Digital Signal Processing

DR

AF T

Lecture Notes and Exam Questions

Convolution Sum Convolution Sum of Two Finite Sequences

January 31, 2006

Consider convolution of h (n) and g (n) (M > N);

y (n) =

DR
Sami Arca

AF T
h (n) , n = 0 . . . M 1 g (n) , n = 0 . . . N 1 .
n

h (k) g (n k) ,

k=0 n

0n M1

k=nM+1 N1

h (k) g (n k) , M n N 1

k=nM+1

h (k) g (n k) , N n N + M 2 0, elsewhere .

Complex Convolution Find complex convolution, V (z), of the following z-transforms. X (z) =

January 31, 2006

Complex convolution is dened as

where is the integration contour and must be chosen within the intersection of region z of convergence of X and Y (w). w

DR
Sami Arca

AF T
|z| 1 , Y (z) = |z| V (z) = 1 2 j X

z z1

2z 2z 1

1 . 2

z Y (w) w1 dw w

EEM 409 Digital Signal Processing Midterm Exam EEM 409 Digital Signal Processing Midterm Exam QUESTIONS Q1.

November 26, 2001 November 26, 2001

-3

quantizer levels -2.25 -0.75 0.75

AF T
xq
2.25 0.75 1.5 3 -1.5

x(n)

Quantizer

x q (n)

-0.75 -2.25

codes 00 01 10

4 1

7 -0.8

x (n)

-2.1

-0.3

1.85

0.8

1.8

-1.2

2.25 11 Find xq (n) and total number of bits to represent xq (n). Q2. n x (n) Q3. n

0 2

1 1

2 3

4 2

5 1

6 3

7 2

Using FFT algorithm nd Fourier Transform of the signal at 8 points.

DR
-1 0 1 1 2 2 2 3 1 4 1 5 2 6 2 7 1 x (n) Q4. X (z) = 4z , (4z 1)2 Sami Arca

Calculate Fourier Transform of the signal at 8 points and plot magnitude and phase of the transform.

ROC = |x| >

1 4

Find x (n) using inverse z-Transform formula. Good Luck Sami Arca

EEM 409 Digital Signal Processing Final Exam EEM 409 Digital Signal Processing Final Exam QUESTIONS Q1.

January 22, 2002 January 22, 2002

-3

-1.5

quantizer levels -2.25 -0.75 0.75

2.25 1 Find xq (n) and codes assigned to the samples. Find total number of bits to represent xq (n). Q2. n 0

DR
1 2 3 4 Q4. X (z) = sion method. Sami Arca

X(6 k) = X (k) x (n) 2 1 -1 1 -2 Q3. Calculate magnitude and phase of the following discrete Fourier transform. 2 X (k) = ej2k/3 2 cos k 1 + j 4 sin k , k = 0, . . . , 5 3 3

Find x (n), a) using inverse z-Transform formula, b) using partial fraction expan-

AF T
xq
2.25 0.75 1.5 3

x(n)

Quantizer

x q (n)

-0.75 -2.25

codes 000 001 01

4 1

x (n)

-2.1

-0.3

1.85

0.8

1.8

-1.2

-0.8

Find Discrete Fourier Transform of the signal at 6 points.

1 1 + , 1 1 0.5z 1 0.5z

ROC = 0.5 < |z| < 2

EEM 409 Digital Signal Processing Final Exam

January 22, 2002

Good Luck

DR
Sami Arca

AF T
Sami Arca 6

EEM 409 Digital Signal Processing Final Exam QUESTIONS

Jan. 24, 2003

Q1) Frequency response of a linear shift invariant system is given as sin () . 1 + cos ()

Find the dierence equation of the system.

Q2)

Q3) Let X () be the Fourier Transform of x (n) = (1/2)n u (n). Let Y (k) be DFT with N = 4 samples of a nite-length signal y (n). We set

DR
Determine y (n). Q4) Find z-transform of x (n) = 1 2
|n|

Sami Arca

AF T
H () = j

Im [z]

Pole-zero plot of system response, H (z),

of a system is as in the gure. H (0) =

Re [z] Find H (z). What can you say about 1.

stability of the system ?

Y (k) = X () |=2k/4

k = 0...3 .

Good Luck.

EEM 409 Digital Signal Processing EEM 409 Digital Signal Processing - December 19, 2003

December 19, 2003

QUESTIONS

Q1) Draw diagram of radix-2 FFT algorithm for 4 points. Find FFT of x = [1, 2, 4, 2].

