ECE503 Digital Signal Processing

Problem Set 04

1. Determine and sketch the manitude and phase response of the following systems:
(a) y(n) = 12 [x(n) + x(n 1)]
(b) y(n) = 12 [x(n + 1) x(n 1)]
(c) y(n) = 21 [x(n) + x(n 2)]
(d) y(n) = 31 [x(n) + x(n 1) + x(n 2)]
(e) y(n) = x(n) x(n 8)
(f) y(n) = 2x(n 1) x(n 2)
(g) y(n) = 14 [x(n) + x(n 1) + x(n 2) + x(n 3)]

2. Consider the filter:

y(n) = 0.9y(n 1) + bx(n)
(a) Determine b so that |H(0)| = 1.

(b) Determine the frequency at which |H()| = 1/ 2.

(c) Is this filter lowpass, bandpass, or highpass?
(d) Repeat parts (b) and (c) for the filter y(n) = 0.9y(n 1) + 0.1x(n).

3. Consider an LTI system with impulse response:

 n
 
1
n u(n)
cos
h(n) =
4
4
(a) Determine its system function H(z).
(b) Is it possible to implement this system using a finite number of adders, multipliers and
unit delays? If yes, How?
1

(c) Provide a rough sketch of |H()| using the pole-zero plot.

(d) Determine the response of the system to the input:
 n
1
x(n) =
u(n)
4

4. A digital filter is characterized by the following properties:

1. It is highpass and has one pole and pne zero.
2. The pole is at a distance r = 0.9 from the origin of the z-plane.
3. Constant signals do not pass through the system.
(a) Plot the pole-zero pattern of the filter and determine its system function H(z).
(b) Compute the magnitude response |H()| and the phase response H() of the filter.
(c) Normalize the frequency response H() so that |H()| = 1.
(d) Determine the input-output relation (difference equation) of the filter in the time domain.
(e) Compute the output of the system if the input is:


x(n) = 2 cos
n + 45 ,
< n <
6
(You can use either algebraic or geometrical arguments.)

5. A causal first-order digital filter is described by the system function:

1 + bz 1
1 + az 1
(a) Sketch the direct form I and direct form II realizations of this filter and find the corresponding difference equations.
H(z) = b0

(b) For a = 0.5 and b = 0.6, sketch the pole-zero pattern. Is the system stable? Why?
(c) For a = 0.5 and b = 0.5, determine b0 , so that the maximum value of |H()| is equal
to 1.
(d) Sketch the magnitude response |H()| and the phase response H() of the filter obtained in part (c). (e) In a specific application it is known that a = 0.8. Does the resulting filter
amplify high frequencies or low frequencies in the input? Choose the value of b so as to improve
the characteristics of this filter (i.e., make it a better lowpass or a better highpass filter).

6. In this problem, we consider the effect of a single zero on the frequency response of a system.
Let z = rej be a zero inside the unit circle (r1). Then
H() = 1 rej ej
= 1 r cos( ) + jr sin( ))
(a) Show that the magnitude response is
|H()| = [1 2r cos( ) + r 2 ]1/2
or, equivalently,
20log10 |H()| = 10log10 [1 2r cos( ) + r 2 ]
(b) Show that the phase response is given as:
z () = tan1

r sin( )
1 r cos( )

(c) Show that the group delay is given as:

g () =

r 2 r cos( )
1 + r 2 2r cos( )

(d) Plot the magnitude |H()|dB , the phase () and the group delay g () for r = 0.7 and
= 0, /2 and .

7. Consider a filter with system function:

H(z) = b0

(1 ej0 z 1 )(1 ej0 z 1 )

(1 rej0 z 1 )(1 rej0 z 1 )

(a) Sketch the pole-zero pattern.

(b) Using geometric arguments, show that for r 1, the system is a notch filter and provide a rough sketch of its magnitude response if 0 = 60 .
(c) For 0 = 60 , choose b0 so that the maximum value of |H()| is 1.
(d) Draw a direct form II realization of the system.
(e) Determine the approximate 3-dB bandwidth of the system.

8. Consider the system:

y(n) = x(n) 0.95x(n 6)
3

(a) Sketch its pole-zero pattern.

(b) Sketch its magnitude response using the pole-zero plot.
(c) Determine the system function of its causal inverse system.
(d) Sketch the magnitude response of the inverse system using the pole-zero plot.

9. Consider a system with impulse response:

h(n) = b0 (n) + b1 (n D) + b2 (n 2D)
(a) Explain why the system generates echoes spaced D samples apart. (b) Determine the
magnitude and phase response of the system. (c) Show that for |b0 + b2 | |b1 |, the locations
of maxima and minima of |H()|2 are at:
=

k
, k = 0, 1, 2, ...
D

(d) Plot |H()| and H() for b0 = 0.1, b1 = 1, and b2 = 0.05 and discuss the results.

10. Openended MATLAB Question: Filtering Audio

In this weeks open-ended question, you will leverage the techniques just investigated in the
lectures dealing with the use of LTI systems as filtering tools. You are given the audio file
panagram.wav, which contains the English-language panagram the quick red fox jumped over
the lazy dog. Furthermore, multiple sinusoidal signals have been added to both left and right
stereo audio channels. The objective of this question is to design a cascade of notch filters to
remove these sinusoidal signals from the audio.
(a) What are the center frequency locations of the sinusoidal signals in both the left and right
stereo audio channels?
(b) What are the FIR filter coefficients for the notch filters that you designed in order to
remove these sinusoidal signals from both stereo channels?
(c) Please provide before-and-after spectrograms of the left and right stereo audio channels
showing the effectiveness of your cascaded notch filters.

Here are several hints to get you started with this problem:

 To load an audio file into MATLAB, use the command

[panagram,fs]=wavread(panagram.wav);.
The audio signal in MATLAB will be in stereo. To verify this, use the command size(panagram),
which should show that this file has two columns with column 1 being left audio and column 2 being right audio.
To play an audio signal in MATLAB, use the command soundsc(panagram,fs);.
To visualize the frequency/time characteristics of a sound, use the command spectrogram.
To design a notch filter using MATLAB, try the fdatool using the following parameters:

Response Type = bandstop.

Design Method = FIR.
Frequency Specifications, Fs = 44100.