For an random signal $X$ (does not matter if it is correlated or uncorrelated), given lag $τ$. If white noise is added to this signal, then can the summation of them be a correlated signal? Also, even if these two are individually uncorrelated. Could somehow be related mutually?


1 Answer 1


Can white noise be correlated to a random singal

Something can be correlated to something else but still be white. For example, let $X(t)$ be a white noise process, and let $Y(t) = 2X(t)$, then clearly $Y$ is still a white noise process, but just as clearly very correlated to $X$.

If white noise is added to this signal, then can the summation of them be a correlated signal?

Say $X_1(t)$ is now some arbitrary process that's not 0, $Y_1(t)$ is white noise, stochastically independent from and hence uncorrelated with $X_1$, and let $Z_1(t)=X_1(t)+Y_1(t)$. Then, clearly $Z$ is correlated with $X$. The same is trivial to prove: Test for correlation:

\begin{align} E\{X_1(t)Z_1(t)\} &= E\{X_1(t)(X_1(t)+Y_1(t))\}\\ &=E\{X^2_1(t)\}+E\{X_1(t)Y_1(t)\}&\text{linearity of expectation}\\ &= E\{X_1^2(t)\}+E\{X_1(t)\}\cdot E\{Y_1(t)\} &\text{$X_1,Y_1$ independent}\\ &=E\{X_1^2(t)\}+E\{X_1(t)\}\cdot 0 &\text{$Y_1$ is white, $\rightarrow$ zero-mean}\\ &=E\{X_1^2(t)\}\\ &\ne 0 \end{align}

Hence, the sum of a random variable and independent white noise is always correlated to the random variable.

Whether the sum is white: that depends fully on $X$.

I'm not sure you're intending to ask this, just in case:

Can I add two white processes $V,W$ and get a non-white processes $U= V+W$?

Sure; let $V(t)$ be a non-zero power white noise process and $W(t)=V(t-1)$,

\begin{align} E\{U(t)U(t+\tau)\}&=E\{(V(t)+V(t-1))(V(t+\tau)+V(t-1+\tau)\}\\ &=E\left\{V(t)V(t+\tau)+V(t)V(t-1+\tau)+\right.\\ &\phantom{E\left\{V(\right.}\left.V(t-1)V(t+\tau)+V(t-1)V(t-1+\tau)\right\}\\ &\ne0 \text{ for } \tau\in\{-1,0,+1\}. \end{align}

That's actually a super common case in the signal processing: your $W$ is the superposition of the transmit signal with a delayed version of itself; the multipath channel. The fact that the signals in these paths correlate is the basis for the multipath channel needing a different treatment than the AWGN channel; the fact is also the foundation of diversity gain methods in MIMO, many if –not even most channel– equalization methods, passive radar, lightning location, necessary for ultrasound imaging to work…


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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