Q2) Using DFT nd convolution of the following sequences. x = [1, 2]

Q3) Design half band low pass discrete-time lter using Fourier series method. Choose lter length as 7 and use hamming window. (Hamming window: 0.54 + 0.46 cos 0

DR
Q4) Find z-transform of x (n) = 1 2
n

AF T
y = [2, 3] w (n) = N1 2n , |n| N1 2 otherwise

u (n). Evaluate the transform at points, zl = A W l ,

l = 0, . . . . 9

where A = e j/6 and W = 0.995ej0.25 .

Good Luck Sami Arca

EEM 409 Digital Signal Processing - December 19, 2003

ANSWERS

A1) Consider that discrete Fourier transfor of sequence g (n) is Gk . Sami Arca 8

EEM 409 Digital Signal Processing

December 19, 2003

N/21

where W = ej2/N and W N/2 = 1. Let,

Hence the transform coecients can be witten as follows,

g0 g1 g2 g3

DR
y1 -1 W1
x0 = 5, x1 = 0, y0 = 3, and y1 = 4. G0 = x0 + x1 G2 = x0 x1 Sami Arca

Here N = 4 and W = ej2/4 = j. g0 = 1, g1 = 2, g2 = 4, and g3 = 2.

AF T
n=0

Gk =

gn + gn+N/2 W kN/2 W kn , k = 0, 1, . . . , N 1

xn = gn + gn+N/2

yn = gn gn+N/2

N/21

G2l =

ln xn WN/2 ,

l = 0, 1, . . . , N/2 1

n=0 N/21 n=0

G2l+1 =

ln (yn W n ) WN/2

x0 x1

G0 G2 G1 G3

y0

-1

-1

W0

-1

Figure 1: Block diagram of four point FFT

=5 =5

G1 = y0 + W y1 = 3 + j4 G3 = y0 W y1 = 3 j4 9

EEM 409 Digital Signal Processing

December 19, 2003

A2) Length of x (n) and x (n) are N = 2 and M = 2 respectively. a (n) = x (n) y (n). Length of a (n) is N + M 1 = 3.
1

DR
X (0) = 3 A (0) = 3 A (1) = 4.5 6.0622 j Inverse DFT of A (k) gives, Sami Arca

Y (0) = 1 Y (1) = 2 3ej2/3 = 2.5 + 2.5981 j Y (2) = 1 + 2ej4/3 = 2.5 2.5981 j A (2) = 4.5 + 6.0622 j

AF T
X (k) = Y (k) = x (n) ej(2/3)nk , k = 0, 1, 2.
n=0 1

y (n) ej(2/3)nk ,

k = 0, 1, 2.

n=0

A (k) = X (k) Y (k) , a (n) = 1 3

2 k=0

k = 0, 1, 2.

A (k) e j(2/3)nk ,

n = 0, 1, 2.

X (k) = 1 + 2ej(2/3)k

Y (k) = 2 3ej(2/3)k

X (1) = 1 + 2ej2/3 = 1.7321 j

X (2) = 1 + 2ej4/3 = 1.7321 j

2, 1, (A (k)) = a (n) = IDFT 6, 0,

n=0 n=1 n=2 otherwise 10

EEM 409 Digital Signal Processing A3)

December 19, 2003

DR
Sami Arca

AF T
/2 < || < lZ 11

1, H () = 0, H ( + 2 l) ,

|| /2

EEM 409 Digital Signal Processing Final Exam EEM 409 Digital Signal Processing Final Exam - January 19, 2004

January 19, 2004

QUESTIONS

Q1) Derive in frequency FFT algorithm and draw block diagram of 4 points. Find FFT of x = [1, 2, 4, 3] using the algorithm.

Q2) The response of an LTI system to input x (n) is y (n). DFT of the input and output signals are X = [2, 3 + 5i, 8, 3 5i] and Y = [12, 4 16i, 16, 4 + 16i] respectively. Find impulse response h (n) of the system.

Q3) The characteristics of an ideal high pass lter is given as, 0, 0 || < 2/3

. Design this lter using Fourier series method. Choose lter length as 7 and use hamming window. (Hamming window:

DR
0 ) Q4) Find z-transform of x (n) = A W l , 1 2
n

l = 0, 1, 2 where A = e j/6 and W = 0.5ej0.25 .

