3-Digital Signal Processing

Semester-V
Digital Signal
Processing
EEE_Semester-V_Ch05.indd 1 7/13/2012 1:08:10 PM

The aim of this publication is to supply information taken from sources believed to be valid and
reliable. This is not an attempt to render any type of professional advice or analysis, nor is it to
be treated as such. While much care has been taken to ensure the veracity and currency of the
information presented within, neither the publisher nor its authors bear any responsibility for
any damage arising from inadvertent omissions, negligence or inaccuracies (typographical or
factual) that may have found their way into this book.
EEE_Sem-VI_Chennai_FM.indd iv 12/7/2012 6:40:43 PM

B.E./B.Tech. DEGREE EXAMINATION,
Nov/Dec 2013
Fifth Semester
Electrical and Electronics Engineering
DIGITAL SIGNAL PROCESSING
Time: Three hours Maximum: 100 marks
Answer ALL questions
PART A (10 × 2 = 20 marks)
1. What is the Nyquist rate for the s/l Xa (t) = 3 cos 600 p t + 2<??> 1800.
⎛ π 30 n ⎞
2. Determine the fundamental period of the s/l cos ⎜
⎝ 105 ⎟⎠
?? .
3. Determine the z-transform & ROC for the s/l

x(n) = d (n – k) + d (n + k)
4. Prove the convdution property of z-transform.
5. Draw the butterfly dig. for decimation in time FFT Algorim<??>.
6. In eight pt decimation in time (DIT), what is the gain signal path that
goes from x(7) to x(2)?
1 + 0.8 z −1
7. Is the given transfer fn H(z) = H ( z ) = represents law <??>
filter or high <??>par filter? 1 − 0.9 z −1
8. The impulse response of an analog filter is given in fig. let h(n) = ha(nT)
where T =1; Determine the system functions.
9. What is meant by b it reversed addressing mode? What the application

for which this addressing mode is prefe<??>.
10. Compare the RISC & CISC processors.
DSP_Nov-Dec 2013-QP.indd 1 4/23/2014 3:52:35 PM

2.2 B.E./B.Tech. Question Papers
Part B (5 × 16 = 80 marks)
11. (a) Determine the response of the following s/m’s to the i/<??> signal.
x(n) = x( n) = { | n|, − 34 ⱕ n ⱕ 3
0, otherwise
(i) x1(n) = x(n – 2) m (n – 3)
(ii) x2 (n) = x(n + 1) μ(n – 1)
(iii) y(n) = (x(n + 1) + x(n) + x(n – 1))
(iv) y(n) = max (x(n + 1), x(n), x(n – 1))
(v) Find the even & odd components of given x(n). (16)
Or
(b) A discrete time systems can be
(i) Static or dynamic
(ii) Linear or non-linear
(iii) Time invariant or time varying
(iv) Stable or unstable
Examine the following system with respect to the properties above

y(n) = x(n) + nx(n + 1) (16)
12. (a) (i) Determine the causal s/l x(n) whose z-transform is given by
1 + z −1
x( z ) = (10)
1 − z −1 + 0.5 z −2
(ii) Determine the z-+/F of the s/l x(n) = cos(w0 n) u(n<??> (6)
Or
(b) Consider the s/m shown in fig. 2 with h(n) = an u(n), –1 < a < 1
determine the response g(n) of the s/m to the excitation x(n) = u(n +
5) – u(n - <??>) (16)

B.E./B.Tech. Question Papers 2.3
13. (a) (i) The 1st five points of the 8 point DFT of a real value sequenced
are {0.25, 0.125, j 0.3018, 0, 0.0125, j 0.0518<??>. Determine
the remaining 3 points.
(ii) Compute the 8-pt DFT of the sequence x = [1, 1, 1, 1, 1, <??>

using decimation in frequence. FFT Alg. (16)
Or
(b) Consider the sequences:
x1(n) = {0, 1, 2, 3, 4} x2(n) = {0, 1, 0, 0, 0} s(n) = {1, 0 <??>
(i) Determine a sequence y(n) so that y(k) = x1(k) x2(k).
(ii) Is these a sequence x3(n) such that s(k) = x1(k) x3(k). (16)
14. (a) Design an FIR linear phase, digital filter apporximating ideal
{
frequence response Hd (ω ) = 1,|0,ωπ /6|ⱕⱕπ /6|ω | ⱕ π
Determine the coefficients of a 25 tap filter based. The window
method with a rectangular window. (16)
Or
(b) (i) Convent the analog filter with system function
s + 0.1
H a ( s) = in to a digital HR filter by mean the
( s + 0.1) 2 + 9
impulse invarianace method.
(ii) Draw the direct form I & direct form II structures for the given
difference equation x(n) = y(n – 1) – 0.5y (n – 2) + x(n) – x(n – 1)
+ x(n + 2). (16)
15. (a) Explain von Neumann, Harvard architecture and modified Harvard
architecture for the computer. (16)
Or
(b) (i) Explain how convdution is performed using a single MA unit.
(ii) Discuss the addressing modes used in programmable DSP?<??>

(16)

B.E./B.Tech. Degree Examination,
MAY/JUNE 2013
Fifth Semester
Time: Three hoursMaximum: 100 marks
PART A (10 × 2 = 20 marks)
1. Given a continuous time signal x(t) = 2 cos500πt. What is the Nyquist
rate and fundamental frequency of the signal?
2. Determine whether x[n] = u[n] is a power signal or an energy signal.
3. Given a difference equation y(n) = x[n] + 3x[n - 1] + 2y[n - 1]. Determine

the system function H(z).
n
4. Find the stability of the system whose impulse response h(n) =  1 
 
u(n). 2
5. Find the discrete Fourier transform for δ [n].
6. Draw the basic butterfly diagram for DIF algorithm.
7. Name two methods for digitizing the transfer function of an analog

filter.
8. List the properties of chebyshev filter.
9. Mention one important feature of Harvard architecture.
10. What is the advantage of pipelining?

Part B (5 × 16 = 80 marks)
11. (a) (i) Given y[n] = x[n2]. Determine whether the system is linear, time
invariant, memory less and causal. (8)
(ii) Determine whether the following is an energy signal or power
signal.(8)
N 
(1) x1[n] = 6 cos  n
 2 
(2) x2[n] = 3(0.5)n u[n].
Or
(b) Starting from first principles, state and explain sampling theorem
both in time domain and in frequency domain. (16)
12. (a) (i) Find the Z-transform and its associated ROC for the following
n −n
−1 1
discrete time signal x[n] =   u[n] + 5   u[- n - 1] (8)
 5  2
(ii) Evaluate the frequency response of the system described by
1
system function H(z) = (8)
1 − 0.5 z −1
Or
1
(b) Using z-transform determine the response y[n] for n ≥ 0 if y[n] =
n 2
y[n - 1] + x[n], x[n] =  1  u(n) y(- 1) = 1. (16)
 3
13. (a) Find the output y[n] of a filter whose impulse response is h[n] =
{1,1,1} and input signal x[n] = {3,-1,0,1,3,2,0,1,2,1} using overlap
save method. (16)
Or
Find the DFT of a sequence x[n] = {1,2,3,4,4,3,2,1}. Using
(b)
decimation in Time (DIT) algorithm. (16)
14. (a) Design and realize a digital filter using bilinear transformation for
the following specifications. (16)
Monotonic pass band and stop band -3.01 dB cut off at 0.5 π rad
magnitude down at least 15dB at ω = 0.75π rad.

Digital Signal Processing (May/June 2013) 2.5
(b) (i)
Consider the causal linear shift invariant filter with system
1 + 0.875 z −1
function H(z) = . Draw the structure
(1 + 0.2 z + 0.9 z −2 )(1 − 0.7 z −1 )
−1
using a parallel interconnection of first and second order

systems.(8)
(ii) Consider the following interconnection of a linear shift invariant
system.
x [n] + w [n] y [n]

+ H2(e jw)
–
h1 [n]
Where x[n] = δ [n]

h2[n] = δ [n - 1]
1 | ω | ≤ π / 2
H2(ejog) = 
0 π / 2 | ω |≤ π
Find the overall impulse response h[n] of the system. (8)
15. (a) Explain various addressing modes of a digital signal processor.(16)
Or
(b) Draw the functional block diagram of a digital signal processor and
explain. (16)

Solutions
May/June 2013
PART A
1. Nyquist rate  2o  1000

Fundamental frequency  o  500
1
2. P  and E  
2
 o  P   the signal x[n] =u[n] is
a power signal
y( z ) 1  3z 1
3. H ( z )  
x( z ) 1  2 z 1
z
4. H ( z)  | z | 2.
z  1/2
1
Pole lies at Z  within the unit circle hence stable.
2
5. x( k )  1
6.
7. Bilinear Transformation
Impulse Invariant Transformation
8. The magnitude response of the cheby stev filter exhibits ripple either in pass band
or in stopband according to type.
9. Program and data memories lie in two separate spaces, permitting fall overlap of
instruction fetch and execution.
10. Instruction fetch, decode, operand read and execute operations are independent,
which allows overall instruction executions to overlap.
PART B
11 a) i)
The system y[n]  x[n2 ] is linear, Time variant, non – causal and system with memory.
11 a) ii)
1)
 
x1 [n]  6 cos  n 
2 
Period N = 4
N 1 2
1
P
N
 x( n)
n 0
P  18  It is a power signal
o

2
E x( n)  
n  
It is not an energy signal
2)
x2 [n]  3(0.5) n u( n)

 x(n)
2
E
n 
E  12  It is an energy signal
N 1
1
 x(n)
2
P  LtN
N n 0
P  0 It is noy a Power signal
11 b) Sampling theorem:
For a band limited signal x(t) with x( jw )  0 w  wm , x(t ) can be uniquely determined by its
samples x( nT ), n  0,  1,  2,  if ws  2wm . Given these samples, we can reconstruct x(t) by
generating a periodic impulse train in which successive impulses have ampliterdes that
are successive sample values. This impulse train is then processed through an ideal
lowpass filter with gain T and cutoff frequency greater than wm and less than ws  wm . The
resulting o/p signal will be exactly equal to x(t).

P (t )    (t  nT )
n 

X s (t )   x( nT ) (t  nT )
n  
1
X s ( jw )  [ x( jw ) * P ( jw )]
2
2 
P ( jw )   ( w  kws )
T k 
1 
X s ( jw )   x( j( w  kws ))
T k 
12 a) i

y( z )   x(n)z
n 
n
 n 1
 1
    z  n  5  (2) n z  n
n 0
 5 n  
 n 
 1
     z  n  5 (2 1 z ) n
n 0
 5 n 1
1 5
x( z )  1

1  0.2 z 1  2 z 1
1
Roc :  z  z
5
12 a) ii)
1 z
H ( z)  z 1

Given 1  0.5 z  0.5
Roc: z  1/2
As ROC includes unit circle H (e jw ) exists.
H (e jw )  H ( z ) z  (e jw )
e jw

e  0.5
jw
cos w  j sin w

cos w  0.5  j sin w
1
H (e jw ) 
(cos w  0.5)  (sin w ) 2
2
sin w
H (e ju )  w  Tan 1
cos w  0.5
12 b)
Given
1
y[n]  y[n  1]  x[n]
2
Taking z-Transform on both the sides
1 1
y[ z ]   z y( z )  y( 1)   x( z )
2
1 z
y( z )   z 1 y ( z )  1 
2 1
z
3
0.5 z
y( z )  
1  0.5 z 1  1
 z   ( z  0.5)
3
0.5 z 3z 2z
y( z )   
z  0.5 z  0.5 1
z
3
Taking Inverse z-Transform
y[n]  0.5 (0.5) n  3(0.5) n  2(1/3) n u[n]
y[n]  3.5(0.5) n  2(1/3) n  u[n]
13 a) overlap save method:

The given sequence can be divided into blocks of data as shown below:
x1 [n]  
0, 0, 3, 1, 0
 x2 [n]  
1, 0,
 1,3,
 2
M 1 zeros 3 data points Two datas Nexts data
from previous points
b c
Similary,
x3 [n]   3, 2, 0,1, 2 x4 [n]  1, 2,1, 0, 0

Since x1 [1], x2 [n] x3 [n] and x4 [n] are of length 5, hence append two zeros in h(n). Then
perform circular convolution
y1 [n]  x1 [n]* h[n]   1, 0,3, 2, 2
y2 [n]  x2 [n]* h[n]   4,1, 0, 4,6
y3 [n]  x3 [n]* h[n]   6, 7,5,3,3
y 4 [ n]  x4 [n]* h( n)  1,3, 4,3,1
h[n] = is obtained by discarding the first M -1 sample from each sequence.
Y [n]   3, 2, 2, 0, 4,6,5,3,3, 4,3,1
13 b)
The twiddle factors associated with the flow graph are

W80  1 W81  e  jz /8  0.707  j 0.707
W82   e  j 2 /8    j
2
W83   e  j 2 /8   0.707  j 0.707

3
X ( k )   20, 5.828  j 2.414, 0, 0.172  j 0.414,

0, 0.172  j 0.414, 0, 5.828  j 2.414
14 a)
2 w
1  Tan 1  2 rad /sec.
T 2
2 w2
2  Tan  4.82 82 rad /sec.
T 2
N= 1.9412  2
2
c   2 rad/sec
(100.3  1)1/ 4
1 s
Ha ( s)  s
S  2s  1
2
2
1
Ha ( s) 
s2  2 2s  4
2 1  z 1
H ( z )  Ha ( s) s 
T 1  z 1
1  2 z 1  z 2
H ( z) 
3.414  0.5858 z 2
14 b) i)
Consider the causal linear shift invariant filter with system function.
1  0.875 z 1
H ( n) 
(1  0.2 z 1  0.9 z 2 )(1  0.7 z 1 )
A  Bz 1 c
 
(1  0.2 z  0.9 z ) (1  0.7 z 1 )
1 2
Solving for A, B, C
A  0.2794 B  0.9265 C  0.7206
14 b) ii)
w[n]  x[n]  h1 [n]

w[n]   [n]   [n  1]
y[n]  w[n]* h2 [n]
y[n]  8[n]   [n  1] * h2 [n]
y[n]  h2 [n]  h2 [n  1]

1
h2 [n] 
2  H

2 (e jw ) e jwn dw
sin n /2
h2 [n] 
n
sin n /2 sin( n  1) /2
h[n]  
n ( n  1)
15 a) Refer 15 b) –A/M10-EC1302
15 b) Refer 15 b) – N/D10-EC1302
NOV/DEC 2012 (Code no: 31219)
Fifth Semester
Time : Three hoursMaximum : 100 marks
Answer ALL questions.
PART A (10 × 2 = 20 marks)
1. Define LTI system.
2. What is meant by Nyquist rate?
3. Relate Z-transform and DFT mathematically.
4. State BIBO stability criterion in Z-transformed domain.
5. How DFT is different than DTFT?
6.
State the number of complex multiplication and complex addition
involve in N-point decimation-in-time FFT algorithms.
7. Define an FIR filter.
8. State the advantages of IIR filter over FIR.
9. What is the need for specialized digital signal processor?

10. How round-off error affects filter design?
PART B (5 × 16 = 80 marks)
11. (a) Derive and explain the sampling and interpolation of discrete-time
signals. Illustrate the sampling of discrete-time signal in frequency
domain using neat diagrams.
Or

(b) (i) Define the error in quantization of sinusoidal signals. (6)

(ii) Derive the signal-to-quantization error (SQNR) in decibel for a
sinusoid signal. (10)
12. (a) (i) Find the inverse Z-transform of
1
X(Z) =
1 − 1.5Z + 0.5Z −2
−1
if (1) ROC : |Z| > 1, (2) ROC : |Z| < 0.5, (3) ROC : 0.5 < |Z| < 1.
(10)
(ii) Find the Z-Transform of a causal and anti-causal signal, and
comment on their ROC. (6)
Or
(b) Find the Discrete Time Fourier Transform of
(i) x(n) = a|n|, -1 < a < 1 (8)
 A, − M ≤ n ≤ M
(ii) x(n) = =  (8)
0, otherwise
13. (a) (i) Determine the 6-point DFT of the signal. (10)
x(n) = {3, 2, 1, 0, 1, 2}.
(ii) Present DFT and IDFT transformation pair in matrix form. (6)
Or
(b) Develop 8-point radix-2 decimation in time algorithm with input in
normal order and output in digit reversed order. Derive the necessary
equations and show the flow diagrams. (16)
14. (a) Design an FIR digital low pass filter with desired system function.
Hd(w) = e-j8w , 0 ≤ |w| ≤ p / 3
= 0, p / 3 <|w| ≤ p.
Use Hamming window with N = 7.
Or
(b)
Design an IIR digital low pass filter to meet the following
requirements
Ripples in passband ≤ 1 dB, Passband cutoff freq. = 4 KHz

Digital Signal Processing (Nov/Dec 2012) 2.11
Ripples in stopband > 40 dB, Stopband cutoff freq. = 6 KHz

Sample rate = 24 KHz.
Use bilinear transformation.
15. (a) Draw and explain the architecture and features of TMS320C54X
processor.
Or
(b) Analyze quantization effects in designing digital filters with suitable
diagram.

Solutions
Nov/Dec 2012
PART A
1. It is Linear Time Invariant System. It should satisfies the following Properties: 1. Homogenity, (2)
Superposition principle. This is a discrete time system it satisfies both the linearity and time invariant
property.
T[a1x1(n) + a2x2(n)] = a1T[x1(n)] + a2T[x2(n)] = a1y1(n) + a2y2(n)
Time Invariant property:
y(n,k) = y(n – k) y(n – k) = T[x(n – k)]
2. The Nyquist rate is the minimum sampling rate required to avoid aliasing, equal to twice the highest
frequency contained within the signal.
fN = 2B = 2 fmax, where B is the highest frequency at which the signal can have nonzero energy.
3.
1  e  jwN N 1
X (K )
X ( z) 
N

k 0
 2 k 
 j 
1 e  N 
4. A linear time-invariant system is BIBO stable if and only if the ROC of the system function includes
the unit circle. A causal linear time-invariant system in BIBO stable if and only if all the poles of
H(Z) are inside the unit circle.
H(z)   n  h(n)


5.
Sr. DTFT DFT

No.
1.  N 1
Equation : X ( )   x ( n)e
n 
 j N
Equation : X (k )   x(n)e  j 2 x kn / N
na
k = 0, 1, … N – 1
2. Here X(w) is continuous function of n Here X(k) is defined for k = 0, 1, … N – 1.
3. DTFT cannot be evaluated on digital computer. DFT can be evaluated on digital computer.
4. DTFT is double sided. DFT is single sided.
5. The sequence X(n) need not be periodic. The sequence X(n) is assumed to be
periodic.
N
6. Number of Complex multiplications: N log10  
2
N N
Number of Complex Addition: log10  
2 2
7. A finite impulse response (FIR) filter is a filter whose impulse response (or response to any finite length
input) is of finite duration, because it settles to zero in finite time.
8. The advantage of IIR filters over FIR filters is that IIR filters usually require fewer coefficients to
execute similar filtering operations, that IIR filters work faster, and require less memory space.
FIR filter: These filters can be easily designed to have perfectly linear phase.
FIR filters can be realized recursively and non-recursively. Greater flexibility to control the shape of their
magnitude response. Errors due to round off noise are less severe in FIR filters, mainly because feedback is
not used.
IIR filter: These filters do not have linear phase. IIR filters are easily realized recursively. Less flexibility,
usually limited to specific kind of filters.
The round off noise in IIR filters is more.
9. Digital signal processing algorithms typically require a large number of mathematical operations to be
performed quickly and repeatedly on a series of data samples. Signals (perhaps from audio or video
sensors) are constantly converted from analog to digital, manipulated digitally, and then converted back to
analog form. Architecture has two separate memories for their instruction and data. It is capable of
simultaneous reading an instruction code and reading or writing a memory or peripheral.
10. Rounding off error is the process of reducing the size of a binary number to finite word size of b-bits
such that the rounded b-bit number is closest to the original unquantized number. The rounding process
consists of truncation and addition.
PART – B
11 a) Sampling Process:

xa(t) = x(t) δT(t)  T (t )    (t  nT )
n 

xa (t )   x(nT ) (t  nT )
n 
t = nT with weight x(nT).

