What is delay?

Delay in plain English

Shout at a cliff and your voice comes back a moment later. Sound travels at about 343 metres a second, so if the cliff is 17 metres away the trip there and back takes a tenth of a second and you hear yourself 100 ms late. A delay effect does the same thing without the cliff. It holds on to whatever you send into it and plays it back after a time you choose.

One repeat on its own is called a single echo or a slapback. Most delays also have a feedback control, which sends some of the echo back into the input so it gets delayed again, and again after that, each time a little quieter. That's how one sung word turns into a trail of repeats that fades away over a few seconds.

Digital delays store the sound as samples in memory and read them back later. Older units did the same job with a loop of tape running past a record head and a playback head, or with a chain of tiny capacitors passing the signal along. The way they lost quality on every repeat is a big part of why people still like the sound of them, and you'll hear that in the filters section further down.

Why use delay?

The most common reason is space. A dry vocal can sound like it's in a cupboard, and reverb fixes that by pushing it back into a room. A delay adds space too, but because the repeats are separate from the voice it stays at the front of the mix. That's why a quiet 1/4 or 1/8 note delay is a common choice under a pop or rap vocal, often set so low that you only notice it when it's muted.

Delay is also rhythm. Synced to the tempo, the repeats become extra notes, and a simple part starts to sound busier than what was played. A dotted 1/8 delay on a guitar or synth line fills the gaps between the notes with echoes, which is how a handful of picked notes can sound like a constant flowing pattern.

And then there are effects. A long feedback tail on the last word of a line and a slapback that makes a voice sound like it was recorded in the 1950s are both just delay with the controls set differently.

Reading the lab

The wide screen at the top is the echo timeline. Time runs left to right, and every repeat is a bar placed at the moment it arrives, as tall as it is loud. The white bar at zero is the original. The top lane is the left speaker and the bottom lane is the right, so the two only differ once you switch on ping-pong. To get those levels, the lab sends a 20 ms burst of noise through the delay and measures what comes back, so the bars include the feedback, the mix and the filters. They leave out ducking, which depends on what the dry sound is doing at each moment, so the whole loop screen shows that instead.

Under it on the left is the whole loop as it comes out of the delay. The pale shape is the dry sound and the coloured shape behind it is the echoes, so you can watch them filling the gaps between phrases. In the middle is one bar of 4/4 at 120 BPM. If a sound lands on beat 1, the dots show where its repeats fall, and the bigger the dot, the louder the repeat.

The panel on the right explains the current settings with numbers worked out from the sound. It gives the delay time in milliseconds and as a note value, how many repeats are louder than −40 dB, and when the echoes drop below −60 dB, which is roughly the point where they've gone. Press A/B to hear the dry sound at the same loudness as the delayed one.

The lead, drums and chords loops are at 120 BPM, which is why the note values line up with them. The vocal and the two guitars weren't recorded to that tempo, so on those you hear the timing of the echoes without a beat to measure it against. Upload your own sound or record one with the Record button and it loops through the lab in place of ours, without leaving your device. The tempo maths below works at any BPM, but the beat grid in the lab assumes 120.

The controls

Most delays have some version of these, from the stock delays in Ableton Live, Logic Pro and FL Studio to a guitar pedal. The names change; a dotted 1/8 at 40% feedback sounds the same on all of them.

Delay time

Delay time is how long each repeat waits, measured in milliseconds. It decides what kind of effect you get more than any other control. Below about 30 ms your ear stops hearing the repeat as a separate event and hears one fuller sound instead, which is the idea behind the Haas effect. Somewhere around 80 to 150 ms you get a slapback, a fast repeat that sounds attached to the original. Past about 200 ms each echo is clearly its own sound.

Short times change the tone; long times change the rhythm. A 250 ms delay on a vocal adds a second, quieter syllable after every word, while a 20 ms delay on the same vocal just thickens it and colours its tone, because the two copies are too close together to hear apart. That colouring is comb filtering.

Syncing delay time to the tempo

Echoes that land on the beat sound like part of the song, and echoes that land just off it sound like a mistake. That's why almost every delay has a sync switch. With it on, you choose a note value and the delay works out the milliseconds from your project tempo.