Q5) Impulse response of a continuous time system is given as, cos ( t/4) , 0, |t| 4 otherwise

. Employ sampling period of T = 0.25sec. Obtain impulse response of discrete time 12

Sami Arca

AF T
H () = 1, 2/3 || < w (n) = 0.54 + 0.46 cos N1 2n , |n| N1 2 otherwise h (t) =

u (n). Evaluate the transform at points, zl =

EEM 409 Digital Signal Processing Final Exam approximation of the system by using impulse invariance method.

January 19, 2004

. Find impulse response hA (t) of the system. Using impulse invariance method, nd impulse response, hD (n), and system function, HD (z) of discrete time approximation of the analog system. The sampling period is T .

DR
Sami Arca

AF T
N

Q6) System function of an LTI system can be written as, Ak s pk

HA (s) =

k=1

Good Luck Sami Arca

13

Digital Signal Processing Midterm Exam QUESTIONS

November 12, 2004

Q1) Find state-space representation for an LTI system given as,

Q2) Transfer function of an LTI system is given as,

Find impulse response of the system by using partial fraction expansion method.

Q3) Find 33 DFT matrix and calculate DFT of x (n) = 3 (n)+ (n 1)+2 (n 2) by using the matrix representation of DFT.

Q4) Design a low-pass lter using Fourier Series method. The bandwidth of the lter is /2 and the length is N = 5. Employ Blackman window,

DR
to reduce ringing eect. Sami Arca

w (n) = 0.42 + 0.5 cos

AF T
y (n) 2y (n 1) + y (n 2) = x (n) . H (z) = 2z2 + 16z 12 . z3 5z2 + 6z 2n N1 4n N 1 + 0.08 cos , n= , , , N 1 N1 2 2

14

EEE 409 Digital Signal Processing Final Exam QUESTIONS

November 12, 2004

Q1) Write Decimation in Time (DIT) Radix-2 FFT algorithm.

Q2) Design half band low pass lter by using 2rd order butterworth approximation.

Q3) Find Fourier transform and energy spectrum of x (n) = 1/ 6 (n)+2/ 6 (n 1) 1/ 6 (n 2).

Q4) x (n) = cos (0.1 n/3) + cos (0.1n) is given. This signal is an input to an LTI lter with impulse response h (n) = a (n) + b (n 1) + a (n 2). The output of the lter is y (n) = cos (0.1n). Find lter coecients.

Q5) Sampling frequency of a discrete signal is increased twice by digital interpolation; y (n) = xup (n) h (n), where

DR
and h (n) is a half band digital lter. Sami Arca

Let x (t) = cos (2t). Sample this signal with sampling frequency f s = 5Hz (sampling

period T s = 1/ f s = 0.2 sec.) and obtain x (n). Using the digital interpolation method and h(n) = 1/ 2 (n) + 1/ 2 (n 1) nd interpolated signal y (n). Calculate y (n) for n = 0.1 (2k + 1) sec, where, k = 0, 1, 2, 3, 4.

AF T
xup = x(n/2), n = 0, 2, 4, . . . 0, otherwise 15

Digital Signal Processing Homework Digital Signal Processing Homework

November 12, 2004

1. MATLAB implementation of Decimation in Time (DIT) Radix-2 FFT (see http://cnx.rice.edu/content/m12016/latest). 2. Design half band low pass lter

(a) by using Fourier series method (length N = 11). Employ Blackman window.

(b) by using 3rd order butterworth approximation

(see http://www-sigproc.eng.cam.ac.uk/ ad2/3f3/3F3-3.doc

and http://www.stanford.edu/class/ee102b/lectures/lecnotes13.pdf ).

3. Find Fourier transform (X () and energy spectrum (X () X ()) of audio signal http://www.members.tripod.com/ buggerluggs/wavs/hello.wav. Calculate Fourier transform at 8192 points.

4. Output of an LTI system with frequency response of H () for an input signal x (n) = cos (a n) + cos (b n) is y(n) = H (a ) cos (a n) + H (b ) cos (b n) .

DR
tion; y (n) = xup (n) h (n), where xup = 0, and h (n) is a half band digital lter. Sami Arca

Let h (n) = F ( 1) [H ()] = a (n) + b (n 1) + a (n 2). Find h (n) such that y (n) = h (n) x (n) = cos (a n). a = 0.1/3 and b = 0.1 .

