005 / ORDER FROM CHAOS

One shared rhythm

Each metronome starts in its own rhythm. A freely moving platform connects them — no conductor, no hidden commands.

Live physics · Web Audio
Experiment running0.0 s
Synchronization15%
R1 · last 30 secondsPhases
↳

Can you stop them synchronizing?

Make the platform heavier or increase the frequency spread. The only link between rhythms is the reaction of their shared support. Synchronization is not guaranteed — that is what makes it interesting.

Theory behind the experimentHow metronomes trade motion through a shared support and when a common rhythm appearsPlain explanation, equations and how the model works

In plain words

A metronome is a pendulum with a mechanism that keeps nudging it so the swing doesn’t die out. Put several metronomes on a solid table and they have no mechanical way to adjust to each other. Put them on a shared board that can move, and a link appears.

A swinging weight gives the board a small push. The board shifts, and the pivots of all the other pendulums accelerate with it. These little pushes gradually change where the metronomes sit relative to each other in their swing cycle. With the right parameters the rhythms lock together. That’s synchronization.

A shared rhythm doesn’t have to mean every weight leans the same way at every moment. The motion can keep a constant offset, for example two groups swinging against each other. Phase tells you where in its cycle a metronome is right now. The same position with opposite velocities means different phases.

Synchronization doesn’t always happen. A board that’s too heavy barely responds to the pushes, and if the metronomes’ natural frequencies differ a lot, a common rhythm struggles to form. Friction, the energy drive and the starting state matter too. So a brief moment of near-perfect agreement doesn’t mean they’ve synchronized for good.

Try it yourself

  1. Compare the default setup with the “Heavy platform” preset. Watch not just whether the weights line up but whether it lasts.
  2. Move the “Mass / length spread” slider to the right. Different lengths give different natural frequencies. See when the agreement becomes unstable.
  3. Watch the synchronization percentage R1R_1 and, under advanced physics, R2R_2. If R1R_1 is low while R2R_2 is close to one, check whether the metronomes have split into two opposing groups.

Physics and equations

For small swings of a single ideal pendulum, the natural angular frequency is roughly

ω0i≈gli\omega_{0i}\approx\sqrt{\frac{g}{l_i}}

Here lil_i is the effective pendulum length and gg the gravitational acceleration. A longer pendulum swings more slowly. On a moving board the frequency comes out slightly higher, but this estimate is enough to see what a length spread does. The full model below doesn’t replace the sine of the angle with the angle.

Let XX be the board’s horizontal displacement and θi\theta_i the angle of pendulum ii from the downward vertical. Its equation is

mili2 θ¨i+milicos⁡θi X¨+miglisin⁡θi=Qi\begin{aligned} m_i l_i^2\,\ddot\theta_i&+m_i l_i\cos\theta_i\,\ddot X\\ &+m_i g l_i\sin\theta_i=Q_i \end{aligned}

mim_i is the swinging weight’s mass and QiQ_i the combined drive-and-friction torque in N·m. The first term is rotational inertia, the third is gravity, and the middle one ties the pendulum to the board’s acceleration. Through X¨\ddot X one metronome affects all the others.

The board has its own equation of motion:

AX¨+∑imili(cos⁡θi θ¨i−sin⁡θi θ˙i2)=−BX˙ \begin{aligned} A\ddot X&+\sum_i m_i l_i\bigl(\cos\theta_i\,\ddot\theta_i-\sin\theta_i\,\dot\theta_i^2\bigr)\\ &=-B\dot X \end{aligned}
A=M+Nmbody+∑imiA=M+N m_{\text{body}}+\sum_i m_i

Here MM is the board’s mass, NN the number of metronomes, mbodym_{\text{body}} the mass of one housing (fixed to the board and moving with it), and BB the board’s drag coefficient in N·s/m. The housings add mass that has to be accelerated along with the board, but they don’t count toward the swinging weights. There’s no spring pulling the board back: it rolls freely and only the drag BX˙B\dot X slows it down.

The energy drive is a smooth model in the spirit of the van der Pol oscillator:

Qi=[μ(1−θi2α2)−b]θ˙iQ_i=\left[\mu\left(1-\frac{\theta_i^2}{\alpha^2}\right)-b\right]\dot\theta_i

μ\mu sets the drive strength and bb the pendulum’s own friction, both in N·m·s (radians are dimensionless). α\alpha is an angle scale, not a target amplitude. With suitable values the mechanism pumps up small swings and damps overly large ones. It’s an approximate stand-in for the escapement; individual teeth aren’t modelled.

Reading the phase measures

Each pendulum’s phase is estimated from its angle and velocity, and two agreement measures are computed from the phases:

φi=atan2⁡ ⁣(−θ˙ig/li, θi)\varphi_i=\operatorname{atan2}\!\left(-\frac{\dot\theta_i}{\sqrt{g/l_i}},\ \theta_i\right)
Rk=∣1N∑j=1Ne ikφj∣R_k=\left|\frac1N\sum_{j=1}^{N}e^{\,ik\varphi_j}\right|

The atan2⁡\operatorname{atan2} function finds a vector’s angle taking its quadrant into account. For a nearly harmonic swing this estimate tracks the stage of the cycle well. R1R_1 and R2R_2 range from 0 to 1. R1R_1 is close to one when the phases match. R2R_2 is close to one both then and when the metronomes split into two groups half a cycle apart. So R2R_2 alone can’t tell overall agreement from two opposing groups.

The synchronization percentage on screen is R1R_1 times 100. It describes the phases right now; it isn’t a probability of synchronizing or a share of “correct” metronomes. Whether the common rhythm is stable only shows in how the R1R_1 history behaves. On the phase circle the dots are individual phases and the line is their mean vector.

How the model works

The metronomes are linked only through the shared board’s horizontal motion. The sound just voices the motion and doesn’t carry any force between metronomes. Ticks play when a pendulum passes through vertical, without snapping to a grid. Each weight is an effective pendulum; the real balance and escapement aren’t modelled separately. The rod and slider in the picture are a familiar way to draw such a pendulum.

All metronomes start with the same amplitude (about 26.5°) and random phases. The board’s initial velocity is chosen so the total momentum is zero. Shuffling the phases changes only the starting phases; the metronomes’ masses and lengths stay the same.

The equations are solved with the fourth-order Runge-Kutta method at a fixed step of 1/240 s. The spread setting changes masses and lengths. The drive and friction add and remove energy, so in the normal mode mechanical energy isn’t supposed to be conserved. The “No drive or damping” mode switches off the drive, the pendulum friction and the board drag, and only there is the energy error meaningful. The board’s displacement isn’t exaggerated in the picture, and the rollers turn to match its motion.

Synchronization can’t be promised for every setting. The fun of the experiment is seeing what helps agreement and what gets in its way.

Sources and further reading

Goldsztein, Nadeau, Strogatz: Synchronization of clocks and metronomes ↗: a study of in-phase and anti-phase synchronization of coupled pendulums. The paper’s model isn’t identical to ours; it’s further reading on the topic.