there is a problem that I should specify a real-valued random process $Y[n]$ such that the autocorrelation function $R_Y[k]$ satisfies $$R_Y[0]=3+u,\ R_Y[1]=R_Y[-1]=-2+u,\ \text{and}\ R_Y[k]=u, |k|>1.$$

First subproblem is to find a feasible set of u with |u|>0. I found region $-\frac{1}{2}\leq u$.

The main problem is to specify a random process $Y[n]$.

Considering that $R_Y[k]=u$ for $|k|>1$, I thought $Y[n]$ is likely to have a DC value. Besides, the autocorrelation function has unique values for k=0,1, I set $$Y[n]=a\,w[n]+b\,w[n-1]+c$$ for constant $a,b,c$ and $w[n]$ is white Gaussian noise with zero mean and variance 1, and has autocorrelation $R_w[k]=\delta[k]$. But the final results becomes $a,b$ are imaginary numbers, which cause complex-valued $Y[n]$.

Is there any way I can get a random process? The problems say a) specify a random process with the given autocorrelation function (that is, specify the stochastic generation mechanism for the process). b) Is there a unique random process with the given autocorrelation function? If the answer is no, identify possible sources of difference between the various random processes that all have this given autocorrelation function.

  • $\begingroup$ Homework? Does the book or your course notes have any mention of processes with autocorrelation $R[k] = u$? How about processes with any autocorrelation function that does not go to zero as $k \to \infty$? $\endgroup$
    – TimWescott
    Jan 14, 2022 at 16:16
  • $\begingroup$ @TimWescott This is from previous exam, but unluckily I do not have a solution.. I've never seen non-diminishing autocorrelation function like this. I do not know what is the referred textbook, sorry. $\endgroup$
    – Junho
    Jan 14, 2022 at 21:50

1 Answer 1


If I read your problem correctly, $R_Y[k]$ is given by


The power spectrum is the DTFT of $R_Y[k]$, which is given by

$$S_Y(\omega)=3-4\cos(\omega)+2\pi u\,\delta(\omega),\qquad -\pi\le\omega<\pi\tag{2}$$

From $(2)$ it is obvious that $S_Y(\omega)\ge 0$ cannot be satisfied, regardless of the value of $u$. Hence, $R_Y[k]$ is not a valid autocorrelation function, and, consequently, there is no solution to your problem.

  • $\begingroup$ Thank you for the answer. Yes, your statement totally make sense. I might need to answer for the problem that such process cannot exist. $\endgroup$
    – Junho
    Jan 14, 2022 at 21:46

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.