In a DAW, use a parametric. It can draw the broad shapes a graphic draws, and it can also aim a narrow cut at one resonance, which a graphic can't. A graphic is the better tool when you need every band under your hand at once, as when you're tuning a PA or pulling feedback out of a monitor wedge mid-show.
The lab above is both. Its ten bells sit at the standard graphic frequencies, so while you only touch their gain it behaves as a graphic, and the moment you move a band or change its Q, that band is parametric.
At a glance
| Property | Graphic EQ | Parametric EQ |
|---|---|---|
| What you control | Gain only, one slider per band | Frequency, gain and Q on every band, and usually the band type |
| Number of bands | Usually 10, 15 or 31 | Usually 3 to 8; some plug-ins allow 24 |
| Where bands sit | Fixed, at standard frequencies such as 63, 125 and 250 Hz | Anywhere from 20 Hz to 20 kHz |
| Band width | Fixed, about an octave or a third of an octave | Anything from several octaves to a thin slice |
| Reading the curve | The sliders draw a rough picture of it | Drawn on screen in plug-ins; read off the knobs on hardware |
| Quickest at | Broad shapes and live adjustments | Finding and fixing one frequency |
| Where you'll find it | Live sound racks, hi-fi, car stereos, music apps | Every DAW, mixing desk channels, mastering |
What is a graphic EQ?
A graphic EQ is a row of bands at fixed frequencies, each with its own slider. Push a slider above the centre line and it boosts that frequency; pull it below and it cuts. The sliders run from low frequencies on the left to high on the right, so their positions draw a rough graph of the EQ curve, and that's where the name comes from.
Most graphics use the same centre frequencies, taken from the ISO 266 list of preferred frequencies. A 10-band graphic has one band per octave: 31, 63, 125, 250 and 500 Hz, then 1, 2, 4, 8 and 16 kHz. A 31-band graphic has one every third of an octave from 20 Hz to 20 kHz, so between 63 and 125 Hz it adds sliders at 80 and 100 Hz. The 15-band size in between puts one every two-thirds of an octave.
The width of each band is fixed as well, and it's matched to the spacing so neighbouring bands meet without gaps. On a 10-band graphic each one is about an octave wide, a Q of roughly 1.41. On a 31-band it's about a third of an octave, a Q near 4.3. You'll meet graphics in live sound racks, such as the Klark Teknik DN360 with its two channels of 30 third-octave sliders from 25 Hz to 20 kHz, and on plenty of hi-fi amplifiers and car stereos. In the studio, the API 560 is a 10-band graphic in the 500-series format.
What is a parametric EQ?
A parametric EQ gives every band three controls, the parameters it's named after. Frequency moves the band anywhere from 20 Hz to 20 kHz, gain sets how far it boosts or cuts, and Q sets how wide it is. Most let you switch a band between a bell, a shelf and a high-pass or low-pass filter as well.
George Massenburg described the design in a paper to the Audio Engineering Society in 1972, and it's now the default almost everywhere you mix. Ableton Live's EQ Eight and Logic's Channel EQ each have eight bands and FabFilter Pro-Q allows up to 24, while the channel EQ on a large mixing desk is usually a four-band parametric.
Q is the control a graphic keeps from you, and it's the one that makes the biggest difference to what a cut sounds like. The figure shows one 9 dB bell at 1 kHz; drag its Q control to see how much of the spectrum it takes with it.
Between the two sits the semi-parametric EQ, a band with frequency and gain but a fixed Q. The swept mid on a small mixing desk is the usual example. It can travel to a problem, but it can't narrow down onto it. If EQ itself is new to you, start with the EQ lesson, which covers what boosting and cutting do to a sound before you choose a layout.
Reading the lab
The lab has ten bells at the 10-band graphic frequencies, each one an octave wide. Select a band from the numbered chips and turn Gain, or drag its point on the big screen straight up or down, and you're moving one slider of a graphic EQ. Drag a point sideways or turn its Q, and that band turns into a parametric one. Band types are locked to bells so the comparison stays fair.
