EE 3512 Signals Spring 2011 Dr.

Obeid 5/9/2011

NAME_____________________

Final Exam
1. (20 points total) Consider the square wave shown here:

x(t) 1 0.5 0 2.5 2 1.5 1 0.5 0 0.5 time (ms) 1 1.5 2 2.5

The magnitude of the Fourier Series of this square wave is given by the expression !! ! !" when ! is odd and !! ! ! when ! is even. Suppose we want to sample this continuous time signal at a sampling rate of !! ! !"!"#. a. (10 points) Which harmonic (i.e. which value of !) will be the first one to alias? Fmax = Fs/2 = 8kHz. Harmonics are at odd integer multiples of 1/T = 1/1ms = 1kHz. Therefore, the first aliased harmonic will be at 9kHz (n=9) b. (5 points) For the harmonic corresponding to ! ! !, what will be the corresponding discretetime frequency !? f=3*1000 = 3000Hz ! = 2*pi*f/Fs = 0.375pi

c. (5 points) For the harmonic corresponding to ! ! !!, what will be the corresponding discrete-time frequency !?

! = 2*pi*11,000/16,000 = 11pi/8 => 5pi/8 = 0.675pi

SHOW ALL WORK IN ORDER TO RECEIVE PARTIAL CREDIT!

Page 1 of 6

EE 3512 Signals Spring 2011 Dr. Obeid 5/9/2011

NAME_____________________

2. (25 points total) Discrete-Time Convolution a. (10 points) Convolve the following two signals:

x[n] 4 3 2 1 0 1 2 3 2 1 0 1 sample (n) 2 3 4 3 2 1 0 1 2 3 2

h[n]

0 1 sample (n)

Y[n] = [3 1 1 0 0] + [0 0 -6 -2 -2] = [3 1 -5 -2 -2]

SHOW ALL WORK IN ORDER TO RECEIVE PARTIAL CREDIT!

Page 2 of 6

EE 3512 Signals Spring 2011 Dr. Obeid 5/9/2011

NAME_____________________

b. (10 points) Suppose you are given the filter ! ! ! !!!!!!!!!!!. Explain how you could determine what kind of filter this is, i.e. high-pass, low-pass, or other. [Note you shouldnt actually try to determine what kind of filter this is, but rather just explain the method of how it could be done in two or three sentences]. Two methods: 1. You could take the discrete time fourier transform of h[n] and plot it with respect to !. 2. You could convolve h[n] with cosines of different frequencies and measure the amplitude of the resulting signals y[n]

c. (5 points) Suppose that when an input signal !! ! ! !!!!! is passed into a filter, the output is ! ! ! !!!!! . What is the filter response to input signal !! ! ! !!!!!!!"? Two methods: 1. Determine h[n] such that x1[n] convolved with h[n] gives y[n]. You will find that h[n] must equal [1 2]. Then convolve x2[n] with h[n] to produce y2[n] = 1 2 -1 -2. 2. Note that x2[n] = x1[n] x1[n-1]. Therefore y2[n] = y[n] y[n-1] = 1 2 -1 -2

SHOW ALL WORK IN ORDER TO RECEIVE PARTIAL CREDIT!

Page 3 of 6

EE 3512 Signals Spring 2011 Dr. Obeid 5/9/2011

NAME_____________________

3. (20 points total) Consider the pole-zero diagram shown below.

10 imaginary 5 0 5 10 200 real

75

b. (5 points) Based on your answer for (a), what kind of filter is this? (i.e. high pass, lowpass, etc). Why? Lowpass filter with two poles. At high frequencies, rolloff will be -40db/dec

c. (10 points) Based on your answer for (a), what would be the filter magnitude at ! ! !""" rads/sec? !"### !"### ! ! ! ! ! !!!"#$ ! ! !"! ! ! ! ! !""! ! ! !"! ! !"""! ! !""! ! !"""

SHOW ALL WORK IN ORDER TO RECEIVE PARTIAL CREDIT!

Page 4 of 6

EE 3512 Signals Spring 2011 Dr. Obeid 5/9/2011

NAME_____________________

4. (15 points) Suppose you are asked to anti-alias a signal using a five-pole filter using either a Butterworth, Chebyshev, or Elliptic filer. Explain the tradeoffs involved in making this decision. When might it make sense to use one type or another? Butterworth has smooth passband but doesnt roll off as quickly in the stopband. In contrast, Cheby and Elliptic filters have passband ripple but roll off much more effectively in the stop band. If your application can tolerate adding some magnitude ripple in the passband, you are better off using Cheby or Elliptic filters. Otherwise, you will have to use a Butterworth.

SHOW ALL WORK IN ORDER TO RECEIVE PARTIAL CREDIT!

Page 5 of 6

EE 3512 Signals Spring 2011 Dr. Obeid 5/9/2011

NAME_____________________

5. (20 points total) The following plot shows the Fourier Transform ! ! of an audio signal !!!! and the transfer function of a filter ! ! that will be used to filter !!!!. The filter output will be !!!!.

1 signal amplitude

x 10

Signal X(f) Magnitude (dB) 20 15 10 5 0 1 10

Filter H(f)

0.5

0 1 10

10 10 freq (Hz)

10

10 10 freq (Hz)

10

a. (5 points) Based on the Fourier Transform !!!!, describe what the signal !!!! might sound like. Although this is a vague question, answer as best as you can in one or two sentences. You only need to look at !!!! to answer this part no need to consider !!!!. The signal appears to have a lot of energy around 100Hz. This is a low frequency in the audio spectrum and therefore we would expect the signal to have a lot of bass tones. In fact, the audio used to generate this figure comes from the hip-hop song Mama Said Knock You Out by LL Cool J which has heavy bass tones throughout. b. (10 points) Sketch as best as possible what the Fourier Transform of the filter output ! ! would look like.

1 signal amplitude

x 10

Signal Y(f)

0.5

0 1 10

10 10 freq (Hz)

10

c. (5 points) How would you expect the filter output ! ! to sound in comparison to the filter input !!!!? High frequency tones get amplified by 10x; low frequency tones also get amplified but by much less than 10x. Overall effect is to make signal sound less bassy than before.

SHOW ALL WORK IN ORDER TO RECEIVE PARTIAL CREDIT!

Page 6 of 6