In the next few lectures, we'll address a very important question:
How many samples are required to accurately reconstruct $f(t)$ when we convert it to the frequency domain, then back to the spatial domain?
We'll first need some background first on "impulse trains".
The Dirac delta is
$\delta(t) = \begin{cases} \infty \;\:\quad \text{if}\ t = 0 \\ 0 \qquad \text{otherwise} \end{cases}$
but has the property that
$\displaystyle \int_{-\infty}^\infty \delta(x)\ dx = 1$
and is drawn as
The Dirac delta can be used to sample a function at the origin, inside an integral:
The Dirac delta can be shifted to sample a function at any point, inside an integral:
"Trains" of impulses are used to sample a continuous function at a regular interval, $T$.
$\displaystyle s_T(t) = \sum_{i=-\infty}^\infty \delta( t - i T )$
Samples can be taken at each impulse:
(note that this is a product of two functions, not the integral of the product)