5. Sampling frequency of a discrete signal is increased twice by digital interpola-

Let x (t) = cos (2t). Sample this signal with sampling frequency fs = 5Hz

(sampling period T s = 1/ fs = 0.2sec.) and obtain x (n). Using the digital 16

AF T
x(n/2), n = 0, 2, 4, . . . otherwise

Digital Signal Processing Homework

November 12, 2004

interpolation method and one of the halfband lters designed in Q2, increase the

6. Let x (n) = xa (n) + xb (n). xa (n) = cos (0.1n) and xb (n) = 2 sin (0.01n). Here xb (n) is unwanted signal which shift the base of xa (n), therefore it is called trend of x (n). We want to remove the trend. Use the linear trend removal algorithm:

segment x (n) into small parts, x (n) = x (n + kN) , n = 0, . . . N 1.

for k th segment nd linear trend ; an + b,

remove the linear trend from the k th segment; yk (n) = xk (n)anb, n = 0 . . . N 1. y (n) =

For this example choose N = 5 and remove the trend of x (n) for n = 0 . . . 299.

DR
Sami Arca

AF T
xk (n kN), where xk (n) =
k k

sampling frequency of x (n) to 10Hz.

yk (n kN) is the output of the algorithm. y (n) xa (n).

17

EE597 Discrete Signal and System Analysis Midterm Exam EE597 Discrete Signal and System Analysis Midterm Exam QUESTIONS

December 20, 2001 December 20, 2001

A linear time-invariant and causal system which is described by y (n) (a + b) y (n 1) + ab y (n 2) = c x (n 2) where |a| < 1 and |b| < 1 is given.

Q1. Find the impulse response of the system.

Q2. Find the frequency response of the system.

Q3. Plot block diagram and ow graph of the system.

Q4. Obtain state-space characterization of the system.

DR
Sami Arca

AF T
18

EE597 Discrete Signal and System Analysis Midterm Exam EE597 Discrete Signal and System Analysis Midterm Exam QUESTIONS

January 30, 2002 January 30, 2002

A linear time-invariant and causal system which is described by

y (n) 2a y (n 1) + a2 y (n 2) = a x (n 1) where a = 0.5, is given. Q1. Find the impulse response of the system using time-domain analysis. Q2. Find the the frequency response of the system.

Q3. Plot block diagram and ow graph of the system.

Q4. Obtain state-space characterization of the system.

Q5. Transfer function of a linear time-invariant system is given as following. Find input-output relation in time domain. Check if the system is causal and stable. 2z 0.25

DR
Sami Arca

AF T
H (z) = z2 19

EEE 409 Digital Signal Processing Midterm Exam QUESTIONS

November 18, 2005

Q1) Design a low-pass lter using Fourier Series method. The bandwidth of the lter is /3 and the length is N = 5. Employ Hamming window,

w (n) = 0.54 + 0.46 cos to reduce ringing eect.

Q2) A sequence of symbols are given ; {a, b, b, c, a, c, b, c, c, b}. The following codes are assigned to the symbols.

a) Find probability of symbols a, b, c. b) Calculate information of the symbols and entropy of the sequence. c) What is the average code length (bits/symbol)?

DR
formula. X (z) = Recall : 1 2 j F (z) dz = Res = q
k

Q3) Find inverse z-transform of the following z-transform using the inverse z-transform z , |z| > 1 (z 1)2

where z = q is multiple pole of F (z) and Res q

Sami Arca

AF T
2n , N1 n= N1 N1 , , , 2 2 a, 00 b, 01 c, 1 Res q
k

1 d1 lim 1 (z q) F (z) ( 1) ! zq dz is the k th resudue.

20

EEE 409 Digital Signal Processing Midterm Exam Q4) Numeric approximation of integral of x (t) is,

November 18, 2005

y (n) =

Since y(n) y(n) and x(n) x(n),

This equation can be considered as a discrete-time system. Find a) the system response, b) the frequency response of the system.

ANSWERS

A1)

DR
0

Sami Arca

AF T
n k=

x ((k 1) ) + x (k) 2

= y ((n 1) ) +

x ((n 1) ) + x (n) 2

y (n) = y (n 1) +

x (n 1) + x (n) 2

/3

g (n) =

1 2 1

jn

1 dw + 2
0

e jn dw

/3

/3

cos (n) dw

1 sin (n/3) n

h (n) = w (n) g (n) h (n) = h (n) 21

EEE 409 Digital Signal Processing Midterm Exam

November 18, 2005

A2)

DR
Sami Arca

AF T
h (0) = 0.3333 h (1) = 0.1489 h (2) = 0.0110 p (a) = p (b) = p (c) = 2 10 4 10 4 10 I (a) = log2 (p (a)) = 2.3219 I (b) = log2 (p (b)) = 1.3219 I (c) = log2 (p (c)) = 1.3219 22

EEE 409 Digital Signal Processing Midterm Exam

November 18, 2005

Coded sequence : 00 01 01 1 00 1 01 1 1 01

A3)

For n 0, z = 1 is pole of F (z) and q = 1 and = 2.

