Let $W(t)$ be continuous time white noise, that is, a wide-sense stationary (WSS) zero-mean Gaussian process with autocorrelation function $R_W (\tau) = σ^2\delta(\tau)$. Calculate the auto correlation function of $X(t) = \int_0^t W(r) dr$.

Can anyone help me understand how they get these integration limits in the following step in the calculation?

enter image description here


By the very definition, we have

$$\delta(\tau) = \begin{cases} 0, & \tau \neq 0, \\ 1, & \tau = 0 \end{cases}.$$


$$\delta(r-q) = \begin{cases} 0, & r \neq q, \\ 1, & r=q \end{cases}.$$

Suppose that $s<t$, i.e. $\min\{s,t\}= s$. For any $q \in (s,t]$ and $r \in [0,s]$ we have


since $r-q<0$. Consequently,

$$\int_0^s \int_s^t \mathbb{E}(W_q W_r) \, dq \, dr = 0.$$

This implies the identity you want to prove.

The argumentation for the case $t \leq s$ works analogously; I leave it to you.

  • $\begingroup$ Yes, we are left with a 1 (times $\sigma^2$) in the inner integral when q = r = min(s,t). And we start at 0 and go along that line to min(s,t). Maybe whats confusing me is that when you look at the double integral i can imagine what is happening but if you look at the inner integral separately the picture Yves is painting isnt working, right? $\endgroup$ – JKnecht Dec 29 '15 at 15:16
  • $\begingroup$ @JKnecht why ask me about the picture Yves is painting? The identity you want to prove is really straight-forward using the considerations I explained in my answer. $\endgroup$ – saz Dec 29 '15 at 18:29
  • $\begingroup$ I tried to explain that i understand what you are saying but it doesnt become straight forward to me because i think about it in a similar way as Yves does. But maybe i am misunderstanding both of you. $\endgroup$ – JKnecht Dec 29 '15 at 20:15
  • $\begingroup$ Yeah, you are right. I read through your post a couple of more times and it really is straight forward and obvious. I dont know why i didnt see it the first couple of times. Maybe because this is my first math course in ten years... $\endgroup$ – JKnecht Dec 29 '15 at 20:45

The integration domain on the left is a rectangle $s\times t$ starting from the origin.

Because of the Dirac $\delta$, that on the right is a line segment along the main diagonal, starting from the origin and extending until the first of $s,t$ is met.

  • $\begingroup$ Thanks for your answer. Im still a little confused tho. Look at my reply to saz. $\endgroup$ – JKnecht Dec 29 '15 at 15:17
  • $\begingroup$ Dont bother...i got it now. $\endgroup$ – JKnecht Dec 29 '15 at 20:45

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.