let your ears and the band solo decide.

Rumble: below about 40 to 80 Hz

Rumble is low-frequency noise that isn't part of the performance: traffic, air conditioning, footsteps, a knock on the mic stand. It's often too low to hear on small speakers, but it still moves the meters and makes compressors react to sounds nobody can hear.

The fix is a high-pass filter set just under the lowest note the instrument plays. The guitar in the lab has a little rumble under 60 Hz, which shows as the pale line rising again at the far left of the spectrum. The lowest note a guitar in standard tuning can play is the E at 82 Hz, so a filter at 60 Hz takes the rumble and leaves every note. On this recording it changes the overall level by only 0.2 dB, which is typical: rumble is usually a small part of the energy, and you filter it so meters and compressors stop reacting to it.

Set the filter by listening. Start low, raise it until the sound begins to thin, then back it off. Kick drums and bass lines need their low end, so leave them alone, or filter only the very bottom at 20 to 30 Hz.

Boom and mud: about 100 to 400 Hz

Boom is too much low end on a sound that should feel lighter, like a guitar recorded close to the microphone or a singer right on top of the mic. Mud sits a little higher, around 200 to 400 Hz, and it's what happens when lots of instruments pile their low mids into the same space until the mix turns cloudy.

A wide cut of 2 to 4 dB is the usual fix. On the guitar in the lab the build-up is unusually strong: about half of the recording's energy sits between 180 and 220 Hz, mostly from notes around the G below middle C. A cut there changes the sound a lot, so a smaller cut goes further than you'd expect. When the whole bottom end is too heavy, a low shelf cut does a similar job across everything below its corner, which the low cut vs low shelf lesson compares with a filter.

Boxiness: about 400 Hz to 1 kHz

A boxy sound has a hollow, cardboard quality, like a voice recorded in a small, bare room or a drum mic picking up more shell than hit. It's usually one area of the lower mids that's a little too strong, so a cut of 2 to 4 dB at a Q of around 2 to 4 often clears it. Sweep slowly here, because the right spot can be a few hundred hertz either way.

Harshness: about 2 to 5 kHz

Human hearing is most sensitive around 2 to 5 kHz, so a small excess here quickly becomes tiring. Bright guitars, close-miked cymbals and some vocals get a hard, piercing edge in this range, and a cut of 1 to 3 dB is often enough. Go gently: clarity and attack live here too, and too much cutting makes a sound dull and distant.

The s and t sounds of a vocal, called sibilance, sit a little higher, around 5 to 10 kHz. A fixed cut there dulls the whole vocal to fix a few syllables, so the better tool is a de-esser, which only cuts while an s is sounding.

Ringing notes and resonances

A resonance is one frequency that rings on louder and longer than the rest, like a tom that hums the same pitch after every hit or a room that makes one bass note boom. They're narrow, so the cut should be narrow too: 6 to 12 dB at a Q of 5 to 10 is common, and because the slice is so thin the overall tone barely changes.

For a frequency that never moves, such as electrical hum, use a notch filter. Mains power runs at 50 Hz in the UK and Europe and 60 Hz in North America. Hum from it shows up at that frequency and often at its multiples, so 100, 150 and 200 Hz may need their own notches.

How to find a frequency to cut

The most common method is boost and sweep. Set a bell to a narrow width, a Q of about 4 to 6, and boost it far more than you ever would in a mix, around 8 to 12 dB. Then slowly move its frequency through the range you suspect while the sound plays. When the bell passes over the problem it jumps out, and whatever you disliked about the sound gets much worse. That's your frequency.

Now take the boost away before you decide anything. A 10 dB narrow boost makes almost any frequency sound ugly, so the sweep can talk you into cutting something that was fine. Listening to the slice on its own helps: if what you hear in the band solo is the problem you noticed, cut it. If it sounds like a normal part of the instrument, leave it.

Once you've decided, flip the gain from a boost to a cut and widen the band a little. Boom and mud usually spread across a wider range than the sweep suggests, so a Q of 1.5 to 3.5 sounds smoother than the narrow one you searched with. Keep the narrow setting for a single ringing note.

A spectrum analyser helps too. Logic's Channel EQ and Ableton's EQ Eight can both draw one behind their curves, and the lab shows one on its big screen. A peak that stands well above its neighbours through the whole loop is worth a listen. Let the analyser point you somewhere and let your ears decide, because a tall peak can simply be the note the song is in.

How to use subtractive EQ, step by step

This order works on almost any track, and the lesson in the lab follows it.

