Digital Signal Processing
Uploaded by
ganapathi7898Digital Signal Processing
Uploaded by
ganapathi7898CONTENTS
S.no Name of chapter Page
no
1. 1. DIGITAL SIGNAL PROCESSING 1
1.1 Advantages of DSP 1
1.2 Disadvantages of DSP 1
1.3 Applications of DSP 2
1.4 Classification of Signals 2
1.5 Operation on Signals: 13
1.6 Classification of Systems 15
1.7 properties of LTI system 17
1.8 Solution of linear constant-co-efficient difference equation 24
1.9 Concept of frequency in continuous-time and discrete-time. 27
2. 2. POWER, ENERGY and CONVOLUTION 34
2.1 Energy and Power 35
2.2 Discrete convolution 37
2.2.1 Overlap add and save method 41
3. 3. DISCTRETE FOURIER SERIES 43
3.1 Properties of DFS 45
46
3.2 Properties of DTFT
51
3.3 Fourier transform of Discrete time signals
4. 4. DISCRETE FOURIER TRANSFORM 54
4.1 DFT as a linear transformation 59
4.2 DFT properties 63
5. 5.FFT 74
5.1 DITFFT 74
5.2 DIFFFT 83
6. 6.DIGITAL FILTER STRUCTURE 87
1
6.1 Direct Form – I 87
6.2 Direct form-II 87
6.3 Cascade Combination of second-order section (CSOS) 88
6.4 Parallel Combination of Second Order Section (PSOS) 90
6.5 Jury – Stability Criterion 91
6.6 Difference between recursive and non-recursive filters 93
6.7 Difference between FIR and IIR filters 93
7. 7.IIR FILTER DESIGN 95
7.1 Butter worth filter design. 95
7.2 Chebyshev filter design 101
7.3 Impulse invariant method 104
7.4 Bilinear transformation 108
8. 8.FIR FILTER DESIGN 114
8.1 Type 1 Sequence 114
8.2 Type 2 Sequence 115
8.3 Type 3 Sequence 116
8.4 Type 4 Sequence 117
8.5 Poles & Zeros of linear phase sequences: 118
8.6 Fourier series Method 120
8.7 Windowing 125
9. 9.OBJECTIVE PAPER-1 135
10. 10.OBJECTIVE PAPER-2 136
11. 11.OBJECTIVE PAPER-3 138
12. 12.OBJECTIVE PAPER-4 140
13. 13.OBJECTIVE PAPER-5 141
14. 14.OBJECTIVE PAPER-6 143
15. 15.OBJECTIVE PAPER-7 144
2
16. 16.OBJECTIVE PAPER-8 146
17. 17.OBJECTIVE PAPER-9 147
148
18.OBJECTIVE PAPER-10
18. 19.OBJECTIVE PAPER-11 151
19. 20. OBJECTIVE PAPER-12 153
20. 21.OBJECTIVE PAPER-13 155
21. 22.OBJECTIVE PAPER-14 157
22. 23.OBJECTIVE PAPER-15 158
23. 24. REFERENCES 162
3
1. DIGITAL SIGNAL PROCESSING
A signal is defined as any physical quantity that varies with time, space or
another independent variable.
A system is defined as a physical device that performs an operation on a
signal.
System is characterized by the type of operation that performs on the signal.
Such operations are referred to as signal processing.
1.1 Advantages of DSP
1. A digital programmable system allows flexibility in reconfiguring the digital
signal processing operations by changing the program. In analog redesign of
hardware is required.
2. In digital accuracy depends on word length, floating Vs fixed point
arithmetic etc. In analog depends on components.
3. Can be stored on disk.
4. It is very difficult to perform precise mathematical operations on signals in
analog form but these operations can be routinely implemented on a digital
computer using software.
5. Cheaper to implement.
6. Small size.
7. Several filters need several boards in analog, whereas in digital same DSP
processor is used for many filters.
1.2 Disadvantages of DSP
1. When analog signal is changing very fast, it is difficult to convert digital
form .(beyond 100KHz range)
2. w=1/2 Sampling rate.
3. Finite word length problems.
4. When the signal is weak, within a few tenths of millivolts, we cannot
amplify the signal after it is digitized.
4
5. DSP hardware is more expensive than general purpose microprocessors &
micro controllers.
6. Dedicated DSP can do better than general purpose DSP.
1.3 Applications of DSP
1. Filtering.
2. Speech synthesis in which white noise (all frequency components present to
the same level) is filtered on a selective frequency basis in order to get an audio
signal.
3. Speech compression and expansion for use in radio voice communication.
4. Speech recognition.
5. Signal analysis.
6. Image processing: filtering, edge effects, enhancement.
7. PCM used in telephone communication.
8. High speed MODEM data communication using pulse modulation systems
such as FSK, QAM etc. MODEM transmits high speed (1200-19200 bits per
second) over a band limited (3-4 KHz) analog telephone wire line.
9. Wave form generation.
1.4 Classification of Signals
I. Based on Variables:
1. f(t)=5t : single variable
2. f(x,y)=2x+3y : two variables
3. S1= A Sin(wt) : real valued signal
4. S2 = A ejwt : A Cos(wt)+j A Sin(wt) : Complex valued signal
S1(t )
5. S4(t)= S 2(t ) : Multichannel signal
S 3(t )
Ex: due to earth quake, ground acceleration recorder
Ir ( x, y, t )
6. I(x,y,t)= Ig ( x, y, t ) multidimensional
Ib( x, y, t )
5
II. Based on Representation:
III. Based on duration.
1. right sided: x(n)=0 for n<N
2. left sided :x(n)=0 for n>N
3. causal : x(n)=0 for n<0
4. Anti causal : x(n)=0 for n 0
5. Non causal : x(n)=0 for n >N
IV. Based on the Shape.
1. (n)=0 n0
=1 n=0
2. u (n) =1 n0
=0 n<0
6
Arbitrary sequence can be represented as a sum of scaled, delayed impulses.
P (n) = a-3 (n+3) +a1 (u-1) +a2 (u-2) +a7 (u-7)
Or
x(n) = x(k ) (n k )
k
n
u(n) =
k
(k ) = (n) + (n-1)+ (n-2)…..
=
k 0
(n k )
3.Discrete pulse signals.
Rect (n/2N) =1 n N
=0 else where.
5.Tri (n/N) = 1- n /N n N
=0 else where.
1. Sinc (n/N)= Sa(n /N) = Sin(n /N) / (n /N), Sinc(0)=1
Sinc (n/N) =0 at n=kN, k= 1, 2…
Sinc (n) = (n) for N=1; (Sin (n ) / n =1= (n))
6.Exponential Sequence
x (n) = A n
If A & are real numbers, then the sequence is real. If 0< <1 and A is +ve,
then sequence values are +ve and decreases with increasing n.
For -1< <0, the sequence values alternate in sign but again decreases in
magnitude with increasing n. If >1, then the sequences grows in magnitude as n
increases.
7.Sinusoidal Sequence
x(n) = A Cos(won+ ) for all n
7
8.Complex exponential sequence
If = ejwo
A = A ej
x(n) = A ej n ejwon
= A n Cos(won+ ) + j A n Sin(won+ )
If >1, the sequence oscillates with exponentially growing envelope.
If <1, the sequence oscillates with exponentially decreasing envelope.
So when discussing complex exponential signals of the form x(n)= A e jwon or
real sinusoidal signals of the form x(n)= A Cos(won+ ) , we need only consider
frequencies in a frequency internal of length 2 such as < Wo < or 0
Wo<2 .
V. Deterministic (x (t) = t x (t) = A Sin(wt))
& Non-deterministic Signals. (Ex: Thermal noise.)
VI. Periodic & non periodic based on repetition.
VII. Power & Energy Signals
Energy signal: E = finite, P=0
Signal with finite energy is called energy signal.
Energy signal have zero signal power, since averaging finite energy over
infinite time. All time limited signals of finite amplitude are energy signals.
Ex: one sided or two sided decaying. Damped exponentials, damped sinusoidal.
x(t) is an energy signal if it is finite valued and x2 (t) decays to zero fasten
1
than as t .
t
8
Power signal: E = , P 0, P Ex: All periodic waveforms
Neither energy nor power: E= , P=0 Ex: 1/ t t 1 E= , P= , Ex: tn
VIII. Based on Symmetry
1. Even x(n)=xe(n)+xo(n)
2. Odd x(-n)=xe(-n)+xo(-n)
3. Hidden x(-n)=xe(n)-xo(n)
1
4. Half-wave symmetry. xe(n)= [x(n)+x(-n)]
2
1
xo(n)= [x(n)-x(-n)]
2
Signal Classification by duration & Area.
a. Finite duration: time limited.
b. Semi-infinite extent: right sided, if they are zero for t < where = finite
c. Left sided: zero for t >
Piecewise continuous: possess different expressions over different intervals.
Continuous: defined by single expressions for all time. x(t) = sin(t)
Periodic: xp (t) = xp (t nT)
T
1
For periodic signals P = x(t ) 2 dt
T 0
9
X rms = P
For non periodic
T
1
P = Lt
To 0
x(t ) 2 dt
To
Xavg = Lt x(t )dt
0
x(t) = A cos( 2 fo t + ) P=0.5 A2
x(t) = A e j( 2 fo t + ) P=A2
1 2 1 2
E= A2 b E= A b E= A b
2 3
Q.
e - t dt =
1
0
10
Q.
1 2 1
Ex = A 0.5T + (-A)2 0.5T = 0.5 A2 T
2 2
Px = 0.5 A2
Q.
