What are unison and detune?

Unison and detune in plain English

On most synths, each note you play comes from one oscillator, the part that repeats a wave shape at the pitch of the key. Turn the unison voice count up to 7 and every key you press starts seven of them at once. If all seven played exactly the same pitch, you'd mostly hear one louder note. Detune is what stops that from happening: it tunes some of the voices a little flat and others a little sharp, so they're never quite together.

A string section works the same way. Twelve violinists playing one line are never exactly in tune or in time with each other, and that spread is a big part of why a section sounds fuller than one violin turned up. Unison gets a similar thickness from a synth by running lots of oscillators side by side.

Most unison sections have a third control, called width or spread, that places the voices across the stereo field. With every voice in the centre you get a thick sound that sits in the middle of the mix. Spread them out and the same note fills the space between the speakers.

Reading the lab

The lab plays its own parts: held chords, a lead riff or a bassline, all in A minor at 120 BPM. Every note is built from as many oscillators as the Voices knob says, each tuned and panned by the Detune and Width knobs.

The big screen shows one held note rendered with your settings, with the left speaker on top and the right one below. The dashed outline is the same note from a single oscillator, which stays flat. With unison on, the pink shape swells and dips as the voices drift in and out of step with each other. Under it, the voice map puts a dot for every oscillator, higher for sharp and lower for flat, and left or right for where it sits between the speakers.

The spectrum shows the note's first 16 harmonics on a straight hertz scale. One oscillator draws each harmonic as a thin spike, and detuned voices smear each spike into a cluster, which gets wider the higher up you look. The coach on the right explains what you're hearing using numbers measured from the rendered note. Press One voice at any point to hear the same part from a single oscillator at the same loudness.

What is detune?

Detune is measured in cents. A cent is a hundredth of a semitone, so 100 cents takes you from A to B flat and 1,200 cents is a whole octave. Cents are a proportion of the pitch, which means the same number of cents is a different number of hertz on every note.

Take the A below middle C, which is 220 Hz. Ten cents sharp of that is 221.27 Hz, so two voices 10 cents apart differ by 1.27 Hz. Two octaves lower, at 55 Hz, the same 10 cents is a difference of only 0.32 Hz. To work it out for any note, the gap in hertz is the frequency × (2^{cents ÷ 1200} − 1).

That gap in hertz is what you hear. Two tones a few hertz apart keep moving in and out of step: when their peaks line up they add together, and half a cycle later they partly cancel. The level rises and falls as many times a second as the tones are hertz apart, an effect called beating. It's what a guitarist listens for when tuning one string against another by ear, where the beats slow down as the strings get closer and stop when they match.

Push the detune further and the beating speeds up. With two voices at ±30 cents, they're 60 cents apart, more than half a semitone, and the wobble runs at nearly eight times a second. By then the chords in the lab have stopped sounding thick and started sounding out of tune.

Detune and the harmonics

A saw wave has harmonics at 2, 3, 4 and more times the note, and every harmonic of a detuned voice is out by the same number of cents. In hertz, that gap grows as you go up. With voices 40 cents apart on a 220 Hz note, the fundamentals are 5.1 Hz apart, but the 16th harmonics, up at 3.5 kHz, are 82 Hz apart. So the bottom of the sound moves slowly while the top fizzes, and on the lab's spectrum each harmonic spreads into a wider cluster the higher it sits.

Detune between two oscillators

Before synths had unison controls, people got the same effect with separate oscillators. The Minimoog has three, and tuning two or three of them a few cents apart is a common way to thicken a Moog bass or lead. Almost every synth still has a fine-tune knob on each oscillator, often labelled detune or fine, that does this for one pair. A unison section does it across many voices from one knob.

What does unison do?

The unison control sets how many voices play each note. Two voices give one regular wobble you can follow by ear. Add more and every pair of voices beats at its own rate: with seven voices spread evenly over ±20 cents, the closest neighbours are 6.7 cents apart and the outermost pair 40 cents apart, so at 220 Hz the pairs beat at six different rates, from 0.85 to 5.1 times a second.

Adding voices makes the level move further, if anything. The coach's swing figure measures how far the level of the held note moves, and in the lab two voices at ±5 cents swing by about 5.6 dB while seven at ±20 cents swing by about 10 dB. The change you hear is in the pattern: two voices pinch at one even rate you could clap along to, while seven move in uneven swells of different sizes, spread across every harmonic at once, and that's what you hear as a shimmer.

Every extra voice also adds level. Voices that start at unrelated points in their cycle add up in power, so seven voices come out about 8.5 dB louder than one. Many synths turn each voice down as you add more to make up for most of that, and the lab turns each one down by the square root of the voice count, so the level stays about the same whether you have 1 voice or 16.

Serum and Vital both allow up to 16 unison voices per oscillator. Each of those voices is a full oscillator, so the processing cost multiplies: a four-note chord with 16 voices is 64 oscillators running at once, and twice that if both oscillators have unison switched on.

Stereo width and spread

Width decides where the voices sit between the speakers. At 0% every voice is in the middle, so both speakers play exactly the same thing. At 100% the lab pans the outermost pair hard left and hard right and spaces the rest between them, alternating sides so each speaker gets a mix of flat and sharp voices. Synths differ in how they spread the voices, but most work along these lines.

