Additive Synthesis Explained: Building Sounds From Sine Waves

What is additive synthesis?

Additive synthesis creates a timbre by adding sine waves together, one per partial, and shaping the level and pitch of each one over time. It works in the opposite direction from subtractive synthesis, which starts with a bright waveform and carves harmonics away with a filter.

The idea comes straight from Fourier's theorem, which states that any sound can be constructed from elementary sinusoids^[1]. Additive synthesis was one of the first computer-music methods, and some of the first high-quality synthetic instrument tones were built with it by Jean-Claude Risset at Bell Telephone Laboratories in the 1960s^[2].

For a producer the practical meaning is simple. Every waveform on your synth is really a recipe of sine waves. A sawtooth, for example, contains every harmonic, and the nth harmonic sits at 1/n the level of the fundamental^[3]. Change the recipe and you change the sound, with no filter involved. If you want the bigger picture of how additive fits next to FM, wavetable and granular methods, the synthesis types overview maps them side by side.

Partials, harmonics and overtones: what is the difference?

A partial is any single sine component in a sound; a harmonic is a partial that sits at an exact whole-number multiple of the fundamental. All harmonics are partials, but not every partial is a harmonic. That distinction decides whether an additive patch sounds like an organ or like a gong.

• Fundamental The lowest partial in a pitched tone. On an A2 note it sits at 110 Hz, and it usually sets the pitch you hear.
• Harmonic A partial at 2x, 3x, 4x the fundamental and so on. For that same 110 Hz note, harmonics land at 220, 330, 440 and 550 Hz.
• Overtone Any partial above the fundamental. The numbering is off by one: the 2nd harmonic is the 1st overtone, which trips up a lot of forum threads.
• Inharmonic partial A partial that falls between the whole-number multiples, such as 2.4x or 5.4x. A few of these make a tone metallic; many of them make it noisy.

Classic synth waveforms are just fixed spectra. Knowing them lets you rebuild any of them by hand in an additive editor, which is the fastest way to train your ear for what each partial contributes. The oscillator waveforms guide covers how these shapes sound in a mix.

Harmonic vs inharmonic partials: why bells sound like bells

Harmonic partials fuse into one clear pitch, while inharmonic partials sit off the whole-number grid and produce metallic, bell, gong and clang tones.

A church bell is the textbook case. Its five lowest partials are named the hum, prime, tierce, quint and nominal. In true-harmonic tuning the hum sits two octaves below the nominal, the prime one octave below, the tierce a minor third above the prime and the quint a perfect fifth above the prime^[6]. That minor-third partial is not part of any harmonic series built on the prime, and it gives the bell its familiar color.

Apple's Alchemy documentation makes the same point from the synth side: sounds with strong inharmonic content, such as bells, change dramatically when you force their partials back into a perfect harmonic series^[7].

Stretching the harmonic series on purpose

You rarely need to type ratios by hand. Harmor's Prism control shifts partials away from their original relationship to the fundamental; a mild setting sounds detuned and an extreme setting sounds metallic^[8]. Treat a control like that as a harmonic-to-metal dial. Small amounts add width and shimmer to pads, large amounts turn a saw into a struck plate.

Which additive synths can you use today?

The three additive engines most producers already own or can try are Alchemy in Logic Pro, Harmor for FL Studio and the Harmonic engine in Arturia Pigments. Each one hides the raw partial count behind friendlier controls, so you can start from a preset and open the partial view when you need it.

The oldest additive instrument in most studios is not a plugin at all. The Hammond tonewheel organ lets the player mix up to nine near-sine tones with drawbars, each set from zero to eight^[5]. Any organ plugin with drawbars is an additive synth with a very focused interface.

How do you build an organ tone with additive synthesis?

Map the nine Hammond drawbars onto partials, set their levels like a drawbar registration, then add a short key click and some movement. On a full-console Hammond the drawbars correspond to harmonics 1, 3, 2, 4, 6, 8, 10, 12 and 16 of the 16' pitch^[5].

1. Start from a single sine
   Initialize the patch so only partial 1 plays. Set the amp envelope to about 0-3 ms attack, full sustain and 60-120 ms release so notes stop like a real organ.
2. Enter the registration
   Turn on partials 1, 2 and 3 at equal level for the classic 888000000 sound, then add partials 4, 6 and 8 at lower levels if the part needs to cut through a busy mix.
3. Add key click
   Give one high partial, around the 16th harmonic, its own envelope with a 5-10 ms decay and zero sustain. That short tick on each note reads as tonewheel click.
4. Add movement
   Put a chorus or rotary-style effect after the synth and a touch of saturation. Pure sines in perfect ratios sound static, so this stage matters more than the partial levels.
5. Voice it for the genre
   For trap and drill, drop it an octave and play short stabs; the trap organ stab guide covers the rhythm and processing side.

How do you design a bell with additive synthesis?

Place a handful of partials at bell-like ratios, give every partial its own decay with the high ones dying first, and add a burst of short inharmonic partials for the strike. The ratios below are converted from the true-harmonic bell intervals described earlier^[6], with the prime as 1.0.

1. Set the core ratios
   Hum 0.5, prime 1.0, tierce about 1.19, quint about 1.5 and nominal 2.0. If your synth only accepts harmonic numbers, use its detune, shift or Prism-style control to move partials off the grid.
2. Shape each decay
   Start with the hum decaying over 3-6 seconds, the prime and tierce over 2-4 seconds and the nominal over 1-2 seconds. In real plucked and struck tones the higher partials generally fade faster, which is why the sound darkens as it rings^[3].
3. Add the strike
   Switch on four to eight upper partials at odd ratios such as 2.7, 3.4 and 4.2, with 20-80 ms decays. They create the clang at the start and then get out of the way.
4. Detune for beating
   Offset one or two partials by 1-3 Hz. The slow beating between them is what makes a synthetic bell feel physical instead of sterile.
5. Place it in space
   Send it to a long plate or hall reverb and cut the reverb below 200-300 Hz so the hum does not smear the low end.

The same approach scales down to trap bells and music-box leads: fewer partials, shorter decays and brighter ratios. For melodic patterns and processing in that style, read how to make trap bells and plucks.

Why is additive synthesis CPU-heavy, and when should you use it?

Additive synthesis is expensive because every partial is its own oscillator with its own envelopes, and a bright low note can need hundreds of them. Julius O. Smith at Stanford's CCRMA notes that the method has been held back from wider use by its computational cost^[1].

Pitch range matters too. Apple explains that high notes hit the limit of audibility after a small number of harmonics, while low bass notes can carry hundreds, and that a flute needs fewer partials than a cello or violin^[7]. The practical fix in Alchemy is to set Num Partials to the lowest multiple of four that still sounds right.

Editing is the other cost. On Kawai's K5, setting the levels of 64 harmonics one by one could take 20 minutes, which is why that synth introduced group editing of odd, even and octave harmonics^[4]. Modern engines go further with odd/even balance, tilt, images and resynthesis.

• Pick additive for Organs, bells, glassy pads with partials moving independently, and resynthesis of a recorded note you want to stretch, morph or de-vibrato.
• Pick subtractive or wavetable for Basses and leads built on filter sweeps. You will get there faster, and wavetable synthesis gives you evolving spectra at a fraction of the CPU.
• Pick FM for Bright inharmonic tones with very few operators. Two or three sine operators in FM synthesis can produce a bell that would take dozens of additive partials.