Check back soon!
Video Answer
Get the answer to your homework problem.
Try Numerade free for 7 days
Input your name and email to request the answer
Like
Report
Request a Custom Video Solution
We will assign your question to a Numerade educator to answer.
Answer Delivery Time
Exercise C1.13: DTFT Computations using Two-Sided SequencesIn this exercise we consider the DTFT of two-sided sequences (including aut ${ }^0$ covariance sequences), and in doing so illustrate some basic properties of autocos variance sequences.(a) We first consider how to use the DTFT to determine $\phi(\omega)$ from $r(k)$ on computer. We are given an ACS:$$r(k)= \begin{cases}\frac{M-|k|}{M}, & |k| \leq M \\ 0, & \text { otherwise }\end{cases}$$Generate $r(k)$ for $M=10$. Now, in MatLAB form a vector $\mathbf{x}$ of length $L=2$ : as:$$\mathrm{x}=[r(0), r(1), \ldots, r(M), 0 \ldots, 0, r(-M), \ldots, r(-1)]$$Verify that $\mathrm{x} f=\mathrm{ft}(\mathrm{x})$ gives $\phi\left(\omega_k\right)$ for $\omega_k=2 \pi k / L$. (Note that the elemen of $x f$ should be nonnegative and real.). Explain why this particular choice $\mathrm{x}$ is needed, citing appropriate circular shift and zero padding properties the DTFT.Note that $\mathrm{xf}$ often contains a very small imaginary part due to comput roundoff error; replacing $x f$ by real ( $x f$ ) truncates this imaginary compone: and leads to more expected results when plotting.A word of caution - do not truncate the imaginary part unless you are su it is negligible; the command $\mathrm{zf}=\mathrm{real}(\mathrm{fft}(\mathrm{z})$ ) when$$\mathrm{z}=[r(-M), \ldots, r(-1), r(0), r(1), \ldots, r(M), 0 \ldots, 0]$$gives erroneous "spectral" values; try it and explain why it doesn't work.(b) Alternatively, since we can readily derive the analytical expression for $\phi(\omega$ we can instead work backwards. Form a vector$$\mathrm{yf}=[\phi(0), \phi(2 \pi / L), \phi(4 \pi / L), \ldots, \phi((L-1) 2 \pi / L)]$$and find $y=i f f t(y f)$. Verify that $y$ closely approximates the ACS.(c) Consider the ACS $r(k)$ in Exercise C1.12; let $a_1=-0.9$ and $b_1=0$, and s $\sigma^2=1$. Form a vector $\mathrm{x}$ as above, with $M=10$, and find $\mathrm{xf}$. Why is? not a good approximation of $\phi\left(\omega_k\right)$ in this case? Repeat the experiment fo $M=127$ and $L=256$; is the approximation better for this case? Why?We can again work backwards from the analytical expression for $\phi(\omega)$. For a vector$$\mathrm{yf}=[\phi(0), \phi(2 \pi / L), \phi(4 \pi / L), \ldots, \phi((L-1) 2 \pi / L)]$$and find $y=$ ifft $(y f)$. Verify that $y$ closely approximates the ACS for large (say, $L=256$ ), but poorly approximates the ACS for small $L$ (say, $L=20$ By citing properties of inverse DTFTs of infinite, two-sided sequences, expla how the elements of $\mathrm{y}$ relate to the $\operatorname{ACS} r(k)$, and why the approximation poor for small $L$. Based on this explanation, give a "rule of thumb" on $t$ l choice of $L$ for good approximations of the ACS.(d) We have seen above that the $\mathrm{fft}$ command results in spectral estimates from 0 to $2 \pi$ instead of the more common range of $-\pi$ to $\pi$. The Matlab command fftshift can be used to exchange the first and second halves of the $f$ ft output to correspond to a frequency range of $-\pi$ to $\pi$. Similarly, fftshift can be used on the output of the ifft operation to "center" the zero lag of an ACS, Experiment with $\mathrm{fftshift}$ to achieve both of these results. What frequency vector w is needed so that the command plot(w, fftshift $(f f t(x))$ ) gives the spectral values at the proper frequencies? Similarly, what time vector $t$ is needed to get a proper plot of the ACS with stem(t,fftshift(ifft $(x f))$ )? Do the results depend on whether the vectors are even or odd in length?
We’ll notify you at this email when your answer is ready.
More Solved Questions in Your Textbook
Not your book?
Spectral Analysis
Chapter 1
Chapter 1
Chapter 2
Chapter 3
Chapter 4
Chapter 5
Chapter 6
Sections
Problem 1 Problem 2 Problem 3 Problem 4 Problem 5 Problem 6 Problem 7 Problem 8 Problem 9 Problem 10 Problem 11 Problem 12 Problem 13 Problem 14