I want to find the pre envelope of

$$x(t) = {\Pi}_{a}(t)\cos(2{\pi}f_{0}t)$$ where

I found the Fourier transform to be
$$ \begin{align} X(f) & \triangleq \mathfrak{F}[{\Pi}_{a}(t)\cos(2{\pi}f_{0}t)] \\ & = a\text{ sinc}(2a(f-f_{0})) + a\text{ sinc}(2a(f+f_{0})) \\ \end{align} $$
So $$X_{+}(f) = 2u(f)X(f)$$
But I don't know how to continue from here. Is there an easier way to find the pre envelope with another method?

  • $\begingroup$ uhm, you sure that $$ X_+(f) = 2u(f)X(f) \quad ?$$ looks like some of $a \ \operatorname{sinc}(2 a (f-f_0))$ could leak into where $f < 0$. i think you might have assume that $|f| \ll f_0$ to say that. $\endgroup$ – robert bristow-johnson Jan 18 '16 at 3:53
  • $\begingroup$ s-mat-pcs.oulu.fi/~ssa/ESignals/sig4_5.htm $\endgroup$ – temp8jfhfhf Jan 18 '16 at 4:19
  • $\begingroup$ i know about the Hilbert transform and the pre-envelope (what i like to call the "complex envelope)". anyway, if $\frac1a \ll f_0$, then you can say that $$ X_+(f) \ \approx \ 2a \ \operatorname{sinc}(2a(f-f_0)) \quad . $$ then what is $x_+(t)$ ? which is, i think the pre-envelope. $\endgroup$ – robert bristow-johnson Jan 18 '16 at 6:38
  • $\begingroup$ also, i asked this at your other question: can you be clear about the meaning of $$ \Pi_a(t) $$ ? how wide is this rectangular function? is it as wide as $a$ or is it $2a$? $\endgroup$ – robert bristow-johnson Jan 18 '16 at 6:43
  • $\begingroup$ It is from $-a$ till $a$. The reverse Fourier of $X_{+}(f)$ seems difficult since it also has $u(f)$, so I will need to make a convolution later, and I would like to avoid that. $x_{+}(t)$ is the pre-envelope $\endgroup$ – temp8jfhfhf Jan 18 '16 at 6:54

The pre-envelope is also called analytic signal. Its Fourier transform is indeed given by the expression in your question:


where $X(f)$ is the Fourier transform of the original signal, and $u(f)$ is the unit step function. Obviously, $X_+(f)$ has only positive frequency components. The analytic signal $x_+(t)$ with its Fourier transform given by $(1)$ is necessarily a complex-valued signal:


where $\mathcal{H}\{\cdot\}$ denotes the Hilbert transform. Note that the analytic signal is not the same as the complex envelope. For a band pass signal $x(t)$, the complex envelope is a low pass signal, whereas the analytic signal is a band pass signal.

From $(2)$, in order to compute the analytic signal you need to compute the Hilbert transform of $x(t)$:


For the given signal you get

$$\hat{x}(t)=\frac{1}{\pi}\int_{-a}^a\frac{\cos(2\pi f_0\tau)}{t-\tau}d\tau\tag{4}$$

The result of $(4)$ can be written in terms of the cosine integral $\text{Ci}(x)$ and the sine integral $\text{Si}(x)$:

$$\hat{x}(t)=\frac{1}{\pi}\left\{\cos(2\pi f_0t)\left[\text{Ci}(2\pi f_0(t+a))-\text{Ci}(2\pi f_0(t-a))\right]+\\\sin(2\pi f_0)\left[\text{Si}(2\pi f_0(t+a))-\text{Si}(2\pi f_0(t-a))\right]\right\}\tag{5}$$

I don't think that anybody expected you to come up with that awful expression. Anyway, for large values of $f_0$, the first term in $(5)$ becomes very small, and the second term converges to $\sin(2\pi f_0)\Pi_a(t)$. So for large $f_0$ you get the expected result

$$x_+(t)\approx x(t)+j\sin(2\pi f_0)\Pi_a(t)=\Pi_a(t)e^{j2\pi f_0 t}\tag{6}$$

Obviously, the respresentation

$$x(t)=\text{Re}\left\{\Pi_a(t)e^{j2\pi f_0 t}\right\}\tag{7}$$

is always valid, but the complex-valued signal $\Pi_a(t)e^{j2\pi f_0 t}$ is no analytic signal, it's just a good approximation of the analytic signal for large values of $f_0$.

Note that for $x(t)=m(t)\cos(2\pi f_0t)$ with $m(t)$ a band-limited function, i.e. $M(f)=0$ for $|f|>B$, the complex-valued signal $m(t)e^{j2\pi f_0 t}$ is an analytic signal, as long as $f_0>B$. The problem with the function given in your question is that $\Pi_a(t)$ is not band-limited.

| improve this answer | |

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.