  
F [ xa (t )]  F   x(nT ) (t  nT ) 
 n  

  x(nT ) F[ (t  nT )]
n 

F [ (t  nT )]    (t  nT )e  jwon dt
n 
F [ (t  nT )]  e  jwonT
11 b) i)
(i) rounding: Discard the excess bit by rounding the resulting number. Quantization error eq(n) is
 
rounding to the range  eq (n)  where ∆ is the distance between two successive quantization level.
z 2
If Xmin and Xmax represent the minimum and maximum value of X(n) and L is the number of quantization
x  xmin
levels, then   max ; xmax – xmin is the dynamic range of the signal.
L 1
(II) truncation: Discard the excess bit.
11 b) (ii) Signal to quantization error (SNQR):
eq(t) = xa(t) – xq(t)
1  2 1 
Pq 
2 
eq (t )dt   cq2 (t ) dt eq (t )  ( /2 )t ,  x  t   ,
 o
1    2
Pq     r 2 dt 
 0  2  12
A2 /3
the quantization step is ∆ = 2A/2b. Hence Pq 
22b
1 Tp A2
  A cos  r 
2
Px  0 dt 
Tp 0 2
The quantity of the output of the A/D converter is usually measured by the signal-to-quantization noise
ratio (SQNR), which provides the ratio of the signal power to the noise power:
Px 3 2b
SQNR   2
Pa 2
Expressed in decibels (dB), the SQNR is
SQNR (dB) 10log10 SQNR = 1.76 + 6.02b

12 a) i) Inverse Z – transform:
1 A B
X (Z )   
1  1.5Z 3  0.5Z 2 Z  0.5 Z  1
zz – 1.5z + 0.5 =0
d B
 Z – 0.5 Z –1  0  
Z  0.5 Z  1
A = –0.5; B = 2
x (n) = –0.5 * (0.5)n u(n) + 2 * (1)n u(n)
z
ROC Z > 0.5 means x( z ) 
z  0.5
The value of Z is placed on outside unit circle. ROC: it is causal system.
For x(n) = 0.5n u(–n) means ROC z < 0.5
The value of Z is placed inside the unit circle. It is anti-causal signal.
12 a) ii) causal and anti – causal signal:
(i) Finite Duration, right sided signal – Causal signal – outside the unit circle –ROC
(ii) Finite duration left side signals – Anticausal signal – Inside the unit circle –ROC
(iii) Finite duration, two sided (non causal signal) – Between the unit circle –ROC
(iv) Infinite duration right sided signal – Causal signal – Inside the Circle.
12 b) Discrete Fourier Transform:
(i) DTFT of the given signal is formulated as below
X (e j )   n  x(n)e j n   n  a|n| e j n

 
1  a2 1  a2
X (e j )  
 e j  e  j  2 1  2a cos   a 2
1  2a  a
 z 
(ii) DTFT of given signal:
 A,  M  n  M
X ( n)  
0, otherwise
 e j  e  j ( N 1)   e j  e j ( N 1) 
X (e j )  A  j   A   j 
 1 e   1 e 
z cos  N  2 cos  ( N  1)
A
z (1  cos  )
13 a) i)
Solution: x(n) = {3, 2, 1, 0, 1, 2}
DFT equation is X ( K )   n  0 x(n)e – j  2 / 6 nk

N 1 5
X(0) = 9, X(3) = 1,
X(1) = 4, X(4) = 0,
X(2) = 0, X(5) = 4
X(K) = {9, 4, 0, 1, 0, 4}
13 a) ii) The formulas for the DFT and IDFT is given by,
N 1
X (k )   x(n)WNkn k  0,1, , N  1
n0
N 1
1
X ( n) 
N
 x(k )W
n 0
 kn
N n  0,1, , N  1
 x(0)   X (0) 
 x(1)   
xN    X   X (1) 
   N
  
   
 x( N  1)   X ( N  1) 
1 1 1  1 
1 W W 2
 W N 1 
 N N N 
WN    WN2 WN4  WN2( N 1) 
 
     
1 WNN 1 WN2( N 1)  WN( N 1)( N 1) 
WN = e–j2x/N
N-point DFT
XN = WNxN
xN  WN1 X N
1 n
IDFT. xN  WN X N
N
1 n
WN1  WN
N
WN WNn  NI N
13 b) DIT FFT algorithm:
Refer 12 b) – CS2403 – N/D10
14 a)
  j 3 
e for  
3
Soln: Given that H d ( )  
0 
  
 3
   

sin (n  3)   
  3 
1. hd ( )  3 e  j 3 e jn d   for n = 0, 1, 2, 4, 5, 6
3 (n  3)
   
sin (n  3)   
  3 
2. for n = 0, 1, 2, 4, 5, 6
(n  3)
1/3 for n = 3
3. hd(0) = 0 hd(1) = 0.1378 hd(2) = 0.2756

hd(3) = 0.333 hd(4) = 0 hd(5) = 0.1378
hd(6) = 0.2756
4. Hamming window function is
 2 n   N 1   N 1 
WHamm (n)  0.54  0.46 cos   for –  n 
 N 1   2   2 
0 otherwise
 2 n
0.54  0.46 cos for n  0,1, M  1
wH (n)   M 1
0 otherwise
5. WHamm(0) = 0.08 = WHamm(–3); WHamm(1) = 0.31 = WHamm(–2) = 0.31; WHamm(2) = 0.77 = WHamm(–1)
WHamm(3) = 1
h( z )  hd (3)   n  0 ( z – n  z n )
2
6.
7. Filter coeffients are h(n) = hd(0) * WHamm(n)
h(0) = h(6) = 0.33; h(1) = h(5) = 0.0264; h(2) = h(4) = 0.085; h(3) = 0
14 b)
Soln: Pass band rippled αp ≤ 1dB

Stop band riple αs ≥ 40dB
Pass band cutoff frequency is 4KHz
Stop band cutoff frequency is 6 KHz
Sampling rate = 24KHz
First obtain the frequency in radiance wp = 2π * 4000 = 8000π rad/sec
Stop band frequency ws = 2π * 6000 = 12000π rad/sec
Sampling Time Ts = 1/fs = 1/24000 sec.
Apply the warping equation
Ωp = (2/T)tan((wp*T)/2) = 48000*tan((8000*π)/24000)*(4.15 × 10–5)) = 1046.4 rad/sec

Ωs = (2/T)tan((ws*T)/2) = 48000*tan((12000*π)/24000)*(4.15 × 10–5)) = 1572 rad/sec
Ωs/Ωp = 1.5
₤2 = .25
Order of the filter:
 1   1 
log  2  1  2  1 
 As  
1   Ap 
N  12
2  s 
log 
  
 p
Cutoff frequency is
 
 
1 p s 
c   1
 1

2 
 1 2N  1 2N
  2  1  2  1 
  Ar   A3  
Poles of HA(s) are given as, pt = ± Ωc ej(N + 2k + j)x/2N k = 0, 1, … N–1
The Transfer function is s-domain
Conversion of the s-domain equation int z – domain using Bilinear transformation.

15 a)
Notes: All registers and data lines are 16-bits wide unless otherwsie specified.
† Not available on all devices.
15 b) Quantization effects in designing digital filters:
1. Rounding effects in multiplication – Refer 14 a) – A/M11 – EC2302

2. Dead band effects – Refer 13 a) ii) – M/J09 – EC1302
3. Overflow effects in addition – Refer 13 a) ii) – N/D09 – EC1302
4. Limit cycle oscillations – i) zero input limit cycle oscillation – Refer 13 a) i) – M/J09 – EC1302
ii) Overflow limit cycle oscillation – Refer 13 a) ii) – N/D09 – EC1302
5. Signal Scaling – Refer 13 b) i) – A/M11 – IT1252
MAY/JUNE 2012
Fifth Semester
Time : Three hoursMaximum : 100 marks
Answer ALL questions.
PART A (10 × 2 = 20 marks)
1. Define Nyquist rate.
2. State and prove the time reversal property of Fourier transform.
3. Consider the signal x(n) = |1| for –1 ≤ n ≤ 1 and 0 for all other values of
n, sketch the magnitude and phase spectrum.
4. Find the convolution for x(n) = {0, 1, 0, 2} and h(n) = {2, 0, 1}.
5. Differentiate IIR and FIR filter.
6. Give relationship between DTFT and Z transform. What is meant by

quantization ever?
7. State warping and give the necessity of prewarping.
8. Define the condition for stability of digital filters.
9. Define periodogram.
10. Define Gibbs phenomena.
PART B (5 × 16 = 80 marks)
11. (a) Check for following systems are linear, causal, time in variant,
stable, static.
 1
(i) y(n) = x  
 2n 

Digital Signal Processing (May/June 2012) 2.13
(ii) y(n) = sin(x(n))

(iii) y(n) = x(n) cos(x(n))

(iv) y(n) = x(–n + 5)

(v) y(n) = x(n) + nx(n + 2). (16)
Or
(b) Compute linear and circular convolution of the two sequence x1(n) =
{1, 2, 2, 2} and x2(n) = {1, 2, 3, 4}.
12. (a) (i) Determine the system function and the unit sample response of
1
the system described by the difference equation y(n) + y(n – 1)
2
+ 2x(n).(8)
(ii) Determine the step response of the system y(n) – αy(n – 1) +
x(n), –1 < α < 1, when the initial condition is y(–1) = 1. (8)
Or
(b) An filter system is described by the difference equation y(n) = x(n) +
x(n – 10).
(1) Compute and sketch its magnitude phase response.
π
(ii) Determine its response to the input x(n) = cos n + 3sin
10
π π 
 n + n . (16)
3 10
13. (a) (i) Derive the computational equation for the 8-point FFT DIT. (8)
(ii) State and prove any five properties of DFT. (8)
Or
(b) Find the X(K) for the given sequence x(n) = {1, 2, 3, 4, 1, 2, 3, 4}.
(16)
14. (a) For the analog transfer function H(s) = 2/(s + 1) (s + 3) determine
H(z) using bilinear transformation. With T = 0.1 sec. (16)
Or
1 π / 4 ≤ | ω | < π
(b) Design an ideal high pass filter with Hd(ejw) =  using
0 | ω | ≤ π / 4
Hamming window with N = 11. (16)

15. (a) (i) Explain the addressing formats in the DSP processors. (8)
(ii) Draw the architecture of the DSP processor and explain. (8)
Or
(b) (i) Explain the functional modes present in the DSP processor. (8)
(ii) Explain about pipelining in DSP. (8)

APRIL/MAY 2011
Fifth Semester
PART A (10 × 2 = 20 marks)
1. State the advantages of FFT over DFTS.
2. What is meant by bit reversal?
3. Why do we go for analog approximation to design a digital filter?
4. Give any two properties of chebyshev filters.
5. State the properties of FIR filters.
6. What is meant by Gibb’s Phenomenon?
7. What is meant by fixed point arithmetic? Give example?
8. Explain the meaning of limit cycle oscillator?
9. State the various applications of DSP.
10. What is echo cancellation?
PART B (5 × 16 = 80 marks)
11. (a) With appropriate diagrams describe
(i) Overlap-save method
(ii) Overlap-add method
Or
(b) Explain radix-2 DIF FFT algorithm. Compare it with DIT-FFT
algorithms.

12. (a) Explain in detail butterworth filter approximation

Or
(b) Explain the bilinear transform method of IR filter design. What is
warping effect? Explain the poles and zeros mapping procedure
clearly.
⎛ 2⎞ ⎛ 2⎞
13. (a) Realize the system function H ( z ) = ⎜ ⎟ z + 1 + ⎜ ⎟ z −1 by linear
⎝ 3⎠ ⎝ 3⎠
phase FIR structure.
Or
(b) Explain the designing of FIR filters using windows.
14. (a) Explain the quantization process and errors introduced due to
quantization.
Or
(b) (i) Explain how reduction of product round-off error is achieved in
digital filters.
(ii) Explain the effects of coefficient quantization in FIR filters.
15. (a) (i) Explain how various sound effects can be generated with the
help of DSP.
(ii) State the applications of multirate signal processing.
Or
(b) (i) Explain how DSP can be used for speech processing.
(ii) Explain in detail about decimation and interpolation.

Solutions
PART A
1. (i) The FFT is used to compute DFT with reduced number of
calculations.
(ii) The FFT reduces the computational time and complexity to com-
pute DFT.
2. In DIT algorithm we can find that for the output sequence to be in a natu-
ral order where the input sequence has to be stored in a shuffled order.
For an eight point DIT algorithm the input sequence is in the order x(0),
x(4), x(2), x(6), x(1), x(5), x(3) and x(7). When N is a power of 2, the
input sequence must be stored in bit-reversal order for the output to be
computed in a natural order.
3. (i) Computational time is less

(ii) Less complexity and simple.
4. (i) The magnitude response of the Chebyshev filters exhibits ripple

either in passband or in stopband according to type.
(ii) The poles of the chebyshev filter lie on an ellipse.
5. (i) FIR filter is always stable

(ii) A realizable filter can always be obtained
(iii) FIR filter has a linear phase response.
6. In FIR filter design by fourier series method (or rectangular window

method) the infinite duration impulse response is truncated to finite dura-
tion impulse response. The abrupt truncation of impulse response intro-
duces oscillations in the passband and stopband. This effect is known as
Gibb’s phenomenon (or Gibb’s oscillation).
7. Those in which every element in the set has the same number of binary
digits and in which every element in the set has the binary point at the
same position, i.e., the binary point is fixed. These representations are
called “fixed-point arithmetic.”
8. In recursive systems when the input is zero or some nonzero constant

value, the nonlinearities due to finite precision arithmetic operations may
cause periodic oscillations in the output. These oscillations are called
limit cycle oscillations.

9. DSP used in
• in spectral analysis
• in channel vocoders
• in homomorphic processing systems
• in speech synthesisets
• in linear prediction systems.
10. Echo cancellation is widely used in data modems and in telephone

exchanges for echo reduction.
PART B
11. (a) (i) Overlap-Save method
Let the length of an input sequence be LS and the length of an
impulse response in M. In this method the input sequence is
divided into blocks of data of size N=L+M−1. Each block con-
sists of east (M−1) data points of previous block followed by L
new data points to form a data sequence of length N=L+M−1.
For first block of data the first M−1 points are set to zero. Thus
the blocks of data sequence are
x1 ( n) = {0, 0, 0,...0, x(0), x( n)... x( L − 1)}.

⎪ ⎪
⎪ ⎪
⎪ ⎪
x2 ( n) = ⎨ x ( L − M + 1) ,... x( L − 1), x ( L ) ... x(2 L − 1) ⎬

⎪ ⎪
⎪ Last ( M-1) data points L new data points ⎪
⎪ from x ( n ) ⎪
⎩ 1 ⎭
⎧ ⎫
⎪ ⎪
x3 ( n ) = ⎨ x ( 2 L − M + 1) ,... x(2 L − 1), x ( 2 L ) ... x(3L − 1) ⎬

⎪⎩ Last ( M −1) data points from x2 ( n) ⎪⎭
L new data points
and so on.
Now the impulse response of the FIR filter is increased in length
by appending L−1 zeros and N-point circular convolution of
xi(n) with h(n) is computed.
(i.e ) yi (n) = xi (n) N h (n) .

Digital Signal Processing (April/May 2011) 5.7
In yi ( n ) , the first (M−1) points will not agree with the lin-
ear convolution of xi ( n) and h( n) because of aliasing, while
the remaining points are identical to the linear convolution.
Hence we discard the first M−1 points of the filtered section
xi ( n) N h ( n) . The remaining points from successive sections
are then abutted to construct the final filtered output.
For example, Let the total length of the sequence LS = 15 and
the length of the impulse response is 3. Let the length of each
block is 5.
Now the input sequence can be divided into blocks as
x1 ( n) = {0,0, x (0 ) , x (1) , x (2)}
M − 1 = 2 Zeros.
⎧ ⎫
⎪ ⎪
x2 ( n) = ⎨ x (1) , x (2) , x (3) , x ( 4 ) , x (5)⎬

⎪⎩last two data points from previous blocks ⎪⎭
⎧last two data

points from previous blocks
⎫
⎪ ⎪
x3 ( n) = ⎨ x ( 4 ) , x (5 ) , x (6) , x (7) , x (8)⎬
⎩⎪ ⎭⎪
x4 ( n) = { x (7) , x (8) , x (9) , x (10 ) , x (11)}
x5 ( n) = { x (10 ) , x (11) , x (12) , x (13) , x (14 )}
x6 ( n) = { x (13) , x (14 ) ,0, 0,0}
Now we perform 5 point circular convolution of xi ( n) and h( n)

by appending two zeros to the sequence h(n). In the output block
yi ( n ) , first M−1 points are corrupted and must be discarded.
⎧ ⎫
⎪ ⎪
y1 ( n) = x1 ( n) Ν h ( n) = ⎨ y (0 ) , y (1) , y (2) , y (3) , y ( 4 )⎬
⎪⎩
1 1 1 1 1
⎪⎭
⎧⎪ ⎫⎪
y2 ( n) = x2 ( n) Ν h ( n) = ⎨ y2 (0 ) , y2 (1) , y2 (2) , y2 (3) , y2 ( 4 )⎬
⎪⎩ ⎪⎭
⎧⎪ ⎫⎪
y3 ( n) = x3 ( n) Ν h ( n) = ⎨ y3 (0 ) , y3 (1) , y3 (2) , y3 (3) , y3 ( 4 )⎬
⎪⎩ ⎪⎭

⎧⎪ ⎫⎪
y4 ( n) = x4 ( n) Ν h ( n) = ⎨ y4 (0 ) , y4 (1) , y4 (2) , y4 (3) , y4 ( 4 )⎬

⎩⎪ ⎭⎪
⎧⎪ ⎫⎪
y5 ( n) = x5 ( n) Ν h ( n) = ⎨ y5 (0 ) , y5 (1) , y5 (2) , y5 (3) , y5 ( 4 )⎬
⎪⎩ ⎪⎭
⎧⎪ ⎫⎪
y6 ( n) = x6 ( n) Ν h ( n) = ⎨ y6 (0 ) , y6 (1) , y6 (2) , y6 (3) , y6 ( 4 )⎬

⎩⎪ discarded ⎭⎪
The output blocks are abutted together to get
⎪⎧ y1 (2) , y1 (3) , y1 ( 4 ) , y2 (2) , y2 (3) , y2 ( 4 ) , y3 (2) , y3 (3) ,⎪⎫

y ( n) = ⎨ ⎬
⎪⎩ y3 ( 4 ) , y4 (2) , y4 (3) , y5 (2) , y5 (3) , y5 (6) , y6 (2) , y6 (3) ⎪⎭

Let the length of the Sequence be Ls and the length of the
impulse response is M. The sequence is divided into blocks of
data size having length L and M−1 zeros are appended to it to
make me data size of L+M−1.
Thus the data blocks may be represented as,
⎧⎪ ⎫⎪
x1 ( n) = ⎨ x (0 ) , x (1) ,.... x ( L − 1) , 0,0,...
⎬
⎪⎩ M −1 zeros appended ⎪⎭
⎧⎪
M −1zeros appended
⎫⎪
x2 ( n) = ⎨ x ( L ) , x ( L + 1) ,.... x (2 L − 1) , 0,0,... ⎬
⎩⎪ ⎭⎪
⎧⎪ ⎫⎪
x3 ( n) = ⎨ x (2 L ) , x (2 L + 1) ,.... x (3L − 1) , 0,0,...
⎬
⎪⎩ M −1 zeros appended ⎪⎭
Now L−1 zeros are added to the impulse response h(n) and
N-point circular convolution is performed. Since each block is
terminated with M−1 zeros, the last M−1 points from each out-
put block must be overlapped and added to the first M−1 points
of the succeeding block. Hence this method is called overlap-
add method.

Let the output blocks are of the form,

y1(n) = {y1(0), y1(1),……., y1(L−1), y1(L),…..y1(N−1)}
y2(n) = {y2(0), y2(1),….. y2(L−1), y2(L),……y2(N−1)}
y3(n) = {y3(0), y3(1),….. y3(L−1), y3(L),……y3(N−1)}
The output sequence is,
y(n) = {y1(0),y1(1),…..y1(L−1),y1(L) + y2(0),…..y1(N−1)+
y2(M−2), y2(M),…..y2(L) + y3(0), y2(L+1) + y3(1)…….y3(N−1)}
(b) DIT-FFT Algorithms

Let x(n) be an 8-point sequence. Therefore N = 8. The samples of
x(n) are, x(0), x(1), x(2), x(3), x(4), x(5), x(6), x(7).
First Stage Computation
In the first stage of computation, two numbers of 4-point sequences
g1(n) and g2(n) are obtained from x(n) as shown below.
⎡ ⎛ N ⎞⎤
g1 ( n) = ⎢ x( n) + x ⎜ n + ⎟ ⎥ = [ x( n) + x( n + 4)]; for n = 0,1, 2,3.
⎣ ⎝ 2 ⎠⎦
when n = 0; g1(n) = g1(0) = x(0) + x(4)

when n = 1; g1(n) = g1(1) = x(1) + x(5)
when n = 2; g1(n) = g1(2) = x(2) + x(6)
when n = 3; g1(n) = g1(3) = x(3) + x(7)
⎡ ⎛ N ⎞⎤
g2 ( n) = ⎢ x( n) − x ⎜ n + ⎟ ⎥ WNn = [ x( n) − x( n + 4)] W8n ;
⎣ ⎝ 2 ⎠⎦
for n = 0,1, 2,3.
when n = 0; g2 ( n) = g2 (0) = [ x(0) − x(4)]W8
0
when n = 1; g2 ( n) = g2 (1) = [ x(1) − x(5)]W8

1
when n = 2; g2 ( n) = g2 (2) = [ x(2) − x(6)]W8

2
when n = 3; g2 ( n) = g2 (3) = [ x(3) − x(7)]W8

3
Second Stage Computation

In the second stage of computation, 2 numbers of 2-point sequences
d11(n) and d12(n) are generated from the samples of g1(n), and another
2 numbers of 2-point sequences d21(n) and d22(n) are generated from
the samples of g2(n) as shown below.

