For a = 3 and b = 2 the perimeter is four times ∫₀^(π/2) √(9sin²t + 4cos²t) dt, and the first link evaluates that integral as 15.8654395893 — twelve places, which is all the panel shows; carried to fourteen digits the same perimeter is 15.865439589291. There is no closed form for it in elementary functions, which is what makes this the archetype of an elliptic integral. Ramanujan's first approximation, π(3(a + b) − √((3a + b)(a + 3b))), becomes π(15 − √99) here and prints 15.865437575564 on the second link, short of the true value by about 2⋅10⁻⁶. The third link is his second formula, π(a + b)(1 + 3h/(10 + √(4 − 3h))) with h = ((a − b)/(a + b))² = 0.04, typed as pi*5*(1+0.12/(10+sqrt(3.88))); it prints 15.865439589251, an error of four units in the eleventh decimal, from one line of arithmetic.
The trap the fourth link exists to spring is the one everyone meets first: 2π√((a² + b²)/2) = 16.019042244414, the circumference of the circle whose radius is the root-mean-square of the two semi-axes. It looks principled and its error is 0.15360, tens of thousands of times Ramanujan's first error, because the approximation's deviation is of order (a − b)² relative to the mean radius — good for near-circles, bad for anything eccentric, and 3 to 2 is quite eccentric. Ramanujan posed both formulas as questions without proof; their error terms were analysed much later, and the second is much the better of the two: its error shrinks far faster as the ellipse approaches a circle, which is why it is the formula people quote. The exact value is a hypergeometric series in h, and no finite combination of rational operations and roots reaches it.
Answer
15.8654395893
How to read it
The perimeter from the integral, at the panel's twelve places. The second link is Ramanujan's first formula, the third his second, the fourth the rms approximation whose error is 0.15360.
Source
Srinivasa Ramanujan's two ellipse-perimeter approximations, published 1914; he called the method empirical and left no derivation, and the error terms were analysed much later; the perimeter integral of the ellipse is the historical reason the class of integrals is named 'elliptic' in Legendre's treatment of 1811
Work it in the calculator
These links open the right calculator with this problem's numbers already typed in. The line above is what the screen should show — check it yourself.