The maths is short. A minute has 60,000 milliseconds, so one beat lasts 60,000 divided by the BPM. At 120 BPM that's 500 ms, which is a 1/4 note. Everything else comes from there:

• a 1/8 note is half a beat, 250 ms at 120 BPM
• a 1/16 is a quarter of a beat, 125 ms
• a dotted note is one and a half times the plain one, so a dotted 1/8 is 250 × 1.5 = 375 ms
• a triplet fits three notes where two would go, so a 1/8 triplet is 500 ÷ 3 ≈ 167 ms

The dotted 1/8 earns its own mention because it's the one beginners find hardest to picture. Each repeat lands three sixteenths after the sound before it. Played under a line of straight eighth notes, the echoes fall in the spaces between them, so the part sounds twice as busy without the timing ever clashing. On the beat grid in the lab, the first three repeats fall off the beat and the fourth lands on beat 4.

Triplets do something different. Two out of every three triplet repeats fall between the lines of the straight grid, so a 1/8 triplet delay on a straight part gives it a swing. Used on purpose that sounds great, and used by accident it sounds sloppy, which is a good reason to know which one you've picked.

Feedback

Feedback, sometimes labelled regeneration or repeats, decides how much of each echo goes back into the delay to make the next one. At 0% you get one repeat. At 50%, each repeat is half the level of the one before, which is 6 dB quieter, so about ten repeats go by before the echoes are 60 dB down. At 70% each one only loses about 3 dB, so the trail lasts around twice as long.

Feedback and delay time work together. At 120 BPM, ten repeats of a 1/16 note delay are over in just over a second; ten repeats of a 1/4 note take five. If a tail is running on into the next section of a song, shorten the time or bring the feedback down, and if you only want the tail on certain words, automate the feedback or the send level instead of leaving it high everywhere.

At 100% or more, every repeat comes back as loud as the last or louder, so the echoes build up until something distorts. Dub engineers rode their tape echoes into that runaway on purpose. Many plug-ins stop just below it or add saturation to keep it under control, and the lab stops at 90% so the repeats always die away.

Mix, and inserts versus sends

The mix control, also called dry/wet, sets the balance between the original and the echoes. In the lab it's a straight crossfade, so 0% is the dry sound only, 50% is equal amounts of both and 100% is echoes only. Plug-ins differ a little in how they handle the middle of that range, but the ends always mean the same.

Where you put the delay changes how you set the mix. On an insert, straight on the track, you'd usually keep it between 15 and 35%. On a send, the delay sits on its own return channel and the track sends a copy of itself there, so you set the delay to 100% wet and use the send knob to decide how much echo you get. The dry sound already comes from the track itself. Sends let several tracks share one delay and let you EQ or compress the echoes without touching the original, which the insert vs send lesson covers in full.

Ping-pong

A ping-pong delay sends the first repeat to one side, the next to the other and so on, so the echoes bounce between the speakers. Inside, the left side of the delay feeds the right and the right feeds the left. The lab sends the input into the left side, so the first repeat comes from the left and every second one from the right.

It's a good way to make a part sound wide without moving the dry sound away from where you panned it. It also holds up in mono: the repeats lose their bounce, but they don't cancel each other out. Keep the feedback moderate, around 30 to 50%, or the bouncing takes over the part. There's more on it in the ping-pong delay lesson.

High cut and low cut on the repeats

Most delays have a high cut and a low cut, and the useful ones put them inside the feedback loop. A filter after the delay would darken every echo by the same amount, but a filter inside the loop acts on the sound every time it goes round, so the first repeat is a little darker, the second darker again and the tenth is a muffled shadow of the original.

That's what tape echoes did on their own. Every pass through the tape lost some top end and added a little noise and saturation, which is why the repeats on a tape echo seem to drift away into the distance. A high cut somewhere between 2 and 5 kHz gets you most of the way there, and the tape delay lesson covers the rest of that sound.

The low cut is less romantic and more useful. A vocal or a synth has plenty of energy between 100 and 300 Hz, and when every repeat brings that back, the low mids of the mix fill up and the bass loses its edges. A low cut of 200 to 300 Hz on the delay leaves the echoes sounding thinner on their own and far easier to hear in the mix. In the lab, set both filters and watch the later bars on the timeline drop faster than the early ones.

Ducking

A ducked delay turns its echoes down while the dry sound is playing and lets them come back up when it stops. On a vocal that means the words stay clear while the singer is singing, and the gaps between lines fill with echoes. You can set a much louder delay this way without it smearing the lyrics.