⎛ N⎞
d11 ( n) = g1 ( n) + g1 ⎜ n + ⎟ = g1 ( n) + g1 ( n + 2); for n = 0,1.
⎝ 4⎠
when n = 0; d11(n) = d11(0) = g1(0) + g1(2)

when n = 1; d11(n) = d11(1) = g1(1)+ g1(3).
⎡ ⎛ N ⎞⎤
d12 ( n) = ⎢ g1 ( n) − g1 ⎜ n + ⎟ ⎥ WN0 = ⎡⎣ g1 ( n) − g1 ( n − 2) ⎤⎦ W4n ;
⎣ ⎝ 4 ⎠⎦ 2
for n = 0,1.
when n = 0; d12 ( n) = d12 (0) = [ g1 (0) − g1 (2)]W40
when n = 1; d12 ( n) = d12 (1) = [ g1 (1) − g1 (3)]W41
d21 ( n) = g2 ( n) + g2 ( n + N 4 ) = g2 ( n) + g2 ( n + 2) ; for n = 0,1.
when n = 0; d21 ( n) = d21 (0) = [ g2 (0) + g2 (2)]

when n = 1; d21 ( n) = d21 (1) = [ g2 (1) + g2 (3)]
d22 ( n) = ⎡⎣ g2 ( n) − g2 ( n + N
4 )⎤⎦ WN
n
= ⎡⎣ g2 ( n) − g2 ( n + 2) ⎤⎦ W4n ;
2
for n = 0,1.
when n = 0; d22 ( n) = d22 (0) = [ g2 (0) − g2 (2)]W40

when n = 1; d22 ( n) = d22 (1) = [ g2 (1) − g2 (3)]W41
Third Stage Computation

In the third stage of computation, 2-point DFTS of the 2-point
sequences d11(n), d12(n), d21(n) and d22(n) are computed.
The 2-point DFT of the 2-point sequence d11(n) is computed as
shown below.
1
DFT {d11 ( n)} = D11 ( k ) = ∑ d11 ( n)W2kn ; for k = 0,1
n= 0
1
when k = 0; D11 (0) = ∑d
n= 0
11
( n)W20 = d11 (0) + d11 (1)
1
when k = 1; D11 (1) = ∑d
n= 0
11
( n)W2n = d11 (0)W20 + d11 (1)W21
= d11 (0)W20 + d11 (1)W21 W20 = ⎡⎣ d11 (0) − d11 (1) ⎤⎦ W20

Smilarey the 2-point DFTS of the 2-point sequences d12(n), d21(n)

and d22(n) are computed and the results are given below
D12 (0) = d12 (0) + d12 (1)
D12 (1) = ⎡⎣ d12 (0) − d12 ( n) ⎤⎦ W20
D21 (0) = d21 (0) + d21 (1)
D21 (1) = ⎡⎣ d21 (0) − d21 ( n) ⎤⎦ W20
D22 (0) = d22 (0) + d22 (1)
D22 (1) = ⎡⎣ d22 (0) − d22 ( n) ⎤⎦ W20 .
Combining the Three Stages of Computation

The final output Dij(k) gives the X(k). The relation can be obtained
as shown below.
X(2k) = G1(k); k = 0,1,2,3 X(2k + 1) = G2(k); k = 0,1,2,3

X(0) = G1(0) X(1) = G2(0)
X(2) = G1(1) X(3) = G2(1)
X(4) = G1(2) X(5) = G2(2)
X(6) = G1(3) X(7) = G2(3)
G1(2k) = D11(k); k = 0,1 G1(2k + 1) = D12(k); k = 0,1
∴ G1(0) = D11(0) ∴ G1(1) = D12(0)
G1(2) = D11(1) G1(3) = D11(1)
G2(2k) = D21(k); k = 0,1 G2(2k + 1) = D22(k); k = 0,1
∴ G2(0) = D21(0) ∴ G2(1) = D22(0)
G2(2) = D21(1) G2(3) = D22(1).
From above relations we get,

X(0) = G1(0) D11(0)
X(4) = G1(2) D11(1)
X(2) = G1(1) D12(0)
X(6) = G1(3) D12(1)
X(1) = G2(0) D21(0)
X(5) = G2(2) D21(1)
X(3) = G2(1) D22(0)
X(7) = G2(3) D22(1).

1 a+b 1
a A=a+b
1
1
WNK
b B = (a + b) WNK
–1 a–b
Basic butterfly or flow graph of DIF FFT
From the above we observe that the output is in bit reversed order.
In radix-2 FFT, the input is in normal order the output will be in bit
reversed order.
Flow graph for 8-point Radix-2 DIF FFT: If we observe the basic
computation performed at every stage, we can arrive at the follow-
ing conclusion.
• In each computation two complex numbers “a” and “b” are
considered.
• The sum of the two complex number is computed which forms a
new complex number “A”.
• Them start complex number “b” from “a” to get the term “a−b”.
The difference term “a−b” is multiplied with the phase factor or
twiddle factor “WNK ” to form a new complex number “B”.
The signal flow graph is also called butterfly diagram since it resembles
a butterfly. In radix-2 FFT, N 2 butterflies per stage are required to repre-
sent the computational process. The butterfly diagram used to compute
the 8-point DFT in a radix-2 DIF FFT can be arrived as shown below.
Flowgraph (or Butterfly diagram) for first stage of computation
1 1
x(0) x(0) + x (4) = g1(0)
1
1
x(1) x(1) + x (5) = g1(1)
1
1 1
x(2) x(2) + x (6) = g1(2)
1
1 1 1
x(3) x(3) + x (7) = g1(3)
1 w 0
8
x(4) (x(0) – x(4)) w80 = g2(0)
−1
1 w81
x(5) −1 (x(1) – x(5)) w81 = g2(1)
1 w82
x(6) (x(2) – x(6)) w82 = g2(2)
−1
1 w 3
8
x(7) (x(3) – x(7)) w83 = g2(3)
−1

Flow graph (or butterfly diagram) for second stage of computation

g1(0) g1(0) + g1(2) = d11(0)
1 1 1
1
g1(1) 1 g1(1) + g1(3) = d11(1)
1 w 4
0
g1(2) [g1(0) – g1(2)]w40 = d12(0)

–1
1 w 4
1
g1(3) [g1(1) – g1(3)]w41 = d12(1)

–1
1 1
g2(0) g2(0) + g2(2) = d21(0)
1
1 1
g2(1) g2(1) + g2(3) = d21(1)
1
w 4
0
g2(2) [g2(0) – g2(2)]w40 = d22(0)

–1
1 w 4
1
g2(3) [g2(1) – g2(3)]w41 = d22(1)

–1
Flowgraph (or butterfly diagram) for third stage of computation

1 1
d11(0) d11(0) + d11(1) = D11(0) = G1(0) = x(0)
w20
d11(1) [d11(0) – d11(1)]w20 = D11(1) = G1(2) = x(4)
–1
1 1
d12(0) d12(0) + d12(1) = D12(0) = G1(1) = x(2)
w20
d12(1) [d12(0) – d12(1)]w20 = D12(1) = G1(3) = x(6)
–1
1 1
d21(0) d21(0) + d21(1) = D21(0) = G2(0) = x(1)
w 2
0
d21(1) [d21(0) – d21(1)]w20 = D21(1) = G2(2) = x(5)

–1
1 1
d22(0) d22(0) + d22(1) = D22(0) = G2(1) = x(3)
d22(1) [d22(0) – d22(1)]w20 = D22(1) = G2(3) = x(1)

–1 w20

The combined flowgraph (or butterfly diagram) of all the three

stages of computation
1 1 1 1 1 1
x(0) x (0)
1 1 1 1 w20
x(1) x (4)
1 1 w40 −1 1 1
x(2) x (2)
−1 w20
1 1 w41
x (3) x (6)
−1
w80 1 w40 −1 1 1
x(4) x (1)
−1 w81 w20
1 1
x (5) x (5)
−1 w82 w40 −1
1 1
x (6) x (3)
−1 −1
x(7) x (7)
−1 w83 −1 w41 −1 w20
The flowgraph (or butterfly diagram) for 8-point DIF radix-2 FFT
Comparison of DIT and DIF

Differences
In DIT, the input is bit-reversed while the output is in natural order.
For DIF, the reverse is true. (ie) input is normal order, while output
is bit-reversed. However both DIT and DIF can go from normal to
shuffled data or vice-versa.
Considering the butterfly diagram, in DIF, the complex multiplica-
tion takes place after the add-subtract operation.
Similarities
Both algorithms require same number of operations to compute
DFT.
Both algorithms require bit-reversal at some place during computation.
12. (a) The popular methods of designing IIR digital filter involves the
design of equivalent analog filter and then converting the analog
filter to digital filter. Hence to design a butterworth IIR digital fil-
ter, first an analog butterworth filter transfer function is determined
using the given specifications. Then the analog filter transfer func-
tion is converted to a digital filter transfer function by using either
impulse invariance transformation or bilinear transformation.
Analog butterworth filter: The analog butterworth filter is
designed by approximating the frequency response using an error

function. The error function is selected such that the magnitude in

maximally flat in the passband and monotonically decreasing in
the stopband.
The magnitude response of low pass filter obtained by this approxi-
mation is given by
1
H a (Ω ) =
2
2N (1)
⎛ Ω⎞
1+ ⎜
⎝ Ωc ⎟⎠
We know that the frequency response H a ( Ω ) of any analog filter is

obtained by letting S = j Ω in the analog transfer function H a ( Ω ) .
Hence substituting Ω by s/j in equation (1) gives the system trans-
fer function.
1 1
∴ H a (S ) H a ( −S ) = 2N = N (2)
⎛ S j⎞ ⎛ S2 ⎞
1+ ⎜ 1+ ⎜ 2 2 ⎟
⎝ Ωc ⎟⎠ ⎝ j Ωc ⎠
In equation (2), when S Ωc is replaced by Sn (i.e. letting

Ωc =1 rad/sec) the transfer function is called normalized transfer
function.
∴ H a ( Sn ) H a ( − Sn ) =
1
1 + ( − Sn2 )
N
The transfer function of equation (3) will have 2N poles which are
given by the roots of the denominator polynomial. It can be shown
that the poles of the transfer function symmetrically lies on a unit
circle in S-plane with angular spacing of π N .
For a stable and casual filter the poles should lie on the left half
of S-plane. Hence the desired filter transfer function is formed by
choosing the N-number of left half poles. When N is even, all the
poles are complex and exist as conjugate pair. When N is odd, one of
the pole is real and other poles are complex and exist as conjugate
pair. Therefore the transfer function of butterworth filters will be
a product of second order factors. The analog filter transfer func-
tion of normalized and unnormalized butterworth low pass filters
are given below.
Normalized butterworth low pass filter transfer function

Let N be the order of the filter.

N /2 1
When N is even, H a ( S ) = π .
K =1 S n2 + bk Sn + 1
N −1
1 2 1
When N is odd, H a ( S ) = π
Sn + 1 K =1 S n2 + bk Sn + 1
⎛ ( 2k − 1) π ⎞
Where bk = 2sin ⎜ ⎟.
⎝ 2N ⎠
Unnormalized butterworth low pass filter transfer function
The Unnormalized transfer function is obtained by replacing
Sn by S/ Ωc , where Ωc is the 3-dB cut off frequency of the low
pass filter.
Let N be the order of the filter.
N /2 Ωc 2
When N is even, H a ( S ) = Kπ=1
S + bk Ωc S + Ω2c
2
N −1
Ωc 2 Ωc 2
When N is odd, H a ( S ) = π
S + Ωc K =1 S + bk Ωc + Ω2c
2
⎛ ( 2k − 1) π ⎞
Where bk = 2sin ⎜ ⎟.
⎝ 2N ⎠
Frequency response of butterworth filter

The frequency response of butterworth filter depends on the
order N. The magnitude response (frequency response) for dif-
ferent values of N are shown in figure. In the below figure, it
can be observed that the approximated magnitude response
approaches the ideal response as the value of N increases,
Order of the Filter
In butterworth filters the frequency response of the filter
depends on the order, N. Hence the order N has to be to satisfy
the given specifications.
Usually the specifications of the filter are given in terms of gain
or attenuation at a passband and stopband frequency.
Let A1 = Gain or magnitude at a passband frequency Ω1.

⏐H (Ω) ⏐
Ideal response
1.0
0.707
N =1
0.5
N=2
N=4
N = 10
ΩC Ω
A2 = Gain or magnitude at a stopband frequency Ω2.

Calculate a parameter N1 using equation 1 and correct it to
nearest integer. Choose N such that N ≥ N1.
1 ⎧⎪⎛ 1 ⎞ ⎛ 1 ⎞ ⎫⎪
log ⎨⎜ 2 − 1⎟ ⎜ A2 − 1⎟ ⎬
N1 =
2 ⎩⎪⎝ 1
A ⎠ ⎝ 1 ⎠ ⎪⎭
(1)
log
(Ω )
2
(Ω )
1
fig magnitude response of butterworth low pass filter for vari-

ous values of ‘N’.
Properties of butterworth filter
• The Butterworth filters are all pole designs (i.e. the zeros of
the filters exist at infinity).
• At the cut off frequency Ωc the magnitude of normalized but-
1 ⎛ 1 ⎞
terworth filter is ⎜⎝ ie H a (Ω ) = = 0.707⎟ .
⎠
2 2
• Hence the dB magnitude at the cutoff frequency will be 3dB
less than the maximum value.
• The filter order N completely specifies the filter.
• The magnitude is maximally flat at the origin.
• The magnitude is monotonically decreasing function of Ω.
• The magnitude response approaches the ideal response as
the value of N increases.

Design procedure for low pass digital butterworth IIR filter

Let A1 = Gain at a passband frequency w1
A2 = Gain at a stopband frequency w2
Ω1 = Analog frequency corresponding to w1
Ω2 = Analog frequency corresponding to w2.
• Choose either bilinear or impulse invariant transformation
• Calculate the ratio of Ω2/Ω1
Ω tan ω2 2
For bilinear transformation, 2 =
Ω1 tan ω1 2
Ω2 ω2
For impulse invariant transformation, =
Ω1 ω1
• Decide the order N of the filter. The order N should be greater
than or equal to N1, where N1 is given by
1 ⎪⎧⎛ 1 1 ⎞ ⎪⎫
log ⎨⎜ 2 − 1 − 1⎟ ⎬
⎪⎩⎝ A2 ⎠ ⎪⎭
2
2 A1
N1 =
⎛Ω ⎞
log ⎜ 2 ⎟
⎝ Ω1 ⎠
choose N such that, N ≥ N1

• Calculate the analog cut off frequency Ωc.
For bilinear transformations,

2 tan ω1 2
Ωc = 1 2N
T⎛ 1 ⎞
⎜⎜ 2 − 1 ⎟⎟
⎝ A1 ⎠
For impulse invariant transformation,
ω1 T
Ωc = 1 2N
⎛ 1 ⎞
⎜⎜ 2 − 1⎟⎟
⎝ A1 ⎠
• Determine the analog transformation of the filter.
• Using the chosen transformation, transform Ha(S) to H(z),
where H(z), is the transfer function of the digital filter.
• Realize the digital filter transfer function H(z) by a suitable
structure.

(b) Bilinear transformation

The Bilinear transformation technique is widely used to design the
high frequency filters like high pass (or) Bandpass filters.
The transformation of a stable analog filter results in a stable digital
filter as conformal mapping of all the poles in the left half of the
S-plane are mapped onto points inside the unit circle of Z-domain.
The system function of analog filter be,
b
H ( s) =
s+a
(1)
Y ( s) b
Let H ( s) = =
X ( s) s + a
⇒ sY(s) + aY(s) = bx(s)
Take laplace transform,
dy(t )
+ a y(t ) = bx(t )
dt
Integrate the above equation between the limits (nT − T) and nT.
nT nT nT
dy(t )
∫
nT −T
dt
dt + a ∫ y(t )dt = b ∫ x(t )dt
nT −T nT −T
The trapezoidal rule for numeric integration is given by,

nT
T
∫
nT −T
a(t )dt =
2
[a( nT ) + a( nT − T )]
aT
y( nT ) − y( nT − T ) + [ y( nT ) + y( nT − T )]
2
bT
= [ x( nT ) + x( nT − T )]
2
Take Z-Transform, then the system function of the digital filter is,
aT bT
Y ( z ) − z −1Y ( z ) + ⎡YT + z −1Y ( z )⎤⎦ = ⎡ X ( z ) + z −1 X ( z )⎤⎦
2 ⎣ 2 ⎣

aT bT
Y ( z ) (1 − z −1 ) + (1 + z −1 )Y ( z ) = (1 − z −1 ) X ( z )
2 2
bT
Y (z) (1 + z −1 )
= 2
X ( z ) 1 − z −1 + aT 1 + z −1
( ) ( )
2
b
H (z) = (2)
2 ⎛ 1 − z −1 ⎞
⎜ ⎟ + a
T ⎝ 1 + z −1 ⎠
Compare equation (1) and (2)
2 ⎛ 1 − z −1 ⎞
S= ⎜ ⎟
T ⎝ 1 + z −1 ⎠
Steps to be followed
• From the given specification, find prewarping analog frequencies
2 w
using formula and Ω = tan .
T 2
• Using the analog frequencies find H(S ) of the analog filter.
• Select the sampling rate of digital filter call it T second per sample.
2 (1 − z −1 )
• Substituting S = in transfer function in step 2.
T (1 + z −1 )
13. (a) Warping Effect

In bilinear transformation the relation between analog and digital
frequencies is non linear. When the s-plane is mapped in to z-plane
using bilinear transformation, this non-linear relationship introduces
distortion in frequency axis, which is called frequency warping.
(b) The desired frequency response Hd(ejw) of a filter is periodic in fre-
quency and can be expanded in a fourier series.
The resultant series is given by,
∞
H d ( e jω ) = ∑ h ( n) e
n =−∞
d
− jω n
π
1
Where hd ( n) =
2π −π
∫ H (e ω ) e ω
j j n
dw

and known as fourier coefficients having infinite length. One

possible way of obtaining FIR filter is to truncate the infinite
⎛ N −1 ⎞
fourier series at n = ± ⎜ ⎟ , where N is the length of the
⎝ 2 ⎠
desired sequence. But abrupt truncations of the fourier series
results in oscillation in the passband and stopband. These oscil-
lations are due to slow convergence of the fourier series and
this effect is known as the Gibb’s phenomenon. To reduce these
oscillations, the fourier coefficients of the filter are modified by
multiplying the infinite impulse response with a finite weighing
sequence w(n) called a window where
⎛ N −1 ⎞
w(n) = w(−n) ≠ 0 for |n| ≤ ⎜ ⎟
⎝ 2 ⎠
⎛ N −1 ⎞
= 0 for |n| > ⎜ ⎟.
⎝ 2 ⎠
After multiplying window sequence w(n) with hd(n), we get a
finite duration sequence h(n) that satisfies the desired magni-
tude response,
⎛ N −1 ⎞
h(n) = hd(n)w(n) for all n ≤ ⎜ ⎟
⎝ 2 ⎠
⎛ N −1 ⎞
= 0 for n > ⎜ ⎟.
⎝ 2 ⎠
The frequency response H(ejw) of the filter can be obtained by
convolution of Hd(ejw) and ω(ejw) given by
π
1
2π −∫π d
H(ejw) = H (e jθ )ω (e j (ω −θ ) )dθ (1)
= H d ( e jω ) * ω ( e jω )
Because both Hd(ejw) and W(ejw) are periodic functions the oper-
ations is often called as periodic convolution. The windowing
technique in shown in figure. The desired frequency response
and its filter coefficients are shown in fig(a) and (b) respectively.
The fig(c) and (d) show a finite window sequence w(n) and its
fourier transform w(ejw). The fourier transform of a window
consists of a central lobe and side lobes. The central lobe con-
sists most of the energy of the window. To get an FIR filter, the

Hd (e jw)
hd (n)
1.0
n
–wc o wc w
Fig (a) Fig (b)
w (n) H (e jw)
wR (e jw)
–p –2p/N o 2p/N p o w
Fig (c) Fig (d) Fig (e)
h(n) = hd(n)w(n) g (n)
N–1
n n
o
– (N – 1) (N – 1)
o
2 2
Fig (f) Fig (g)
Fig Windowing Technique
sequence hd(n) and w(n) are multiplied and a finite length of

non-casual sequence h(n) is obtained. The fig(f) and (e) show
h(n) and its fourier transform H(ejw). The frequency response
H(ejw) is obtained using equation (1). The realizable sequence
g(n) in fig(g) can be obtained by shifting h(n) by α number of
N −1
samples, where α = .
2
From the equation (1), we get that the frequency response of
the filter H(ejw)depends on the frequency response of window
w(ejw). Therefore, the window, chosen for truncating the infinite
impulse response should have some desirable characteristics.
They are
(1) The central lobe of the frequency response of the window
should contain most of the energy and should be narrow.

