Contradiction in definition of AWGN?

80 Views Asked by At

I have read the definitions of Additive White Gaussian Noises as,

$S_x(f) = \frac{N_0}{2}$

This gives me (after an inverse Fourier Transform),

$R_x(\tau) = \frac{N_0 \delta(\tau)}{2}$

Now, $R_x(0) = E(X^2(t)) = \sigma^2$ (the variance of the random variable, a finite quantity)

However, on putting $\tau = 0$ in $R_x(\tau)$, I obtain an undefined value. What is the flaw in this argument?