We can solve the wave equation u_tt = u_xx on the interval [0,1]
with the Dirichlet boundary conditions u(0,t) = u(1,t) = 0 and the
initial conditions u(x,0) = 1 and u_t(x,0) = 0 in two different ways.
Using the method of reflections, we find the solution u(x,t)
to be piecewise constant in different regions of its domain bounded
by characteristic lines (notice that the solution has a singularity
along these characteristics).
Using the general series solution, if we know that
1 = Sum_(n odd) 4/(n pi) sin(n pi x)
  
  
(picture)
then
u(x,t) = Sum_(n odd) 4/(n pi) sin(n pi x) cos(n pi t)
These two solutions are the same (away from the singularities)!
For example, at t = 1/4, we found from the method of reflections
that u(x, 1/4) = 0 if 0 < x < 1/4 and if 3/4 < x < 1, but
u(x, 1/4) = 1 if 1/4 < x < 3/4. The series solution at t =1/4 is
u(x,1/4) = Sum_(n odd) 4/(n pi) sin(n pi x) cos(n pi / 4)
Here are some pictures
of the sum of only the first N terms of this series (for various N).
Notice that the more terms we use,
the closer the sum becomes to the right answer.
Let phi(x) be the function that equals 1 on (1/4,3/4) and is 0
on the rest of the interval [0,1] (basically, phi(x) is
the function u(x,1/4) above.)
Here's another
picture
of the function first 20 terms in Fourier series
that converges to phi(x) (at all the points where it's continuous).
Notice that at x=1/4 and at x=3/4, the series is actually converging to
the number 1/2 (the average of 1 and 0)! Of course, if we plot the
sine series
on the whole real line, it is approximating the odd periodic extension of
phi(x): picture.
The convergence is not uniform! Notice for
points close to x=1/4, it takes a very long time for the series to converge
to 0 or 1. Here's a table that
illustrates the convergence at 1/4 and at some points near 1/4.