Ducked delay in plain English
A delay plays back whatever you feed it a set time later. That sounds great in the gaps, when the echo answers a line that has just finished, and gets in the way everywhere else, because the echo of one word arrives while the next word is being sung. Two copies of a voice singing different words at once are harder to follow than one.
Ducking fixes that by turning the echoes down whenever the dry sound is playing. The delay keeps running the whole time, with the same delay time and feedback, but what comes out of it is held a few decibels lower while the singer sings. As soon as they stop, the echoes rise back to full level and fill the space.
This page is about that one control. If you're new to delay itself, the delay lesson covers delay time, feedback and mix first.
Why duck a delay?
Without ducking, a vocal delay forces a trade-off. Set it loud enough to hear clearly at the end of a line and it clutters the middle of the next one. Turn it down until the words are clear again and it all but vanishes in the gaps, which is where you wanted it. You end up with a timid delay that does neither job well.
A ducked delay lets the level be different in the two places. You can set the echoes loud, because they only reach that level when nothing else is in the way, and the same setting stays out of the way while the part plays. The engineering word for one sound hiding another is masking, and ducking stops the echoes and the part they come from masking each other.
It's also less work than the alternative, which is riding the delay return by hand or automating it up at the end of every line. Ducking does that automatically and follows the performance, so a held note keeps the echoes down for longer and a long gap lets them through for longer.
Reading the lab
The lab above is the same delay as the main delay lesson, set up to show ducking. The Duck knob sets how far the echoes come down when the dry sound is at its loudest, from Off up to 24 dB, and the Release knob sets how fast they come back up once it stops. Sync, time and ping-pong are locked here, so the delay always follows the loop's 120 BPM, and every step uses a 1/4 note, which is 500 ms.
The screen to watch is the whole loop, the first one in the second row. The pale shape is the dry sound and the coloured shape behind it is the echoes. When ducking is on, an amber line appears underneath: it drops when the echoes are being turned down and rises when they're let back up. Hover over any point, or press and drag on a phone, and the lab tells you how far the echoes are being turned down at that moment. The echo timeline at the top leaves ducking out, because it measures the repeats with a short burst of noise and nothing playing afterwards, so there's nothing for the ducking to react to.
The lab's ducker listens to the dry sound going into the delay. It starts to turn the echoes down when the dry sound's peaks pass −40 dBFS and reaches the full Duck amount at −10 dBFS, so quieter passages are ducked less than loud ones. It reacts within about 50 ms, and at the default 150 ms release the echoes take about half a second to come all the way back after a loud sound stops dead. The threshold and attack are fixed in the lab; the Release knob is the one control that changes. Real delay plug-ins and compressors let you set all three, which the next sections cover.
The Right now panel reads its numbers from the sound that's playing. With ducking on, it tells you the deepest duck in the loop and how much of the loop the echoes spend more than 3 dB down, which is the quickest way to tell whether a part has any gaps for the echoes to come back in.
How much to duck
The Duck amount, called depth or range on some delays, is how far the echoes drop at the loudest moments. Say your echoes are set so the first repeat sits 3.5 dB under the dry sound, as in step 1 of the lab. With 6 dB of ducking, that repeat sits 9.5 dB under the dry sound while the part plays, and back at 3.5 dB under in the gaps. With 12 dB it's 15.5 dB under while the part plays, which is low enough that you mostly hear it as space around the notes.
For a vocal delay, 6 to 12 dB is the usual range. Below about 4 dB the difference between ducked and unducked is hard to hear in a full mix. Past 15 dB the echoes are almost silent under the voice and jump up in every gap, which you start to hear as the delay switching on and off. That can be a deliberate effect on a sparse part, and on most parts it's too much.
The bigger change comes from what ducking lets you do with the rest of the delay. In step 5 the mix goes up to 60% and the feedback to 55%, so the first echo is 3.5 dB louder than the dry lead. Without ducking that would bury the part. With 12 dB of ducking the notes still come through, and the rests carry seven repeats above −40 dB.
Threshold, attack and release on a ducker
A ducker is a compressor whose sidechain, the signal it listens to, is the dry sound, while the audio it turns down is the echoes. So it has the controls a compressor has, and on a delay with built-in ducking you'll usually find at least two of them. Ableton Live's Echo, for example, has a Ducking switch on its Character tab with Threshold and Release controls.
Threshold
The threshold is how loud the dry sound has to be before the echoes start to duck. Set it too high and only the loudest words duck the echoes, so they pop up again on every quieter word. Set it too low and the breaths, the reverb on the dry track or the bleed in a live recording keep the ducker working in the gaps, so the echoes never come back. Aim for the point where the gain reduction meter falls back to zero between lines.
Attack
On a ducked delay the attack matters less than you'd think. The first echo of a word arrives a whole delay time after the word, 500 ms here, so by then the ducker has long since reacted. What attack decides is how quickly the echoes still ringing from the last line get out of the way when the next line starts. A fast attack, 1 to 10 ms, clears them before the first syllable has finished. A slower attack lets the tail of the old echoes overlap the start of the new line for a moment.