1 1 2
Ey = [ A2 0.5T] 2 = A T
3 3
1 2
Py = A
3
x(t) = A ejwt is periodic
T
1
Px = x(t ) 2 dt = A2
T 0
x(2t -6 ): compressed by 2 and shifted right by 3 OR shifted by 6 and
compressed by 2.
x(1-t): fold x(t) & shift right by 1 OR shift right and fold.
x(0.5t +0.5) Advance by 0.5 & stretched by 2 OR stretched by 2 & advance
by 1.
11
(t 2 ) t 2
y (t) = 2 x [- ] = 2 x[ ] 2 x( t + ) ; 5 + =-1; - + =1 =>
3 3 3
= -1/3 ; = 2/3
Area of symmetric signals over symmetric limits (- , )
Odd symmetry:
x0 (t) dt =0
Even symmetry:
xe (t) dt = 2
0
xe (t) dt
Xe (t) +Ye (t): even symmetry.
Xe (t) Ye (t): even symmetry.
Xo (t) +Yo (t): odd symmetry.
Xo (t) Xo (t): even symmetry.
Xe (t) +Yo (t): no symmetry.
Xe (t) Yo (t): odd symmetry.
1
Xe(n)= [x(n)+x(-n)]
2
1
Xo (n) = [x (n)-x (-n)]
2
Area of half-wave symmetry signal always zero.
Half wave symmetry applicable only for periodic signal.
F0 = GCD ( f1,f2)
T = LCM (T1, T2)
Y(t) = x1(t) + x2(t)
Py= Px1+Px2
Y(t)rms = Py
12
U(0) = 0.5 is called as Heaviside unit step.
X(t) = Sin(t) Sin( t)
= 0.5 cos (1- )t – 0.5 cos (1+ ) t
W1=1-
W2=1+ almost periodic OR non periodic.
Px = 0.5[0.52 +0.52] =0.25 W
Area of any sinc or Sinc 2 equals area of triangle ABC inscribed within the main
lobe.
Even though the sinc function is square integrable ( an energy signal) , it is not
1
absolutely integrable( because it does not decay to zero faster than )
t
(t) = 0 t 0
= t=0 ( )d
=1
An impulse is a tall narrow spike with finite area and infinite energy.
The area of impulse A (t) equals A and is called its strength. How ever its
hight at t=0 is .
13
= 2 (t) – 2e-t u(t)
2 e-t (t) = 2 (t)
1
[ [t- ]] = (t )
2 2
I2= cos(2 t ) (2t 1)dt
4
= cos(2 t )0.5 (t 0.5)dt = 0.5 cos(2 t) at t=-0.5 = -
4
0.5
x1(t) = x(t)
k
(t-kts ) =
k
x(kts) (t-kts)
x1(t) is not periodic.
The doublet
’(t) =0 t 0
14
= undefined t=0 ' (t )dt 0
’ (-t) = - ’ (t) then Odd
function.
1
[ [t- ]] = (t )
Differentiating on both sides
1
’ [ [t- ]] = ' (t )
With =-1
’ (-t) = - ’ (t)
d
[ x(t ) (t )] = x’ (t) (t- ) + x (t) ’ (t- )
dt
= x’ ( ) (t- ) + x (t) ’ (t- )-----------1
Or
d d
[ x(t ) (t )] = [ x( ) (t )] = x ( ) ’ (t- ) -----------2
dt dt
1=2
x’ ( ) (t- ) + x (t) ’ (t- ) = x ( ) ’ (t- )
x (t) ’ (t- ) = x ( ) ’ (t- ) - x’ ( ) (t- )
x (t) ’ (t- ) dt =
x ( ) ’ (t- ) dt -
x’ ( ) (t- ) dt
= 0- x’ ( ) = - x’ ( )
Higher derivatives of (t) obey n(t) = (-1)n n(t) are alternately odd and
even, and possess zero area. All are eliminating forms of the same sequence
that generate impulses, provided their ordinary derivatives exits. None are
absolutely integrable. The impulse is unique in being the only absolutely
integrable function from among all its derivatives and integrals (step, ramp etc)
What does the signal x(t) = e-t ’(t) describe?
x(t) = ’ (t) – (-1) (t) = ’ (t) + (t)
2
I = [(t 3) (2t 2)] 8 cos( t ) ' (t 0.5)]dt
2
15
d
= 0.5 (t-3) t 1 - 8 [cos t ] t 0.5
dt
= 23.1327 Answer.
1.5 Operation on Signals:
1. Shifting.
x(n) shift right or delay = x(n-m)
x(n) shift left or advance = x(n+m)
2. Time reversal or fold.
x(-n+2) is x(-n) delayed by two samples.
x(-n-2) is x(-n) advanced by two samples.
Or
x(n) is right shift x(n-2), then fold x(-n-2)
x(n) fold x(-n) shift left x(-(n+2)) = x(-n-2)
Ex:
x(n) = 2, 3 , 4 , 5, 6, 7 .
Find 1. y(n)=x(n-3) 2. x(n+2) 3. x(-n) 4. x(-n+1) 5. x(-n-2)
1. y(n)= x(n-3) = { 0 ,2,3,4,5,6,7} shift x(n) right 3 units.
2. x(n+2) = { 2,3,4,5, 6 ,7} shift x(n) left 2 units.
3. x(-n) = { 7,6,5, 4 ,3,2} fold x(n) about n=0.
4. x(-n+1) = { 7,6, 5 ,4,3,2} fold x(n), delay by 1.
5. x(-n-2) = { 7,6,5,4,3,2} fold x(n), advanced by 2.
3. a. Decimation.
Suppose x(n) corresponds to an analog signal x(t) sampled at intervals Ts. The
signal y(n) = x(2n) then corresponds to the compressed signal x(2t) sampled at Ts
and contains only alternate samples of x(n)( corresponding to x(0), x(2), x(4)…).
We can also obtain directly from x(t) (not in compressed version). If we sample it
1
at intervals 2Ts (or at a sampling rate Fs = ). This means a two fold reduction
2Ts
16
in the sampling rate. Decimation by a factor N is equivalent to sampling x(t) at
intervals NTs and implies an N-fold reduction in the sampling rate.
b. Interpolation.
y(n) = x(n/2) corresponds to x(t) sampled at Ts/2 and has twice the length of
x(n) with one new sample between adjacent samples of x(n).
The new sample value as ‘0’ for Zero interpolation.
The new sample constant = previous value for step interpolation.
The new sample average of adjacent samples for linear interpolation.
Interpolation by a factor of N is equivalent to sampling x(t) at intervals Ts/N
and implies an N-fold increase in both the sampling rate and the signal length.
Ex: Decimation Step interpolation
{ 1 , 2, 6, 4, 8} { 1 , 6, 8} { 1 , 1, 6, 6, 8, 8}
n 2n n n/2
Step interpolation Decimation
{ 1 , 2, 6, 4, 8} { 1 , 1,2,2,6, 6,4,4,8, 8} { 1 , 2, 6, 4, 8}
n n/2 n 2n
Since Decimation is indeed the inverse of interpolation, but the converse is not
necessarily true. First Interpolation & Decimation.
Ex: x(n) = { 1 1, 2, 5, -1}
x(n/3) = { 1,0,0, 2 2,0,0,5,0,0,-1,0,0} Zero interpolation.
= { 1,1,1, 2 ,2,2,5,5,5,-1,-1,-1} Step interpolation.
4 5 2 1
= { 1, , , 2 , 3,4,5,3,1,-1, - ,- } Linear interpolation.
3 3 3 3
4. Fractional Delays.
M
It requires interpolation (N), shift (M) and Decimation (n): x (n - ) = x (
N
( Nn M )
)
N
2n 1
x(n) = {2, 4, 6 , 8}, find y(n)=x(n-0.5) = x ( )
2
17
g(n) = x (n/2) = {2, 2, 4, 4, 6 , 6, 8,8} for step interpolation.
n 1
h(n) =g(n-1) = x( ) = {2, 2, 4, 4 , 6, 6,8,8}
2
2n 1
y(n) = h(2n) = x(n-0.5) = x( ) = {2, 4 , 6, 8}
2
OR
g(n) = x(n/2) = {2,3,4,5, 6 ,7,8,4} linear interpolation.
h(n) = g(n-1) = {2,3,4, 5 , 6, 7,8,4}
g (n) = h(2n)={3,5,7,4}
1.6 Classification of Systems
1. a. Static systems or memory less system. (Non Linear / Stable)
Ex. y(n) = a x (n)
= n x(n) + b x3(n)
= [x(n)]2 = a(n-1) x(n)
y(n) = [x(n), n]
If its o/p at every value of ‘n’ depends only on the input x(n) at the same value
of ‘n’
Do not include delay elements. Similarly to combinational circuits.
b. Dynamic systems or memory.
If its o/p at every value of ‘n’ depends on the o/p till (n-1) and i/p at the same
value of ‘n’ or previous value of ‘n’.
Ex. y(n) = x(n) + 3 x(n-1)
= 2 x(n) - 10 x(n-2) + 15 y(n-1)
Similar to sequential circuit.
2. Ideal delay system. (Stable, linear, memory less if nd=0)
Ex. y (n) = x(n-nd)
nd is fixed = +ve integer.
18
3. Moving average system. (LTIV ,Stable)
m2
y(n) = 1/ (m1+m2+1) x(n k )
k m1
This system computes the nth sample of the o/p sequence as the average of
(m1+m2+1) samples of input sequence around the nth sample.