(2) The highest side lobes level of the frequency response should
be small.
(3) The side lobes of the frequency response should decrease in
energy rapidly as w tends to p.
14. (a) Quantization process

The common methods of quantization process are
(1) Truncation
(2) Rounding.
Errors introduced due to quantization (truncation and
Rounding)
If the quantization method is truncation, the number is approxi-
mated by the nearest level that does not exceed it. In this case the
error xT − x is negative or zero when xT is truncation value of x and
it is assumed |x |≤ 0.
The error made by truncating a number to b bits following the binary
point satisfies the inequality,
0 ≥ xT − x > −2 − b (1)
For example, consider the decimal number 0.12890625. Its binary

equivalent is 0.00100001. If we truncate the binary number to 4 bits,
we have xT = (0.0010)2 whose decimal value is 0.125.
Now the error ( xT − x ) = −0.00390625. Which is greater than
−2 − b = −2 −4 = −0.0625 satisfying the inequality given in eqn 1.
The equation (1) holds for both sign-magnitude, one’s complement
and two’s complement if x > 0. If x < 0, we have to find whether the
equation (1) holds for all types of representations.
Consider first the two’s complement representations, the magnitude
of the negative number is
b
x = 1 − ∑ ci 2− i
i =1
If we truncate to number to ‘N’ bits then

N
xT = 1 − ∑ ci 2− i
i =1

The change is magnitude

b N
xT − x = ∑ ci 2 − i − ∑ ci 2 − i
i =1 i =1
b
= ∑ ci 2 − i (2)
i =1
≥ 0.
From equation (2) we find that due to truncation the change in
magnitude is positive, which implies that error is negative and
satisfy the inequality.
0 ≥ xT − x > −2 − b (3)
For one’s complement representation the magnitude of negative

number with b bits is given by
b
x = 1 − ∑ ci 2 − i − 2 − b
i =1
When number is truncated to N bits, then

N
xT = 1 − ∑ ci 2 − i − 2 − N
i =1
The change in magnitude due to truncation is

b
xT − x = ∑ ci 2 − i − (2 − N − 2 − b )
i =1
<0
Therefore the magnitude decreases with truncation which
implies that error is positive and satisfy the inequality.
0 ≤ xT − x < 2 − b (4)
The equation (4) holds for sign magnitude representation also.
In floating point systems the effect of truncation is visible only
in the mantissa. Let the mantissa is truncated to N bits.
If
x = 2c. M , then
x T = 2c. M T
Error e = xT − x = 2c ( M T − M ) .

From equation (3), with two’s complement representation of

Mantissa, we have
0 ≥ M T − M > −2 − b
0 ≥ e > −2 − b 2c (5)
xT − x e
we define relative error ε = = .
x x
Now equation (5) can be written as
0 ≥ ε x > −2 − b 2c
or 0 ≥ ε 2c M > −2 − b 2c (or ) 0 ≥ ε M > −2 − b
1
If M = , the relative error is maximum.
2
Therefore, 0 ≥ ε > −2.2− b
1
If M = − , the relative error range is
2
0 ≤ ε < −2.2− b
If one’s complement representation the error for truncation of
positive values of the mantissa is
0 ≥ M T − M > −2 − b (or ) 0 ≥ e > −2 − b ⋅ 2c

with e = ε x = ε 2 M and M T = 1 2,
c
We get the maximum range of the relative error for positive

M as
0 ≥ ε > −2.2 − b (6)
For negative mantissa values the error is
0 ≤ M T − M < 2 − b (or ) 0 ≤ e < 2c ⋅ 2 − b
with M = −1/2. The maximum range of the relative error for

negative M is
0 ≥ ε > −2.2− b which is the same as positive M(6)
The probability density function p(e) for truncation of fixed
point and floating point numbers are shown in below

Fixed point P(e) is complement and

P(e) 2b/2 sign magnitude
2 b
Two’s complement
– 2–b o e – 2–b o 2–b e
Floating point P(e)

P(e)
Two’s
is complement and
complement
2b /4 2b /2 sign magnitude
– 2.2–b o 2.2–b e – 2.2–b o e
In fixed point arithmetic the error due to rounding a number to b

bits produces an error e = xT − x which satisfies the inequality.
2− b 2− b
− ≤ xT − x ≤ (7)
2 2
This is because with rounding, if the value lies half way between
two levels, it can be approximated to either nearest high level or
by the nearest lower level. For fixed-point number (equation (7))
satisfies regardless of whether sign-magnitude, one’s complement
is used for negative numbers.
In floating-point arithmetic, only the mantissa is affected by
quantization.
If x = M ⋅ 2c and xT = M T ⋅ 2c then
e = xT − x = ( M T − M ) 2c
(8)
But for rounding

2− b 2− b
− ≤ MT − M ≤ (9)
2 2
Using equation (8), equation (9) can be written as,
2− b 2− b
−2 − c ⋅ ≤ x T − x ≤ 2c ⋅
2 2
−b −b
2 2
(or ) − 2− c ⋅ ≤ ∑ x ≤ 2c ⋅
2 2

Fixed point
Floating point
P (e)
2b 2b P (e)
2
e
– 2–b o 2–b
2 2 – 2–b o 2–b e
Fig (a) Fig (b)
We have x = 2c ⋅ M
2− b
then − 2 − c ⋅ ≤ Σ 2c ⋅ M ≤ 2c ( 2 − b 2 )
2
2− b 2− b
Which gives ≤ ∑⋅ M ≤
2 2
The mantissa satisfies 1 2 ≤ M < 1 If M = 1/2 we get the maxi-
mum range of relative error −2 − b ≤ ∑ < 2 − b.
The probability density function for rounding is shown in figure
(a) and (b).
(b) (i) Product Quantization

In fixed point arithmetic, the product of two b-bit numbers
result in a number of length 2b-bits. If the word length of the
register used to store the result in b-bits then it is necessary to
quantize the product (result) to b-bits. In realization structures
of IIR system, multipliers are used to multiply the signal by
constants. The output of the multipliers (i.e.) the products are
quantized to finite word length in order to store them in registers
and to be used in subsequent calculations. The error due to the
quantization of the output of multiplier is referred to as product
quantization error.
In digital system the product quantization is performed by round-
ing due to the following desirable characteristics of rounding.
• In rounding the error signal is independent of the type of
arithmetic employed.
• The mean value of error signal due to rounding is zero.
• The variance of error signal due to rounding is the least.

In product quantization, the error analysis is necessary to define

the noise transfer function, which depends on the structure of
the digital network. The noise transfer function (NTF) is defined
as transfer function from the noise source to the filter output.
(i.e. NTF is the transfer function obtained by beating the noise
source as actual input).
The model of the multiplier of a digital network using fixed
point arithmetic is shown in figure. The multiplier is consid-
ered as an infinite precision multiplier. Using an adder the error
signal is added to the output of multiplier so that the output of
adder is equal to the quantized output. Therefore the output of
finite word length multiplier can be expressed as
Quantized product = Q[ax (n)] = ax (n)+ e(n)
where, ax (n) = unquantized product and
e (n) = product quantization error signal
The product quantization error signal is treated as a random
process with uniform probability density function.
Input Quantization
For processing of analog signal using a digital system the analog
signal has to be digitized by A/D converter. The A/D converter
consists of sampler and quantizer. The sampler will samples
the value of analog signal at uniform intervals to produce a
sequence of unquantized values of the signal. The quantizer
will quantize the analog value and produce the corresponding
binary codes. The process of assigning binary number to quan-
tized analog value is also called coding.
The two types of errors that are produced by A/D conversion
process are quantization errors and saturation errors. The quan-
tization error is due to representation of the sampled signal by a
fixed number of digital levels (quantization levels). The satura-
tion error occurs when the analog signal exceed the dynamic
range of A/D converter.
In sign magnitude or one’s complement representation of binary
numbers using b-bits binary code, we can generate 2b − 1 dif-
ferent binary numbers. In two’s complement representation of
binary numbers using b-bits binary code (including sign bit),
we can generate 2b different binary numbers. If the range of
analog signal and be quantized is R then the quantization step
size q is given by

R
Quantization step size, q = b – (for two’s complement
representation). 2
Usually the analog signal is sealed such that the magnitude of
quantized signal is less than or equal to one.
In such case the range of analog signal to be quantized is –1 to
+1 therefore R = 2.
Let x(n) = Unquantized sample of the signal and
xq(n) = Quantized sample of the signal.
Now the quantization error in defined as,
Quantization error, e(n) = xq(n) – x(n).
(ii) Quantization of Filter Coefficients

In the realization of FIR and IIR filters in hardware or in soft-
ware, the accuracy with which filter coefficients can be speci-
fied in limited by the word length of the register used to store
the coefficients. Usually the filter coefficients are quantized to
the word size of the register used to store them either by trunca-
tion or by rounding.
The location (or the value) of poles and zeros of the digital
filters directly depends on the value of filter coefficients. The
quantization of the filter coefficients will modify the value of
poles and zeros and so the location of poles and zeros will be
shifted from the desired location. This will create deviations, in
the frequency response of the system. Hence we obtain a filter
having a frequency response that is different from the frequency
response of the filter with unquantized coefficients.
The sensitivity of the filter frequency response characteristics
and quantization of the filter coefficients it minimized by real-
izing the filter having a large number of poles and zeros as
an interconnection of second-order sections. This leads to the
parallel form and cascade form realization in which the basic
building blocks are first order and second order sections. It is
possible to prove that the coefficient quantization has less effect
in cascade realization when compared to parallel realization.
15. (a) (i) The Digital signal processing has made an impact on several
aspects of audio engineering, which encompasses recording,
storage, transmission and reproduction of signals. The signals

Audio Multiplexer
APC Bit Packing Encoder and
Inputs and Lowpass
Unit and Buffer Modulator
Filters
Error Correction
Bit Generator
To Recording
Head
From Playback
Head
Audio Demultiplexer DAC Error Detection Ring Decorder and

Outputs and Lowpass Unit and Correction Buffer Demodulator
include natural and electronic music, theoretical performance,

natural sounds, cine songs.
The above figure shows the block diagram of a digital tape
recorder. Each input signal is applied to a low pass filter, which
eliminates high frequency noise. The output from the filter is sam-
pled and converted into digital words. The big streams are multi-
plexed. In addition, bit for timing, for correcting errors, passing
checking and block scaling are introduced. A modulator is used
to convert the composite bit stream into a series of analog pulses.
During reproduction, the signal is decoded unpacked and dis-
tributed to each output.
The advantages of digital recording are,
(1) High signal to noise ratio limited only by the analog to
digital converter.
(2) Absence of harmonic distortions.
(3) No interference cross task.
(4) Elimination of amplitude variations due to changes in mag-
netization of the tape.
(ii) There are many cases where multirate signal processing is used.
They are
(1) In high quality data acquisition and storage systems
(2) In audio signal processing.
(3) In video, PAL and NTSC run at different sampling rates.
Therefore to watch an American program in Europe, one
needs a sampling rate converter.

(4) In speech processing to reduce the storage space or the

transmitting rate of speech data
(5) In transmultiplexers.
(6) Narrowband filtering for feral ECG and EEG.
(b) (i) There are three main areas in speech processing: speech syn-
thesis, speech recognition and speech coding. In speech
Synthesis, a machine is developed which can accept as input a
piece of English text and convert it to natural sounding speech.
Applications of speech synthesis include speech output from
computers, interrogating database from an ordinary telephone,
permitting a doctor in remote location to access medical records
stored in a central computer and reading machine for the visu-
ally challenged.
In speech recognition, a system is produced which can rego-
nise a speech from any speaken of a given language. The main
application areas for speech recognition are telephone-banking,
direct control of machines by human voice, quality control,
voice input to computer for document creation.
Speech coding is concerned with the development of techniques
which exploits the redundance in the speech signal, in order to
reduce the number of bits required to present it. The main appli-
cation areas for speech coding are voice mail systems, cordless
telephone channel, narrow-band cellular radio, military com-
munications are secrecy missions.
(ii) Decimation
The sampling rate of a discrete-time signal x(n) can be reduced
by a factor M by taking every M-th value of the signal. The
block diagram representation of downsampler (or) decima-
tor is shown in figure. The quadratic symbol in below figure
with arrow pointing downwarder is called decimator (or) down
samples. The output signal y(n) is a downsampled signal of the
input signal x(n) can be represented by y(n) = x(Mn).
x (n) ↓m y (n) = x (Mn)
Upsampling (or) Interpolation

The sampling rate of a discrete-time signal can be increased
by a factor L by placing L−1 equally spaced zeros between

each pair of samples. Mathematically upsampling is repre-

sented by
⎧ ⎛n⎞ ⎫
⎪x , n = 0, ± L, ±2 L....⎪
y( n) = ⎨ ⎜⎝ L ⎟⎠ ⎬
⎪0 ⎪
⎩ , otherwise. ⎭
The block diagram representation of the interpolator (or)
upsamples is shown in below figure.
x (n) ↑L y (n)

NOV/DEC 2010
Fifth Semester
PART A (10 × 2 = 20 marks)
1. Obtain the circular convolution of the following sequences.
2. How many multiplications and additions are required to compute N-point

DFT using radix-2 FFT?
3. What is prewarping?
4. What is the advantage of direct form II realization when compared to

direct form I realization?
5. Give the equations for Hamming window and Blackman window
6. Determine the transversal structure of the system function

H(z) = 1 + 2z−1 − 3z−2 − 4x−3.
7. What is truncation?
8. What is product Quantization error?
9. What is decimation?
10. What is sub band coding?

PART B (5 × 16 = 80 marks)
11. (a) (i) Compute the eight-point DFT of the sequence
⎧1 1 1 1 ⎫
x ( n ) = ⎨ , , , , 0, 0, 0, 0 ⎬
⎩2 2 2 2 ⎭
(ii) Explain overlap-add method for linear FIR filtering of a long
sequence.
Or
(b) (i) Compute the eight-point DFT of the sequence.
⎧1, 0 ≤ n ≤ 7
x( n) = ⎨
⎩0, otherwise
By using the decimation-in-frequency FFT algorithm.
(ii) Summarize the properties of DFT.
12. (a) Determine the system function H(z) of the chebyshev’s low pass
digital filter with the specifications.
Or
(b) Obtain the direct form I, direct form II, cascade and parallel form
realization for the system
y(n) = −0.1y(n−1) + 0.2y(n−2) + 3x(n) + 3.6x(n−1) + 0.6x(n−2)
13. (a) (i) Design an ideal high pass filter with a frequency response
⎧ π
⎪⎪1 for 4 ≤| w| ≤ π
H d (e jw ) = ⎨
⎪0 for |w| ≤ π
⎪⎩ 4
Find the values of h(n) for N = 11 using Hamming window
Find H(z) and determine the magnitude response.
Or
(b) (i) Determine the coefficients {h(n)}of linear phase FIR filter of
length M = 15 which has a symmetric unit sample response and
a frequency response that satisfies the condition.
⎛ 2π k ⎞ ⎧1 for k = 0,1, 2,3

Hr = ⎜ =
⎝ 15 ⎟⎠ ⎨⎩0 for k = 4,5,6, 7

(ii) Obtain the linear phase realization of the system function
14. (a) Discuss in detail the errors resulting from rounding and truncation
Or
(b) Explain the limit cycle oscillations due to product round off and
overflow errors.
15. (a) Explain the polyphase structure of decimator and interpolator

Or
(b) Discuss the procedure to implement digital filter bank using multi-
rate signal processing.

Solutions
PART A
1. x(n) = {1,2,1}; h(n)={1, −2,2}
Solution: Given x(n) = {1, 2, 1}
h(n) = {1, −2, 2}
y(n) = x(n) ⊕ h(n).
⎡ y ( 0 )⎤ ⎡ x ( 0 ) x (2) x (1) ⎤ ⎡ h (0 )⎤
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ y (1) ⎥ = ⎢ x (1) x (0 ) x (2)⎥ ⎢ h (1) ⎥
⎢⎣ y (2)⎥⎦ ⎢⎣ x (2) x (1) x (0 )⎥⎦ ⎢⎣ h (2)⎥⎦
⎡1 1 2⎤ ⎡ 1 ⎤ ⎡ 3 ⎤
= ⎢⎢2 1 1 ⎥⎥ ⎢⎢ −2⎥⎥ = ⎢⎢ 2 ⎥⎥
⎢⎣1 2 1 ⎥⎦ ⎢⎣ 2 ⎥⎦ ⎢⎣ −1⎥⎦
∴ y ( n) = {3, 2, −1}
N
2. log 2 N multiplication and N log 2 N additions are needed to compute
2
N-point DFT using Radix-2 FFT
3. The effect of the non-linear compression at high frequencies can be com-

pensated. When the desired magnitude response is piece-wise constant
over frequency, this compression can be compensated by introducing a
suitable prescaling or prewarping the critical frequencies by using the
formula,
2 w
Ω= tan .
T 2
4. Number of delay elements is reduced in direct form-II realization

compared to direct form-I realization.
5. Hamming window:
⎛ 2π n ⎞
WH ( n ) = 0.54 + 0.46 cos ⎜ ⎟
⎝ N −1 ⎠

Blackman window:
⎛ 2π n ⎞ ⎛ 4π n ⎞
WB ( n ) = 0.42 + 0.5cos ⎜ ⎟ + 0.08cos ⎜ ⎟
⎝ N −1 ⎠ ⎝ N −1 ⎠
6. Given, H(z) = 1 + 2z−1−3z−2−4z−3
x(n) z –1 z –1 z –1
1 2 –3 –4
y(n) + + +
7. Truncation is the process of discarding all bits beyond the nth bit.
eg. let x = 0.011010100
If it is truncated to 4 bits, then x = 0.0110
8. Any b-bit number is multiplied by another b-bit number, then we can
get 2b-bit result. If the length of the register is b-bits, then the result is
to be quantized to b-bits, If it is realized, then multipliers are used. The
error due to the quantization of the output of the multiplier is known as
product quantization error.
9. The process of reducing the sampling rate by a factor D is known as
decimation.
10. Sub band coding is a process used to do the signal compression.
PART B
11. (a) (i) Using the radix-2 decimation-in-time-algorithm.
1 1 1/2 1 1 1 1 1
x (0) = 1/2 x (0) = 2
1/2 1 1 0.5 – j0.5 1 1 x (1) = 0.5 –

x (4) = 0
w 20 –1 j1.207
1 1 1/2 w4 0
0 1 1
x (2) = 1/2 x (2) = 0
–1 0.5 +j
w2 0
1/2 w4 0
0.5 1 1 x (3) = 0.5 –
x (6) = 0
–1 –1 j0.207
1 –1 1/2 1 1 1 w 80
x (1) = 1/2 x (4) = 0
w 81 –1
w 20 1/2 1 1 0.5 – j 0.5 x (5) = 0.5 +
x (5) = 0
–1 j0.207
–1
1 1 1/2 w4 0
0 w 82
x (3) = 1/2 x (6) = 0
–1 –1
w2 0
1/2 0.5 + j 0.5
0 x (7) = 0.5 +
x (7) = 0
–1 w 41 –1 w 83 –1 j1.207

Where
W20 = 1
W40 = 1
π
W40 = e − j2π ×1 4 = cos − jsin π 2 = − j;W80 = 1
2
W81 = e − j2π ×1/8 = 0.707 − j0.707;W82 = e − j2π ×2/8 = j
W83 = −0.707 − j0.707.
∴ X ( k ) = {2,0.5 − j1.207,0,0.5 − j0.207,0,0.5 + j0.207,0,
0.5 + j1.207}
(ii) Refer answer 11(a)(ii) from April/May 2011 Question paper.
(b) (i) DIF Algorithm

The twiddle factors associated with butterflies can be found as,
W80 = 1; W81 = 0.707 − j0.707; W82 = − j; W83 = −0.707 − j0.707.
∴ x (k) = {8,0,0,0,0,0,0,0}
1 2 4 1 1 w80
x(0) = 1 x(0) = 8
1 −1
1 2 1 1 4 1 w80
x(1) = 1 x(4) = 0
w80 −1
1 2 1 0 1 1
x(2) = 1 x(2) = 0
−1 w2
w80
1 2 1 0 8
x(3) = 1 x(6) = 0
0 −1 1 −1 1
0 w 8 1 0 1
x(4) = 1 x(1) = 0
−1 1 w80
0 w81 1 0
x(5) = 1 x(5) = 0
−1 w82 w80 −1
0 0 1 1
x(6) = 1 x(3) = 0
−1 −1 w2
0 0 8
x(7) = 1 x(7) = 0
−1 w83 −1 −1 w80

(ii)
Discrete time Discrete fourier
Property signal transformation
Linearity a1x1(n) + a2x2(n) a1x1(k) + a2x2(k)
Periodicity x(n + N) = x(n) x(k+N) = x(k)
Circular time shift x(n−m)N x(k)e−j2πkm/N
Time reversal x(N−n) x(N−k)
Conjugation x*(n) x*(N−K)
Circular frequency x(n)ej2πmn/N x(k−m)N
shift
12. (a) ap=1dB ripple in the pass band 0≤w≤0.2p.

as= 15dB ripple in the stop band 0.3p≤w≤p.
using bilinear transformation (assume T = 1sec).
Solution: Given data ap = 1dB; wp = 0.2π; as=15dB;ws=0.3p
Prewrap frequency values. Since we intend to employ the bilin-
eared transformation method, we must prewrap these frequen-
cies. The prewraped values are given by,
2 w 0.2π
Ωp= tan P = 2 tan = 0.65
T 2 2
2 w 0.3π
Ωs = tan s = 2 tan = 1.02
T 2 2
Value of N,
1.5−1
10 0.1α s − 1 cosh −1 10
−1
cosh 10 0.1−1
10 0.1α p − 1
N≥ ≥ = 3.01
Ω ⎛ 1.02 ⎞
cosh −1 s cosh −1 ⎜
Ωp ⎝ 0.65 ⎟⎠
Let us rate N = 4.

