I try to express the relation between:

  • the delay obtained by maximizing the cross-correlation between two signals $x$ and $y$ noted $\tau_{x,y}$ and
  • the phase velocity noted $v_p(\omega)$

In the case where:

  • The signals $x$ and $y$ are spaced by a constant distance $\Delta$
  • The signal $y$ is a shifted version - at speed $v_p(\omega)$ - of the signal $x$, i.e.$Y(\omega)=X(\omega)e^{-j\omega\Delta/v_p(\omega)}$ ; with $X(\omega)$ and $Y(\omega)$ the Fourier transforms of $x$ and $y$

With these assumptions, the delay that corresponds to the maximum correlation between $x$ and $y$ is expressed as:

$$\tau_{x,y} = \text{max}_t \int_{-\infty}^{\infty} |X(\omega)|^2e^{j\omega(t+\Delta/v_p(\omega))}d\omega $$

I can't manage to go any further. Is it possible to go further and express $\tau_{x,y}$ as a function of $v_p(\omega)$? Perhaps by adding the assumption of white signal, i.e. $|X(\omega)|=1$?


1 Answer 1


For periodic power signals, the relationship $$\tau_{x,y} = \max_t(R_{xy})$$ is not unequivocal. Thus, restricting the considered signals to finite energy signals is required. But even then, it will probably be impossible to find an elegant functional relationship between $\tau_{x,y}$ and $v_p$, as it strongly depends on the signals involved. The trivial case without dispersion is simple and you might find classes of signals with a mathematically elegant relationship even with dispersion but in general it will not get simpler than the equation you derived. I would love to be proved wrong, as this is quite interesting.

  • $\begingroup$ Thank you for your help. I have the feeling that for finite energy signals, the delay obtained by correlation can be considered as a weighted mean of the delays associated to the phase velocity, i.e. $$\tau_{x,y} \approx \frac{\int_{-\infty}^{\infty} |X(\omega)|^2 \frac{\Delta}{v_p(\omega)}}{\int_{-\infty}^{\infty} |X(\omega)|^2} $$ But I don't know how to show it... $\endgroup$
    – User327201
    Oct 20, 2022 at 9:41
  • $\begingroup$ I have a similar gut feeling. Showing this, if possible, won't be easy. An entry point could be to start with a monochromatic wave packet, arguing that any signal can be expressed as a sum of sines, and then add degrees of freedom successively. $\endgroup$
    – Max
    Oct 20, 2022 at 10:07

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.