If M1=0; M2=5
5
y(7) = 1/6 [ x (7 k ) ]
k 0
= 1/6 [x(7) + x(6) + x(5) + x(4) + x(3) + x(2)]
y(8) = 1/6 [x(8) + x(7) + x(6) + x(5) + x(4) + x(3)]
So to compute y (8), both dotted lines would move one sample to right.
4. Accumulator. ( Linear , Unstable )
n
y(n) = x(k )
k
n 1
= x(k ) + x(n)
k
= y(n-1) + x(n)
x(n) = { …0,3,2,1,0,1,2,3,0,….}
y(n) = { …0,3,5,6,6,7,9,12,12…}
O/p at the nth sample depends on the i/p’s till nth sample
Ex:
x(n) = n u(n) ; given y(-1)=0. i.e. initially relaxed.
19
1 n
y(n) = x(k ) + x(k )
k k 0
n n
n( n 1)
= y(-1) + x(k ) = 0 +
k 0
n
k 0
=
2
5. Linear Systems.
If y1(n) & y2(n) are the responses of a system when x1(n) & x2(n) are the
respective inputs, then the system is linear if and only if
[ x1(n) x 2(n)] = [ x1( n)] + [ x 2(n)]
= y1(n) + y2(n) (Additive property)
[ax(n)] = a [ x(n)] = a y(n) (Scaling or Homogeneity)
The two properties can be combined into principle of superposition stated as
[ax1(n) bx 2(n)] = a [ x1( n)] + b [ x 2(n)]
Otherwise non linear system.
6. Time invariant system.
Is one for which a time shift or delay of input sequence causes a corresponding
shift in the o/p sequence.
y(n-k) = [ x(n k )] TIV
TV
7. Causality.
A system is causal if for every choice of no the o/p sequence value at index n=
no depends only on the input sequence values for n no.
y(n) = x(n) + x(n-1) causal.
y(n) = x(n) + x(n+2) + x(n-4) non causal.
8. Stability.
For every bounded input x(n) Bx < for all n, there exists a fixed +ve finite
value By such that y(n) By < .
1.7 PROPERTIES OF LTI SYSTEM.
1. x(n) = x(k ) (n k )
k
20
y(n) = [ x(k ) (n k ) ]
k
for linear
x(k ) [ (n-k)] for time invariant
k
x(k )h(n k ) = x(n) * h(n)
k
Therefore o/p of any LTI system is convolution of i/p and impulse response.
y(no) = h(k ) x(no k )
k
1
= h(k ) x(no k ) + h(k ) x(no k )
k k 0
= h(-1) x(n0+1) + h(-2) x(n0+2)……….+h(0) x(n0) + h(1) x(n0-1) + ….
y(n) is causal sequence if h(n) =0 n<0
y(n) is anti causal sequence if h(n) =0 n 0
y(n) is non causal sequence if h(n) =0 |n|>N
Therefore causal system y(n) = h( k ) x ( n k )
k 0
n
If i/p is also causal y(n) = h( k ) x ( n k )
k 0
2. Convolution operation is commutative.
x(n) * h(n) = h(n) * x(n)
3. Convolution operation is distributive over additive.
x(n) * [h1(n) + h2(n)] = x(n) * h1(n) + x(n) * h2(n)
4. Convolution property is associative.
x(n) * h1(n) * h2(n) = [x(n) * h1(n)] * h2(n)
5 y(n) = h2 * w(n) = h2(n)*h1(n)*x(n) = h3(n)*x(n)
21
6
h (n) = h1(n) + h2(n)
7 LTI systems are stable if and only if impulse response is absolutely
summable.
y (n) = h( k ) x ( n k )
k
h(k ) x(n k )
k
Since x (n) is bounded x(n) bx<
y (n) Bx
k
h(k )
S=
k
h(k ) is necessary & sufficient condition for stability.
8 (n) * x(n) = x(n)
9 Convolution yields the zero state response of an LTI system.
10 The response of LTI system to periodic signals is also periodic with
identical period.
y(n) = h (n) * x(n)
= h( k ) x ( n k )
k
y (n+N) = h( k ) x ( n k N )
k
22
put n-k = m
= h ( n m) x ( m N )
m
= h ( n m) x ( m)
m
m=k
= h( n k ) x ( k )
k
= y(n) (Ans)
Q. y (n)-0.4 y(n-1) =x (n). Find causal impulse response? h(n)=0 n<0.
h(n) = 0.4 h(n-1) + (n)
h(0) = 0.4 h(-1) + (0) =1
h(1) = 0.4 h(0) = 0.4
h(2) = 0.42
h(n) = 0.4n for n 0
Q. y(n)-0.4 y(n-1) = x(n). find the anti-causal impulse response? h(n)=0 for n
0
h(n-1) = 2.5 [h(n)- (n) ]
h(-1) = 2.5 [h(0)- (0) ] = -2.5
h(-2) = -2.52 . …….. h(n) = -2.5n valid for n -1
Q. x(n)={1,2,3} y(n)={3,4} Obtain difference equation from i/p & o/p
information
y(n) + 2 y(n-1) + 3 y(n-2) = 3 x(n) + 4 x(n-1) (Ans)
Q. x(n) = {4,4,}, y(n)= x(n)- 0.5x(n-1). Find the difference equation of the
inverse system. Sketch the realization of each system and find the output of each
system.
Solution:
The original system is y(n)=x(n)-0.5 x(n-1)
The inverse system is x(n)= y(n)-0.5 y(n-1)
y (n) = x (n) – 0.5 x(n-1)
23
Y (z) = X (z) [1-0.5Z-1]
Y ( z)
=1-0.5 Z-1 System
X ( z)
Inverse System
y (n) – 0.5 y(n-1) =x(n)
Y (z) [1-0.5 Z-1] = X (z)
Y ( z)
[1-0.5 Z-1] -1
X ( z)
g (n) = 4 (n) - 2 (n-1) + 4 (n-1) - 2 (n-2) = 4 (n) + 2 (n-1) - 2 (n-2)
y (n) = 0.5 y(n-1) + 4 (n) + 2 (n-1) – 2 (n-2)
y (0) = 0.5y(-1) + 4 (0) = 4
y(1) = 4
y(2) = 0.5 y(1) - 2 (0) = 0
y(n) = {4, 4} same as i/p.
Non Recursive filters Recursive filters
N N
y(n) =
k
ak x(n-k) y(n) =
k 0
ak x(n-k) –
k 1
bk y(n-
for causal system k)
Present response is a function of
=
k 0
ak x(n-k)
24
For causal i/p sequence the present and past N values of the
N
excitation as well as the past N
y(n) =
k 0
ak x(n-k)
values of response. It gives IIR o/p
Present response depends
only on but not always.
present i/p & previous i/ps but not future y(n) – y(n-1) = x(n) – x(n-3)
i/ps. It gives FIR o/p.
1
Q. y(n) = [x (n+1) + x (n) + x (n-1)] Find the given system is stable or not?
3
Let x(n) = (n)
1
h(n) = [ (n+1) + (n) + (n-1)]
3
1
h(0) =
3
1
h(-1) =
3
1
h(1) =
3
S= h(n) < therefore Stable.
Q. y(n) = a y(n-1) + x(n) given y(-1) = 0
Let x(n) = (n)
h(n) = y(n) = a y(n-1) + (n)
h(0) = a y(-1) + (0) = 1 = y(0)
h(1) = a y(0) + (1) = a
h(2) = a y(1) + (2) = a2 . . . . . . . h(n) = an u(n) stable if a<1.
1
y(n-1) = [ y(n) – x(n)]
a
25
1
y(n) = [ y(n+1) – x(n+1)]
a
1
y(-1) = [ y(0) – x(0)]=0
a
y(-2) = 0
1
Q. y(n) = y(n-1) + x(n) for n 0
n 1
=0 otherwise. Find whether given system is time variant or not?
Let x(n) = (n)
h (0) = 1 y(-1) + (0) = 1
h(1) = ½ y(0) + (1) = ½
h(2) = 1/6
h(3) = 1/24
if x(n) = (n-1)
y(n) = h(n-1)
1
h(n-1) = y(n) = h(n-2) + (n-1)
n 1
n=0 h(-1) = y(0) = 1 x 0+0 =0
n=1 h(0) = y(1) = ½ x 0 + (0)= 1
n=2 h(1) = y(2) = 1/3 x 1 + 0 = 1/3
h(2) = 1/12
h (n, 0) h(n,1) TV
Q. y (n) = 2n x(n) Time varying
1
Q. y (n) = [x (n+1) + x (n) + x (n-1)] Linear
3
Q. y (n) = 12 x (n-1) + 11 x(n-2) TIV
Q. y (n) = 7 x2(n-1) non linear
Q. y (n) = x2(n) non linear
Q. y (n) = n2 x (n+2) linear
Q. y (n) = x (n2) linear
Q. y (n) = ex(n) non linear
26
Q. y (n) = 2x(n) x (n) non linear, TIV
(If the roots of characteristics equation are a magnitude less than unity. It is a
necessary & sufficient condition)
Non recursive system, or FIR filter are always stable.