Axis of the Envelope

We know
ε = 100.1α p − 1 = 0.508
μ = ε −1 + 1 + ε −2 = 4.17
⎛ μ1/ N − μ −1/ N ⎞ ⎛ ( 4.17)1/4 − ( 4.17)−1/4 ⎞
a = Ωp ⎜ ⎟ = 0.65 ⎜ ⎟
⎝ 2 ⎠ ⎝ 2 ⎠
= 0.237
⎛ μ1/ N + μ −1/ N ⎞ ⎛ ( 4.17)1/4 + ( 4.17)−1/4 ⎞
b = Ωp ⎜ ⎟⎠ = 0.65 ⎜ ⎟
⎝ 2 ⎝ 2 ⎠
= 0.6918
φk =
π
+
(2k − 1) π , k = 1,2,3, 4
2 2N
φ1 = 112.5°, φ 2 = 157.5°, φ3 = 202.5°; φ4 = 247.5°
The poles are

Sk = a cos φk + jb sin φk ; k = 1, 2, 3, 4
S1 = a cos φ1 + jb sin φ1 = 0.237 cos 112.5° + j 0.6918 sin 112.5 °
= −0.0907 + j 0.639
S = a cos φ2 + jb sin φ2 = 0.237 cos 157.5 ° + j 0.6918 sin 157.5 °
2
= −0.2189 + j 0.2647
S3 = a cos φ3 + jb sin φ3 = 0.237 cos 202.5 ° + j 0.6918 sin 202.5 °
= −0.2189 – j 0.2647
S = a cos φ4 + jb sin φ4 = 0.237 cos 247.5° + j 0.6918 sin 247.5°
4
= –0.0907 – j 0.639
The denominator polynomial of
H(s) = [(s + 0.0907)2 + (0.639)2][(s + 0.2189)2 + (0.2647)2]
= (s2 + 0.1814s + 0.4165) (s2 + 0.4378s + 0.118)
As N is even, the numerator of H(s) =
⎛ ( 0.4165 )( 0.118 ) ⎞
⎜ ⎟ = 0.04381
⎝ 1+ ε 2 ⎠

The transfer function H(s) =

0.04381
( s + 0.1814 s + 0.4165)( s2 + 0.4378s + 0.118)
2
The z-transform of the digital filter,
2 ⎛ 1 − z −1 ⎞
H ( z ) = H ( s)/ s = ⎜ ⎟
T ⎝ 1 + z −1 ⎠
0.04381
H ( z) =
⎛ 4 ⎛1− z ⎞ −1 2
⎛ 1 − z −1 ⎞ ⎞
⎜ ⎟
⎜ 1 ⎝ 1 + z −1 ⎠ + 0.1814 × 2 ⎜⎝ −1 ⎟
+ 0.4165⎟
⎝ 1+ z ⎠ ⎠
⎛ 4 ⎛ 1 − z −1 ⎞ 2 ⎛ 1 − z −1 ⎞ ⎞
⎜ 1 ⎜⎝ 1 + z −1 ⎟⎠ + 0.4378 × 2 ⎜⎝ 1 + z −1 ⎟⎠ + 0.118⎟
⎝ ⎠
0.4381(1 + z −1 )
4
=
(4.7794 − 7.1668 z −1 + 4.0538 z −2 ) (4.9936 − 7.764 z −1 + 3.2424 z −2 )
0.01836 (1 + z −1 )
4
H ( z) =
(1 − 1.499 z + 0.8482 z −2 )(1 − 1.5548 z −1 + 0.6493z −2 )
−1
(b) Y(z) (1 + 0.1z−1−0.2z−2) = x(z)(3 + 3.6z−1 + 0.6z−2)
y( z ) 3 + 3.6 z −1 + 0.6 z −2
=
x( z ) 1 + 0.1z −1 − 0.2 z −2
Direct Form I
3
x(z) + + y(z)
z –1 z –1
3.6 – 0.1
+ +
z –1 z –1
0.6 0.2

Direct Form II
x(z) + + y(z)
– 0.1 z –1 3.6
+ +
–1
z
0.2 0.6
Cascade Structure
3 z 2 + 3.6 z + 0.6 3( z + 0.2)( z + 1)
H ( z) = =
z 2 + 0.1z − 0.2 ( z − 0.4)( z + 0.5)
3 1 1
x(z) + + + + y(z)
z –1 z –1
1
0.4 0.2 – 0.5

H 1(z) H 2(z)
1 + 0.2z −1 1 + z −1
Where H1 (z ) + and H 2 ( z ) =
1 − 0.4 z −1
1 + 0.5 z −1
Parallel Structure
H ( z) A B C
= + +
Z Z Z − 0.04 Z + 0.5
3 ( z + 0.2)( z + 1) A ( z − 0.4 )( z + 0.5) + B ( z )( z + 0.5) + (( z )( z − 0.4 ))
=
z ( z − 0.4 )( z + 0.5) z ( z − 0.4 )( z + 0.5)
z = 0.4, B (0.4 )(0.9) = 3 (0.4 + 0.2)(0.4 + 1)
B=7
z = −0.5C ( −0.5)( −0.9) = 3 ( −0.3)( −0.5 + 1)
C = −1
If z = 0, A ( −0.4 ) = 3 (0.2)(1)

A = −1.5
−1.5 z 7z 1z
∴ H ( z) = + −
z z − 0.4 z + 0.5
7 1
H ( z ) = −1.5 + −
1 − 0.4 z −1
1 + 0.5 z −1
– 1.5
7
x(z) + + + + y(z)
z –1
0.4 –1
+ +
z –1
– 0.5
13. (a) (i) The Hamming window sequence is given by,

2π n − ( N − 1) ( N − 1)
WH ( n) = 0.54 + 046 cos for ≤n≤
N −1 2 2
0 for otherwise
The window sequence for N = 11 is given by

πn
WH ( n) = 0.54 + 0.46 cos for − 5 ≤ n ≤ 5
5
0 for otherwise
WH (0 ) = 0.54 + 0.46 = 1
π
WH (1) = WH ( −1) = 0.54 + 0.46 cos = 0.912
5
2π
WH (2) = WH ( −2) = 0.54 + 0.46 cos = 0.682
5
3π
WH (3) = WH ( −3) = 0.54 + 0.46 cos = 0.398
5
4π
WH ( 4 ) = WH ( −4) = 0.54 + 0.46 cos = 0.1678
5
WH (5) = WH ( −5) = 0.54 + 0.46 cos π = 0.08

To find hd(n)
π
1
hd ( n) =
2π ∫π H (e ) e
−
d
jw jwn
dw
1 ⎡⎢ ⎤
π 4 π
= ∫ e iwn dw + ∫ e iwn dw ⎥
2π ⎢ π π ⎥
⎣ 4 ⎦
=
1
2π jn {
[e jwn ]−−ππ / 4 + [e jwn ]π π 4 }
1 ⎡ πn⎤
= ⎢ sin π n − sin ⎥
πn ⎣ 4 ⎦
For n = 0, hd(0) = (1−1/4) = 3/4.

π
sin π − sin
hd (1) = hd ( −1) = 4 = −0.225
π
π
sin 2π − sin
hd (2) = hd ( −2) = 2 = −0.159
2π
3π
sin 3π − sin
hd (3) = hd ( −3) = 4 = −0.075
3π
sin 4π − sin π
hd ( 4 ) = hd ( −4 ) = =0
4π
5π
sin 5π − sin
hd (5) = hd ( −5) = 4 = 0.045
5π
The filter coefficients using hamming window sequence are
h ( n) = hd ( n)WH ( n) for − 5 ≤ n ≤ 5
0 for otherwise
h (0 ) = hd ( n)WHn (0) = (1)(0.75) = 0.75
h (1) = h ( −1) = hd (1)WHn (1) = ( −0.225)(0.912) = −0.2052
h (2) = h ( −2) = hd (2)WHn (2) = ( −0.159)(0.682) = −0.1084
h (3) = h ( −3) = hd (3)WHn (3) = ( −0.075)(0.398) = −0.03

h ( 4 ) = h ( −4 ) = hd (4)WHn ( 4 ) = (0 )(0.1678) = 0
h (5) = h ( −5) = hd (5)WHn (5) = ( −0.045)(0.08) = 0.0036
The transfer function of the filter is given by
5
H ( z ) = h (0 ) + Σ ⎡⎣ h ( n) ( z −1 + z n )⎤⎦
n =1
= 0.75 − 0.2052 ( z −1 + z ) − 0.1084 ( z −2 + z 2 )

−0.03 ( z −3 + z 3 ) + 0.0036 ( z −5 + z 5 ) ,
The transfer function of the realizable filter is
H1 ( z ) = z −5 H ( z )
= 0.0036 − 0.03 z −2 − 0.1084 z −3 − 0.20527 z −4 + 0.752 z −5
−0.2052 z −6 − 0.1084 z −6 − 001084 z −7
−0.03 z −8 + 0.036 z −10
The filter coefficients of casual filter are
h (0 ) = h (10 ) = 0.0036; h (1) = h (9) = 0; h (2) = h (8) = −0.03;

h (3) = h (7) = −0.1084; h ( 4 ) = h (6) = −0.2052; h (5) = 0.75
( N −1)/2
H (e jw ) = ∑ a (n) cos wn
n= 0
⎛ N − 1⎞
a(0) = h ⎜ = h (5) = 0.75
⎝ 2 ⎟⎠
⎛ N −1 ⎞
a ( n ) = 2h ⎜ − n⎟
⎝ 2 ⎠
a (1) = 2h(5 − 1) = 2h(4) = −0.4104
a (2) = 2h(5 − 2) = 2h(3) = −0.2168
a(3) = 2h(5 − 3) = 2h(2) = −0.06
a ( 4 ) = 2h(5 − 4) = 2h(1) = 0
a (5) = 2h(5 − 5) = 2h(0) = 0.0072.
H (e jw ) = 0.75 − 0.4104 cos w − 0.2168cos 2w
− 0.06cos3w + 0.0072cos5w

W
(in degrees) 0 15 30 45 60 75 90
( )
H e jw
0.07 0.125 0.28 0.497 0.7168 0.88 0.9668
( )
| H e jw | dB −23.1 −18 −11 −6.07 −2.89 −1.1 −0.29
105 120 135 150 165 180

0.9945 1 1.0026 1.003 1 1.0108
−0.0478 0 0.0229 0.028 0 0.093
⏐H (e jw)⏐dB 0
–5
– 10
– 15
– 20
– 25
n
Q p /2 w
13. (b) (i) | H ( k ) |= 1 for 0 ≤ k ≤ 3 and 12 ≤ k ≤ 14

= 0 for 4 ≤ k ≤ 11
⏐H (K )⏐
K
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
given data,
⎛ N − 1⎞
Q (k ) = − ⎜ πk
⎝ N ⎟⎠
−14π k
= 0≤k ≤7
15

given data,
⎛ N − 1⎞
Q (k ) = − ⎜ πk
⎝ N ⎟⎠
−14π k
= 0≤k ≤7
15
14π k
and Q ( k ) = 14π − for 8 ≤ k ≤ 14
15
14 π k
H (k ) = e − j 15 for k = 0, 1, 2, 3
= 0 for 4 ≤ k ≤ 11
=e − j14π ( k −15) 15
for 12 ≤ k ≤ 14
N −1
⎡ ⎤
H (0 ) + 2 ∑ Re ( H ( k )ej 2π nk /15 )⎥
2
1⎢
h ( n) =
n⎢ k =1
⎥
⎢⎣ ⎥⎦
1 ⎡ 7
⎤
= ⎢
15 ⎣
1 + 2 ∑
k =1
e ( − j14π k /15) ⎥
⎦
1 ⎡ 3
2π k (7 − n) ⎤
= ⎢
15 ⎣
1 + 2 ∑ cos ⎥
k =1 15 ⎦
1 ⎡ 2π ( 7 − n ) 4π (7 − n) 6π ( 7 − N ) ⎤
h ( n) = ⎢1 + 2 cos + 2 cos + 2 cos ⎥
15 ⎣ 15 15 15 ⎦
h ( 0 ) = h (14 ) = −0.05; h ( 0 ) = h ( 3) = 0.041; h ( 4 ) = h ( 0 ) = −0.1078
h ( 2 ) = h (12 ) = −0.0666; h ( 3) = h (11) = −0.0365; h ( 5 ) = h ( 9 ) = 0.034
h ( 6 ) = h ( 8 ) = 0.3188; h ( 7 ) = 0.466
1 1 −1 −2 1 −3 1 1
H (z) = + z + z + z + z −4 + z −5 + z −6 .
2 3 4 3 2
x(z) z –1 z –1 z –1
+ + +
z –1 z –1 z –1
1
1/2 1/3 1/4
y(z) + + +

14. (a) Refer answer 14(a) from April/May 2011 Question paper.
(b) When a stable IIR digital filter is excited by a finite input sequence,
that is constant, the output will ideally decay to zero. However,
the nonlinearities due to the finite – precision arithmetic opera-
tions often cause periodic oscillations to occur in the output. Such
oscillations in recursive systems are called zero input limit cycle
oscillation.
Consider a first order IIR filter with difference equation.
Y(n) = x(n) + ay(n − 1).
1
Let us assume α = and the data register length is 3 bits plus a sign
2 n=0
⎧0.875 for
bit. If the input is x( n) = ⎨
⎩ 0 for otherwise
and rounding applied after the arithmetic operation the table 1 illus-
trates the limit cycle behavior. Here Q[.] represents the rounded
operation. From table 1, we find that n ≥ 3 the output remains con-
1
stant and gives as steady output causing limit cycle behavior.
8
When a = − 1 , we can see from table 2, that the output oscillates
2
between 0. 125 and –0.125.
Table 1
y(n=x(n)+
n x(n) y(n−1) ay(n−1) q[ay(n−1)] q[ay(n−1]
0 0.875 0.0 0.0 0.000 7/8
1 0 7/8 7/16 0.100 1/2
2 0 1/2 1/4 0.010 1/4
3 0 1/4 1/8 0.001 1/8
4 0 1/8 1/16 0.001 1/8
5 0 1/8 1/16 0.001 1/8
7/8
3/8
1/4
1/8 1/8
1/8

Table 2
y(n=x(n)+
n x(n) y(n−1) ay(n−1) q[ay(n−1)] q[ay(n−1]
0 0.875 0 0 0.000 7/8
1 0 7/8 −7/16 1.100 −1/2
2 0 −1/2 1/4 0.010 1/4
3 0 1/4 −1/8 1.001 −1/8
4 0 −1/8 1/16 0.001 1/8
5 0 1/8 −1/16 1.001 −1/8
6 0 −1/8 1/16 0.001 1/8
Note that beyond n = 4 the value of [ay(n−1] is 16, and is

∓
binary .000100 and 0.000100 which when rounded gives 1.001
exhibiting oscillatory.
Deadband: The limit cycles occurs as a result of the quanti-
zation effects in multiplications. The amplitudes of the output
during a limit cycle are confined to a range of values that is
caused the deadband of the filter.
Let us consider a single pole IIR system, whose difference
equation is given by,
y(n)= ay(n−1) +x(n), n>0.
After rounding the product term we have,
yq (n) = Q[ay(n − 1)] + x(n).
During the limit cycle oscillations,
Q[ay (n − 1)] = y(n − 1) for a > 0
= − y(n − 1) for a < 0.
By defining of rounding we have
2− b
Q ⎡⎣α y ( n − 1) ⎤⎦ − α y ( n − 1) ≤
2
2− b
∴ y ( n − 1) − α y ( n − 1) ≤
2
−b
1 22
y ( n − 1) ≤
1− α
The above equation defines the deadband for the first order filter.

In addition to limit cycle oscillations causing by rounding the

result of multiplication, there are several types of limit cycle
oscillations caused by addition, which make the filter output
oscillate between maximum and minimum amplitudes. Such
limit cycles have been referred to as overflow oscillation.
An overflow in addition of two or more binary numbers occurs
when the sum exceeds the word size available in the digital
implementation of the system.
F(n) 1
–1
Fig(a) Transfer characteristics of an adder
Let us consider two positive numbers n1 and n2.

n1 = 0.111 = 7 8
n2 = 0.110 = 6 8
n1 + n2 = 1.101 → − 5 8 in sign magnitude.
F(n) 1
–1
Fig(b) Saturation adder transfer characteristics
In the above example, when two positive numbers are added the
sum is wrongly interpreted as a negative number.
The transfer characteristics of an adder is shown below, where
n is the input to the adder and f(n) is the corresponding output.

In figure (a), we can find the overflow occurs if the total input is
out of range (–1,1). This problem can be eliminated by modify-
ing the adder characteristics as shown in figure (b). Here when
an overflow is detected, the sum of adder is set equal to the
maximum value.
15. (a) Polyphase Structure of Decimator

In polyphase realization of FIR filters, the transfer function H(z) is
decomposed into M branches given by
m −1
H (z) = ∑z −m
pm ( z m )
m= 0
⎡⎣( N +1 m)⎤⎦
Where pm ( z ) = ∑ h ( M n + m) z − n
n= 0
The z-transform of an infinite sequence is given by

∞
H (z) = ∑ h ( n) z
n =−∞
−n
In this case H ( z ) can be decomposed into M-branches as

∞
H (z) = ∑ h ( n) z
n =−∞
−n
∞
where pm ( z ) = ∑ h (rM + m) z
r =−∞
−r
m −1 ∞
H (z) = ∑ ∑Z −m
h ( rM + m) z − rM
m = 0 r =−∞
m −1 ∞
=∑ ∑ h (rM + m) z ( − rM + m)
m = 0 r =−∞
Let h ( rM + m) = pm ( r )
m −1 ∞
⇒ H (z) = ∑∑
m = 0 r = −∞
pm ( r ) z − (rM + m)
m −1 ∞
Y (z) = ∑∑
m = 0 r = −∞
pm ( r ) x ( z ) z − (rM + m)
m −1 ∞
y ( n) = ∑∑ pm ( r ) x [ n − ( rM + m) ]
m = 0 r = −∞

Let xm ( r ) = x ( rM − m) then
m =1 ∞
y ( n) = ∑ ∑ p (r ) x (n − r )
m m
m = 0 r =−∞
m =1
= ∑ pm ( n) * xm ( n)
m= 0
m =1
= ∑ y ( n)
m= 0
Where ym ( n) = pm ( n) + xm ( n)
The operation pm ( n) * xm ( n) is known as polyphase convolu-

tion, and the overall process is polyphase filtering.
If M = 3 then
2
y ( n) = ∑ y ( n) = y ( n) + y ( n) + y ( n)
m= 0
m 0 1 2
= p0 ( n) * x0 ( n) + p1 ( n) * x1 ( n) + p2 ( n) * x2 ( n)
We know that xm ( n) can be obtained first by delaying x ( n) by

m units, and then down sampling by a factor M. Next ym ( n)
can be obtained by convolving xm ( n) with pm ( n). The structure
of a polyphase decimator with 3 branches and a sampling rate
reduction by a factor three is shown in fig(a).
For a general case of M branches and a sampling rate reduc-
tion by a factor M, the structure of polyphase decimator is
fig(b). The splitting of x(n) into the low rate-sub-sequence
x0 ( n ) , x1 ( n ) ,.... xm −1 ( n ) is often represented by a commutator.
In the configuration shown in fig(b), the input values x(n) enter
the delay chain at high state. Then the M down sampler sends
the group of M input values to M filters at time n = mM.
For example at time n = 0, the value of x0 (0 ) = x (0 ) , x1 (0 ) = x ( −1),
x2 (0 ) = x ( −2) ..... x M −1 (0 ) = x ( − M + 1) are sent.
In fig(c), to produce the output y(0), the commuta-
tor must rotate in counter clockwise directing start-
ing from m = m−1,…..m = 2, m = 1, m = 0 and give the
input values x ( − M + 1) ,... x ( −2 ) , x ( −1) , x ( 0 ) to the filters
pm −1 (0 ) ,.... p2 (2) , p1 ( n) , p0 ( n) .

x 0(n)
x(n) M P0(n) + y(n)
Z –1
x 1(n)
M P1(n) +
Z –1 x 2(n)
M P2(n)
Fig(a) Polyphase structure of a 3 branch decimator
x(n) M P0(n) + y(n)
Z –1
M P1(n) +
Z –1
M P2(n) +
M P3(n)
Fig(b) Polyphase structure of a M-branch decimator
m = 0 x0 (n) y0(n)
P0(n) + y(n)
Fx
Rate Fx y1(n) Rate Fy =
x(n) P1(n) + M
m=1
m=2 y2(n)
P2(n) +
m = m –1
ym–1(n)
Pm–1(n)
Fig(c) Polyphase decimator with a commutator
Polyphase Structure of Interpolator