DR
Res2 = 1 z = 0, F (z) = zm Resm = 0 Sami Arca

For n = m and m > 0 there is two poles at z = 0 and at z = 1. Consider pole at 1 (z 1)2

AF T
H = p (a) I (a) + p (b) I (b) + p (c) I (c) = 1.5219 bits A = 16 62+41 = = 1.6 10 10 x (n) = = 1 2 j 1 2 j X (z) zn1 dz zn dz (z 1)2

F (z) =

zn (z 1)2

d 1 (z 1)2 F (z) = lim nzn1 = n lim z1 dz z1 1!

(1)m+1 dm1 m (z F (z)) = m! dzm1 (z 1)m+1

(1)m+1 1 dm1 lim m1 (zm F (z)) = lim m z0 (m 1) ! z0 dz (z 1)m+1

= m 23

EEE 409 Digital Signal Processing Midterm Exam Consider the pole at z = 1,

November 18, 2005

Then the inverse transform for any n is,

A4)

DR
= = = For continuous case,
t

and the system function is

Sami Arca

AF T
Res2 = 1 1 d m (z 1)2 F (z) = lim m+1 = m lim z1 dz z1 z 1! Resm + Res2 = 0 0 1 x (n) = nu (n) Y (z) = z1 Y (z) + z1 X (z) + X (z) 2 H (z) = Y (z) 1 + z1 = 2 1 z1 X (z) H () = H (z) 2 2 ( j) cot () 2 z = e j 1 + ej 1 ej ej/2 2 cos () ej/2 2 j sin () y (t) = x () d

24

EEE 409 Digital Signal Processing Midterm Exam

November 18, 2005

Equating the system functions of discrete and continuous systems, 1 1 + z1 = s 2 1 z1

results with

Using this relation it is possible to transform a continuous time system to a discrete time system. This transformation is called as bilinear transform.

DR
Sami Arca

AF T
s= 2 1 z1 1 + z1 25

H (s) =

1 s

EEE 409 Digital Signal Processing Final Exam QUESTIONS

January 24, 2006

Q1) Design a high-pass lter using Fourier Series method. The passband of the lter is [/2, ] and the length is N = 5. Employ Hamming window,

w (n) = 0.54 + 0.46 cos to reduce ringing eect.

Q2) A sequence of symbols are given ; {0, 1, 2, 2, 2, 2, 2, 3, 3, 3, 3, 0, 0, 2, 2}. The following codes are assigned to the symbols. 0, 000 1, 001 2, 1

a) Find probability of symbols 0, 1, 2, 3. b) Calculate information of the symbols and entropy of the sequence. c) What is the average code length (bits/symbol)?

DR
expansion method. X (z) = 2 H (s) H (s) = Sami Arca

Q3) Find inverse z-transform of the following z-transform using the partial fraction

Q4) n th order analog Butterworth low pass lter is dened as, 1

n

where H (s) is the system function and a is the bandwith of the lter in rad/sec. First 26

AF T
2n , N1 n= N1 N1 , , , 2 2 3, 01 z (4 z 3) , (2 z 1) (4 z 1) |z| > 1 2 1 + (1) s a
2n

EEE 409 Digital Signal Processing Final Exam order Butterworth lter is,

January 24, 2006

The system function is,

since the pole s = a resides in the left side of the s-plane. Using bilinear transform; 2 1 z1 T 1 + z1

nd discrete approximation of the rst order analog Butterwort lter for the bandwith a = 2 rad/sec and sampling interval T = 2 sec. Employ pre-warping; a T 2 arctan T 2

DR
lter. Sami Arca

where d is discrete equivalent of analog frequency a in rad/sec. To avoid warping 2 a T in H (s). First nd discrete equivalent H (z) a should be replaced by arctan T 2 of H (s) using the bilinear transform and then write dierence equation of the discrete

AF T
H (s) H (s) = 1 = s 1 a 1 1 s s 1+ 1 a a
2

H (s) =

1+

s a

s=

d =

27