Q. y (n) + 2 y2(n) = 2 x(n) – x(n-1) non linear, TIV
Q. y (n) - 2 y (n-1) = 2x(n) x (n) non linear, TIV
Q. y (n) + 4 y (n) y (2n) = x (n) non linear, TIV
Q. y (n+1) – y (n) = x (n+1) is causal
Q. y (n) - 2 y (n-2) = x (n) causal
Q. y (n) - 2 y (n-2) = x (n+1) non causal
Q. y (n+1) – y (n) = x (n+2) non causal
Q. y (n-2) = 3 x (n-2) is static or Instantaneous.
Q. y (n) = 3 x (n-2) dynamic
Q. y (n+4) + y (n+3) = x (n+2) causal & dynamic
Q. y (n) = 2 x ( n )
If =1 causal, static
<1 causal, dynamic
>1 non causal, dynamic
1 TV
Q. y (n) = 2(n+1) x (n) is causal & static but TV.
Q. y (n) = x (-n) TV
1.8 Solution of linear constant-co-efficient difference equation
Q. y(n)-3 y (n-1) – 4 y(n-2) = 0 determine zero-input response of the system;
Given y(-2) =0 & y(-1) =5
Let solution to the homogeneous equation be
yh (n) = n
n - 3 n-1 - 4 n-2 =0
n-2[ 2 - 3 - 4] =0
= -1, 4
27
yh (n) = C1 1n + C2 2n = C1(-1)n + C2 4n
y(0) = 3y(-1) +4 y(-2) = 15
C1+ C2 =15
y (1) = 3y (0) +4 y (-1) = 65
-C1+4C2 = 65 Solve: C1 = -1 & C2=16
y(n) = (-1)n+1 + 4n+2 (Ans)
If it contain multiple roots yh(n) = C1 1n + C2 n 1n + C3 n2 1n
or 1n [C1+ nC2 + n2 C3….]
Q. Determine the particular solution of y(n) + a 1y(n-1) =x(n)
x(n) = u(n)
Let yp (n) = k u(n)
k u(n) + a1 k u(n-1) =u(n)
To determine the value of k, we must evaluate this equation for any n 1
k + a1 k =1
1
k=
1 a1
1
yp (n) = u(n) Ans
1 a1
x(n) yp(n)
1. A K
2. Amn Kmn
3. Anm Ko nm + K1nm-1 + …. Km
4. A Coswon or A Sinwon K1 Coswon + K2 Sinwon
5 1
Q. y(n) = y(n-1) - y(n-2) + x(n) x(n) = 2n n 0
6 6
Let yp (n) = K2n
5 1
K2n u(n) = K 2n-1 u(n-1) - K 2n-2 u(n-2) + 2n u(n)
6 6
For n 2
5 1
4K = (2K) - K +4 Solve for K=8/5
6 6
28
8 n
yp (n) = 2 Ans
5
Q. y(n) – 3 y(n-1) - 4 y(n-2) = x(n) + 2x(n-1) Find the h(n) for recursive
system.
We know that yh (n) = C1 (-1)n + C2 4n
yp (n) =0 when x(n) = (n)
for n=0
y(0) - 3y(-1) - 4 y(-2) = (0) + 2 (-1)
y(0) =1
y(1) = 3 y(0) +2 = 5
C1 + C2 =1
1 6
-C1 + C2 =5 Solving C1 = ; C2 =
5 5
1 6
h(n) = [ (-1)n + 4n ] u(n) Ans
5 5
OR
h(n) – 3 h(n-1) -4 h(n-2) = (n) + 2 (n-1)
h(0) = 1
h(1) =3 h(0) + 2 = 5
plot for h(n) in both the methods are same.
Q. y(n) – 0.5 y(n-1) = 5 cos 0.5n n 0 with y(-1) = 4
yh(n) = n
n – 0.5 n-1 =0
n-1 [ -0.5] =0
=0.5
yh(n) = C (0.5)n
yp(n) = K1 cos 0.5n + K2 sin 0.5n
yp(n-1) = K1 cos 0.5(n-1) + K2 sin 0.5(n-1)
= - K1 sin 0.5n - K2cos 0.5n
yp(n) - 0.5 yp(n-1) = 5 cos 0.5 n
29
= (K1 + 0.5 K2) cos 0.5 n -(0.5 K1 – K2) sin 0.5n
K1 + 0.5 K2 = 5
0.5 K1 – K2 =0 Solving we get: K1= 4 & K2=2
yp(n) = 4 cos 0.5 n + 2 sin 0.5n
The final response
y (n) = C (0.5)n + 4 cos 0.5 n + 2 sin 0.5n
with y(-1) = 4
4 = 2C-2
i.e. C=3
y (n) = 3 (0.5)n + 4 cos 0.5 n + 2 sin 0.5n for n 0
1.9 Concept of frequency in continuous-time and discrete-time.
1) xa (t) = A Cos ( t)
x (nTs) = A Cos ( nTs)
= A Cos (wn)
w = Ts
= rad / sec w = rad / Sample
F = cycles / sec f = cycles / Sample
2) A Discrete- time – sinusoid is periodic only of its f is a Rational number.
x (n+N) = x (n)
Cos 2 f0 (n+N) = Cos 2 f0 n
K
2 f0N = 2 K => f0 =
N
Ex: A Cos ( )n
6
30
w= =2 f
6
1
f= N=12 Samples/Cycle ; Fs= Sampling Frequency; Ts =
12
Sampling Period
Q. Cos (0.5n) is not periodic
Q. x (n) = 5 Sin (2n)
1
2 f = 2 => f = Non-periodic
Q. x (n) = 5 Cos (6 n)
2 f = 6 => f = 3 N=1 for K=3 Periodic
6 n
Q. x (n) = 5 Cos
35
6 3
2 f = => f = for N=35 & K=3 Periodic
35 35
Q. x (n) = Sin (0.01 n)
0.01
2 f = 0.01 => f = for N=200 & K=1 Periodic
2
Q. x (n) = Cos (3 n) for N=2 Periodic
fo = GCD (f1, f2) & T = LCM (T1, T2) ------- For Analog/digital signal
[Complex exponential and sinusoidal sequences are not necessarily periodic in
2
‘n’ with period ( ) and depending on Wo, may not be periodic at all]
Wo
N = fundamental period of a periodic sinusoidal.
3. The highest rate of oscillations in a discrete time sinusoid is obtained
when w = or -
31
Discrete-time sinusoidal signals with frequencies that are separated by an
integral multiple of 2 are Identical.
Fs Fs
4. - F
2 2
- Fs 2 F Fs
-
Ts Ts
- Ts
Therefore - w
5. Increasing the frequency of a discrete- time sinusoid does not necessarily
decrease the period of the signal.
n
x1(n) = Cos ( ) N=8
4
3 n
x2(n) = Cos ( ) N=16 3/8 > 1/4
8
2 f = 3 /8
3
=> f =
16
1
6. If analog signal frequency = F = samples/Sec = Hz then digital frequency
Ts
f=1
32
W= Ts
2 f = 2 F Ts => f =1
2 F = 4 ;
2 f = /4
1 1
F= ; T= 8 ; f= N=8
8 8
7. Discrete-time sinusoids are always periodic in frequency.
Q. The signal x (t) = 2 Cos (40 t) + Sin (60 t) is sampled at 75Hz. What is
the common period of the sampled signal x (n), and how many full periods of x (t)
does it take to obtain one period of x(n)?
33
F1 = 20Hz F2 = 30Hz
20 4 K1 30 2 K 2
f1 = f2 =
75 15 N1 75 5 N 2
The common period is thus N=LCM (N1, N2) = LCM (15, 5) = 15
The fundamental frequency Fo of x (t) is GCD (20, 30) = 10Hz
1
And fundamental period T = 0.1s
Fo
Since N=15
1
1sample ---------- sec
75
15
15 sample ----------- ? => 0.2 S
75
So it takes two full periods of x (t) to obtain one period of x (n) or GCD (K1,
K2) = GCD (4, 2) = 2
Frequency Domain Representation of discrete-time signals and systems
For LTI systems we know that a representation of the input sequence as a
weighted sum of delayed impulses leads to a representation of the output as a
weighted sum of delayed responses.
Let x (n) = ejwn
y (n) = h (n) * x (n)
= h(k ) x(n k ) h(k ) ejw (n-k)
k k
=e jwn
h(k ) e-jwk
k
Let H (e ) =jw
h(k )
k
e-jwk is the frequency domain representation of the
system.
y (n) = H (ejw) ejwn ejwn = eigen function of the system.