By transposing the decimator structure shown in fig(b). We can
obtain polyphase structure for interpolator, which consists of a
set of L sub-filters connected in parallel as shown in fig(c). Here

the polyphase components of impulse response are given by,

pm ( n ) ....h ( nL + m ) , m = 0,1, 2,...L − 1
x(n) P0(z m) M + y(n)
Z –1
P1(z m) M +
Z –1
P2(z m) M +
Pm –1(z m) M
Fig(a)
x(n) M P0(z ) + y(n)
Z –1
M P1(z ) +
Z –1
M P2(z ) +
M Pm –1(z)
Fig(b)
y0(n)
x(n) P0(n) L + y(n)
y1(n) Z –1
P1(n) L +
Z –1
y2(n)
P2(n) L +
yL–1(n) Z –1
PL–1(n) L
Fig(c)

x(n)
P0(n)
P1(n)
y(n)
P2(n)
PL–1(n)
Fig(d)
Where h(n) is the impulse response of anti-imaging filter. The
output of L sub-filters can be represented as
ym ( n ) = x ( n ) pm ( n ) , m = 0,1, 2,...L − 1
By upsampling with a factor L and z-m the polyphase compo-

nents are produced from ym ( n ) . These polyphase components
are all added together to produce the polyphase output signal
y(n). The output y(n) also can be obtained by combining the
signals xm ( n ) using a commutator shown in fig(d).
(b) Filter banks are of two types, analysis filter bank and synthesis
filter bank.
Analysis filter bank: The M-channel analysis filter bank
is shown in figure. It consists of M subfilters. The individual
subfilter H k ( z ) is known as analysis filter. All the subfilters
are equally spaced in frequency and each have the same band-
width. The spectrum of the input signal x ( e ) lies in the range
jw
0 ≤ W ≤ π . The filter bank splits the signal into sub bands each
π
having a bandwidth . The filter H 0 ( z ) is lowpass, H1 ( z ) to
M
H M −2 ( z ) are bandpass and H M −1 ( z ) is highpass. As the spec-
π
trum of the signal is bandlimited to , the sampling rate can
M
be reduced by a factor M. The downsampling moves all the sub-
π
band signals into baseband 0 ≤ W ≤ .
M

Synthesis Filter Bank: The M-channel synthesis filter band

is dual of M-channel analysis filter bank. In this case each
U M ( z ) is fed to an upsamplet. The upsampling process pro-
duces the signal U M ( z M ) . These signals are applied to filters
GM ( z ) and finally added to get the output signal x̂ (z). The filter
G0 ( z ) to Gm−1 ( z ) have the same characteristics as the analysis
filters H 0 ( z ) to H m −1 ( z ) .
Subband Coding Filter Bank: If we combine the analysis fil-

ter bank of fig(a) and synthesis filter of fig(b), we obtain an
M-channel subband coding filter bank as shown in fig(c).
x(z) H0(z ) M U0(z)
H1(z ) M U1(z)
H2(z ) M U2(z)
Hm–1(z ) M Um–1(z )
Fig(a) Analysis filter bank
U0(z ) G0(z)
>
M + x(z)
U1(z )
M G1(z) +
U2(z )
M G2(z) +
Um–1(z )
M Gm–1(z)
Fig(b) Synthesis filter bank

H0(z) G0(z)
>
x(z) M M + x(z)
H1(z) M M G1(z) +
H2(z) M M G2(z) +
Hm–1(z) M M Gm–1(z)
Fig(c) Subband coding filter bank
The analysis filter bank splits the broadband input sig-

nal x ( n ) into M non-overlapping frequency band signals
x0 ( z ) , x1 ( z ) .... x M −1 ( z ) of equal bandwidth. These outputs
are coded and transmitted. The synthesis filter bank is used to
reconstruct output signal x̂ ( n ) which should approximate the
original signal. The application of subband coding is in speech
signal processing.

APRIL/MAY 2010
Fifth Semester
PART A (10 × 2 = 20 marks)
1. Define even and odd signals.
2. State the disadvantage of DSP over all processing.
3. Check whether the following system is time-invarient, y(n) = nx2(n)
4. State and prove initial value theorem.
5. Define convolution property of continuous time and discrete time signals.
6. What is the method of finding IDFT through DFT?
7. What is aliasing? How it can be eliminated?
8. Define acquisition time and aperture time.
9. Obtain the digital filter transfer function of the following analog filter
using impulse invarient transform.
10. What is wrapping effect?
PART B (5 × 16 = 80 marks)
11. (a) (i) Develop the algorithm for radix 2, 8 point DIT-FFT method.
(ii) Find the DFT of the sequence x(n) = {0,1,2,3} using DIT-FFT
algorithm.
Or

(b) With appropriate diagrams describe

(i) Overlap-save method
12. (a) (i) Find the inverse z-transform of X(z) =

( z 2 + z)
X ( z) = , ROC z > 3
( z − 1)( z − 3)
by partial fraction method.
(ii) Determine the z-transform of x(n) = n(−1)n u(n)
(iii) Find the transfer function of the following
LTI system: y(n) = y(n−1) − 0.5 y(n−2) + x(n) + x(n−1)
Or
(b) (i) State and prove convolution theorem in z-transform.
(ii) Find the z-transform and ROC of the signal.
13. (a) (i) State and prove Parseval’s theorem in continuous time fourier
transform.
(ii) Show the relationship b/w DFT and DTFT.
(iii) Find the output of an LTI system whose impulse response is
h(n) = {1, 1, 1} and input signal is x(n) = {3, −1, 0, +1}, using
circular convolution.
Or
(b) (i) Determine the IDFT of the following sequence using DIF-FFT
method.
X(K) = {20, −5.828 + j2.414, 0, − 0.172 − j 0.414, 0, − 0.172 −
j0.414, 0, −5.828 + j 2.414}
14. (a) Explain in detail the concept of sampling, recovery of signal and
discrete tome processing of continuous time signals.
Or
(b) (i) Explain with block diagram, the functioning of serial-parallel
subranging and ripple A/D convertor.
(ii) Explain any one type D/A convertor with schematic diagram.

15. (a) (i) Apply bilinear transform to the transfer function.

(ii) Design a digital butterworth filter satisfying the following con-
straints using bilinear transform. Assume T = 1 sec.
π
0.9 ≤ H (e jω ) ≤ 1 for 0 ≤ ω ≤
2
3π
H (e jω ) ≤ 0.2 for ≤ω ≤π
4
Or
(b) (i) Explain briefly the theory of Chebyshev app of digital filter
design.
(ii) Design a Chebyshev filter for the following specification using
bilinear transformation (T = 1S)
0.8 ≤ H (e jω ) ≤ 1 0 ≤ ω ≤ 0.2π
H (e jω ) ≤ 0.2 0.6π ≤ ω ≤ π

Solutions
PART A
1. A discrete time signal x(n) is said to be symmetric (even) signal if it
satisfies the condition.
x(−n) = x(n) for all n.
The signal is odd signal if it satisfies the condition
x(−n) = −x(n) for all n.
2. • System complexity
• Bandwidth limited by sampling rate
• Power consumption
3. y( n) = nx 2 ( n)
y( n, K ) = T [ x( n − K )] = ( n) x 2 ( n − K )
y( n − K ) = T [ x( n − K )] = ( n − K ) x 2 ( n − K )
y( n, K ) ≠ y( n − K )
∴ The system is time varient.
4. If X+(z) = z(x(n)), then

x(0) = lim X + ( z )
z →∞
Proof:
∞
X + ( z ) = ∑ x ( n) z − n
n= 0
As z → ∞, all the term varies except x(0), which proves the theorem.
∞
i.e. lim X + ( z ) = lim ∑ x( n) z − n = x(0)
z →∞ z →∞
n= 0
5. Convolution property of discrete time signals

y(n) = x(n) * h(n)
* → Convolution operation
Convolution property of continuous time signals
∞
y( K ) = ∑
K = −∞
x( K ) h( n − K )

N −1
6. DFT : x( K ) = ∑ x( n)WNnK ; 0 ≤ K ≤ N − 1
n= 0
1 N −1
IDFT : x( n) =
N K =0
∑ X ( K ) WN− nK ;0 ≤ n ≤ N − 1
where WN = e − j 2π / N
7. The super imposition of high frequency component on the low frequency

component is known as frequency aliasing. To avoid aliasing the sam-
pling frequency must be greater than twice the highest frequency present
in the signal
F ≥ 2fm
8. Acquisition time: Acquisition time is the interval between the release of

the hold state (imposed by the input circuitry of a track-and-hold) and the
instant at which the voltage on the sampling capacitor settles to within
1LSB of a new input value.
Aperture time: Aperture time (tAD) in an ADC is the interval between
the sampling edge of the clock signal (the rising edge of the clock signal
in the figure) and the instant when the sample is taken. The sample is
taken when the ADC’s track-and-hold goes into the hold state.
1
9. H (S ) =
( S + 1)( S + 2)
1 A B
H (S ) = = +
( S + 1)( S + 2) S + 1 S + 2
A( S + 2) + B( S + 1) = 1 ⇒ A = 1 B = −1
1 1
H a (S ) = −
S +1 S + 2
1 1
H ( z) = −
1− e zPK T −1
1− e z PK T −1
1 − e −2 z −1 − 1 + e −1 z −1
=
(1 − e −1 z −1 )(1 − e −2 z −1 )
0.2317 z −1
=
1 − 0.502 z −1 + 0.049 z −2
10. Relation b/w the analog and digital frequencies
2 ω
Ω= tan
T 2

for larger values of w the relationship is non linear.

This non linearity introduces distortion in the frequency axis. This is
known as wrapping effect.
PART B
11. (a) (i) DIT-FFT: It is a method in which we divide a time factor N.
Algorithm:
1. No. of input samples N = 2m where m = log2N
8
x(n)
N/2 N/2
4 x e(n) x o(n) 4
N/4 N/4 N/4 N/4
x ee(n) x eo(n) x oe(n) x oo(n)
N −1
x( K ) = ∑ x( n) WNnK
n= 0
N N
−1 −1
2 2
x( K ) = ∑ x(2n) W
n= 0
N
nK
2
+ WNK ∑ x(2n + 1) W
n= 0
N
nK
2
2. Input sequence is bit reversed and taken.

3. No. of stage is given by m.
N
4. Each stage consists of butterflies.
2
N
5. No. of complex multiplication. log 2 N
2
6. No. of complex addition. N log 2 N
I - Stage:
xe(0) = x(0) x0(0) = x(1)
xe(1) = x(2) x0(1) = x(3)
xe(2) = x(4) x0(2) = x(5)
xe(3) = x(6) x0(3) = x(7)

⎧ ⎛ n ⎞
⎪ xe ( K ) + WN x0 ( K ) ⎜⎝ for 0 ≤ K ≤ 2 − 1⎟⎠
K
⎪
X (K ) = ⎨
⎪ x ⎛ K − N ⎞ − W K x ⎛ K − N ⎞ for N ≤ K ≤ N − 1
⎪⎩ e ⎜⎝ 2⎠
⎟ N 0 ⎜
⎝ ⎟
2⎠ 2
II - Stage:
⎧ N
⎪⎪ xee ( K ) + WN xe 0 ( K ) for 0 ≤ K ≤ 4 − 1
2K
X e (K ) = ⎨
⎪ x ( K ) − W 2( K − N / 4) x ( K − N /4) for N ≤ K ≤ N − 1
⎪⎩ ee N e0
4 2
⎧ N
⎪⎪ x0 e ( K ) + WN x00 ( K ) 0 ≤ K ≤ 4 − 1
2K
X 0 (K ) = ⎨
⎪ x ( K ) − W 2( K − N / 4) x ( K − N /4) N ≤ K ≤ N − 1
⎪⎩ 0 e N 00
4 2
III - Stage:
X ee ( K ) = DFT [ X ee ( n)] = ∑
K = 0,1
xee ( n)WNnK/ 4
X e 0 ( K ) = DFT [ X e 0 ( n)] = ∑x
k = 0,1
e0
( n)WnnK
/4
X 0 e ( K ) = DFT [ x0 e ( n)] = ∑x
k = 0,1
0e
( n)WNnK/ 4
x00 ( K ) = DFT [ X 00 ( n)] = ∑

K = 0,1
x00 ( n)WNnK/ 4
x(0) x eo(0) x e(0) x(0)

x(4) x eo(1) x e(1) x(1)
x(2) –1 x eo(0) –1 x e(2) x(2)
x(6) x eo(1) –1 x e(3) x(3)
x(1) –1 x oe(0) x o(0) –1 x(4)
x(5) x oe(1) x o(1) x(5)
x(3) –1 x oo(0) –1 x o(2) –1 x(6)
x(7) x oo(1) –1 x o(3) –1 x(7)
–1 –1
I st Stage II nd Stage III rd Stage

(ii) N = 4 ⇒ 2m = N ⇒ m = 2 ⇒ 2 stages
x(0) = 0 0+2=2 2+4=6
x(2) = 2 0 – 2 = –2 –2 + j2 = –2 + 2j
x(1) = 1 –1 1+3=4 –1 2 – 4 = –2
x(3) = 3 1 – 3 = –2 1 –1 –2 – j2 = –2 – 2j
–1 –j
I/P I Stage II Stage

0 2 6
2 −2 −2 + 2j
1 4 −2
8 −2 −2 − 2j
(b) Refer answer 11(a) from April/May 2011 Question paper.
12. (a) (i) Converting the above improper rational function into sum of a
constant and proper rational function we get,
5z − 3
X ( z) = 1 +
( z − 1)( z − 3)
5z − 3 C C
= 1 + 2
( z − 1)( z − 3) z − 1 z − 3
5z − 3
C1 = ( z − 1) = −1
( z − 1)( z − 3) z =1
5z − 3
C2 = ( z − 3) =2
( z − 1)( z − 3) z =3
X ( z) 1 2
∴ = +
z z −1 z − 3
z 2 − 3z − z + 3
z2 − 4x + 3
z 2z
X ( z) = +
z −1 z − 3
x( n) = −u( n) + 23n u( n)
∞ ∞
z
(ii) z [( −1) n u( n)] = ∑ ( −1) n z − n = ∑ ( − z −1 ) n =
n= 0 n= 0 z −1

By using differentiation property

d ⎛ z ⎞ −z
z [n( −1) n u( n)] = − z ⎜⎝ ⎟⎠ =
dz z + 1 ( z + 1)2
(iii) Taking z transform on both sides
y( z ) = z −1 y( z ) − 0.5 z −2 y( z ) + x( z ) + z −1 x( z )
y( z ) − z −1 y( z ) + 0.5 z −2 = X ( z ) + z −1 X ( z )
Y ( z ) [1 − z −1 + 0.5 z −2 ] = X ( z )(1 + z −1 )
Y ( z) 1 + z −1 z ( z + 1)
= = 2
X ( z ) 1 − z + 0.5 z
−1 −2
z − z + 0.5
(b) (i) If x(z) = z{x(n)} and H(z) = z{h(n)} then

Z(x(n) * h(n)) = x(z)H(z)
Where x(n) * h(n) denotes the linear convolution of sequence.
Proof:
∞
We have y(n) = x(n) * h(n) = ∑
K = −∞
x( K )h( n − K )
and
∞ ∞
⎡ ⎤
Y ( z ) = z{ y( n)} = ∑ ⎢⎣ ∑
n = −∞ K = −∞
x( K )h( n − K ) ⎥ z − n
⎦
∞ ∞
= ∑ ∑ x( K ) z
n = −∞ k = −∞
−K
h( n − K ) z − ( n − K )
Interchange the order of summation

∞ ∞
y( z ) = ∑
K = −∞
x( K ) z − K ∑ h(n − k )z
n = −∞
−( n− k )
Replace ( n − K ) by l
∞ ∞
Y ( z) = ∑
K = −∞
x( K ) z − K ∑ h(l ) z
l = −∞
−l
= H ( z) X ( z)
(ii) x( n) = −b n u( n − 1)
∞ ∞
z
z[ −b n u( n)] = ∑ ( −b) n z − n = ∑ ( −bz −1 ) n =
n= 0 n= 0 z+b
z
∴ X ( z) =
z+b

The ROC is z > b . It represents the interior of the circle hav-

ing radius b.
Im (z)
Zplane
ROC
Rel (z)
13. (a) (i) If F [ x( n)] = X (e jω )

then
∞ π
1
E= ∑ [ x(n)]2 =
n = −∞ 2π −∫π
( X (e jω )2 )dω
Proof :
∞ ∞
E= ∑ ( x(n))
n = −∞
2
= ∑ x(n) x ∗ (n)
n = −∞
π ∗
∞
⎡ 1 ⎤
= ∑ x ( n) ⎢ ∫ X (e jω
)e jω n
⎥ dω
n = −∞ ⎣ 2π −π ⎦
π
1 ⎡ ∞ ⎤
E= ∫π X *
(e jω ) ⎢ ∑ x( n)e − jω n ⎥ dω
2π − ⎣ n = −∞ ⎦
π π
1 1
=
2π ∫π
−
X * ( e jω ) X ( e jω ) d ω =
2π ∫π ( X (e ω )) dω
−
j 2
(ii) DTFT: IT is a transform of discrete sequence.

Since the time domain it discrete the spectrum is periodic.
Relationship b/w DFT and DTFT:
∞
⎡ K ⎤ 1 ⎡ K − iN ⎤
XK = X ⎢ = ∑ X⎢
⎣ NT ⎦ T i = −∞ ⎣ NT ⎥⎦
⎥
1 ⎡ K ⎤
= X⎢ K = 0,1,.....N − 1.
T ⎣ NT ⎥⎦

(iii) x(n) = {3,−1,0,1}

h(n) = {1,1,1,0}
Using matrix approach
⎡1 0 11 ⎤ ⎡ 3 ⎤ ⎡3 + 0 + 0 + 1⎤ ⎡ 4 ⎤
⎢1 1 01 ⎥ ⎢ −1⎥ ⎢3 − 1 + 0 + 1 ⎥ ⎢ 3 ⎥
⎢ ⎥⎢ ⎥ = ⎢ ⎥=⎢ ⎥
⎢1 1 10 ⎥ ⎢ 0 ⎥ ⎢3 − 1 + 0 + 0 ⎥ ⎢ 2 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣0 1 11 ⎦ ⎣ 1 ⎦ ⎣0 − 1 + 0 + 1 ⎦ ⎣0 ⎦
∴ y( n) = x( n) N h( n) = {4,3, 2, 0}
(b) (i) X(K) = {20, −5.828 + j2.414, 0, − 0.172 − j 0.414, 0, − 0.172 −

j0.414, 0, −5.828 + j 2.414}
w 80 = 1
w 18 = 0.707–j0.707
w 28 = –j
20 20 19.465 w 38 = 0.707–j0.707
x *(0) = 20
w 0 –j5.121
x *(1) = –5.828 8
–6.535–j3.121 –0.535–j5.121
–j2.414 w 18 –1 20.535+j5.121
0 20
x *(2) = 0
w 28 –1 7.465–j1.121
x *(3) = –0.172 6–j2 –12.535–j1.121 32.535+j121
+j0.414 w 38 –1 –1 –1
20 20 20–j0.293
x *(4) = 0
x *(5) = –0.172 –1 –5.656–j3.121 –j0.293 20+j0.293
+j0.414 –1
0 20 31.313–j5.949
x *(6) = 0
–1 –1
x *(7) = –5.828 5.656+j2.828 11.313–j5.949 8.687+j5.945
–j2.414 –1
–1 –1
x(K) S1 S2 o/p
20 10 5 1
−5.828 + j2.414 −3 − j −3 2
0 0 5 3
−0.172 − j0.414 −3 + j 1 4
0 10 5 4
−0.172 − j0.414 −1 − j3 −1 3
0 0 5 2
−5.828 + j 0.214 −1 + j3 3 1

14. (a) Sampling:

Sampling is a process of converting a continuous-time signal to a dis-
crete time signal. It contains a sampling switch that closes for a very
short interval of time τ, during which the signal presents at the output.
Xa(t ) Xn = (nt )
F = 1/T
Xa(t )
Xa(nT )
nT
T 2T 3T 4T 5T 6T 7T 8T 9T 10T 11T 12T
xa(t) → continuous signal

Xa(j Ω) → spectrum of continuous signal
The Fourier Transform
∞
X a ( j Ω) = ∫ x (t )e
−∞
a
− j Ωt
dt
∞
1
x( n) = xa ( nT ) =
2π ∫X
−∞
a
( j Ω)e jΩnt d Ω
π
1
we know x( n) =
2π ∫π X (e ω )e ω dω
−
j j n

π
⎛ jω 2π K ⎞ jω n
∞
1 1
xn =
2π ∫−π T ∑
K = −∞
Xa ⎜
⎝ T
+j
T ⎠
⎟ e dω
⎛ jω 2π K ⎞
∞
1
∴ X ( e jω ) =
T
∑
K = −∞
Xa ⎜
⎝ T
+j
T ⎠
⎟
1 ∞
⎛ 2π K ⎞
X ( e jω ) = ∑ X a ⎜ jΩ + j ⎟
T K = −∞
⎝ T ⎠
Recovery of analog signal:

From sampling theorem it is clear that if we sample band limited
analog signal xa(t) above its nyquist rate, then we can reconstruct
xa(t) from its sample x(n).
∞
⎛ 2π k ⎞
T X (e jΩt ) = ∑
K = −∞
X a ⎜ jΩ + j
⎝ T ⎠
⎟
Digital Analog
DAC Interpolator LPP
i/p o/p
Digital to Analog Conversion
(b) (i) • Counting A/D convertor

• Successive approximation A/D convertor
• Flash A/D convertor
• Over sampling Sigma-Delta A/D convertor
1. Counting A/D converter
Vi
+
Analog i/p Voltage
–
Comparator
Vref D/A Convertor
N-bit
Clk
N-bit Convertor
AN Gate Start