Release
Release is the control that shapes how a ducked delay sounds. It sets how fast the echoes come back up once the dry sound stops, and it has two jobs that pull against each other. It has to be long enough that the echoes stay down through the short gaps between words, which are often only a tenth or two of a second; shorter than those gaps, and the echoes jump up between syllables and the delay flutters. It also has to be short enough that the echoes come back in the longer gaps between lines, or they'll stay ducked right through.
On a vocal, 150 to 300 ms is a good place to start. You can also work from the tempo: one beat lasts 60,000 ÷ BPM milliseconds, so 500 ms at 120 BPM. A release of around half a beat means the echoes are mostly back up by the time a 1/4 note delay plays its first repeat of the last word, one beat after the singer stops.
The figure below runs the lab's own ducker on a test tone. Move the dry level and the recovery time changes with it, because the lab measures the dry sound with a meter that falls at a fixed rate in decibels. A louder sound has further to fall before it drops under −40 dBFS, so the echoes take longer to come back. A compressor with an ordinary release curve does much the same: the more gain reduction it's applying when the sound stops, the longer it takes to get back to zero, which is why a ducker can feel slower after a loud line than after a quiet one.
The Release knob in the lab shows both sides of that trade. On the lead, cutting the release from 150 to 40 ms lets the echoes back into more of the small gaps, so they spend 63% of the loop more than 3 dB down instead of 74%. At 600 ms the lead's rests are too short for them to recover at all: they never climb back above −5 dB, and the delay is ducked for the whole loop.
Ducking needs gaps
Ducking only helps a part that stops. Every gap is a chance for the echoes to come back up, and a part with no gaps keeps them ducked the whole time, so they end up at one lower level, the same as you'd get by turning the delay down.
The loops in the lab show both cases. The acoustic guitar strums for about six and a half seconds and then stops for about three. While it strums, the ducking moves with each strum, and when it stops the echoes ring out at full level. The vocal is one long phrase with hardly a breath in it, so the echoes spend the whole loop 5 to 12 dB down and never get their moment.
For a part like that vocal, filters inside the delay's feedback loop do more good than ducking. A high cut around 5 kHz and a low cut around 300 Hz make every repeat darker and thinner than the voice, so even when the echoes overlap the words you can tell the two apart. The delay feedback lesson explains why a filter inside the loop gets stronger on every repeat. A short duck on top keeps the echoes a little further back while the words are sung.
Dense parts like a busy drum loop or a guitar that never lets up behave the same way. Pick Drums or Guitar in the lab with 12 dB of ducking and you'll see the amber line hardly moves.
How to set up a ducked delay, step by step
If your delay has its own Duck control, use that and go straight to step 4. Otherwise these steps work in any DAW where a compressor can take its sidechain from another track, and all the main DAWs can.
- Put the delay on a return trackSet the delay to 100% wet and feed it from the vocal with a send. On a return, the dry vocal stays on its own track, which you need for the next step.
- Add a compressor after the delayIt has to come after the delay on the return. A compressor before the delay would duck the sound going in, and the echoes of it would still come out half a second later, right on top of the next line.
- Feed the compressor's sidechain from the dry vocalIn Ableton's Compressor, open the Sidechain section with the small triangle in the device's title bar, switch Sidechain on and choose the vocal track under Audio From. In Logic, pick the vocal from the Side Chain menu at the top of the plug-in window.
- Set a ratio of 4:1 or more and a fast attackA high ratio makes the echoes drop firmly as soon as the vocal starts. An attack of 1 to 10 ms clears the old echoes before the first syllable is over.
- Lower the threshold until the meter shows 6 to 10 dBWatch the gain reduction meter while the vocal sings, and check it falls back to zero in the gaps between lines.
- Set the release by earStart at 200 ms. Shorten it if the echoes stay down through the gaps, and lengthen it if they flutter up between words.
- Set the return level lastWith the ducking working, turn the return up until the echoes sound right in the gaps. It will usually end up higher than you'd have dared without ducking.
The sidechain compression lesson has a lab for the compressor half of this, with a kick ducking a bass and the threshold, attack and release all on knobs.
Ducking a reverb
A long reverb has the same problem as a delay. The tail of one word is still sounding when the next word starts, so on a busy vocal a 2 or 3 second hall can fog up the words even when its level sounds right in the gaps. The fix is the same as steps 1 to 6 above: put the reverb on a return, add a compressor after it and key the compressor's sidechain from the dry vocal.
Reverb usually needs less ducking than a delay, 3 to 6 dB of gain reduction where a delay takes 6 to 12, because a smooth reverb tail is less distracting than a clear repeat of a word. A slightly longer release, 300 to 500 ms, lets the tail swell gently at the end of a line instead of jumping up. Short rooms and plates under about a second rarely need ducking, since their tails are over before they get in the way.