H (ejw) = eigen value
Q. Find the frequency response of 1st order system y (n) = x (n) + a y (n-1)
(a<1)
34
Let x (n) = ejwn
yp (n) = C ejwn
C ejwn = ejwn + a C ejw (n-1)
C ejwn [1-ae-jw] = ejwn
1
C=
[1 ae jw ]
1 1
Therefore H (ejw) = =
[1 ae jw ] 1 a(cos w j sin w)
1
H (e jw ) =
1 2a cos w a 2
aSinw
H (e jw ) Tan 1 ( )
1 aCosw
1
Q. Frequency response of 2nd order system y(n) = x(n) - y (n 2)
2
x (n) = e jwn
y (n) ce jwn
p
1 jw( n 2 )
c e jwn = e jwn - ce
2
35
1 1 20 16Cos2 w
c e jwn (1+ e 2 jw ) = e jwn c= c
2 1
1 e 2 jw 5 4Cos2 w
2
Sin 2 w
c tan 1
2 Cos2 w
36
2. POWER, ENERGY and CONVOLUTION
Continuous Time ot = onTs = won Discrete Time
Periodic f (t) = N 1 2
c e
jK n
N
c
k
jkot Periodic xp(n) =
k e k 0
k
DTFS
Non periodic Periodic Ck =
T
1 N 1 2
Ck =
T
0
f (t )e jKot dt 1
x p ( n )e
j
N
nK
N n 0
2nTs
1 jK
k=0 to N-1
T
x ( n )e
NTs
T = N Ts
t = n Ts : dt = Ts
Non-Periodic f(t) = Non – Periodic x(n) =
2
1 1
F (w)e X (w)e
jt
d jwn
dw
2
2 0
Non-Periodic F(w) = Periodic X(w) = x(n)e
n
jwn
X(w) = FT of DTS
f (t )e jt dt
37
2.1 Energy and Power
2 2
1
=
x ( n ) x ( n ) x ( n )*
x ( n ) X *
( w) e jwn
dw
E =
n n n 2 0
=
2
1 jwn
0
*
X ( w) x ( n ) e dw
2 n
=
2
1
X
*
( w) X ( w)dw
2 0
2
1
= 2 X ( w)
dw
2 2
1
E= X (w)
x ( n ) dw -------- Parsval’s Theorem
Therefore:
n 2
N 2
1
P =
Lt
N 2 N 1
n N
x ( n) for non periodic signal
N 1 2
1
= N
n 0
x ( n) for periodic Signal
2
1 N 1 1 N 1 N 1
* j N nk
=
N n 0
x ( n) x ( n) x ( n) C k e
*
N n 0 k 0
2
N 1 1 N 1 j
C x ( n)e
* nk
N
k
=
k 0 N n 0
N 1 2 N 1 2
Therefore P= Ck 0
k E=N Ck
k 0
38
Ex: Unit step
N
1
P = Lt
N 2 N 1 n 0
u 2
( n)
N 1 1
= Lt 2 N 1 2
N
Power Signal
E=
jwon
Ex: x (n) = Ae
N 2
1
P= Lt
N 2 N 1 n N
Ae jwon
1
= Lt A 2 [1 1 ........]
N 2 N 1
A2 (2 N 1)
Lt A2
=
N 2N 1 it is Power Signal and E =
Ex: x (n) = n u(n) neither energy nor power signal
Ex: x (n) = 3 (0.5)n n0
9
x (n) 9(0.25)n
2
12 J n
1
E=
n n 0 1 0.25 note: [
n 0 1
]
2n
Ex: x (n) = 6 Cos whose period is N=4 x (n) = {
4
6 ,0,6,0 }
1 3 2 1
P=
4 n 0
x ( n )
4
[36 36] 18W
2 n
j
4
Ex: x (n) = 6 e whose period is N = 4
39
2
1 3 1
P=
4 n 0
x ( n )
4
[36 36 36 36] 36Watts
2.2 DISCRETE CONVOLUTION
It is a method of finding zero input response of linear Time Invariant
system.
Ex: x(n) = u(n)
h(n) = u(n)
y(n) = u(k )u(n k )
k
u(k) = 0 k<0
40
u(n-k) = 0 k>n
n
u (k )u (n k )
n
k 0
= 1 = (n+1) u(n) = r(n+1)
k 0
Q. x(n) = an u(n) and h(n) = an u(n) a<1 find y(n)
n
y(n) = k 0
ak an-k = an (n+1) u(n)
Q. x(n) = u(n) and h(n) = n u(n) <1 find y(n)
n
y(n) =
k
k u(k) u(n-k) =
k 0
k = (1- n+1) / (1- )
The convolution of the left sided signals is also left sided and the convolution
of two right sided also right sided.
n n
Q. x(n) = rect ( )=1 N
2N
=0 else where
n
h(n) = rect ( )
2N
y(n) = x(n) * h(n)
= [u (n+N) – u (n-N-1)] * [u (n+N) – u (n-N-1)]
= u (n+N) * [u (n+N) – u (n-N-1)] – u (n-N-1)* [u (n+N) – u (n-N-1)]
= u (n+N) * u (n+N) – 2 u (n+N)*u (n-N-1)] + u (n-N-1) * u (n-N-1)
= r(n+2N+1) – 2r(n) + r(n-2N-1)
n
= (2N+1) Tri ( )
2N 1
n n
Tri ( ) = 1- for n N
N N
=0 elsewhere.
41
Q. x(n) = {2,-1,3}
h(n) = { 1,2,2,3} Graphically Fold-shift-multiply-sum
y(n) =
1 2 2 3
2 2 4 4 6
-1 -1 -2 -2 -3
3 3 6 6 9
y(n) = { 2,3,5,10,3,9}
Q. x(n) = {4, 1
,3} h(n) = { 2,5, 0
,4}
2 5 0 4
4 8 20 0 16
1 2 5 0 4
3 6 15 0 12
y(n) = { 8,22,11,31,4, 12 } Note that convolution starts at n=-3
42
Q)
h(n): 2 5 0 4
x(n): 4 1 3
_________________________________
8 20 0 16
2 5 0 4
6 15 0 12
____________________________________
y(n): 8 22 11 31 4 12
Q. Convolution by sliding step method: h(n) = 2
, 5, 0, 4 ; x(n)= 4 , 1, 3
i) 2 5 0 4 ii) 2 5 0 4
3 1 4 3 1 4
___________________ _________________________
y(0) = 8 2 20 y(1) = 2+20 =
22
iii) 2 5 0 4 iv) 2 5 0 4
3 1 4 3 1 4
________________________ _______________________
6 5 0 y(2) = 11 15 0 16 y(3) = 31
v) 2 5 0 4 Vi) 2 5 0 4
3 1 4 3 1 4
________________________
_______________________
43
0 4 y(4) = 4 12 y(5) = 12
If we insert zeros between adjacent samples of each signal to be convolved,
their convolution corresponding to the original convolution sequence with zeros
inserted between its adjacent samples.
Q. h(n) = 2 , 5, 0, 4 ; x(n)= 4 , 1, 3 X(z) = 2z3+5z2+4 ; X(z) =
4z2+z+3
Their product Y(z) = 8z5+22z4+11z3+31z2+4z+12
y(n) = 8 ,22,11,31,4,12
h(n) = 2 , 0, 5, 0, 0, 0, 4 ; x(n) = 4, 0, 1, 0, 3
H(z) = 2z6+5z4+ 4 ; X(z) = 4z4+z2+3
Y(z) = 8z10+22z6+31z4+4z2+12 y(n) = { 8,0,22,0,11,0,31,0,4,0,12}
Q. Compute the linear convolution of h(n)={1,2,1} and x(n) = { 1, -1, 2, 1 ,2, -
1, 1, 3, 1} using overlap-add and overlap-save method.
h (n): 1 2 1
x (n): 1 -1 2 1 2 -1 1 3 1
x1(n): 1 -1 2
x2(n): 1 2 -1
x3(n): 1 3 1
____________________________________________________________
y1(n) = (h (n)*x1(n))1 1 1 3 2
y2(n) = 1 4 4 0 -1
y3(n) = 1 5 8 5
1
y(n) = { 1 1 1 4 6 4 1 4 8 5
1} OVER LAP and SAVE method
h (n): 1 2 1 0 0 (N2=3)
x1(n): 1 -1 2 1 2 (N3+N2-1) = 5
44
x2(n): 1 2 -1 1 3
x3(n): 1 3 1 0
0
y1(n) = 1 1 1 4 6 5 2
y2(n) = 1 4 4 1 4 7 3
y3(n) = 1 5 8 5
1
y(n) = {1 1 1 4 6 4 1 4 8 5
1}
45
3. DISCTRETE FOURIER SERIES
Q. Determine the spectra of the signals
a. x(n) = Cos 2 n
wo = 2
1
fo = is not rational number
2
Signal is not periodic.