2. Successive approximation A/D converter
Vi S/H + Control Logic

Clk
– and SAR
bN–1
bo n bit
DAC
3. Over sampling Sigma - Delta A/D convertor
Analog
i/p N-bit
+ Integrator 1 bit ADC Mth Band ↓M
o/p
Digital
LPF
1 bit DAC
4. Flash A/D convertor
Vi
V6
V5 E
n
c
V4 o
d
i
n Digital
V3 g Code
L
o
V2 g
i
c
V1
V0

(ii) Over sampling Sigma - Delta D/A converter.
Digital 1 bit O/A Analog Analog

↑L Lth Band LPF
i/p Converter LPF o/p
Intepolator
2
15. (a) (i) H ( S ) = By assuming T = 0.15
( S + 1)( S + 2)
A B
H (S ) = +
S +1 S + 2
2
A = ( S + 1)
( S + 1)( S + 2) S = −1
A= 2
2
B = ( S + 2)
( S + 1)( S + 2) S = −2
B = −2
2 2 2 2
H (S ) = − = −
S + 1 S + 2 S − ( −1) S − ( −2)
N
CK
H (S ) = ∑
K =1 S − PK
N
CK
H ( z) = ∑
K =1 1 − e
PK T −1
z
P1 = −1 P2 = −2
2 2
H ( z) = −
1 − e −T z −1 1 − e −2T z −1
0.173 z −1
H ( z) =
1 − 1.722 z −1 + 0.732 z −2
(ii) Given data:

1 1
= 0.9 = 0.2
1+ ε 2
1+ λ2
π 3π
ωP = ωS =
2 4

The analog frequency ratio is

2 ω 3π
ΩS tan S tan
= T 2 = 8 = 2.414
ΩP 2 ω π
tan P tan
T 2 4
The order of the filter
log λ ε
N≥
ΩS
log ΩP
N =3
ΩP
ΩC = 2.54 radians = 1
Σ N
2 ω
ΩP = tan P = 2 rad/sec.
T 2
The T.F of 3rd order normalized Butterworth filter

1
H (S ) =
( S + 1)( S 2 + S + 1)
H a ( S ) = Replace S → S 2.5 in H ( s)
1
H a (S ) =
⎞ ⎡⎛ S ⎞ ⎤
2
⎛ S S
⎜⎝ + 1⎟ + ⎢⎜ ⎟ + + 1⎥
2.5 ⎠ ⎣⎝ 2.5 ⎠ 2.5 ⎦
15.625
=
S 3 + 5S 2 + 12.5S + 15.625
2 ⎛ 1 − z −1 ⎞
H ( Z ) = H (S ) S = ⎜ ⎟
T ⎝ 1 + z −1 ⎠
0.22(1 + 2 z −1 + 2 z −2 + z −3 )
H (Z ) =
1 + 0.29 z −1 + 0.02 z −2 + 0.03 z −3
(b) (i) Steps to design Chebyshevic LPF

1. From the given specification find the order of filter N.
2. Round off it to the next higher integer.
3. Using the following formulas find the values of a and b,
which are minor and major axis of ellipse respectively,

⎡ μ 1N − μ − 1N ⎤
a = ΩP ⎣ ⎦
2
⎡ μ 1N + μ − 1N ⎤
b = ΩP ⎣ ⎦
2
where
μ = ε −1 + ε −2 + 1
ε = 100.1α − 1P
4. Calculate the poles of Chebyshev filter which die on an

ellipse by using the formula
SK = a cos φ K + jb sin φ K K = 1, 2,......N

π ⎛ 2 K − 1⎞
where φ K = +⎜ π K = 1, 2,......N
2 ⎝ 2 N ⎟⎠
5. Find the denominator polynomial of the transfer function

using the above poles.
6. The number of the transfer function depends on the value
of N.
(ii) Given ω S = 0.6π , ω P = 0.2π

1
= 0.2 ⇒ λ = 4.899
1+ λ2
1
= 0.8 ⇒ ε = 0.75
1+ ε2
2 ω
ΩP = tan P = 0.6498
T 2
2 ωS
Ω S = tan = 2.752
T 2
cosh −1 λ ε
N= = 1.208
ΩS
cosh −1 ΩP
N =2
μ = ε −1 + 1 + ε −2 = 0

⎡ μ 1N − μ − 1N ⎤
a = ΩP ⎢ ⎥ = 0.3752
⎢⎣ 2 ⎥⎦
⎡ μ N + μ− N ⎤
1 1
b = ΩP ⎢ ⎥ = 0.75
⎣⎢ 2 ⎦⎥
π (2 K − 1)π
φK = + K = 1, 2
3 2N
SK = a cos φ K + jb sin φ K
S1 = −0.2653 + j 0.53
S2 = −0.2653 − j 0.53
Denominator
H ( S ) = ( S + 0.2653)2 + (0.53)2
= S 2 + 0.5306S + 0.3516
0.3516
For N even Numerator H ( S ) is 1
= 0.28
[1 + (0.75)2 ] 2
0.28
H (S ) =
S 2 + 0.5306S + 0.3516
Bilinear transformation
2 ⎛ 1 − z −1 ⎞
H (Z ) = H (Z ) S = ⎜ ⎟
T ⎝ 1 + z −1 ⎠
0.28(1 + z −1 )2
H (Z ) =
5.4128 − 7.298 z −1 + 3.29 z −2
0.052(1 + z −1 )2
H (Z ) =
1 − 1.3480 z −1 + 0.608 z −2

NOV/DEC 2009
Fifth Semester
PART A (10 × 2 = 20 marks)
1. Derive the necessary and sufficient conditions for an LTI system to be
BIBO stable.
2. What is a shift invariant system?
3. List any two properties of fourier transform.
4. Draw the basic butterfly diagram for the computation in the radix-2
decimation-in-frequency FFT algorithm.
5. Compare IIR and FIR filters.

1
6. Convert H ( s) = into a digital filter using approximation of deriva-
s +1
2
tives with T = 0.1 sec.
7. Define Gibb’s phenomenon.
8. State the condition satisfied by linear phase FIR filter.
9. Define sampling rate conversion.
10. Define sub-band coding.
PART B (5 × 16 = 80 marks)
11. (a) (i) The impulse response of a linear time invariant system is
h(n) = {1, 2, 1, −1}. Determine the response of the system to
the input signal x(n) = {1, 2, 3, 1}.

(ii) Determine the range of values of the parameter ‘a’ for which
the linear time-invariant system with impulse response h(n) =
an u(n) is stable.
Or
(b) (i) State and explain the properties of Z-transform.
1
(ii) Determine the inverse Z-transform of X ( z ) =
1 − 1.5 z −1 + 0.5 z −2
If (1) ROC: z > 1 (2) ROC: z < 0.5 (3) ROC: 0.5 < z < 1.
12. (a) (i) Compute 8 point DFT using DIF FFT radix 2 algorithm.
(ii) Mention the differences and similarities between DIT and DIF
FFT algorithms.
Or
(b) (i) List the steps involved for the radix-2 DIT-FFT algorithm.
Explain.
(ii) Using DIT FFT radix 2 algorithm convolve x(n) = {1, −1, 2}
and h(n) = {2, 2}.
13. (a) Explain in detail the steps involved in the design of IIR filter using
bilinear transformation.
Or
(b) Determine the cascade and parallel realization for the system,
described by the system function.
⎛ 1 ⎞⎛ 2 ⎞
10 ⎜1 − z −1 ⎟ ⎜1 − z −1 ⎟ (1 + 2 z −1 )
⎝ 2 ⎠⎝ 3 ⎠
H (z) =
⎛ 3 −1 ⎞ ⎛
1 −1 ⎞ ⎛ 1 −2 ⎞
⎜⎝1 − z ⎟⎠ ⎜⎝1 − z ⎟⎠ ⎜⎝1 − z + z ⎟⎠
−1
4 8 2
14. (a) Design a FIR low pass filter having following specifications
H d ( e jw ) = 1for −π 6 ≤ w ≤ π 6
0 for otherwise
and given that N = 7 using
(i) Hanning window
(ii) Hamming window
(iii) Blackman window
Or

(b) Determine the coefficients of a linear phase FIR filter of length

M = 15 which has a symmetric unit sample response and a fre-
quency response that satisfies the conditions
⎛ 2π k ⎞
Hr ⎜ = 1for k = 0,1,2,3
⎝ 15 ⎟⎠
0.4 for k = 4
0 for k = 5,6,7
15. (a) Discuss the effect of finite word length.

Or
(b) Explain multirate digital signal processing.

Solutions
PART A
1. Condition for an LTI system to be BIBO stable:
∞
∑
K =−∞
h( K ) < ∞.
When this condition is satisfied, then the system will be stable. The above
condition states that the LTI system is stable if its unit sample response
is absolutely summable.
2. A relaxed system H is time invariant or shift invariant if and only if

H y(t ) implies that, x(t − m) H y(t − m)
x(t ) ⎯⎯→ ⎯⎯→
for every input signal x(t) and every time shift m.
(i.e.) in time invariant systems, if y(t) = H{x(t)} then
y(t − m) = H {x(t − m)} .
3. (i) Linearity
Let F { x,(t )} = x1 ( j Ω); F {x2 (t )} = x2 ( j Ω).
The linearity property of fourier transform says that,
F {a1 x1 (t ) + a2 x2 (t )} = a1 x1 ( j Ω) + a2 x2 ( j Ω).
(ii) Time shifting

The time shifting property of fourier transform says that, if
F {x(t )} = X ( j Ω) then
F { x(t − t0 )} = e − jΩ0t ⋅ X ( j Ω).
1 a+b 1
4. a A=a+b
1
1
WNK
b B = (a – b) WNK
–1 a–b

5. FIR filter IIR filter
1. These filters can be easily designed These filters do not have

to have perfectly linear phase. linear phase.
2. FIR filters can be realized IIR filters are easily realizes
recursively and non-recursively. recursively.
3. Greater flexibility to control the Less flexibility, usually limited
shape of their magnitude response. to specific kind of filters.
4. Errors due to round off noise are The round off noise in IIR
less severe in FIR filters, mainly filters are more.
because of feedback not used.
6. For approximation of derivatives,
2 ⎛ 1 − z −1 ⎞
s= ⎜ ⎟
T ⎝ 1 + z −1 ⎠
H ( s)
∵ H ( z) = .
2 ⎛ 1 − z −1 ⎞
s= ⎜ ⎟
T ⎝ 1 + z −1 ⎠
1
= 2
4 ⎛ 1 − z −1 ⎞
⎜ ⎟ +1
T 2 ⎝ 1 + z −1 ⎠
0.01 (1 + z −1 )2
=
4(1 − z −1 )2 + T 2 (1 + z −1 )2
0.01(1 + z −1 )2
=
4(1 + z −2 − 2 z −1 ) + 0.01(1 + z −2 + 2 z −1 )
0.01(1 + z −1 )2
=
4 + 4 z −2 − 8 z −1 + 0.01 + 0.01z −2 + 0.02 z −1
0.01(1 + z −1 )2
H ( z) =
4.01z −2 − 7.98 z −1 + 4.01
7. Refer answer 6 from April/May 2011 Question paper.

8. The necessary and sufficient condition for linear phase characteristic in

FIR filter is the impulse response h(n) of the system should have the
symmetrical property,
h( n) = h( N − 1 − n)
where N is the duration of the sequences.
9. Sampling rate conversion is the process of converting the sequence x(n)

which is got from sampling the continuous time signal x(t) with a period
T1 of another sequence y(K) obtained from sampling x(t) with a period T 1.
10. In sub-band coding, the input signal is first split into number of non-
overlapping frequency bands by bandpass filters. The output of each
bandpass filter is decimated or down sampled by a factor m.
PART B
11. (a) (i) The output response y(n) = x(n) * h(n).
∴ z{y(n)} = z{x(n) * h(n)}
∴ Y(z) = X(z) . H(z)
∞
X ( z) = ∑ x ( n) z
n = −∞
−n
2
= ∑ x(n)z
n = −1
−n
(from given)
= x( − 1)z1 + x(0) + x(1) z −1 + x(2) z −2

X ( z ) = z + 2 + 3 z −1 + 1z −2
∞
Similarly , N ( z ) = ∑ h(n) z
n = −∞
−n
2
= ∑ h(n)z
n = −1
−n
= h( −1) z1 + h(0) z 0 + h(1) z −1 + h(2) z −2

= z + 2 + z −1 − z −2
∴ Y ( z ) = [ z + 2 + 3 z −1 + z −2 ][ z + 2 + z −1 − z −2 ]
= z 2 + 2 z + 1 − z −1 + 2 z + 4 + 2 z −1 − 2 z −2 + 3 z +0
+ 6 z −1 + 3 z −2 − 3 z −3 + z −1 + 2 z −2 + z −3 − z −4
Y ( z ) = z 2 + 4 z + 8 + 8 z −1 + 3 z −2 − 2 z −3 − z −4

Compare the above equation with the standard equation

∞
Y ( z) = ∑ y ( n) z
n = −∞
−n
= ........... + y( −2) z 2 + y( −1) z1 + y(0) + y(1) z −1 +

y(2) z −2 + y(3) z −3 + y(4) z −4 + ........
∴ y( n) = {1, 4,8,8,3, −2, −1}
(ii) If the system is stable, it should satisfy the condition,

∞
∑
n =−∞
h( n) < ∞
∞ ∞ ∞
∴ ∑ h( n) = ∑ a n u ( n) = ∑ a n
n =−∞ n =−∞ n= 0
Using infinite geometric series sum formula,

∞
1
∑a
n= 0
n
=
1− a
for the range 0 < a < 1
(b) (i) (1) Linearity property

The linearity property of Z-transform states that the Z-transform
of linear weighted combination of discrete time signal is equal
to similar linear weighted combination of Z-transform of indi-
vidual discrete time signals.
Let z{x1(n)} = X1(z) and z{x2(n)} = X2(z) then
z{a1x1(n) + a2x2(n)} = a1X1 (z) + a2X2 (z).
(2) Shifting property
The shifting property of Z-transform states that Z-transform of
a shifted signal shifted by m-units of time is obtained by multi-
plying Zm to Z-transform of unshifted signal.
Let z{x(n)} =X(z)
Then z{x(n-m)} = z-mX(z) and
z{x(n+m)}= zmX(z)
(3) Multiplication by ‘n’ (or Differentiation in Z-domain)
If z{x(n)}=X(z)
dX ( z )
Then z{nx(n)} = − z
dz

m
⎛ d⎞
In general, z{nm × ( n)} = ⎜ − z ⎟ × ( z ).
⎝ dz ⎠
(4) Multiplication by an exponential sequence,
an (or) Scaling in Z-domain
If z{x(n)} = X(z), then z{an x(n)} = X(a−1z)
(5) Time reversal
If z{x(n)} = X(z), then z{x(−n)} = X(z−1)
(6) Conjugation
If z{x(n)} = X(z), then z{x*(n)} = X*(z*)
(7) Convolution theorem
If z{x1(n)} = X1 (z), and z{x2 (n)} = X2(z)
Then z{x1(n)*x2(n)} = X1(z) X2(z)
(8) Correlation property
If z{x(n)} = X(z) and z{y(n)} = y(z),
Then z{rxy(m)} = X(z)Y(z−1)
∞
Where rxy ( m) = ∑ x(n)y(n − m).
n =−∞
(ii) Given,
1
X ( z) =
1 − 1.5 z −1 + 0.5 z −2
z2
X ( z) = 2
z (1 − 1.5 z −1 + 0.5 z −2 )
X ( z) z z
= 2 =
z z − 1.5 z + 0.5 ( z − 1)( z − 0.5)
A B
= +
z − 1 z − 0.5
∴ Using partial fraction expansion technique,
Z = A(z − 0.5) + B(z − 1)
Case (1)
Put z = 0.5
0.5 = B ( −0.5)
-1=B

Case (2)
Put Z = 1
1 = 0.5 A
1
=A
0.5
2=A
X ( z) 2 −1
∴ = +
z z − 1 z − 0.5
z z
X ( z) = 2 ⋅ − (1)
z − 1 z − 0.5
(i) For ROC: z > 1
Formulae:
⎧ z ⎫
z −1 ⎨ ⎬ = a n u( n) for ROC: z > a
⎩ z − a⎭
⎧ z ⎫
z −1 ⎨ ⎬ = − a n u( − n − 1) for ROC: z < a
⎩ z − a⎭
∴ From equation (1),
⎧ z ⎫ ⎧ z ⎫
x( n) = 2 z −1 ⎨ ⎬ = z −1 ⎨ ⎬
⎩ z − 1⎭ ⎩ z − 0.5 ⎭
= 2(1) n u( n) − (0.5) n u( n)
x( n) = [2 − (0.5) n ] u( n) for n ≥ 0
(ii) For ROC: z < 0.5

⎧ z ⎫ −1 ⎧ z ⎫
x( n) = 2 z −1 ⎨ ⎬−z ⎨ ⎬
⎩ z − 1⎭ ⎩ z − 0.5 ⎭
= 2( −(1) n u( − n − 1)) − ( −(0.5) n u( − n − 1))
= −2u( − n − 1) + (0.5) n u( − n − 1)
x( n) = [ −2 + (0.5) n ] u( − n − 1) for n ≤ −1
(iii) ROC: 0.5 < z < 1

⎧ z ⎫ −1 ⎧ z ⎫
x( n) = 2 z −1 ⎨ ⎬−z ⎨ ⎬
⎩ z − 1⎭ ⎩ z − 0.5 ⎭
= 2( −(1) n u( − n − 1)) − (0.5) n u( n)
x( n) = −2u( − n − 1) − (0.5) n u( n)

12. (a) (i) x(n) = {0, 1, 2, 3, 4, 5, 6, 7}

1 4 1 1 12 1 1
x(0) = 0 x(0) = 28
1
6 1 1 16 1
x(1) = 1 x(4) = –4
1 1 –1
8 1 –4 1 1
x(2) = 2 x(2) = –4 +4j
1 –1 1
1 10 1 –4 1
x(3) = 3 x(6) = –4–4j
–1 –j –1
1 – 4 +4j 1 1
x(4) = 4 x(1) = –4+j9.656
–1 –4 1 1
–4 j 5.656 1
x(5) = 5 x(5) = –4+– j1.656
–1 0.707 –j 0.707 1 1 –1
1 1
x(6) = 6 x(3) = –4+j1.656
–1 –4 –j –1 –4 – 4j 1
1
x(7) = 7 x(7) = –4– j9.656
–1 –4 –0.707 – –1 –5.656 –j –1
j 0.707
∴ X(K) = {28, −4 + j9.656, −4 + 4j, −4 + j1.656, −4, −4 − j1.656, −4 − 4j, −4 − j9.656}
(ii) Refer answer 11(b) from April/May 2011 Question paper.

(b) (i) • Let the number of input samples be N = 2M, where M is an
integer.
• The input sequence is shuffled through bit-reversal.
• The number of stages in the flow graph is M = log2N.
N
• Each stage consists of butterflies.
2
• Inputs/outputs for each butterfly are separated by 2m − 1
samples, where ‘m’ represents the stage index. (ie) for first
stage m = 1 and second stage m = 2.
• The number of complex addition is N log2 N.
N
• The number of complex multiplication is log2 N.
2
• The twiddle factor exponents are a function of the stage
index K,
K
− j 2π ×
WNK =e N
(ii) The no. of. Samples in the o/p sequence are

N1 + N2 − 1 = 3 + 2 −1 = 4
∴ x(n) = {1, −1, 2, 0}
h(n) = (2, 2, 0, 0}.

To determine X(K):
1 3 1 1
x(0) = 1 2 = x(0)
1
–1 1
x(2) = 2 –1+j = x(1)
1 –1 1
x(1) = –1 1 –1
4 = x(2)
1 1 –1
–1
x(3) = 0 –1– j = x(3)
1 –1 –j –1
∴ X(K) = {2, −1 + j, 4, −1−j}
1 2 1 1
h(0) = 2 4 = H(0)
1
2 1
h(2) = 0 2–2j = H(1)
1 –1 1
h(1) = 2 1 2
0 = H(2)
1 1 –1
2
h(3) = 0 2+2j = H(3)
1 –1 –j –1
∴ H(K) = {4, 2 − 2j, 0, 2 + 2j}
∴ Y(K) = X(K)H(K) = {8, 4j, 0, −4j}

1 8 1 1
Y(0) = 8 8 = y(0)
1
8 1
Y(2) = 0 0 = y(1)
1 –1 1
Y(1) = 4j 1 0
8 = y(2)
1 1 –1
8j
Y(3) = –4j 16 = y(3)
1 –1 +j –1
1
∴ y ( n) = {8, 0,8,16} = {2, 0, 2, 4} .
4
13. (a) Refer answer 12(b) from April/May 2011 Question paper.

