Given a discrete-time (DT) sequence $g[n]$, I want to represent it as a continuous-time (CT) signal. I can do this by representing this sequence as a weighted sum of Dirac delta impulses. Would it make a difference if I pass the DT signal through a DT filter first and then represent it as a weighted sum of Dirac impulses or pass the CT signal through a CT filter. The two cases are as follows:

Case 1: DT sequence converted first to CT signal and passed through CT filter.

The DT signal can be represented as a CT signal as: $$g(t)=\sum_kg[k]\delta(t-kT)\tag{1}$$ If $g(t)$ in $(1)$ is the input to a continuous-time LTI system with impulse response $h(t)$, the output is given by $$y(t)=\int h(\tau)g(t-\tau)d\tau = \int h(\tau)\left(\sum_kg[k]\delta(t-\tau-kT)\right)d\tau$$ or $$y(t) =\sum_kg[k]\int h(\tau)\delta(t-\tau-kT) d\tau$$ or $$y(t) = \sum_kg[k]h(t-kT)\tag{2}$$

Case 2: DT sequence passed through DT filter and then converted first to CT signal.

The DT signal is passed through a DT filter $h[n] = h(nT)$ to obtain a DT signal $z[n]$: $$z[n]=\sum_kg[k]h[n-k]\tag{3}$$ This can be represented as a CT signal as $$z(t)=\sum_kz[k]\delta(t-kT)\tag{4}$$

Is $z(t)$ same as $y(t)$? Can I express one in terms of the other? Any help wuld be greatly appreciated? -ryan

  • $\begingroup$ Your equation (2) is not accurate. You have to use a convolution integral, not a sum. $\endgroup$ – Jazzmaniac Oct 10 '16 at 9:21
  • $\begingroup$ @Jazzmaniac Thanks forthe response. If I use the convolution integral, wouldn't it still evaluate to a sum as g(t) is expressed as a sum? $\endgroup$ – r2d2 Oct 10 '16 at 9:39
  • $\begingroup$ @Jazzmaniac I edited the question to explain equation (2). $\endgroup$ – r2d2 Oct 10 '16 at 9:59
  • $\begingroup$ The two cases can never be the same because in the first case you end up with a weighted sum of CT impulse responses, whereas in the second case you still have a discrete-time signal, which becomes a weighted sum of Dirac impulses, but which remains zero between the sample points. $\endgroup$ – Matt L. Oct 10 '16 at 10:46
  • $\begingroup$ @MattL.: Ok, I see the difference now. Yes, they can't be the same. But if you see my derivation in my answer, why it "seems" to hold true. What is incorrect there? $\endgroup$ – r2d2 Oct 10 '16 at 11:14

I believe (2) and (4) are same. One can write $z(t)$ as

$$z(t) = \sum_k z[k]\delta(t-kT) = \sum_k \left(\sum_l g[l]h[k-l]\right)\delta(t-kT)$$

But $h[k-l] = h((k-l)T)$. Thus, $$z(t) = \sum_l g[l] \left(\sum_k h((k-l)T) \delta(t-kT) \right) = \sum_l g[l] \left(\sum_k h(kT-lT) \delta(t-kT) \right) $$ or $$z(t) = \sum_l g[l] h(t-lT)$$ which, by change of variables, is same as (2).


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.