1
$\begingroup$

I want to calculate Fourier transform of the sampled signal in two ways. Let $$s(t) = \sum_{k = -\infty}^{\infty}\delta(t - kT)$$And $z(t) = x(t)s(t)$. So we have $$z(t) = \sum_{k = -\infty}^{\infty}x(kT)\delta(t - kT)$$Directly computing Fourier transform leads to $$\mathcal{F}(z(t)) = \sum_{k = -\infty}^{\infty}x(kT)e^{-j\omega kT} $$ On the other hand, if we use $\mathcal{F}(x(t)s(t)) = \frac{1}{2\pi}X(j\omega)\star S(j\omega)$ we have $$\mathcal{F}(z(t)) = \frac{1}{T}\sum_{k = -\infty}^{\infty}X(j(\omega - k\frac{2\pi}{T}))$$ How these are equivalent to each other mathematically? I mean how we can prove $$\frac{1}{T}\sum_{k = -\infty}^{\infty}X(j(\omega - k\frac{2\pi}{T})) = \sum_{k = -\infty}^{\infty}x(kT)e^{-j\omega kT}$$ And is there any intuition for this connection in terms of other subjects like sampling theorem or DTFT?

$\endgroup$

1 Answer 1

3
$\begingroup$

The equality

$$\frac{1}{T}\sum_{k = -\infty}^{\infty}X(j(\omega - k\frac{2\pi}{T})) = \sum_{k = -\infty}^{\infty}x(kT)e^{-j\omega kT}\tag{1}$$

is an instance of Poisson's sum formula. The term on the right-hand side of $(1)$ is just the Fourier series representation of the periodic function on the left-hand side of $(1)$.

The samples $x(kT)$ of the time domain signal are basically the Fourier series coefficients of the periodic spectrum of the sampled signal.

The significance of $(1)$ is that it shows that the discrete-time Fourier transform (DTFT) $X_d(e^{j\Omega})$ of the sequence $x_d[k]=x(kT)$ equals a periodized version of the Fourier transform of the corresponding continuous-time signal $x(t)$:

$$X_d(e^{j\Omega})=\sum_{k=-\infty}^{\infty}x_d[k]e^{-jk\Omega}=\frac{1}{T}\sum_{k=-\infty}^{\infty}X\left(\frac{j(\Omega-2\pi k)}{T}\right),\qquad\Omega=\omega T\tag{2}$$

Clearly, if $X(j\omega)$ is band-limited with a maximum frequency $\omega_c<\pi/T$, then the shifted spectra in the sum on the right-hand side of $(2)$ won't overlap, i.e., there is no aliasing and the signal $x(t)$ can be reconstructed without error from its samples $x(kT)$. That means that the basic form of the sampling theorem is implicit in Eq. $(2)$.

$\endgroup$
2
  • $\begingroup$ Thanks a lot. Is there any connection to DTFT in this case? $\endgroup$
    – S.H.W
    Jul 5, 2020 at 14:30
  • $\begingroup$ Of course there is: the DTFT of the discrete signal $x_d[n]=x(nT)$ is equal to the expression in Eq. (1). $\endgroup$
    – Matt L.
    Jul 5, 2020 at 14:59

Your Answer

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

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