(b) Cascade realization

H ( z ) = H1 ( z ) H 2 ( z ) H 3 ( z )
10 − 5 z −1
where H1 ( z ) =
3
1 − z −1
4
2
1 − z −1
H2 (z) = 3
1 −1
1− z
8
1 + 2 z −1
and H 3 ( z ) =
1
1 − z −1 + z −2
2
x(n)
10
+ + + + + +
1
Z –1 Z –1 Z –1
3/4 –5 1/8 2/3 + 1 2
Z –1
–1/2
Parallel realization
A B Cz −1 + D
H (z) = + +
3 1 1
1 − z −1 1 − z −1 1 − z −1 + z −2
4 8 2
(10 − 5 z −1 ) ⎛⎜⎝1 − z −1 ⎞⎟⎠ (1 + 2 z −1 )

2
3
=
⎛ 3 −1 ⎞ ⎛
1 −1 ⎞ ⎛
1 −2 ⎞
⎜⎝1 − z ⎟⎠ ⎜⎝1 − z ⎟⎠ ⎜⎝1 − z + z ⎟⎠
−1
4 8 2
A = 0.133 ⎫
B = −0.52 ⎪⎪
⎬ using partial fraction expansion method
C = 1.3867 ⎪
D = −0.213⎪⎭
0.133 0.52 1.3867 − 0.213 z −1

∴ H (z) = − +
3 1 1
1 − z −1 1 − z −1 1 − z −1 + z −2
4 8 2

1.33
+
Z –1
3/4
–5/2
+ +
x(n)
Z –1
3/4
+ + + y(n)
13.87
Z –1
+ +
Z –1
14. (a) H d ( e jw ) = 1 for −π ≤ w ≤ π

6 6
π
6
1
∴hd ( n) =
2π −
∫π 1⋅ e jwn
dw
6
π
1 ⎛ e jwn ⎞ 6
=
2π ⎜⎝ jn ⎟⎠ − π
6
1 ⎛e jπ n 6
− e − jπ n 6 ⎞
=
π n ⎜⎝ 2j ⎟⎠
1 ⎛ π n⎞
hd (n) = ⎜ sin ⎟⎠
πn ⎝ 6
for N = 7,
1 ⎛ π n⎞ 1
when n = 0, hd ( 0 ) = lt ⎜ sin ⎟⎠ = 6 = 0.1666
πn ⎝
n→ 0 6
1 ⎛ π⎞
when n = 1, hd (1) = ⎜ sin ⎟ = 0.1591 = hd ( −1)
π ⎝ 6⎠
1 ⎛ π⎞
when n = 2, hd ( 2) = ⎜ sin ⎟⎠ = 0.1378 = hd ( −2)
2π ⎝ 3
1 ⎛ π⎞
when n = 3, hd (3) = ⎜ sin ⎟⎠ = 0.1061 = hd ( −3)
3π ⎝ 2

1. Hanning window
2nπ ⎛ N − 1⎞ ⎛ N − 1⎞
wc ( n) = 0.5 + 0.5cos for − ⎜ ≤n≤⎜
N −1 ⎝ 2 ⎟⎠ ⎝ 2 ⎟⎠
πn
= 0.5 + 0.5cos for − 3 ≤ n ≤ 3
3
when n = 0, wc ( 0 ) = 1
when n = 1, wc (1) = wc ( −1) = 0.75
when n = 2, wc ( 2) = wc ( −2) = 0.25
when n = 3, wc (3) = wc ( −3) = 0
∴ h ( n) = hd ( n) × wc ( n)
when n = 0, h (0 ) = 0.1666 × 1 = 0.1666 = h (0 )

when n = 1, h (1) = 0.1591 × 0.75 = 0.1193 = h ( −1)
when n = 2, h (2) = 0.1378 × 0.25 = 0.0344 = h ( −2)
when n = 3, h (3) = 0 = h ( −3)
For symmetrical impulse response and ‘N’ is odd,

N −3
N −1
⎛ N − 1⎞ −⎛⎜⎝ 2 ⎞⎟⎠ 2
H (z) = h⎜ ⎟ z + ∑ h ( n) ( z − n + z −( N −1− n) )
⎝ 2 ⎠ n= 0
H ( z ) = h (3) z −3 + h (0 ) (1 + z −6 ) + h (1) ( z −1 + z −5 ) + h (2) ( z −2 + z −4 )

∴Y ( z ) = h (3) z −3 X ( z ) + h (0 ) ( X ( z ) + z −6 X ( z ))
+ h (1) ( z −1 X ( z ) + z −5 X ( z )) + h (2) ( z −2 X ( z ) + z −4 X ( z ))
X(z) Z –1 Z –1 Z –1
+ + +
Z –1 Z –1 Z –1
0.1666 0.1193 0.0344 h (3) =0
+ + + Y(z )

2. Hamming window
2π n ⎛ N − 1⎞ ⎛ N − 1⎞
w H ( n) = 0.54 + 0.46 cos for − ⎜ ⎟ ≤n≤⎜
N −1 ⎝ 2 ⎠ ⎝ 2 ⎟⎠
πn
= 0.54 + 0.46 cos for − 3 ≤ n ≤ 3
3
when n = 0, w H ( 0 ) = 1
when n = 1, w H (1) = w H ( −1) = 0.77
when n = 2, w H ( 2) = w H ( −2) = 0.31
when n = 3, w H (3) = w H ( −3) = 0.08
∴ h ( n) = hd ( n) × w H ( n)
when n = 0, h (0 ) = 0.1666 × 1 = 0.1666

when n = 1, h (1) = h ( −1) = 0.159 × 0.77 = 0.1224
when n = 2, h (2) = h ( −2) = 0.1378 × 0.31 = 0.0427
when n = 3, h (3) = h ( −3) = 0.1061 × 0.08 = 0.0085
for Symmetrical Impulse response and ‘N’ is odd
Y ( z ) = h (3) z −3 X ( z ) + h (0 ) ( X ( z ) + z −6 X ( z ))
+ h (1) ( z −1 X ( z ) + z −5 X ( z )) + h (2) ( z −2 X ( z ) + z −4 X ( z ))
X(z) Z –1 Z –1 Z –1
+ + +
Z –1 Z –1 Z –1
0.1 0.1224 0.0427 0.0085
+ + + Y(z )

3. Blackman window
2π n 4π n
w B ( n) = 0.42 + 0.5cos + 0.08cos
N −1 N −1
⎛ N − 1⎞ ⎛ N − 1⎞
for − ⎜ ≤n≤⎜
⎝ 2 ⎟⎠ ⎝ 2 ⎟⎠
πn 2π n
w B ( n) = 0.42 + 0.5cos + 0.08cos ;−3 ≤ n ≤ 3
3 3
n = 0, w B ( 0 ) = 1
n = 1, w B (1) = w B ( −1) = 0.63
n = 2, w B ( 2) = w B ( −2) = 0.13
n = 3, w B (3) = w B ( −3) = 0
∴h ( n) = hd ( n) × w B ( n)
n = 0, h ( 0 ) = 0.1666 × 1 = 0.1666
n = 1, h (1) = 0.159 × 0.63 = 0.1002 = h ( −1)
n = 2, h ( 2) = h ( −2) = 0.1378 × 0.13 = 0.0179
n = 3, h (3) = h ( −3) = 0
For symmetrical impulse response and ‘N’ is odd,
Y ( z ) = h (3) z −3 X ( z ) + h (0 ) ( X ( z ) + z −6 X ( z ))
+ h (1) ( z −1 + X ( z ) + z −5 X ( z )) + h (2) ( z −2 X ( z ) + z −4 X ( z ))
X(z) Z –1 Z –1 Z –1
+ + +
Z –1 Z –1 Z –1
0.1666 0.1002 0.0179 h (3) =0
+ + + Y(z )

⎧e − jw∞
⎪
(b) H d ( w ) = ⎨0.4 e − jw∞
⎪0 ⎧ − j ⎛⎝⎜ 215π k ⎞⎠⎟ ⎛⎝⎜ M2−1⎞⎠⎟
⎩ k = 0,1,2,3
⎪e ,
2π k ⎪⎪ − j⎜
⎛ 2π k ⎞ ⎛ M −1⎞
H ( K ) = H d (w )
⎟⎜ ⎟
w= = ⎨0.4 e ⎝ 15 ⎠ ⎝ 2 ⎠ , k=4
M ⎪0, k = 5,6,7
⎪
⎪⎩
⎧e − j14π k 15 , k = 0,1,2,3
⎪
∴ H ( K ) = ⎨0.4e − j14π k 15
, k=4
⎪0, k = 5,6,7
⎩
M −1
⎧ ⎫
1 ⎪ 2
j 2π kn M ⎪
∴ h ( n) =
M
⎨ ( ) ∑
H 0 + 2 Re ( ()
H k e )⎬
⎪ k =1 ⎪
⎩ ⎭
1 ⎧ 4
⎫
= ⎨1 + 2∑ Re ( H ( k ) e j 2π kn M )⎬
15 ⎩ k =1 ⎭
1 ⎧ ⎛ − j14π − j2nπ
⎞ ⎛ − j 28π − j 4nπ ⎞
= ⎨1 + 2 Re ⎜ e 15 ⋅ e 15 ⎟ + 2 Re ⎜ e 15 ⋅ e 15 ⎟
15 ⎩ ⎝ ⎠ ⎝ ⎠
⎛ − j 42π − j 6nπ ⎞ ⎛ − j 56π − j 8nπ ⎞ ⎫
+2 Re ⎜ e 15 ⋅ e 15 ⎟ + 2 Re ⎜ e 15 ⋅ e 15 ⎟ ⎬
⎝ ⎠ ⎝ ⎠⎭
1 ⎧ ⎛ 2π 4π
= ⎨1 + 2 cos ⎜⎝ (7 − n)⎞⎟⎠ + 2 cos ⎛⎜⎝ (7 − n)⎞⎟⎠
15 ⎩ 15 15
⎛ 6π ⎞ ⎛ 8π ⎞⎫
+2 cos ⎜ ( 7 − n)⎟ + 2 cos ⎜ ( 7 − n)⎟ ⎬ for n = 0 to14
⎝ 15 ⎠ ⎝ 15 ⎠⎭
when n = 0, h ( 0 ) = h (14 ) = −0.0141
when n = 1, h (1) = h (13) = −0.0019
when n = 2, h ( 2) = h (12) = 0.04
when n = 3, h (3) = h (11) = 0.0122
when n = 4, h ( 4 ) = h (10 ) = −0.0914
when n = 5, h (5) = h (9) = −0.0181
when n = 6, h (6) = h (8) = 0.3133
when n = 7, h ( 7) = 0.52

15. (a) In the case of FIR filters there are no limit cycle oscillations, if the
filter is realized in direct or cascade form, since these structure have
no feedback. However recursive realization of FIR system such as
the frequency. Sampling structures are subject to the above problems.
Let us consider a linear shift invariant 8/m with unit sample response
h(n) which is non-zero over the interval −0 ≤ n ≤ N − 1.
The convolution sum is given as,
N −1
y ( n) = ∑ h ( k ) x ( n − k ).
k =0
In the direct form realization structure, if the rounding is used, the

noise value at each multiplies can be assumed uniformly distributed
2 −2 b 2 −2 b
between ± with zero mean value and variance of
2 12
1. The sources lk ( n) are white sources.
2. The errors are uniformly distributed.
3. The error samples are uncorrelated with the input and each other.
Because the noise sources are assumed independent, the variance of
output noise is,
N × 2 −2 b
σε2 =
12
For an LTI system we can find the least upper bound on the output
sequence y(n) as,
N −1
y ( n) ≤ xmax ∑ h ( n)
n= 0
Where xmax is the maximum value of input sequence, inorder to avoid

overflow we have S0 |y(n)| < 10 for all n. This means the scaling factor,
1
50 < N −1
xmax ∑ h( n)
n= 0
2 −2 b ⎡ N −2i 2 ⎤
∴ The output is, σ ei2 = ∑ g (n)⎥⎦
2 ⎢⎣ n= 0 i
and the total output noise variance
μ
2 −2 b ⎡ M N −2i 2 ⎤
σ e2 = ∑ σ ei2 = ∑ ∑ g (n)
i =1 20 ⎢⎣ i =1 n= 0 k ⎥⎦

15. (b) These systems that use single sampling rate from A/D convert to
D/A converter are known as angle rate systems. The discrete time
systems that process data at more than one sampling rate are known
as multirate systems.
There are many cases where multirate signal processing is used, they are:
1. In high quality data acquisition and storage system.
2. In audio signal processing for ex CD is sampled at 44.1 KHz but
DAT in 48 KHz.
3. In video PAL and NTSC run at different sampling rates. Therefore
to watch an American Program in Europe, one needs a Sampling
rate converter.
4. In speech processing to reduce the storage space or the transmit-
ting rate of speech data.
5. In transmultiplexers
6. Narrowband filtering for fetal ECG and EEG.
Two basic operations in multirate signal processing are decimation
and interpolation. Decimative reduces the sampling rate, where as
interpolation increase the sampling rate.

B.E./B.Tech. DEGREE EXAMINATION
.0%&-1"1&3

Time: Three hoursMaximum: 100 marks
Answer ALL Questions
PART A (10 × 2 = 20 marks)
1. What is an LTI system?
2. What is meant by aliasing effect?
3. What is ROC of Z transform? State its properties.
4. Define discrete-time Fourier transform pair for a discrete sequence.
5. Find the 4-point DFT of the sequence x(n) = {1,1}.
6. What is FFT? What is its advantage?
7. What is the need for employing window for designing FIR filter?
8. What is Warping effect? What is its effect on frequency response?
9. What is pipelining? What are the different stages in pipelining?
10. What is the function of parallel logic unit in DSP processor?
Part B (5 × 16 = 80 marks)
11. (a) (i) What is causality and stability of a system? Derive the necessary
and sufficient condition on the impulse response of the system
for causality and stability. (8)

Digital Signal Processing 2.7
(ii) Determine the stability for each of the following linear systems :
∝
(1) y2(n) = ∑ (3 / 4)
k =0
k
x( n − k ) (3/4)k x(n - k)
∝
(2) y2(n) = ∑2
k =0
k
x( n − k ) 2k x(n - k)(8)
Or
(b) (i) What is meant by energy and power signal? Determine whether
the following signals are energy or power or neither energy nor
power signals.
n
(1) x2(n) =  1  n(n)
 2
π
(2) x2(n) = sin  n
6 
 πr π
(3) x2(n) = e  + 
 3 6
(4) x4(n) = e2x u(n)(12)
(ii) What is meant by sampling? Explain sampling theorem. (4)
n −n
12. (a) (i) Find the Z transform and its ROC of x(n) =  −1 n(n) + 5  1 
 5  2
u(- n - 1) (6)
 1
(ii) A system is described by the different equation y(n) =   y(n - 1)
 2
n
= 5x(n). Determine the solution, when the input x(n) =  1  u(n)
 5
and the initial condition is given by y(- 1) = 1, using Z transform.
(10)
Or
(b) (i) Determine the impulse response of the system described by the
 1
difference equation y(n) = y(n -1) -   y(n - 2) + x(n) + x(n - 1)
 2
using Z transform and discuss its stability. (10)
(ii) Find the linear convolution of x(n) = {2,4,6,8,10} with h(n) =
{1,3,5,7,9}.(6)
13. (a) (i) State and prove convolution property of DFT. (6)
(ii) Find the inverse DFT of

{
X(K) = 7, − 2 − j 2 , − j , 2 − j 2,1, 2 + j 2 , j , − 2 + j 2 (10) }
Or
(b) (i)
Derive decimation-in-time radix-2 FFT algorithm and draw
signal flow graph for 8-point sequence. (8)
Using FFT algorithm, compute the DFT of x(n) =
(ii)
{2,2,2,2,1,1,1,1}.(8)
14. (a) (i) Obtain cascade and parallel realization for the system having
difference equation.
y(n) + 0.1y(n - 1) - 0.2y(n - 2) = 3x(n) + 3.6x(n - 1) + 0.6x(n - 2)
(8)
(ii) Design a length 5 FIR band reject filter with a lower cut-off
frequency of 2KHz, an upper cut-off frequency of 2.4 KHz, and
a sampling rate of 8000 Hz using Hamming window. (8)
Or
(b) (i) Explain impulse invariant method of designing IIR filter. (6)
(ii) Design a second order digital low pass Butterworth filter with
a cut-off frequency 3.4 KHz at a sampling frequency of 8 KHz
using bilinear transformation. (10)
15. (a) (i) Draw the block diagram of Harvard architecture and explain.
(8)
(ii) Explain the advantage and disadvantages of VLIW architecture.
(8)
Or
(b) Write short notes on :
(i) Memory mapped register addressing
(ii) Circular addressing mode
(ii) Auxiliary registers.

3-Digital Signal Processing

Uploaded by

3-Digital Signal Processing

Uploaded by

Semester-V

EEE_Semester-V_Ch05.indd 1 7/13/2012 1:08:10 PM

EEE_Sem-VI_Chennai_FM.indd iv 12/7/2012 6:40:43 PM

3. Determine the z-transform & ROC for the s/l

4. Prove the convdution property of z-transform.

5. Draw the butterfly dig. for decimation in time FFT Algorim<??>.

9. What is meant by b it reversed addressing mode? What the application

10. Compare the RISC & CISC processors.

DSP_Nov-Dec 2013-QP.indd 1 4/23/2014 3:52:35 PM

(i) x1(n) = x(n – 2) m (n – 3)

(ii) x2 (n) = x(n + 1) μ(n – 1)

(iii) y(n) = (x(n + 1) + x(n) + x(n – 1))

(iv) y(n) = max (x(n + 1), x(n), x(n – 1))

(ii) Linear or non-linear

(iii) Time invariant or time varying

(iv) Stable or unstable

Examine the following system with respect to the properties above

DSP_Nov-Dec 2013-QP.indd 2 4/23/2014 3:52:36 PM

(ii) Compute the 8-pt DFT of the sequence x = [1, 1, 1, 1, 1, <??>

(i) Determine a sequence y(n) so that y(k) = x1(k) x2(k).

(ii) Discuss the addressing modes used in programmable DSP?<??>

DSP_Nov-Dec 2013-QP.indd 3 4/23/2014 3:52:36 PM

2. Determine whether x[n] = u[n] is a power signal or an energy signal.

3. Given a difference equation y(n) = x[n] + 3x[n - 1] + 2y[n - 1]. Determine

5. Find the discrete Fourier transform for δ [n].

6. Draw the basic butterfly diagram for DIF algorithm.

7. Name two methods for digitizing the transfer function of an analog

8. List the properties of chebyshev filter.

9. Mention one important feature of Harvard architecture.

10. What is the advantage of pipelining?

EEE_Semester-V_Ch02.indd 3 4/19/2014 3:11:00 PM

EEE_Semester-V_Ch02.indd 4 4/19/2014 3:11:01 PM

using a parallel interconnection of first and second order

x [n] + w [n] y [n]

Where x[n] = δ [n]

EEE_Semester-V_Ch02.indd 5 4/19/2014 3:11:02 PM

1. Nyquist rate  2o  1000

It is not an energy signal

P  0 It is noy a Power signal

13 a) overlap save method:

x3 [n]   3, 2, 0,1, 2 x4 [n]  1, 2,1, 0, 0

The twiddle factors associated with the flow graph are

W83   e  j 2 /8   0.707  j 0.707

X ( k )   20, 5.828  j 2.414, 0, 0.172  j 0.414,

w[n]  x[n]  h1 [n]

2. What is meant by Nyquist rate?

3. Relate Z-transform and DFT mathematically.

4. State BIBO stability criterion in Z-transformed domain.

5. How DFT is different than DTFT?

7. Define an FIR filter.

8. State the advantages of IIR filter over FIR.

9. What is the need for specialized digital signal processor?

EEE_Semester-V_Ch02.indd 9 4/19/2014 3:11:03 PM

(b) (i) Define the error in quantization of sinusoidal signals. (6)

EEE_Semester-V_Ch02.indd 10 4/19/2014 3:11:04 PM

Ripples in stopband > 40 dB, Stopband cutoff freq. = 6 KHz

EEE_Semester-V_Ch02.indd 11 4/19/2014 3:11:04 PM

T[a1x1(n) + a2x2(n)] = a1T[x1(n)] + a2T[x2(n)] = a1y1(n) + a2y2(n)

Time Invariant property:

y(n,k) = y(n – k) y(n – k) = T[x(n – k)]

H(z)   n  h(n)

Sr. DTFT DFT

The round off noise in IIR filters is more.

t = nT with weight x(nT).

(II) truncation: Discard the excess bit.

11 b) (ii) Signal to quantization error (SNQR):

eq(t) = xa(t) – xq(t)

SQNR (dB) 10log10 SQNR = 1.76 + 6.02b

x (n) = –0.5 * (0.5)n u(n) + 2 * (1)n u(n)

The value of Z is placed on outside unit circle. ROC: it is causal system.

For x(n) = 0.5n u(–n) means ROC z < 0.5

The value of Z is placed inside the unit circle. It is anti-causal signal.

12 a) ii) causal and anti – causal signal:

12 b) Discrete Fourier Transform:

(i) DTFT of the given signal is formulated as below

X (e j )   n  x(n)e j n   n  a|n| e j n

3. Given a difference equation y(n) = x[n] + 3x[n - 1] + 2y[n - 1]. Determine

7. Name two methods for digitizing the transfer function of an analog

(b) (i) Define the error in quantization of sinusoidal signals. (6)

Ωp = (2/T)tan((wpT)/2) = 48000tan((8000π)/24000)(4.15 × 10–5)) = 1046.4 rad/sec

6. Give relationship between DTFT and Z transform. What is meant by

⎧last two data