The pale line on the big screen is the spectrum of the original loop and the teal shape is the loop after the EQ. The pink line is the EQ curve, and the tinted hill under it is the selected band on its own, which shows how far each bell spreads into its neighbours. Hover over the curve to read the boost or cut at any frequency. The Right now panel describes what the band you last touched is doing, using numbers measured on the loop.
A/B switches the EQ off and matches the level, so a boost can't win by being louder. Listen to this band plays only the slice of sound the selected band covers, and pressing a name on the range strip along the bottom of the screen plays that range of the spectrum on its own.
Cutting a problem between two sliders
A graphic is quick right up to the moment the problem sits between two sliders. Say an acoustic guitar has a boxy ring at 350 Hz. On a 10-band graphic the nearest bands are 250 Hz and 500 Hz, half an octave either side, so the only thing you can do is pull both down.
With both at −6 dB, the curve does reach −6 dB at 350 Hz. It's also −7.1 dB at 250 Hz and at 500 Hz, and it stays more than 3 dB down from about 170 Hz to 750 Hz. That's over two octaves of the guitar's body gone to fix one ring.
On a parametric you move one bell to 350 Hz and raise its Q to 4. The same 6 dB cut now runs from about 310 Hz to 400 Hz at half depth, and 250 Hz and 500 Hz are each down by less than 1 dB. Raise the Q to 8 and you can go to −9 dB over an even thinner slice, roughly 330 Hz to 370 Hz.
A 31-band graphic gets closer, with sliders at 315 Hz and 400 Hz, but each one is still about a third of an octave wide and neither sits on 350 Hz. For a single resonance, such as a ringing tom or a whistle in a room recording, that gap is the reason to use a parametric. The notch filter lesson covers the narrowest cuts of all.
Neighbouring bands add up
The bands on a graphic overlap so the curve stays smooth, and that has a side effect: the curve isn't always where the sliders say it is. Push 2 kHz and 4 kHz both up 6 dB on an octave graphic and the curve peaks at about 7.1 dB at each of them, because each bell still adds about 1.1 dB an octave from its centre. Halfway between, at 2.8 kHz, it sits near 5.9 dB.
One extra decibel is easy to live with. Push five neighbouring sliders the same way and the extra piles up, until the row of sliders is a poor picture of what you're hearing. That's one reason some graphics use proportional Q, where a band is wide for small moves and gets narrower the further you push its slider, so big moves overlap less. API builds the 560 that way. The bells in this lab are constant Q, meaning the width stays the same at any gain, which is how many graphics and most parametric plug-ins behave.
A parametric has overlap too, whenever two bands sit close together. The difference is that a plug-in draws the summed curve on screen, so you see the total instead of adding it up from slider positions.
Graphic EQs in live sound
Live sound is where graphics are still most common. A 31-band graphic on each monitor mix lets an engineer hear a frequency start to feed back and pull the nearest slider down in about a second, mid-song, without paging through a screen. Tuning a PA to a room works the same way: if a measurement microphone shows a 4 dB peak around 125 Hz, the 125 Hz slider takes it out.
Every band is on the front panel at once, so there's no selecting a band before you adjust it, and another engineer can glance at the rack and see what's been done. Digital mixing consoles usually offer both kinds now, and for one stubborn feedback frequency the parametric is the better choice, because it can go narrower.
Broad tone shaping on either
For wide, gentle moves the two are close to equal. Take 3 dB out at 250 Hz and add 3 dB at 4 kHz and the loop gets a little less thick and a little more forward, whichever EQ you do it on. A graphic gets there with two quick slider moves, and a parametric with two bells at a low Q.
The shape people most often draw on a home stereo is the smile: low and high sliders up, the middle ones down. It makes music sound fuller at low volume, because ears lose sensitivity to bass and treble as the level drops. That's a fair choice for listening. In a mix it scoops out the midrange where the vocal and most instruments carry their tone, so it's rarely the right move on a track.
Parametric EQ in a mix
Mixing is mostly small, targeted changes to one sound at a time, and problem frequencies rarely land on a graphic slider. Mud on a bass might sit at 180 Hz and harshness on a vocal at 3.2 kHz, and neither is a band on a 10-band graphic. On a parametric you'd put one bell at 180 Hz on the bass and another at 3.2 kHz on the vocal, each only as wide as the problem.