Its spectra content consists of the single frequency
b. x (n) = Cos n after expansion x(n)={ 1,0.5,-0.5,-1,-0.5,0.5}
3
1
fo = N=6
6
2
1 5 j
Ck = x(n)e
nk
6
k=0 to 5
6 n 0
1 j k
2 4 5
j k j k j k
jk
Ck = x ( 0) x (1) e 3
x ( 2) e 3
x (3) e x ( 4) e 3
x (5) e 3
6
Co = x(0) x(1) x(2) x(3) x(4) x(5) = 0
1
For k=0
6
Similarly
K=1 C1 = 0.5 , C2 = 0 = C3 = C4 , C5 = 0.5
Or
2 2 2
1 j 6n 1 j 5
C
n j kn
= Cos n e
6
x (n) + e = k e 6
3 2 2 k 0
46
2 4 6 8 10
j n j n j n j n j n
6 6 6 6 6
= Co+C1 e +C2e + C3 e +C4 e +C5 e
1
By comparison C1=
2
2 5 6 10n
j n j 2 n j
6 6 6
Since e =e =e
1
C5
2
c. x (n) = {1,1,0,0}
nk
1 3 j 2
Ck= 4
4
x ( n )e k=0, 1, 2, 3
n 0
2k
1 j
1 1e 2
= 4
1
c0 2 ; c1
1
4
1 j ; c2 0 ; c3
1
4
1 j
1
Co & C0 = 0
2
2
c
1
4 & C1 = 4
c 2
0 & C2 undefined
2
c 3
& C3 =
4 4
47
3.1 PROPERTIES OF DFS
1. Linearity
DFS ~x1 (n) Ck1
DFS ~x2 (n) Ck 2
DFS
a ~
x1 ( n) bx2 (n) aCk1 bCk 2
~
2. Time Shifting
2mk
DFS x ( n m) e
j
~ N
Ck
3. Symmetry
DFS ~x (n) C
* *
k Ck =
N 1 2nk
1 j
N
~
x ( n )e
n 0
N
N 1
~
x (n) C
* *
k x ( n ) Ck e
~ j2
N
nk
DFS
k 0
~x ( n) ~
x * (n) 1
DFS
Re ~
x ( n) DFS C
k C *
k C ke
2 2
DFS
~ ~ (n) 1
*
j Imx (n) DFS
~ x ( n) x
C k C *
k C ko
2 2
~x (n)
If is real then
48
~ ~x (n) ~
x * (n)
xe (n)
2
~ ~x (n) ~
x * (n)
xo (n)
2
DFS
~
x e ( n )
1
C k2
C * k ReC k
DFS
~
x o ( n )
1
2
C k C *
k j ImC k
Periodic Convolution
N 1 ~
1
~
x ( m ) x 2 ( n m ) C k 1C k 2
DFS
m 0
If x(n) is real
C k C * k
Re[ C k ] Re[ C k ]
Im[ Ck ] Im[ Ck ]
C k C k
C k Ck
3.2 PROPERTIES OF FT (DTFT)
1. Linearity
y (n) = a x1 (n) + b x2 (n)
Y (e jw ) = a X1(e jw ) + b X2(e jw )
49
2. Periodicity
j ( w 2 )
H (e ) = H (e jw )
3. For Complex Sequence
h (n) = hR(n) + j hI(n)
H (e jw ) = [ h
n -
R (n) j h I (n) ] [Cos(wn) - j Sin(wn)]
[ h
n -
R (n) Cos(wn) h I (n)Sin(wn) = HR (e jw )
[ h
n -
I (n) Cos(wn) h R (n)Sin(wn) = HI (e jw )
H (e jw ) = H R (e jw ) jH I (e jw )
= H R2 (e jw ) H I2 (e jw ) H (e jw ) H * (e jw )
H I (e jw ) 1
H (e ) tan jw
jw
H R (e )
4. For Real Valued Sequence
H (e ) jw
= h(n)e
n
jwn
= h(n)Cos(wn) j h(n)Sin(wn)
n n
= H R (e jw ) jH I (e jw ) -------------------- (a)
H (e jw
) = h(n)e
n
jwn
= h(n)Cos(wn) j h(n)Sin(wn)
n n
50
= H R (e jw ) jH I (e jw ) -------------------- (b)
From (a) & (b)
H R (e jw ) H R (e jw )
H I (e jw ) H I (e jw )
Real part is even function of w
Imaginary part is odd function of w
H (e jw ) H * (e jw )
=>
H (e jw ) H (e jw )H * (e jw ) H * (e -jw )H(e -jw ) H(e -jw )
Magnitude response is an even function of frequency
H I (e -jw ) -1 H I (e
jw
)
H (e - jw
) tan
-1
- jw
tan jw
H (e jw )
H R (e ) H R (e )
Phase response is odd function.
5. FT of a delayed Sequence
FT [h (n-k)] = h ( n k )e
n
jwn
Put n-k = m
= h ( m)e
m
jw( m k )
e jwk
jwk
= e h(m)e jwm = H (e jw )
m
6. Time Reversal
x (n) X (w)
x (-n) X (-w)
51
F T [x (-n)] = x ( n )e
n
jwn
Put –n = m
x (
m
m ) e jwm
X ( w)
7. Frequency Shifting
x(n) e jwo n X (w-wo)
F T [x (n) e jwo n
]=
n
x (n) e jwo n e-jwn
=
n
x (n) e j ( w wo ) n = X (w-wo)
8. a. Convolution
x1 (n) * x2 (n) X1(w) X2(w)
n
[x1 (n) * x2 (n) ] e-jwn =
n k
[ x1 (k) x2 (n-k) ] e-jwn
Put n-k = m
=
n
x1 (k)
m
[x2(m)] e-jw (m+k)
=
n
x1 (k) e -jwk
m
[x2(m)] e-jwm
= X1(w) X2(w)
1
b. [X1(w) * X2(w)] x1 (n) x2 (n)
2
9. Parsevals Theorem
52
1
n
*
x1(n) x2 (n) =
2
[X1(w) X2*(w)] dw
dX ( w)
n x (n) j
dw
10. F T of Even Symmetric Sequence
H (e
jw
)=
n
h (n) e-jwn
1
=
n
h (n) e-jwn + h (0) + n 1
h (n) e-jwn
Let n = -m
=
m 1
h (-m) ejwm + h (0) +
n 1
h (n) e-jwn
Let h (-m) = h (m) for even
Therefore = h (0) + 2
n 1
h (n) Cos (wn) is a real valued function of
frequency
0 ; H (e jw ) >0
; H (e jw ) <0
11. F T of Odd Symmetric Sequence
For odd sequence h (0) = 0
H (e jw ) =
n 1
h (n) [e-jwn - ejwn ]
= -j 2
n 1
h (n) Sin (wn) HI (e jw ) is a imaginary valued
function of freq. and a odd function of w
i.e, H (e jw ) = - H (e jw )
53
H (e jw ) = HI (e jw ) for HI (e jw ) > 0
= - HI (e jw ) for HI (e jw ) < 0
H (e jw ) For w over which HI (e jw ) > 0
2
= for w over which HI (e jw ) < 0
2
2
1
12. x(0) =
2 X (w)dw
0
Central Co-ordinates
X (0) = x(n)
n
13. Modulation
X (w w0 ) X (w w0 )
Cos (won) x (n)
2 2
3.3 FOURIER TRANSFORM OF DISCRETE TIME SIGNALS
X (w) =
n
x (n) e-jwn
F T exists if x(n)
n
The FT of h (n) is called as Transfer function
1
Ex: h (n) = for -1 n 1
3
=0 otherwise
1
1
)= 3
jwn
e 1 e jw = 1 2Cos( w)
jw e 1 jw 1
Sol: H (e =
n 1 3 3
54
w
0 1
1
2 3
1
-
3
Ex: h (n) = an u (n)
)=
H (e jw a n
e jwn
n 0
1
= (ae
n 0
jw n
) =
1 ae jw
Q. x(n) = n n u(n) <1
d 1
n u(n)
n
j jw
dw 1 e
e jw
= (1 e jw ) 2
1
Hint: n u(n)
n 0
n
e jwn
= (e jw ) n
n 0
=
1 e jw
Q. x(n) = n 0nN
Or
x(n) = n [ u(n) – u(n-N)]
= n u(n) – N n-N u(n-N) Using Shifting Property
55
jwN
1 e
N
[ ]
X(w) =
1 e jw
1 e jw
1 (e jw ) N
=
1 e jw Ans
Q. x(n) = n
1 two sided decaying exponential
x(n) = n u(n) + -n u(-n) - (n) using folding property
1 1 1 2
1 =
=
1 e jw 1 e jw 1 2Cosw 2
Q. x (n) = u (n) Since u (n) is not absolutely summable
1
we know that u (t) ( w) jw
1
(w)
Similarly X (w) =
1 e jw +
56
4. DFT (Frequency Domain Sampling)
The Fourier series describes periodic signals by discrete spectra, where as the
DTFT describes discrete signals by periodic spectra. These results are a consequence of the
fact that sampling on domain induces periodic extension in the other. As a result, signals that
are both discrete and periodic in one domain are also periodic and discrete in the other. This
is the basis for the formulation of the DFT.
Consider aperiodic discrete time signal x (n) with FT X(w) = x ( n )e
n
jwn
Since X (w) is periodic with period 2 , sample X(w) periodically with N equidistance
2
samples with spacing w .
N
K = 0, 1, 2…..N-1
2
2k j
x(n)e
Kn
X N
N n
The summation can be subdivided into an infinite no. of summations, where each sum
contains
N 1 2
2 j
2k
x(n)e
1 Kn
j
............ x(n)e
Kn N
X N
+
N n N n 0
2 N 1 2
j
x(n)e
Kn
N
..................
n N
57
lN N 1 2
j
n
Kn
N
x(n)e
=
lN
l
Put n = n-lN
x(n lN )e
N 1 2
j K ( n lN )
N
=
l n 0
2
j
l
N 1 Kn
=
x(n lN )e N
n 0
N 1 2
j
Kn
N
X(k) = xp(n) e
n 0
N 1 2
j Kn
N
We know that xp(n) = Ck e n= 0 to N-1
k 0
N 1 2
Ck=
1
N
n 0
xp(n) e
j
N
Kn
k=0 to N-1
1
Therefore Ck= X(k) k=0 to N-1
N
N 1
2
j Kn
IDFT ------------ xp (n) =
1
N
k 0
X(k) e N
n = 0 to N-1
This provides the reconstruction of periodic signal xp(n) from the samples of spectrum
X(w).
The spectrum of aperiodic discrete time signal with finite duration L<N, can be exactly
2k
recovered from its samples at frequency W k= .
N
Prove: x(n) = xp (n) 0 n N-1
58
Using IDFT
N 1 2
1 j Kn
x (n) = X(k) e N
N k 0
N 1 N 1 2
1 j Kn
X (w) = [ X (k) e N
] e-jwn
n 0 N k 0
N 1 N 1 2
jn ( w
1 K)
= X (k) [ e N
]
k 0 N n 0
N 1
1
If we define p(w) = e-jwn
N n 0
wN
Sin
N 1
1 1 e jwN 2 e jw 2
= =
N 1 e jw w
NSin
2
N 1
2k
Therefore: X (w) =
k 0
X (k) P(w-
N
)
2k
At w = P (0) =1
N
2k
And P (w- ) = 0 for all other values
N
N 1 N 1
2k
X (w) = X(k) = X( )
k 0 k 0 N
Ex: x(n) = an u(n) 0<a<1
2k
The spectrum of this signal is sampled at frequency W k= . k=0, 1…..N-1, determine
N
reconstructed spectra for a = 0.8 and N = 5 & 50.