Pre-delay solves part of the same problem in a different way. It holds the reverb back by somewhere between 20 and 100 ms, which keeps the start of each word dry and clear, but once the reverb starts it plays at full level while the singer carries on. Ducking holds it down for as long as the singer is singing. The two work well together, and the pre-delay lesson has its own lab. The lab on this page runs a delay only, so to hear reverb on its own, use the lab in the reverb lesson.
Starting settings for ducked delay
These are places to start, picked so the ducking is doing something you can hear. Release here means the ducker's release, on a built-in duck control or on a sidechained compressor after the delay.
| Use | Delay | Feedback | Ducking | Release |
|---|---|---|---|---|
| Slow vocal, long gaps | 1/4 | 30 – 45% | 8 – 12 dB | 200 – 300 ms |
| Fast vocal or rap | 1/8 | 15 – 30% | 6 – 10 dB | 100 – 200 ms |
| Lead synth or guitar solo | Dotted 1/8 | 30 – 50% | 6 – 12 dB | 150 – 300 ms |
| Part with no gaps | Any | 20 – 35% | 3 – 6 dB | 150 – 300 ms |
| Long vocal reverb | n/a | n/a | 3 – 6 dB | 300 – 500 ms |
Common ducked delay mistakes
Putting the compressor before the delay
A compressor keyed from the vocal and placed before the delay turns down what goes into the delay, so the echoes come out ducked half a second late, exactly when the vocal has moved on. It has to sit after the delay, where it can turn the echoes down at the moment they clash.
Keying the compressor from the wrong signal
If the sidechain is left on its default, most compressors listen to the audio passing through them, which here is the echoes. That just compresses the delay. The sidechain has to come from the dry vocal track.
Ducking so deep the delay sounds switched
At 15 dB or more the echoes almost disappear under the part and leap up in every gap. Unless that's the effect you're after, back the ducking off to 6 to 10 dB and listen for echoes you can still just hear under the voice.
A release so short the echoes flutter
Under about 100 ms the ducker lets go in the tiny gaps between syllables, so the echoes bob up and down through every line. If the delay sounds jittery when the vocal is busy, lengthen the release.
Expecting ducking to fix a part with no gaps
On a part that never stops, ducking just lowers the delay. Filter the repeats and lower the feedback, or keep the delay for a throw on the last word of a line, sent only at that moment.
Train your ears
The quiz under the lab sets a hidden amount of ducking on the lead or the acoustic guitar, none, 6 dB or 18 dB, and asks which one you're hearing. The controls and screens are covered and every round is level-matched, so the clue is how far the echoes drop under the notes and how they come back in the rests. With none they stay at one level and run on under the next notes; at 18 dB they all but vanish under the notes and swell up in each rest.
Outside the quiz, go to step 1 in the lab and move the Duck knob slowly up from Off while it plays. Listen for the point where the notes come clear of their own echoes; that's usually the place to stop.
Questions people ask
Is ducked delay the same as sidechain compression?
Ducking is the effect and sidechain compression is the usual way to get it. A Duck control built into a delay is a small compressor that listens to the delay's own dry input and turns the echoes down. Doing it yourself with a compressor on the delay's return, keyed from the dry track, gives you the same result with more control over threshold, attack and release.
How do I duck a delay that has no duck control?
Put the delay on a return track at 100% wet and add a compressor after it. Set the compressor's sidechain input to the dry vocal track, then use a ratio of 4:1 or more, an attack of 1 to 10 ms and a release of 150 to 300 ms. Lower the threshold until the gain reduction meter shows 6 to 10 dB while the singer is singing and falls back to zero in the gaps.
What release should a ducked delay have?
Long enough that the echoes stay down through the short gaps between words, and short enough that they come back up in the longer gaps between lines. On a vocal, 150 to 300 ms is a good start. Below about 100 ms the echoes jump up between syllables and flutter; much past 500 ms they may not recover before the next line starts.
Should I duck the reverb as well?
On a busy vocal with a long reverb, often yes. A compressor on the reverb return, keyed from the dry vocal, taking 3 to 6 dB off while the singer is singing, keeps the words clear and lets the reverb bloom at the ends of lines. Short rooms and plates under about a second rarely need it.
Why does my ducked delay disappear completely?
The part probably never stops long enough for the ducking to let go, or the ducking is too deep. Check the gain reduction meter: if it never gets back to zero, raise the threshold or shorten the release. Also check the sidechain is fed from the dry track. A compressor listening to the delay return itself only squashes the echoes.
Can I duck a delay from the kick instead of the vocal?
Yes. Key the compressor on the delay return from the kick and the echoes dip on every beat, the same way a pad pumps in a dance track. It's a different effect from vocal ducking: the echoes move with the rhythm of the drums, whatever the vocal is doing.