The other thing a parametric can do is sweep. You set a narrow boost, move it slowly across the range until something rings out, then turn the boost into a cut. That only works with a band that moves, and it's the fastest way to learn what each part of a sound is doing. The subtractive EQ lesson walks through it in detail, and frequency ranges tells you where to start looking.
How to find and cut a problem frequency
This is the parametric method, with the graphic version after it. Keep your monitors at a moderate level, because a sweeping boost can get loud and piercing.
- Set a narrow boostOne bell at +8 to +10 dB with a Q of about 4 to 6. In the lab, any band will do once you drag it sideways.
- Sweep slowly while the sound playsDrag the band across the range where you think the problem is. Stop when one spot rings out or sounds worse than everywhere else.
- Check it on its ownUse Listen to this band, or your EQ's solo, to hear only that slice. If it's the sound you were chasing, you've found it.
- Turn the boost into a cutStart at −3 to −6 dB. Widen the Q if the ring is still there, narrow it if the sound loses body.
- Compare at matched levelSwitch the band off and on. If you can't hear what improved, make the cut smaller or take it out.
On a graphic you can't sweep, so pull the sliders nearest the problem down one at a time and keep the one that helps most. If the problem is between two, lower both by less than you would one, so the extra from their overlap doesn't take out more than you meant.
Common mistakes
Treating the sliders as the curve
Neighbouring sliders add up, so a run of sliders all at +4 dB can give you +5 or +6 dB in the middle of the run. On a graphic, judge by ear; on a parametric, look at the summed curve the plug-in draws.
Pushing every slider up
If most of the sliders on a graphic are above the centre line, you've mostly turned the volume up, and the level can clip the next thing in the chain. Cut what's too loud instead, and use the output control to bring the level back.
Narrow boosts on a parametric
A high Q is for cuts. A big, narrow boost makes a sound whistle at one pitch, which is useful while you sweep for a problem and almost never what you want to keep. Boosts sound more natural wide, with a Q under about 2.
Choosing by loudness
Boosts make a sound louder, and louder wins quick comparisons. Compare with the level matched, which is what the A/B in the lab does for you.
Training your ears to pick the slider
Using a graphic well comes down to knowing which slider to reach for, and that's a listening skill. The quiz under the lab boosts one of five ranges by 9 dB with the screens covered and the level matched, then asks which one it was. Eight rounds take a few minutes, and the same skill tells you where to start sweeping on a parametric.
Questions people ask
Which is better, a parametric or a graphic EQ?
For mixing and mastering, a parametric, because it can put a band at any frequency and make it as wide or narrow as the job needs. A graphic is better when speed and seeing every band at once count for more than precision, as with live monitors, PA tuning and hi-fi.
Can a parametric EQ do everything a graphic EQ can?
If it has at least ten bands, it can draw the same curve. Set one bell per octave at the graphic frequencies with a Q of about 1.41 and you have a 10-band graphic, which is how the lab on this page is set up. What you give up is the speed of having every slider on the front panel at once.
Why is it called a graphic EQ?
Because the sliders sit side by side from low to high frequency, so their positions draw a rough graph of the EQ curve. You can read the shape from across a room, which is a big part of why live sound engineers like them.
How many bands does a graphic EQ have?
Usually 10, 15 or 31. A 10-band graphic has one band per octave from 31 Hz to 16 kHz. A 31-band has one every third of an octave from 20 Hz to 20 kHz, and a 15-band has one every two-thirds of an octave. Smaller ones with five to seven bands turn up on guitar pedals and car stereos.
What is a semi-parametric EQ?
An EQ band with frequency and gain controls but a fixed width. The swept mid on many small mixing desks is one. It can move to a problem frequency but it cannot narrow down onto it the way a full parametric band can.
Is a graphic EQ good for mixing?
For broad tone changes it works fine. For cutting one resonance, such as a ringing snare or a harsh frequency in a vocal, it's the wrong tool, because the problem rarely sits exactly on a slider.