Spreading them works because each speaker now plays a different group of voices, and those groups beat at different moments. The left and right channels stop matching, and two channels that don't match sound wide. The lab's correlation meter measures how alike they are: +1 is mono, and with seven voices at ±20 cents and full width the rendered note reads about 0.42. The stereo width lesson covers how width works and how to judge it.

Unison in mono

Many phone speakers, a lot of Bluetooth speakers and some club systems play in mono, which means the left and right channels get added together. A sound in the centre loses nothing when that happens. Wide unison does lose something, because the two channels hold different voices: added together, they combine only in power, the width disappears, and the sound drops in level against everything that was in the middle.

With seven voices at ±20 cents and 100% width, the lab measures the mono sum 1.5 dB lower than the stereo version. Even a decibel or two can push a part back in a mix, and the width you balanced on headphones is gone. The mono compatibility lesson has the full method for checking a mix.

Random or reset phase

Each voice is a repeating wave, and its phase is the point in the cycle where it starts when you press a key. Analogue oscillators run all the time, so each note catches them wherever they happen to be. Software synths usually let you choose between a random starting point for every voice on every note and a reset, where every voice starts from the same point.

With a reset, the voices begin each note in step. Every voice peaks at the same moment, so the note starts with a hard edge, and as the voices drift apart and back together you hear a sweep much like a flanger. In the lab, seven voices with the phase reset start 4.9 dB above the level they settle to. That can suit a pluck, where you want every note to hit the same way, but on held chords the sweep comes back at every chord change, so leave random phase on for pads and chords.

What is a supersaw?

A supersaw is a stack of detuned saw waves played as one sound. The name comes from the Super Saw oscillator on Roland's JP-8000, which played seven detuned saws from a single oscillator. It became the sound of trance leads in the late 1990s, and it's still the starting point for a lot of big EDM chords and leads. Any synth with unison can make one.

How to make a supersaw

1. Start with one sawOpen the filter all the way, set a short attack and a release of around 300 ms, and play a chord.
2. Set the unison to 7 voicesSeven is the original count and a good balance between thickness and processing power.
3. Raise the detune until it shimmersSomewhere around ±15 to 25 cents. If the chord starts to sound sour, back it off.
4. Open the width70 to 100% for a sound that fills the stereo field.
5. Add something in the middleA second oscillator with a single, undetuned saw, or a sine an octave down, keeps the pitch clear and gives the sound a solid centre that survives in mono.
6. Tame the topA low-pass filter between about 8 and 12 kHz takes the edge off the fizz, and a high-pass around 150 Hz keeps it out of the bass's way.
7. Check it in monoIf it drops away or sounds hollow, narrow the width a little or turn up the centre layer.

Unison on bass

Low notes are where unison causes the most trouble. Spread wide, it puts different low end in each speaker, and when those channels are summed the voices beat against each other right where the kick and the bass need to hold still. The detune behaves differently down there too: at 55 Hz, voices 20 cents apart are only 0.64 Hz apart, so a held bass note swells and fades over about a second and a half and can sound like someone is riding its fader.

For a bass that sounds thick and still holds up in mono, keep the voices few and the width at or near zero. The other way is to split the sound: a single voice or a sine for the low end, in the centre, with a detuned layer on top that's high-passed somewhere between 100 and 200 Hz. That's a common way to handle a Reese bass, the detuned-saw bass named after Kevin Saunderson's alias Reese and his 1988 track “Just Want Another Chance”. There's more on making the bottom of a track sit together in kick and bass.

Starting settings for common sounds

These are places to start. Each synth spreads its voices a little differently, and some, Vital among them, show unison detune as a percentage of their own range instead of in cents, so trust the sound over the number.

Unison, chorus and doubling

A chorus effect gets a similar thickness from a single sound. It mixes the sound with copies that pass through short delays, somewhere between about 10 and 30 ms, whose time is swept up and down by an LFO; sweeping the delay bends each copy slightly sharp and then flat. Unison makes its copies at the source, with separate oscillators at fixed offsets, so it can be far thicker, though it only exists inside a synth. Chorus works on anything, including a recorded guitar or vocal.

Double tracking, where a singer or guitarist records the same part twice and both takes play together, is the acoustic version of the same idea. The takes drift a little apart in pitch and timing, and those small differences make the part sound bigger.

Common unison mistakes

Judging it by level

On many synths adding voices also makes the patch louder, and louder sounds better at first. Compare unison against one voice at the same loudness, which is what the One voice button in the lab does, before you decide it's helping.

Detuning chords until they're sour

Wide detune can sound exciting on one note and wrong on a chord, where every note's cluster rubs against the others. If the chords start to sound out of tune, the detune is too wide for the part.

Wide unison on the bass

It sounds huge on headphones and loses its weight on a phone or a mono system. Keep the low end narrow and put the width in a layer above it.

Unison on every part

If every synth in the track is a wide, detuned stack, they all compete for the sides of the mix, and nothing sounds wide because there's nothing narrow to compare it with. Give the width to one or two parts and keep the rest focused.

Train your ears

The quiz under the lab plays the chords with hidden settings, and each round asks about the voice count, the detune or the width, with the levels matched so you can only go by ear. Headphones make the width rounds easier, and your best score is saved in your browser.

Questions people ask

Keep going