1. Home
  2. Learn
  3. Synth basics

SOUND DESIGN

How to Make a Bell Sound on a Synth (the FM Way)

Gelato Audio

A convincing synth bell takes two sine waves and FM synthesis: tune the modulator to an in-between ratio like 3.5 or 1.4 times the carrier, give the volume an instant attack and a long decay with no sustain, and let the FM amount fade faster than the volume. The strike comes out bright and clangy, then settles into a clean ring. Add reverb and play it from middle C up.

Why FM synthesis sounds like a bell

Most instruments make overtones at whole-number multiples of the note. A string playing 200 Hz also rings at 400, 600 and 800 Hz, and your ear fuses that stack into one clear pitch. Struck metal breaks the pattern. A tuned church bell has a hum an octave below its note, a tierce a minor third above it, a quint a fifth above and a nominal an octave above, plus quieter partials higher up. That minor third is a big part of why a bell sounds faintly sad.

The other half of a bell is time. The strike is bright and crowded. The high partials die first, and a softer tone is left ringing.

FM handles both with two sine waves. The modulator shakes the carrier’s pitch hundreds of times a second, and that shaking creates new partials, called sidebands, at the carrier frequency plus and minus whole multiples of the modulator frequency. When the modulator is a whole-number multiple of the carrier, the sidebands land on the harmonic series and you get brass and reed tones. When the ratio falls between whole numbers, the sidebands land in the gaps, and the ear hears metal. The FM amount (also called the index) sets how many sidebands you hear, so an envelope on it can turn a bright strike into a pure tail. For carriers, modulators and algorithms in more depth, read FM synthesis explained simply.

John Chowning, who developed FM at Stanford, gave a bell recipe in his 1973 paper for the Journal of the Audio Engineering Society: a 200 Hz carrier, a 280 Hz modulator (a 1:1.4 ratio), an index that starts at 10 and falls to 0, and one 15-second exponential decay shaping both the loudness and the index. At full strength the lowest partials land at 80, 200, 360, 480, 640, 760, 920 and 1,040 Hz, with more above. Of those eight, only 200 sits on the harmonic series of a 200 Hz note. As the index falls, the sidebands thin out until the 200 Hz carrier rings on its own. Yamaha licensed the idea and built the DX7 on it in 1983.

FM bell settings, step by step

These settings work on any synth with two FM operators or an FM input, from a DX7 clone to the simpler FM synth apps for iPhone. The carrier is the oscillator you hear. The modulator is the one that bends it, usually kept out of the output.

  1. Start from an init patch. Set both oscillators to sine waves. Open the filter all the way or switch it off, since FM will do the brightening.
  2. Set the ratio. Put the modulator at 3.5 times the carrier for a tubular bell, or 1.4 times for Chowning’s heavier church bell. If your synth tunes oscillators in semitones and cents, 3.5 is +22 semitones and -31 cents, and 1.4 is +6 semitones and -17 cents.
  3. Shape the volume. On the carrier’s amp envelope (ADSR), set attack to 0 to 2 ms, decay to 4 to 6 seconds, sustain to 0 and release to 3 seconds or more, so the bell keeps ringing after you lift your finger.
  4. Shape the FM. Give the FM amount its own envelope: 0 ms attack, zero sustain, and a decay of 0.5 to 1.5 seconds, well under the volume’s. This is the step that makes it a bell: the strike clangs and the tail is close to a pure sine. If your synth has only one envelope, send it to the FM amount as well. That is Chowning’s method, one curve for both, and it suits a slow, heavy church bell.
  5. Set the peak amount. Raise the FM amount until the strike clangs, then back off a notch. Too much and the attack turns to fizz. If your synth can send velocity to FM amount, do it, so harder hits come out brighter, the way a mallet works on metal.
  6. Add space. A reverb with a 2 to 4 second decay at about 25 percent mix puts the bell in a room. An eighth-note echo at low feedback suits chimes.
  7. Play it in the right range. Most bells sit best from middle C (C4) up, and a church bell can go down to about G3. Much lower, the partials crowd together into a growl, so turn the FM amount down as you go lower.

In Gelato Synth, set both Shape sliders to sine and Osc B’s Ratio to x3.5, raise Amount on the FM page, turn up Envelope to FM for the bright attack and purer tail, then pull Sustain to zero on the Env page. Envelope to FM ties the FM to that same Env curve, which is Chowning’s one-envelope approach. The Sunny Bell preset is a classic FM bell worth taking apart.

In Vital, put sine waves in oscillators 1 and 2, choose FM <- Osc 2 for oscillator 1, and set oscillator 2’s Transpose and Tune to the semitone and cent figures above. In Serum, FM is one of the oscillator’s warp modes, and the same ratio rules apply.

