Possible simple formula for ellipse circumference

In summary, the conversation discusses a potential method for calculating the circumference of an ellipse by cutting a cylinder with an angled plane and constructing a cylinder that fits the ellipse. The formula for this method is given as 4 * sqrt(a^2(pi^2/4 - 1) + b^2). However, upon further examination and testing with online calculators, it is discovered that this formula gives different results and may not be correct. Other formulas and methods are discussed, including one used by online calculators (c = pi / a * sqrt((a^2 + b^2) / 2)). The conversation also delves into the history of attempts to find a formula for calculating the circumference of an ellipse, with references to
  • #1
ktoz
171
12
TL;DR Summary
ellipse circumference = 4 * sqrt(minor^2(pi^2/4 - 1) + major^2)
While fighting a CAD program, today, I might have stumbled on a potential way to easily calculate the circumference of an ellipse. I checked my method against a half a dozen online ellipse calculators and while my formula gives different results, I can't see where I'm making any logical errors. Perhaps someone can take a look and see if they can spot the error(s).

Basically the method starts with the realization that if you cut a cylinder with an angled plane, the resulting face is an ellipse. Given that, its possible to construct a cylinder that naturally fits any given ellipse. From there, the length of one quadrant of the ellipse is just a matter of getting the length of the hypotenuse of a triangle wrapped around the cylinder and multiplying by 4.

Given
a = minor radius
b = major radius

The wrapped quadrant triangle can be computed like this

Use the minor radius as the radius of the wrapping cylinder
triangle base length, q = pi * a / 2

Use the major radius as the hypotenuse of a triangle that starts at the center of the cylinder base
and intersects the cylinder face above the base
quad triangle height, h = sqrt(b^2 - a^2)

Compute the hypotenuse of the wrapped quad triangle
triangle hypotenuse, c = sqrt(q^2 + h^2)

Multiply by 4 to get the circumference
ellipse circumference = 4 * c

Wrap it all together, simplify and you get the following formula

ellipse circumference = 4 * sqrt(a^2(pi^2/4 - 1) + b^2)

I checked out several examples in a CAD program and, at least to the resolution of the program, the ellipses constructed by cutting derived cylinders in the above way, all matched perfectly.

As I said at the start, this formula gives a different answer than online calculators that use more advanced methods, but the concept is so simple, I don't see where I might be going wrong.

Does the logic hold up? Or does someone see my error(s)?
 
Last edited:
Mathematics news on Phys.org
  • #2
If you unwrap the cylinder then the 1/4 ellipse is not a triangle because the "hypothenuse" is not a straight line. You can see this e.g. from the connection point at the semi-major axis where the two lines connect at a 90 degree angle.
 
  • #3
mfb said:
If you unwrap the cylinder then the 1/4 ellipse is not a triangle because the "hypothenuse" is not a straight line. You can see this e.g. from the connection point at the semi-major axis where the two lines connect at a 90 degree angle.
Hmmm, right you are. Seemed so promising … at 3:00 am. Thanks for pointing out the flaw.

Back to designing what I was originally working on in the CAD program. 🙂

Guess a moderator can mark this as closed.
 
Last edited:
  • Like
Likes hutchphd
  • #4
Dont feel bad, just enjoy the Aha moment and realize that all great ideas have likely been thought of before.
 
  • Like
Likes Keith_McClary
  • #5
jedishrfu said:
Dont feel bad, just enjoy the Aha moment and realize that all great ideas have likely been thought of before.

I'm not making the mistake of thinking I'm right, but I revisited the problem, today, by plugging a bunch of (a, b) pairs into two online ellipse calculators and managed to reverse engineer their algorithm. They seem to be using c = pi / a * sqrt((a^2 + b^2) / 2) (a is major axis, b is minor). Every example I checked was very close to their result. (9 decimal places)

Armed with this formula, I checked other online calculators and found they were giving values roughly 2x larger than the first ones. Logically, these ones seems better, because I've read that the problem isn't easy and c = pi / a * sqrt((a^2 + b^2) / 2) is pretty easy. I wonder how sites using it get away without being swamped with corrections.
 
  • #6
You are in good company being fascinated by this problem. Apparently also Euler, and before him Fagnano, explored these formulas,( given apparently by integrals of reciprocals of square roots of quartic polynomials). Unsuccessful at rationalizing these integrals by means of substitutions analogous to those which worked for trig functions, (computing arc length of circles), Fagnano discovered that, although one could not readily compute the length of a given arc of an ellipse, nonetheless given one point (and a base point), one could compute the further point at which the arclength would be doubled! From that Euler obtained an "addition formula" for elliptic integrals, and started the theory of abelian integrals. The story is pursued in vol. 1 of a magnificent trilogy by Carl L. Siegel, Topics in Complex Analysis, which I found quite challenging but quite rewarding too.

(Looking back, Fagnano actually worked with the arc of a lemniscate, but apparently a closely related problem.)
 
Last edited:
  • Like
Likes uart, olivermsun and ktoz
  • #7

Similar threads

Replies
2
Views
2K
Replies
1
Views
2K
Replies
3
Views
3K
Replies
4
Views
6K
Replies
2
Views
3K
Replies
10
Views
3K
Back
Top