1
$\begingroup$

The signal $$x(t)\;\;\;\;0\leq t\leq 0.2s $$

We know that the CTFT of $x(t)=0$ when $|w|>2*\pi*10^4$

We sample $x(t)$ in sample space of $$T=25\mu s$$ or $$F_s=1/T=40000Hz$$and we get a series with length of 8000.

$$x[n]=x(nT)\;\;\;\;0\leq n \leq 7999 $$

We calculate the DFT of $x[n]$ in size 8000 also.

What is the relation between the CTFT to the DFT

$$X^F(\omega) \rightarrow X_N^d[k]$$

The answer is:

$$ \begin{equation} X_N^d[k]=\begin{cases} \dfrac{N}{T}X^F(10\pi k), & 0\leq k\leq 3999\\ \\ \dfrac{N}{T}X^F(10\pi (k-8000)), & 4000\leq k\leq 7999 \end{cases} \end{equation} $$

$\endgroup$
13
  • $\begingroup$ Do you also submitted the answer? I am bit confused about your objective. $\endgroup$
    – jomegaA
    Feb 3, 2020 at 18:23
  • $\begingroup$ I have the answer but I don’t know how to solve the problem. $\endgroup$
    – ori
    Feb 3, 2020 at 18:42
  • $\begingroup$ The relationship between the spectrum of $X(f)$ of discrete time signal and the spectrum $X_a(F)$ of the analog signal is $$X(f)=F_s\sum_{k=-\infty}^{\infty}X_a\left[(f-k)F_s\right]$$ On right-hand side is the periodic repetition of the scaled spectrum, scaled with period $F_s$ $\endgroup$
    – jomegaA
    Feb 3, 2020 at 19:00
  • $\begingroup$ Proof is in Digital Signal Processing - Proakis $\endgroup$
    – jomegaA
    Feb 3, 2020 at 19:07
  • $\begingroup$ @jomegaA I don't understand how to get to the answer from what you wrote here. $\endgroup$
    – ori
    Feb 4, 2020 at 6:25

1 Answer 1

1
$\begingroup$

Let's give it a shot.

We know that during sampling, we obtained a discretized version, $x(nT_s)$ of a continuous time signal $x(t)$ sampled every $T_s$ seconds. I will denote their corresponding Fourier Transforms as $X_{DT}(e^{j\omega})$ and $X_{CT}(j\Omega)$, respectively, according to the majority of the bibliography. Indices denote discrete time (DT) and continuous-time (CT) Fourier Transforms.

The Discrete time Fourier Transform (DTFT) of the sampled signal is given by $$X_{DT}(e^{j\omega}) = \frac{1}{T_s}\sum_{l=-\infty}^{+\infty}X_{CT}\Big(j\Big(\frac{\omega}{T_s} - l\frac{2\pi l}{T_s}\Big)\Big)$$

Assuming that there is no aliasing (like in your case), over one period this turns into $$X_{DT}(e^{j\omega}) = \frac{1}{T_s}X_{CT}\Big(j\frac{\omega}{T_s}\Big)$$

Due to the $2\pi$-periodic nature of the DTFT and the fact that we're dealing with real-valued signals, we can rewrite this as $$X_{DT}(e^{j\omega}) = \left\{\begin{array}{ll} \frac{1}{T_s}X_{CT}\Big(j\frac{\omega}{T_s}\Big), & 0 \leq \omega < \pi \\ \frac{1}{T_s}X_{CT}\Big(j\frac{\omega-2\pi}{T_s}\Big), & \pi \leq \omega < 2\pi \\ \end{array}\right.$$

The Discrete Fourier Transform (DFT) is an $N-$point uniformly sampled version of the DTFT, so $$X_{DFT}[k] = X_{DT}(e^{j\omega})\Big|_{\omega = 2\pi k/N}$$ for $0 \leq k\leq N-1$.

Combining the last two equations we have $$X_{DFT}[k] = \left\{\begin{array}{ll} \frac{1}{T_s}X_{CT}\Big(j\frac{2\pi k}{NT_s}\Big), & 0 \leq k < N/2 \\ \frac{1}{T_s}X_{CT}\Big(j\frac{2\pi (k-N)}{NT_s}\Big), & N/2 \leq k \leq N-1 \\ \end{array}\right.$$

Except for the constant $N$ in your results - that I have failed to extract - I think what I wrote will help you.

$\endgroup$
1
  • $\begingroup$ The constant N is for approximately result. thanks! $\endgroup$
    – ori
    Feb 5, 2020 at 13:46

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.