Tubular bells, glass and music box variations

Every one of these is the same two sines with different numbers. Start from the patch above and change these settings, then trim by ear:

SoundRatioFM decayVolume decay
Church bell1:1.43 to 6 s8 to 15 s
Tubular bell1:3.50.5 to 1.5 s4 to 6 s
Glass or celesta1:70.1 to 0.3 s1.5 to 3 s
Music box1:530 to 80 ms0.8 to 1.5 s
Vibraphone1:40.2 to 0.4 s2 to 4 s

Church bell. Play it around G3 to G4; Chowning’s 200 Hz carrier sits just above G3. The long FM decay lets the clang fade slowly, and a reverb of 4 seconds or more puts it in a tower.

Tubular bell. Play from C4 to C6. For a slow shimmer in the tail, layer a second copy of the patch with its carrier detuned by 3 cents. At middle C the two copies beat about once every two seconds.

Glass or celesta. A 1:7 ratio is a whole number, so the tone stays in tune, and the far-off sidebands add a clean, high ping on top. Keep the FM amount low and play from C5 to C7.

Music box. The 30 to 80 ms FM decay is the plink. Shorten it until each note has a tiny tick at the front and then goes pure, keep the amount low, and play from C5 up.

Vibraphone. Add a volume LFO at 4 to 6 Hz to copy the motor-driven tremolo of a real vibraphone.

For any mallet sound, a 10 to 20 ms burst of white noise at the start adds the click of the hit. Gelato Synth has a Burst slider on its Noise page for exactly this.

Making a bell sound without FM

FM is the shortest route, but any synth that can make partials off the harmonic series can ring.

Ring modulation. A ring modulator multiplies two oscillators and outputs their sum and difference frequencies in place of the originals. Sines at 200 and 290 Hz come out as 90 and 490 Hz, two partials that don’t line up as harmonics. With triangle or square waves, every pair of harmonics makes its own sum and difference, and the result is a denser, gong-like clang. Put the bell envelope from step 3 on it. Vital has this inside the oscillator as RM <- Osc.

Stacked sines. This takes five sine oscillators, so it suits an additive synth, a modular or a few layered patches. Tune them to a real bell’s main partials: the hum an octave below the note (-12 semitones), the note itself, a minor third up (+3), a fifth up (+7) and an octave up (+12). Give each higher partial a shorter decay than the one below it, and detune each by a few cents so the stack doesn’t sound like a minor chord on an organ. You get a rounder, gentler bell than FM gives.

Two oscillators on an analog-style synth. A single saw, square or wavetable repeats the same shape every cycle, so its partials sit on the harmonic series by design, and a filter can only take partials away, never move them. Two oscillators get around that. Mix two square waves a tritone apart (+6 semitones) under the bell envelope, and since their harmonic series never line up, you get a hard, clanky metal tone. Wavetable synths such as Vital and Serum can make bells the same way, with FM or ring mod between oscillators. A resonant filter hit with a fast envelope gives a woody mallet or marimba tone instead, which is closer to a pluck; how to make a pluck synth covers that recipe.

Fixing a synth bell that sounds harsh or thin

  • Harsh, fizzy strike. The FM amount is too high or its decay is too long. Cut the amount by a quarter or halve the FM decay.
  • Sounds like an organ. Sustain is above zero on the volume envelope. Set it to 0 so the note has to fade.
  • Sounds like a plain sine beep. The FM amount is too low, or its envelope is so short you never hear the clang. Raise the peak amount first.
  • Grainy top notes. Some digital FM synths alias when a high note meets a big FM amount. Scale the amount down across the keyboard, or map it to velocity and play the top octave softly.
  • Cuts off when you let go. Release is too short. Set it close to the decay time.
  • Chords sound like a pileup. Each note brings its own off-series partials, so a four-note bell chord turns to mush fast. Play single notes, octaves or fifths, or drop the FM amount for chord parts and save the full clang for melodies.

Questions

How do you make a bell sound on a synth?

Run two sine waves in FM, with the modulator tuned to an in-between multiple of the carrier such as 3.5 or 1.4. Give the volume an instant attack, zero sustain and several seconds of decay, and let the FM amount die away sooner, so each note starts bright and rings out pure.

What FM ratio makes a bell sound?

Start with 1:3.5 for a tubular bell or 1:1.4 for a heavy church bell, the ratio John Chowning used in his 1973 FM paper. A ratio that falls between whole numbers puts the overtones off the harmonic series, and the ear hears metal. Whole-number ratios such as 1:7 give cleaner glass and celesta tones.

Can you make a bell sound without FM?

Yes. Ring modulation, a stack of sine waves tuned to a real bell's partials, or two square waves a tritone apart will all ring. They work because their partials sit off the harmonic series. Give them the same envelope as an FM bell: instant attack, long decay, no sustain.

What is FM synthesis good for?

Sounds whose brightness changes fast within one note: bells, electric pianos, mallets, metallic hits and punchy basses. It is the method behind the Yamaha DX7, which came out in 1983 and was all over 1980s records.