1
X (w) =
1 ae jw
1
X (wk) = 2
k=0, 1, 2… N-1
j k
1 ae N
0
xp (n) = x(n LN )
l
a
l
n lN
59
an
= a n a lN 0 n N-1
l 0 1 aN
Aliasing effects are negligible for N=50
If we define aliased finite duration sequence x(n)
xˆ (n) x p (n) 0 n N-1
=0 otherwise
N 1 N 1
Xˆ ( w) xˆ ( n )e jwn
n 0 = x
n 0
p (n)e jwn
N 1 N 1
an
=
1
e jwn = (ae jw n
)
n 0 1 a 1 aN
N
n 0
1 1 a N e jwN
Xˆ ( w)
1 a N 1 a e jw
2k
j N
2K 1 1 a N
e N
Xˆ
N 1 a
N 2K
j
1 a e
N
60
1 2K
= j 2k = X
1 ae N N
2k
Although Xˆ ( w) X ( w), the samples at Wk= are identical.
N
1 1
Ex: X (w) = & X (k) = j
2
1 a e jw
k
1 a e N
Apply IDFT
2nk
j
N 1
1 e N
x (n) = N j
2k using Taylor series expansion
k 0
1 ae N
1 N 1 j 2Nnk r j 2Nkr
= N e
k 0
a e
r 0
1 N 1 j 2k ( nN r )
=
N
a e
r
r 0 k 0
=0 except r = n+mN
x (n) = a
m 0
n mN
= a
n
(a
m 0
N
)m
an
=
1 aN
The result is not equal to x (n), although it approaches x (m) as N becomes .
Ex: x (n) = {0, 1, 2, 3} find X (k) =?
3 2k
j
x ( n )e
n
4
X (k) =
n 0
X (0) = x(n) = 0+1+2+3 = 6
n 0
61
3 2
j
x ( n )e
n
4
X (1) = = -2+2j
n 0
X (2) = -2
X (3) = -2-2j
4.1 DFT as a linear transformation
2
j
Let WN e N
N 1
X (k) =
nk
x(n)W N k = 0 to N-1
n 0
N 1
1
x (n) =
N
X (k )W
k 0
nk
N n = 0, 1…N-1
X (0)
x ( 0)
X (1)
x (1)
Let xN = XN =
x( N 1) X ( N 1)
1 1 1 1
1 W 1 W 2
W N( N 1)
N N
1 WN2 WN4 WN2( N 1)
WN =
1 WNN 1 WN2( N 1) ( N 1)( N 1)
WN
The N point DFT may be expressed in matrix form as
DFT IDFT
1
XN = WN xN xN = W N* X N
N
62
x N WN1 X N
1. W NK N W NK
N
1 1 K
W N W N* 2. WN 2
WNK
N
Ex: x (n) = {0, 1, 2, 3}
1 1 1 1 1 1 1 1 1 1 1 1
1 W 1 W 2 W 3 1 W 1 W 2 W 3 1 j 1 j
4 4 4 4 4 4
DFT W4 = 1 W 2 W 4 W 6 = 1 W4 W4 W42
2 0 = 1 1 1 1
1
4 4 4
9
1 j 1 j
3 6 3 2
1 W4 W4 W4 1 W4 W4 W4
1 1 1 1 0 6
1 j 1 j 1 2 2 j
X4 = W4 x4 = 1 1 1 1 2 = 2
1 j 1 j 3 2 2 j
IDFT
0
1 1 1 1 6 1
1 j 1 j 2 2 j
1 * 1
x4 = WN X N = 2 = 2 Ans
4 1 1 1 1
4
1 j 1 j 2 2 j 3
Q.
x (n) = { 1,0.5}
h (n) = { 0.5,1 }
63
Find y (n) = x (n) h (n) using frequency domain. Since y (n) is periodic with period 2.
Find 2-point DFT of each sequence.
X (0) = 1.5 H (0) = 1.5
X (1) = 0.5 H (1) = -0.5
Y (K) = X (K) H (K)
Y (0) = 2.25 Y (1) = -0.25
Using IDFT y (0) = 1; y (1) = 1.25
~y (n) h~(n) ~x (n) ~ ~
h (k ) x (n k )
k
~
=
x (k )h (n k )
~
k
~
~
y (0) = ~
x (
k
k ) h (k )
~ ~ ~ ~
= x (0)h (0) x (1)h (1)
= 1 * 0.5 + 0.5 * 1 = 1
~
~
y (1) = ~
x (
k
k ) h (1 k )
64
~ ~ ~ ~
= x (0)h (1) x (1)h (0)
= 1 * 1 + 0.5 * 0.5 = 1.25
~ ~
y ( 2) = ~
x (
k
k ) h (2 k )
~ ~ ~
= x (0)h (2) ~
x (1)h (1)
= 1 * 0.5 + 0.5 * 1 = 1
~
y (n) = {1, 1.25, 1, 1.25…..}
Q. Find Linear Convolution of same problem using DFT
Sol. The linear convolution will produce a 3-sample sequence. To avoid time
aliasing we convert the 2-sample input sequence into 3 sample sequence by padding with
zero.
For 3- point DFT
X (0) = 1.5 H (0) = 1.5
2
j
2 j
3
X (1) = 1+0.5 e 3
H (1) = 0.5+ e
4 4
j j
3
X (2) = 1+0.5 e H (2) = 0.5+ e 3
Y (K) = H (K) X (K)
Y (0) = 2.25
2 4
j j
3 3
Y (1) = 0.5 + 1.25 e + 0.5 e
4 8
j j
3 3
Y (2) = 0.5 + 1.25 e + 0.5 e
Compute IDFT
2kn
1 2
Y ( k )e
j
3
y(n) = 3 k 0
65
y(0) = 0.5
y(1) =1.25
y(2) =0.5
y(n) = { 0.5, 1.25, 0.5} Ans
4.2 PROPERTIS OF DFT
1) Linearity
If h(n) = a h1(n) + b h2(n)
H (k) = a H1(k) + b H2(k)
2) Periodicity H(k) = H (k+N)
~
3)
h ( n) h(n mN )
m
4) y(n) = x(n-n0)
2kn0
j
N
Y (k) = X (k) e
5) y (n) = h (n) * x (n)
Y (k) = H (k) X (k)
6) y (n) = h(n) x(n)
Y (k) =
1
H (k ) X (k )
N
7) For real valued sequence
N 1
2kn
H R (k ) h(n)Cos
n 0 N
N 1
2kn
H I ( k ) h( n) Sin
n 0 N
a. Complex conjugate symmetry
h (n) H(k) = H*(N-k)
h (-n) H(-k) = H*(k) = H(N-k)
66
i. Produces symmetric real frequency components and anti symmetric
N
imaginary frequency components about the DFT
2
N
ii. Only frequency components from 0 to need to be computed in order
2
to define the output completely.
b. Real Component is even function
HR (k) = HR (N-k)
c. Imaginary component odd function
HI (k) = -HI (N-k)
d. Magnitude function is even function
H (k ) H ( N k )
e. Phase function is odd function
H (k ) H ( N k )
f. If h(n) = h(-n)
H (k) is purely real
g. If h(n) = -h(-n)
H (k) is purely imaginary
8. For a complex valued sequence
x*(n) X*(N-k) = X*(-k)
N 1
DFT [x(n)] = X(k) = x(n)W
n 0
nk
N
N 1
X*(k) = x
n 0
*
( n )W N nk
N 1
*
X (N-k) = x
n 0
*
( n )W Nnk = X*(-k)
N 1
DFT [x*(n)] = x
n 0
*
( n )W Nnk = X*(N-k) proved
Similarly DFT [x*(-n)] = X*(k)
67
9.Central Co-ordinates
N 1
1 N 1
x (0) =
N
k 0
X (k ) N
x( )=
2
1
N
(1)
k 0
k
X (k ) N=even
N 1 N 1
X (0) = x ( n ) (1)
N n
X( )= x ( n) N=even
n 0 2 n 0
10. Parseval’s Relation
N 1 N 1
2 2
N
x ( n) X (k )
n 0 k 0
N 1
Proof: LHS N x ( n) x * ( n)
n 0
N 1
1 N 1
nk
=N *
x ( n ) N X ( k )WN
m 0 k 0
N 1
N 1
= X
k 0
*
( k ) x ( n)W Nnk
n 0
N 1 N 1 2
= X
k 0
*
( k ) X (k) =
k 0
X (k )
11.Time Reversal of a sequence
x((n)) N x( N n) X ((k )) N X ( N k )
Reversing the N-point seq in time is equivalent to reversing the DFT values.
