Autocorrelation function is $$R_{xx}(\tau)=\frac{20}{1+2\tau^2}$$ So at $\tau=0$$$R_{xx}(0)=20=E[X(t)X(t)]=E[X^2(t)]$$ The variance is $$\mathrm{Var}[X(t)]=E[X^2(t)]-E^2[X(t)]=20-E^2[X(t)]$$ As $X(t)$ is WSS (wide sense stationary) the mean is a constant. Is there any way to find its numerical value?

  • $\begingroup$ In your second formula $E[X^2(t)]=20$, whereas in your third formula $E[X^2(t)]=6$. $\endgroup$ – Matt L. Sep 28 '16 at 7:23
  • $\begingroup$ Edited that error. $\endgroup$ – Don Sep 28 '16 at 15:52
  • $\begingroup$ Got it. As If $\lim_{\tau \to \infty} R_{xx}(\tau)=C$ Then $C=E^2[X(t)]$ where $C$ is a constant $\endgroup$ – Don Sep 28 '16 at 16:07

The limit $\lim_{\tau\to\infty} R_x(\tau)$, if it exists, equals $E^2[X(t)]$ and so $E[X(t)]=0$ in this case.

More generally, the mean of a WSS process is nonzero only if the power spectral density has an impulse at the origin. This can be applied to periodic autocorrelation functions such as $\cos(\omega_0t)$ pointed out in @MattL's comment. If the Fourier series for a periodic autocorrelation function has a nonzero DC term, the mean is nonzero.

| improve this answer | |
  • 3
    $\begingroup$ This is of course true for this example, but it might be useful to add that the mean cannot always be determined in this way from the autocorrelation function. E.g., the autocorrelcation function of the random process $X(t)=\cos(\omega_0t+\phi)$ (with $\phi$ a random phase uniformly distributed in $[0,2\pi)$) equals $R_X(\tau)=\frac12\cos(\omega_0\tau)$. Consequently, the limit doesn't exist. The mean, however, is $E[X(t)]=0$. $\endgroup$ – Matt L. Sep 28 '16 at 18:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.