loading...
running...
The science
Design rule. Pendulum i (i = 0 ... n-1) is tuned to finish exactly Ni = Nb + i swings in the cycle time Tc. Its period is therefore Ti = Tc/Ni. For small swings a simple pendulum has T = 2π√(L/g), so the string length is
Li = g (Tc / 2π Ni)2 = g / ωi2, ωi = 2π Ni / Tc
All bobs start together. Each one then gains a fixed fraction of a swing on its neighbour every second, so the line through the bobs shows a travelling sine wave, then a standing wave, then apparent chaos, and after exactly Tc everything is back in step because every pendulum has completed a whole number of swings.
Coherence. With phase φi = atan2(-θ'/ωi, θ) the order parameter
R = | (1/n) Σi ejφi |
is 1 when all phases agree and close to 0 when they are spread evenly round the circle (the phasor wheel in the 2D page). It is the same quantity used for coupled oscillators.
Why the amplitude matters. The small-angle period is only the first term of an exact series. The true period of a pendulum released from angle A is
T(A) = 4 √(L/g) K(k), k = sin(A/2), K(k) = π/(2 AGM(1, √(1-k2))) ≈ (π/2)(1 + A2/16 + ...)
At 60 degrees the period is about 7 % longer. Because all pendulums are lengthened by the same factor the pattern would survive, but the extra swings fall on different integers, so a cycle no longer closes. The "Exact nonlinear, same lengths" setting integrates θ'' = -(g/L) sin θ with a fourth-order Runge-Kutta scheme (step at most 2.5 ms) and shows this loss of coherence. The "compensated lengths" setting uses Li = g (Ti / 4K)2, so that the exact period equals Ti and the cycle closes again at any amplitude.
Checks (run in the page). (1) RK4 period of a 90 degree pendulum against 4√(L/g) K(sin 45°); (2) after one 60 s cycle with 0.5 degree release all 15 pendulums are back at the release angle; (3) energy drift of the integrator at 60 degrees over 100 s; (4) coherence at t = Tc for compensated lengths at 90 degrees is 1, whereas the uncompensated set falls below 0.9.
Simplifications. Point-mass bobs on massless inextensible strings, no air drag, no pivot friction, no coupling through the rail, constant g = 9.81 m/s2. In a real rack a bob of radius r should be measured from the pivot to its centre of mass, and drag slowly shrinks the amplitude, which is why real demonstrations lose contrast after a few cycles. The 3D scene scales the strings to fit the rack (a 60 s cycle needs strings of 21 to 34 cm); the readouts give the true lengths.