0% found this document useful (0 votes)
184 views12 pages

Digital Signal Processing Midterm 2

This document is a midterm exam for a digital signal processing course. It contains 3 questions assessing understanding of linear time-invariant systems, the discrete Fourier transform and its fast algorithms, and linear convolution using the cyclic prefix and DFT. Students are asked to determine properties of filter cascades, describe FFT algorithms, compute complexities, derive relationships between input/output sequences and their DFT representations, and determine recoverability of filter coefficients from outputs.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
184 views12 pages

Digital Signal Processing Midterm 2

This document is a midterm exam for a digital signal processing course. It contains 3 questions assessing understanding of linear time-invariant systems, the discrete Fourier transform and its fast algorithms, and linear convolution using the cyclic prefix and DFT. Students are asked to determine properties of filter cascades, describe FFT algorithms, compute complexities, derive relationships between input/output sequences and their DFT representations, and determine recoverability of filter coefficients from outputs.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

EE 123 University of California, Berkeley

Anant Sahai March 15, 2007

Digital Signal Processing


Midterm 2

Name: SID:

Instructions

• Total time allowed for the exam is 80 minutes

• Please write your name and SID on every page of the exam

• Some useful formulas:

– N point Discrete Fourier Transform (DFT)

X
N −1
2πk
n
X[k] = x[n]e−j N

n=0

– Inverse Discrete Fourier Transform (IDFT)

1 NX
−1
2πn
x[n] = X[k]ej N k
N k=0
1. (40 points) Let H(ejω ) be the frequency response of a discrete time LTI filter. The filter is
arranged in the following cascade combination given below:

2 H(e jw ) 2

a. (10 pts) Is the system given in the above figure linear? Justify your answer by giving
a brief proof or a counterexample.

b. (10 pts) Is the system given in the above figure time-invariant? Justify your answer
by giving a brief proof or a counterexample.
Now consider the following alternate cascade formation:

2 H(e jw ) 2

c. (10 pts) Is the system given in the above figure linear? Justify your answer by giving
a brief proof or a counterexample.

d. (10 pts) Is the system given in the above figure time-invariant? Justify your answer
by giving a brief proof or a counterexample.
2. (80 points) Let x[n] be a given time sequence. Assume that the length of the sequence is
M = 3ν . The goal is to compute its M point DFT coefficients X[k]’s.

a. (10 pts) Find the total number of complex multiplications and additions required to
compute X[k]’s directly from the definition of the DFT.
b. (20 pts) Describe in detail a fast algorithm to compute the DFT of x[n].
Hint: Think of a recursive algorithm similar to the decimation in time FFT algorithm.
c. (10 pts) Find the total number of complex multiplications and additions required to
compute X[k]’s using the algorithm from part (b).
d. (20 pts) Now assume that the length of the sequence x[n] is M = 2ν1 3ν2 . For this
case, describe a fast algorithm to compute the DFT. Also, find the total number of complex
multiplications and additions required under this algorithm.
Let x[n] and y[n] be two M = 3ν point sequences. The goal is to compute the linear convolution
of x[n] and y[n], i.e., z[n] = x[n] ∗ y[n].

e. (10 pts) Find the total number of complex multiplications and additions required to
compute z[n] using the DFT algorithm in part (b).
f. (10 pts) Find the total number of complex multiplications and additions required to
compute z[n] using the FFT algorithm discussed in class, i.e., the FFT algorithm for powers
of two. Hint: The question asks for a linear convolution, which is different from a circular
convolution.
3. (40 points) Let h[n] be the impulse response of a LTI filter. Assume that h[n] = 0 for n < 0
and n ≥ L. Let d[n] be an N (N > L) point data sequence, i.e., d[n] = 0 for n < 0 and n ≥ N .
Denote this sequence by the vector

d = [d[0], d[1], · · · , d[N − 1]]

Create a sequence x[n] by adding a L − 1 point cyclic prefix to d[n]. This can be represented by
the following vector

x = [d[N − L + 1], d[N − L + 2], · · · , d[N − 1], d[0], d[1], · · · , d[N − 1]]

Let w[n] denote the output of the filter when the input is x[n]. We ignore the first L − 1 symbols
of the output sequence w[n] and collect the next N symbols. Let y[n] denote this output of
length N , which can be represented as

y = [w[L], w[L + 1], · · · , w[N + L − 1]]

d[n] Cyclic x[n] w[n] Remove y[n]


D[k] IDFT h[n] DFT Y[k]
prefix prefix

Figure 1: Overall system

a. (10 pts) Compute the sequence y[n] as a function of d[n] and the filter impulse response
h[n].
b. (10 pts) Let Y [k], k = 0, 1, · · · , N − 1 denote the N point DFT of the sequence y[n].
Determine the sequence Y [k] as a function of H[k] and D[k], the N point DFTs of h[n] and d[n]
respectively (See Fig. 1).
c. (20 pts) Let N be a multiple of L, i.e., N = mL for some integer m > 0. Suppose that
the DFT coefficients D[mk], k = 0, 1, · · · , L − 1 are known to be equal to 1. In this case can you
recover the filter coefficients h[n] from the y[n] sequence. How would you do so?

You might also like