The short answer
A delay copies your voice and plays the copy a little after you. That is the entire machine. Three controls decide what the result gets called:
- How late the copy plays, the delay time. Under about 30 milliseconds you hear one thicker voice. Past about 100 you hear a separate repeat.
- Whether the copy repeats, the feedback. Feeding the output back in makes each copy get copied again.
- Whether the delay time keeps sliding back and forth while you talk, which is what turns a static effect into a moving one.
Everything below is those three. A phaser is the one relative that builds its notches a different way, and the last section says where it parts company.
How late the copy plays
At 25 ms, the copy arrives so soon after you that your ear refuses to hear it as a second event. It fuses the two into one thicker voice, and that is a double. At 90 ms, the copy is late enough to hear on its own but still lands inside the syllable you are saying, which is slapback, the sound on a lot of fifties records. At 380 ms, the copy arrives after you have finished the syllable, so you hear your own voice come back: an echo.
One knob walks you through all three. Nothing about the machine changed between them, only how long the copy waited.
Whether the copy repeats
Turn the feedback up and the copy is fed back into the delay, so it gets copied again, and that copy gets copied. On an echo at 380 ms that is the familiar result: your voice repeating, quieter each time, until it falls below what you can hear.
Take the same feedback down to a delay time of 8 ms and something else happens. Copies now arrive 125 times a second, because a thousand divided by eight is a hundred and twenty five, and anything repeating 125 times a second is a note at 125 Hz. Your voice stops sounding like a voice with echoes and starts sounding like one buzzing note. That is the robot voice, built from a delay rather than from the ring modulator people assume.
Whether the delay time slides
The third control makes the delay time a little longer, then a little shorter, over and over while you talk. That movement changes the pitch of the copy, and the reason is the same one that makes a siren drop in pitch as it passes you. While the delay is getting longer the copy is being stretched out, so it comes back lower. While it is getting shorter the copy is being squeezed, so it comes back higher.
Point that movement at each of the sounds above and you get the rest of the family:
- Flanger. Slide the delay time of the robot, with an even shorter delay, and the note it makes slides up and down. That sweep is the jet sound.
- Chorus. Slide the delay time of the double, and the copies drift slightly out of tune with you, the way two people singing the same line are never exactly together.
- Vibrato. Turn your dry voice off so only the copy is heard, then slide its delay time. With nothing to compare against, your whole voice moves up and down instead.
Where the phaser parts company
A phaser sits in the same family of sweeping effects and is built differently. It does not delay the signal; it passes it through allpass filters, which let every frequency through at full volume and change only its timing. Mixing that result against the dry signal cancels a few narrow slices, and sweeping those slices is the swirl.
That difference is audible once you know it. A flanger cancels at many evenly spaced points, because a delay repeats at even multiples, and it sounds metallic. A phaser cancels at a handful of uneven points, and it sounds smoother. Few uneven notches and a soft swirl mean phaser; many even notches and a jet whoosh mean flanger.
The same four nodes in Patchd
In Patchd the Delay, Chorus, Flanger and Vibrato are four separate nodes in the Studio voice-fx rack, so you reach for the one you want by name rather than building it. They are the same delay underneath, and the Delay node is where you can see the whole range for yourself: Time runs from 1 to 2000 ms, which covers the double, the slapback, the echo and the robot without leaving one control.
Feedback runs 0 to 100 percent and defaults to 35, which is the repeating echo most people picture. Drop Time to single digits with that feedback up and the same node is a robot. Nothing is hidden: you are turning the two knobs this page describes.
The takeaway
One delay line, three decisions. How late the copy plays gives you a double at 25 ms, a slapback at 90 and an echo at 380. Letting the copy repeat gives you the dying echo at long times, and a buzzing note at 8 ms where the repeats arrive 125 times a second. Letting the delay time slide gives you a flanger from the robot, a chorus from the double, and a vibrato when you mute the dry voice. The phaser is the cousin that makes its notches out of phase instead.
Patchd is a Windows audio mixer with a full effects rack on every channel, and the four delay nodes live in the Studio voice-fx rack. It is in open beta now, and if you clear the terms your invitation arrives straight away. The Studio tier is $5.99 a month, $39.99 a year, or $119.99 once for a lifetime license, and the real-time engine is identical on both tiers.