Hi '56, There are a couple of things going on here. [1] The sinc function is the the fourier transform of a single rectangular pulse. You can approximate that with an fft, but only if you make the pulse width fairly narrow compared to the total width in the time domain. [2] You are multiplying by a cosine function, which affects the result in the frequency domain.
If you temporarily make f = 0 (so that the cosine function is a constant), go to a narrower rectangle
signal = cos(2*pi*f*t).*(A.*(heaviside(t-10e-9)-heaviside(t-12e-9)));
and plot abs(y) as before, you will see a sinc function, almost. (It's not exactly sinc and looks different anyway because the effect of abs puts all of its oscillations above the x axis). If you then change f back to its original value you will get a pair of waveforms in the frequency domain, each of which looks somewhat sinc-like.
