I have a question regarding an alternating pulse train signal, $p(t)$, and a system which looks like this:

Alternating Pulse Train


Immediately, I'm supposed to be able to extrapolate that $$ P_t = P_1(t)-P_1(t-\Delta) $$

  • How do you obtain this?
  • How would you get the Fourier transform of this pulse train from there?
  • $\begingroup$ The first point you can obtain just by inspection (I'm assuming that $p_1(t)$ is the time-domain representation of the positive-going pulse train. As a hint on the second, take advantage of the linear property of the Fourier transform. $\endgroup$
    – Jason R
    Dec 8 '13 at 13:35
  • $\begingroup$ This looks a lot like a homework problem, and if so, please add the homework or self-study tag. Hint: $\displaystyle P(t) = \sum_{n=-\infty}^\infty (-1)^n\delta(t-n\Delta)$ where $\delta(t)$ is the unit impulse (or delta function if that is what your instructor calls it, shame on him). $\endgroup$ Dec 8 '13 at 15:26
  • $\begingroup$ Still a bit confused - if p<sub>1</sub>(t) is the positive going pulse train, I'm just subtracting a negative pulse train delayed by Δ? $\endgroup$ Dec 8 '13 at 19:03
  • $\begingroup$ If $\displaystyle Q(t) = \sum_{n=-\infty}^\infty \delta(t-2n\Delta)$ is a periodic impulse train of period $2\Delta$, then $Q(t) - Q(t-\Delta)$ is a periodic impulse train of period $2\Delta$ in which the impulses alternate in polarity, that is, it is the $P(t)$ shown in the figure you have included. I have no idea what that $p_t = p_1(t) - p_1(t-\Delta)$ in your question means. $\endgroup$ Dec 8 '13 at 21:43

Since this is a periodic signal, we have to first represent the signal $x(t)$ in terms of its Fourier Series and then compute its Fourier transform

$$\text{Representing } x(t) \text{ in terms of its Fourier series }$$

Let us assume $\Delta=T$

enter image description here


$$x(t)=\sum_{n=-\infty}^\infty x_ne^{j2\pi \frac{n}{T_0}t}=\sum_{n=-\infty}^\infty x_ne^{j\pi nt}$$

$$x_n=\frac{1}{T_0}\int_{\alpha}^{\alpha +T_0} x(t)e^{-j2\pi \frac{n}{T_0}t}dt=\frac{1}{2T}\int_{0T^-}^{2T^-} x(t)e^{-j2\pi \frac{n}{2T}t}dt$$

$$\frac{1}{2T}\int_{0T^-}^{T^-} \delta (t)e^{-j\pi \frac{n}{T}t}dt+\frac{1}{2T}\int_{1T^-}^{2T^-} -\delta (t-T)e^{-j\pi \frac{n}{T}t}dt$$

$$=\frac{1}{2T}(1)-\frac{1}{2T}e^{-j\pi \frac{n}{T}T}=\frac{1}{2T}(1-e^{-j\pi n})=\frac{1}{2T}(1-(-1)^n)$$ $$x(t)=\sum_{n=-\infty}^\infty \frac{1}{2T}(1-(-1)^n)e^{j\pi\frac{n}{T}t} \tag 1$$

$$\text{Determing Fourier Transform of } x(t) \text{ from its Fourier series representation}$$

Let $n=2k+1$, then (1) becomes $$=> x(t)=\sum_{k=-\infty}^\infty \frac{1}{2T}(2)e^{j\frac{(2k+1)}{T}\pi t}$$

$$X(jw)=\int_{-\infty}^\infty x(t)e^{-jwt}dt$$ $$=\int_{-\infty}^\infty \sum_{k=-\infty}^\infty \frac{1}{2T}(2)e^{j\frac{(2k+1)}{T}\pi t}e^{-jwt}dt$$

$$\frac{1}{T}\sum_{k=-\infty}^\infty \int_{-\infty}^\infty [1.e^{j\frac{(2k+1)}{T}\pi t}]e^{-jwt}dt$$

$$\text{ Since } F[1]=2\pi\delta (w) \text{ and } F[e^{j\Omega_0t}]=2\pi \delta(\omega-\Omega_0) \text{ Therefore}$$

$$X(jw)=\frac{1}{T}\sum_{k=-\infty}^\infty 2\pi \delta(w-\frac{(2k+1)}{T}\pi)$$

$$=\frac{2\pi}{T}\sum_{k=-\infty}^\infty \delta(w-\frac{(2k+1)}{T}\pi) \tag 2$$

$$X(f)=\frac{2\pi}{T}\sum_{k=-\infty}^\infty \delta(2\pi f-\frac{2(2k+1)}{2T}\pi)$$

$$\text{Since } \delta(ax)=\frac{1}{|a|}\delta(x)$$

$$X(f)=\frac{2\pi}{|2\pi|T}\sum_{k=-\infty}^\infty \delta( f-\frac{(2k+1)}{2T})$$

$$=\frac{1}{T}\sum_{k=-\infty}^\infty \delta( f-\frac{(2k+1)}{2T})$$ Let $2k+1=m$ $$X(f)=\frac{1}{2T}\sum_{m=-\infty}^\infty (1-(-1)^m)\delta(f-\frac{m}{2T}) \tag 3$$

  • $\begingroup$ I think your variable of integration is wrong - should it not be performed over $t$, rather than $n$? I also found this really hard to follow due to your use of notation - I would expect to see $c_n$ or $X[k]$ to describe the Fourier coefficients. Using $x_n$ suggests some discrete time series to me. Do you agree? $\endgroup$
    – Speedy
    Feb 23 '17 at 10:27
  • $\begingroup$ Yes true. I have corrected it now. I have used the conventions used in 'Communication Systems Engineering' by Proakis and salehi(chapter 2,Frequency Domain Analysis of Signals and Systems). Please note that $c_n$ is just the same as $x_n$, just the varibles are changed. $\endgroup$
    – Soumee
    Feb 23 '17 at 13:35
  • $\begingroup$ Please note that while you are using $$x(t)=\sum_{-\infty}^\infty c_ne^{jnw_0t};c_n=\frac{1}{T_0}\int_{\alpha}^{\alpha +T_0} x(t)e^{-jnw_0t}dt;\text{ where }w_0=\frac{2\pi}{T_0};T_0\text{ being the time peroiod of the given periodic signal}$$ I am using $$x(t)=\sum_{n=-\infty}^\infty x_ne^{j2\pi \frac{n}{T_0}t};x_n=\frac{1}{T_0}\int_{\alpha}^{\alpha +T_0} x(t)e^{-j2\pi \frac{n}{T_0}t}dt;w_0=\frac{2\pi}{T_0}$$ Both are the same. Just use $c_n$ in place of $x_n$ and you are good to go. $\endgroup$
    – Soumee
    Feb 23 '17 at 13:36
  • 2
    $\begingroup$ @Speedy I have tried to resolve your doubts. Check the comments above. Should you have any other doubts, do let me know. $\endgroup$
    – Soumee
    Feb 23 '17 at 14:07

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.