Some delays have a duck control built in. Where they don't, put a compressor after the delay on its return channel and feed the dry vocal into the compressor's sidechain, which is the same trick as in sidechain compression. In the lab, the Duck knob sets how far the echoes drop when the dry sound is at its loudest, and the amber line under the whole loop shows the ducking moment by moment. Hover over it to read the numbers. There's more on setting it up in the ducked delay lesson.

Slapback delay

Slapback is a single fast repeat, usually somewhere between 80 and 150 ms, with little or no feedback. It's the sound of a lot of 1950s rock and roll and rockabilly, where it came from tape machines: the playback head sat a short distance after the record head, so the echo arrived a fixed time later. You can hear it on Elvis Presley's early recordings at Sun Studio.

It works well on vocals and on guitars that need to sound bigger without a long tail. The repeat comes so soon that it sounds like a thickening right behind each note. Leave it free of the tempo, since the exact time is less important than how it sits with the part, and you'll often find somewhere between 90 and 120 ms that feels right. The slapback delay lesson goes into it further.

Starting settings for common delays

These are places to begin. Each one is clearly audible on its own, and you'll usually bring the level down once the rest of the track is playing.

How to set up a delay, step by step

This works for a vocal delay on a send, which is the setup you'll use most. For an insert, skip the send and use the mix knob in its place.

1. Put the delay on a return channelSet its mix to 100% wet, then turn up the send from the vocal until you can clearly hear the echoes.
2. Pick a note valueSwitch on sync and start with a 1/4 note. Try a 1/8 or a dotted 1/8 and keep whichever sits with the groove without cluttering it.
3. Set the feedbackRaise it until the echoes carry across the gap after a line, and stop before they run into the next one. Somewhere between 20 and 40% is typical.
4. Filter the repeatsBring the low cut up to 200 or 300 Hz and the high cut down to 3 to 5 kHz. The echoes should get out of the way of the dry vocal.
5. Add ducking if the words blurDuck the echoes by 6 to 12 dB, or put a sidechained compressor on the return, so they bloom in the gaps.
6. Set the level in the full mixPull the send down until you'd only miss the delay if it were muted. Then mute it and check.

Delay vs reverb

Both add a sense of space, and they're built from the same idea. A delay gives you a few distinct repeats that you can count. A reverb is thousands of reflections arriving so close together that you hear them as one smooth wash, and an algorithmic reverb is built from many short delays feeding into each other.

In a mix the difference is distance. Reverb pushes a sound back into a room, while a delay can add just as much space and keep the sound at the front, because the dry part stays clean and the echoes happen after it. Plenty of vocals use both: a short reverb for the room and a filtered, synced delay for depth. The reverb vs delay comparison goes through when to pick which, and the reverb lesson has its own lab.

Common delay mistakes

Echoes louder than the part

Delays are easy to set too loud on headphones in a quiet room. Once the drums and bass come in, loud repeats blur the words and the timing. Set the level in the full mix, and if you can only just hear the echoes, that's often the right amount.

Full-range repeats

Every echo carries the low end of the original, and those layers add up. If the mix gets muddy as soon as the delay goes on, put a low cut of 200 to 300 Hz on it before you touch anything else.

A delay time that fights the groove

A free time that's a few milliseconds off a note value drifts against the beat and makes a tight part sound loose. Use sync for rhythmic parts. And if you change the tempo of the project, check any delays you set by hand in milliseconds, because they won't follow.

Tails that never stop

High feedback on a long note value can leave echoes hanging over the next section. Automate the feedback or the send so the long tails happen only where you want them, like the last word before a drop.

Training your ears to hear delay times

Hearing the difference between a 1/8 and a 1/4 is easy after a few minutes. Telling a dotted 1/8 from a 1/8 triplet takes practice, and it's worth having, because it lets you pick the right note value without trying every one. The quiz under the lab plays a loop at 120 BPM with a hidden note value, covers the controls and screens, and asks which one it is from three choices. Every round plays at the same loudness, so only the timing can give it away. There are eight rounds, and your best score is saved in your browser.

Tap your foot to the loop while you listen. Echoes that land with your foot are a 1/4, ones that land exactly halfway are a 1/8, and ones that seem to tumble between your taps are usually the dotted 1/8.

Questions people ask

Keep going