1. Listen in the mix firstDecide what's wrong with the sound while everything else is playing. A problem you can only hear in solo usually isn't a problem.
2. Clear the rumbleAdd a high-pass filter and raise it until the sound starts to thin, then back it off a little. Skip this on kick and bass.
3. Find the problemBoost a narrow bell by 8 to 12 dB and sweep it through the range you suspect. Stop where the problem gets worst.
4. Check it on its ownSolo the band, or turn the boost off and listen again, and make sure what you found is the thing you wanted to remove.
5. Cut, starting smallFlip the gain to a cut of 2 to 4 dB and widen the band to suit the problem. Go deeper only on a narrow resonance.
6. Match the level and compareBring the output up by the level the cuts took away, then switch the EQ off and on. If you can't hear what improved, make the cut smaller or take it out.

Choosing how much to cut, and how wide

Match the width of the cut to the width of the problem. Broad problems like mud and boom spread over an octave or more, so they want a wide bell and a modest cut. A resonance is one frequency, so it wants a narrow bell that can go deep without taking anything else.

A cut also makes the whole sound quieter, by an amount that depends on how much of the sound lived in that range. Suppose 40% of a sound's energy sits in the range you cut by 6 dB. Six decibels down is a quarter of the power, so that 40% becomes 10%, and the sound as a whole drops to 70% of its power, about 1.5 dB quieter. The same cut on a range holding 2% of the energy changes the overall level by less than a tenth of a decibel.

Keep that level drop in mind when you judge the result. Louder sounds better for the first few seconds, so a cut that made the sound 1.5 dB quieter can lose a quick comparison even when it improved the tone. Turn the output or the channel fader up to match before you compare, or use the A/B in the lab, which does it for you.

Using cuts to make room in a mix

Subtractive EQ works on the relationship between sounds as well as on single sounds. When a vocal gets lost behind a guitar, turning the vocal up makes the mix louder and leaves the guitar in the way. Cutting 2 to 3 dB from the guitar where the vocal carries its words, somewhere between 1 and 4 kHz, lets the vocal through without touching it.

Choose which part gets the cut by asking which one the song needs in that range. A pad under a vocal rarely needs its upper mids. A kick and a bass line that share the same low notes usually work better when one of them owns the deepest part and the other is cut there, which the kick and bass lesson covers in detail.

Small cuts on several parts add up. If the guitars, the keys and a pad all have similar amounts of low mids, taking 2 dB out of each lowers the total around 300 Hz by 2 dB. A 6 dB cut on just one of them lowers it by only about 1.2 dB, and leaves that one part sounding hollow.

Cuts and compression

Put corrective cuts before the compressor where you can. A compressor reacts to the level of everything that goes into it, including rumble and boom, so if you leave a boomy 200 Hz in a vocal the compressor turns the voice down on every low note. Cut first and the compressor responds to the part of the sound you're keeping. The EQ before or after compression lesson covers the cases where the other order works better.

Common subtractive EQ mistakes

Cutting until the sound is thin

Every cut takes something away, and a few of them stacked up can leave a sound with no body. One step in the lab plays the guitar with a 160 Hz high-pass and a 12 dB cut, and it comes out small and papery. Stop as soon as the problem is gone, and if you're not sure, make the cut smaller.

Believing the sweep

Boost and sweep makes every frequency sound like a problem for a moment. Always turn the boost off and listen again before you cut, and if you can't hear the problem with the boost gone, there's nothing to fix.

Cutting the same frequencies on everything

Charts of where to cut are a starting point. The last step in the lab puts the guitar's EQ on a vocal, where it does almost nothing useful, because this vocal has almost nothing at 200 Hz and its low mids sit higher, around 250 to 300 Hz.

Doing it all in solo

A guitar with a big low-mid cut can sound thin on its own and sit perfectly in the mix, and the opposite happens too. Use solo to find a problem, then decide how much to cut with the rest of the track playing.

Using EQ to fix the wrong thing

If a sample needs 15 dB of cuts to fit, a different sample is usually the better fix. EQ can't remove room echo either, because the echo covers the same frequencies as the sound itself. Moving the microphone, or picking another sound, beats a stack of deep cuts.

Training your ears to find problems

Subtractive EQ depends on hearing where a problem is, and you can practise that. The quiz under the lab boosts one of five ranges by 9 dB and asks which one, with the screens covered and the level matched. It's the same listening you do in a boost and sweep, so getting faster at the quiz makes you faster at finding things to cut. Your best score out of eight is saved in your browser.

Between rounds, try the method on the other sounds in the lab. The chords have their low-mid build-up higher than the acoustic guitar, around 250 Hz, and the second vocal higher again, around 320 Hz. Finding each one by ear, before you look at the screen, is good practice.

Questions people ask

Keep going