0
$\begingroup$

As I see in this slides, Fourier series representation for discrete-time signal $s[n]$ with period $N$ is $\sum_{k = 0}^{N-1} c_k e^{j2\pi k n / N}$

According to Wiki, Fourier series representation for continuous-time signal $s(x)$ with period $P$ is $s_{\infty}(x) = \sum_{n = -\infty}^\infty c_n e^{j 2\pi n x / P}$

My question is that why in discrete-time Fourier series representation, they only sum up from $0$ to $N - 1$ instead of from $-\infty$ to $\infty$ ? Is the slides above gives a wrong formula ?

P.s: in my opinion, the more terms we have, the more accurate our Fourier series is

$\endgroup$
1
  • $\begingroup$ Hi! not a duplicate but the answer is highly related with this one... $\endgroup$
    – Fat32
    Commented Dec 18, 2018 at 10:50

1 Answer 1

0
$\begingroup$

Because the discrete-time basis functions $e^{j2\pi kn/N}$ are not only $N$-periodic in $n$ (as is the signal $s[n]$), but also in $k$. So adding terms for $k\ge N$ does not add anything new, simply because

$$e^{j2\pi (k+N)n/N}=e^{j2\pi kn/N}e^{j2\pi n}=e^{j2\pi kn/N}$$

$\endgroup$
5
  • $\begingroup$ According to your answer, in case of continuous-time signals, if the period $P$ is an integer, we only have to sum up from $0$ to $P$, is it right ? Because $e^{j 2\pi n (x + P) / P} = e^{j 2\pi n x / P} e^{j 2\pi n}$ $\endgroup$ Commented Dec 18, 2018 at 10:55
  • $\begingroup$ @HOANGGIANG: My answer is about the discrete-time basis functions, so I'm not sure I understand your question. Integrate what? $\endgroup$
    – Matt L.
    Commented Dec 18, 2018 at 10:56
  • $\begingroup$ I've edited my comment $\endgroup$ Commented Dec 18, 2018 at 10:57
  • $\begingroup$ @HOANGGIANG: In order to compute the Fourier coefficients of a continuous-time $P$-periodic function, we have to integrate the function (multiplied by the basis functions) over an interval of length $P$, no matter whether $P$ is an integer or not. $\endgroup$
    – Matt L.
    Commented Dec 18, 2018 at 10:57
  • $\begingroup$ @HOANGGIANG: No, because $e^{j2\pi (n+P)x/P}\neq e^{j2\pi nx/P}$ since $x$ is real-valued. $\endgroup$
    – Matt L.
    Commented Dec 18, 2018 at 11:01

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.