x( N n) N 1 j 2k
x( N
n
DFT n)e N
n 0
Let m=N-n
N 1 j 2k
( N m )
= x ( m)e
n 0
N
m=1 to N = 0 to N-1
N 1 j 2k
x ( m) e
m
N
=
m 0
68
N 1 j 2m
( N k )
= x(m)e
m 0
N
= X(N-k)
12.Circular Time Shift of a sequence
j 2k
l
x ( n l ) N X ( k )e N
N 1 j 2k
DFT x(n l ) N x(n l ) N e
n
N
n 0
l 1 j 2k N 1 j 2k
x(n l )
n
x(n l )
n
N
= N e + N e N
n 0 n l
N 1 j 2k
j 2k
x ( N n l )e
l 1 n
x( N n l )
n N
N
= e +
n 0 n l
Put N+n-l = m
N 1 j 2k 2 N 1l j 2k
( ml ) ( m l )
= x ( m) e
m N l
N
+
m N
x ( m) e N
N to 2N-1-l is shifted to N 0 to N-1-l
Therefore 0 to N-1 = (0 to N-1-L) to ( N-L to N-1)
N 1 2k
j ( ml )
Therefore
x(m)e
m 0
N
2k 2 k
N 1 j j l
x ( m) e
m
N N
=
m 0
e
2k
j l
N
= X(k) e RHS
13.Circular Frequency Shift
69
2l
j n
x ( n )e N
X (k l ) N
2l N 1 2l 2k
j
x ( n )e
j n j n n
N N
DFT x ( n ) e
N
= e
n 0
N 1 2n
j ( k l )
= x ( n )e
n 0
N
= X (k l ) N RHS
14.x(n) X(k)
k
{x(n), x(n), x(n)…….x(n)} M X ( )
m
(m-fold replication)
n
x( ) { X (k ), X (k ),...... X (k )} (M- fold replication)
m
2, 3, 2, 1 8, -j2, 0, j2
Zero interpolated by M
{2, 3, 2, 1, 2, 3, 2, 1, 2, 3, 2, 1} {24, 0, 0, -j6, 0, 0, 0, 0, 0, j6, 0, 0}
15.Duality
x(n) X(k)
X(n) N x(N-k) 0 K N 1
N 1 2
1
X ( )e
j n
N
x(n) = N 0
N 1 2
1 ( N k )
j
x(N-k) =
X ( )e N
N 0
N 1 2
1 j
k
=
X ( )e N
N 0
70
N 1 2
j
k
N x(N-k) =
X ( )e N
0
N 1 2k
j
X ( n )e
n
N
= = DFT [ X(n) ] LHS proved
n 0
16.Re[x(n)]
X ep (k ) 1
X ep (k ) = X ((k )) N X * ((k )) N
2
j Im[x(n)]
X op (k )
xep (n) Re[X(k)]
xop (n) j Im[X(k)]
xep (n) Even part of periodic sequence = 1 x(n) x(( n)) N
2
xop (n) Odd part of periodic sequence = 1 x(n) x(( n)) N
2
N 1
Proof: X(k) =
nk
x ( n )W N
n 0
N 1
X(N-k) = x(n)W
n 0
nk
N X (( k )) N
N 1
X*(k) = x
n 0
*
( n )W nk
N
N 1
X (N-k) =
*
x *
( n)WNnk X * (( k )) N
n 0
71
X (( k )) N X * (( k )) N 1 N 1
2
x(n) x * (n) WNnk
2 n 0
= DFT of [Re[x (n)]] LHS
N 1 N 1
17. x
n 0
1
*
( n) x ( n) =
2
1 X
N k 0
1 ( k ) X 2* ( k )
Let y(n) = x1 (n) x2* (n)
Y(k) =
1
N
X 1 ( k ) X 2* ( k )
X
N 1
1
= 1 (l ) X 2* ( k l )
N l 0
X
N 1
1
Y(0) = 1 (l ) X 2* (l )
N l 0
Using central co-ordinate theorem
N 1
Y(0) =
*
x 1 ( n ) x 2 ( n)
n 0
N 1 N 1
Therefore
*
x 1 ( n) x ( n)
2 =
1 X 1 ( k ) X 2* ( k )
n 0 N k 0
QUESTIONS
1 Q. (i) {1,0,0,…….0} (impulse) {1,1,1…..1} (constant)
(ii) {1,1,1,……1} (constant) ) {N,0,0,…….0} (impulse)
j 2k
N 1 n
1 jN2k 1 ( N
) N
e
N
(iii)
n
j 2k j 2k
k 0 1 N
1 e N
2nko
(iv) Cos
N
(k k o ) (k ( N k o )
N 2
(Impulse pair)
72
Or Cos (2 nf ) = Cos (wn)
j 2nko j 2nko )
e N
e N
Sol. x(n) =
2
j 2nko j 2n ( N k o )
e N
e N
=
2
We know that 1 N (k )
j 2nKo
x ( n )e N
X ( K Ko)
x(n)
N
(k k o ) (k ( N k o )
2
I. Inverse DFT of a constant is a unit sample.
II. DFT of a constant is a unit sample.
2 Q. Find 10 point IDFT of
X(k) = 3 k=0
=1 1k9
Sol. X(k) = 1+2 (k )
1
= 1+ 10 (k )
5
1
x(n) = + (n) Ans
5
3 Q. Suppose that we are given a program to find the DFT of a complex-valued sequence
x(n). How can this program be used to find the inverse DFT of X(k)?
N 1
X(k) =
x
n 0
( n )W nk
N
73
N 1
x(n) =
1
N k 0
X ( k )W nk
N
N 1
N x*(n) =
*
X (k )W Nnk
k 0
1. Conjugate the DFT coefficients X(k) to produce the sequence X*(k).
2. Use the program to fing DFT of a sequence X*(k).
3. Conjugate the result obtained in step 2 and divide by N.
4 Q. xp(n) = { 1 , 2, 3, 4, 5, 0, 0, 0}
(i) fp(n) = xp(n-2) = { 0
, 0, 1, 2, 3, 4, 5, 0}
(ii) gp(n) = xp(n+2) = { 3 , 4, 5, 0, 0, 0, 1, 2}
(iii) hp(n) = xp(-n) = {1, 0, 0, 0, 5, 4, 3, 2}
5 Q. x(n) = {1, 1, 0, 0, 0, 0, 0, 0} n = 0 to 7 Find DFT.
1 j 2k j k
x ( n )e
n
8 4
X(k) =
n 0
=1+ e k = 0 to 7
X(0) = 1+1 = 2
j
4
X(1) = 1+ e = 1.707 - j 0.707
j
2
X(2) = 1+ e = 1- j
j 3
4
X(3) = 1+ e = 0.293 - j 0.707
X(4) = 1-1 = 0
By conjugate symmetry X(k) = X*(N-k) = X*(8-k)
X(5) = X*(3) = 0.293 + j 0.707
74
X(6) = X*(2) = 1+j
X(7) = X*(1) = 1.707 + j 0.707
X(k) = { 2 , 1.707 - j 0.707, 0.293 - j 0.707, 1-j, 0, 1+j, 0.293 + j 0.707, 1.707 + j
0.707 }
6 Q. x(n) = {1, 2, 1, 0} N=4
X(k) = {4, -j2, 0, j2}
(i) y(n) = x(n-2) = {1, 0, 1, 2}
j 2k
( no 2 )
4
Y(k) = X(k) e = 4, j2, 0, -j2
(ii) X(k-1) = {j2, 4, -j2, 0}
j 2
ln
N
IDFT x(n) e
jn
2
= x(n) e = {1, j2, -1, 0}
(iii) g(n) = x(-n) = 1, 0, 1, 2
G(k) = X(-k) = X*(k) = {4, j2, 0, -j2}
(iv) p(n) = x*(n) = {1, 2, 1, 0}
P(k) = X*(-k) = {4, j2, 0, -j2}* = {4, -j2, 0, j2}
(v) h(n) = x(n) x(n)
= {1, 4, 1, 0}
H(k) =
1
X (k ) X (k ) = 1 [ 24, -j16, 0, j16] = {6, -j4, 0, j4}
4 4
(vi) c(n) = x(n) x(n)
= {1, 2, 1, 0} {1, 2, 1, 0} = {2,4,6,4}
C(k) = X(k)X(k) = {16, -4, 0, -4}
(vii) s(n) = x(n) x(n) = {1, 4, 6, 4, 1, 0, 0}
S(k) = X(k) X(k) = {16, -2.35- j 10.28, -2.18 + j 1.05, 0.02 + j 0.03, 0.02 - j 0.03, -2.18 -
j 1.05, -2.35 + j 10.28}
x ( n) 1 4 1 0 6
2
(viii)
75
1 1
[16 4 4] 6
2
X ( k )
4 4
76
5.FFT
N 1
= x ( n)W
nk
X(k) N 0 K N 1
n 0
N 1
nk nk
= { Re[x(n)] + j Im[x(n)] } { Re( WN ) + j Im( WN ) }
n 0
N 1 N 1
=
n 0
Re[x(n)] Re( W ) - nk
N n 0
Im[x(n)] Im( WN ) +
nk
N 1
j{ Im[x(n)] Re( WN ) + Im( WN )Re[x(n)]}
nk nk
n 0
2
Direct evaluation of X(k) requires N complex multiplications and N(N-1) complex
additions.
4 N2 real multiplications
{ 4(N-1) + 2} N = N(4N-2) real additions
The direct evaluation of DFT is basically inefficient because it does not use the symmetry
N
K nk
W NK N WN
nk
& periodicity properties W N 2
WN &
5.1 DITFFT:
N N
1 1
2
+
2
X(k) = x ( 2n)W
( 2 n 1) k
2 nk
N x(2n 1)W N
n 0 n 0
(even) (odd)
N
N
1 1
2
x
2
= xe (n)WN2nk
n 0
+ W N
K
n 0
o (n)W N2 nk
N N
1 1
2
x
2
= e N / 2 + W NK
x (
n 0
n )W nk
n 0
o (n)W Nnk/ 2
K
= Xe(k) + W N Xo(k)
77
Although k=0 to N-1, each of the sums are computed only for k=0 to N/2 -1, since Xe(k)
& Xo(k) are periodic in k with period N/2
N
K K
For K N/2 WN 2
=- W N
X(k) for K N/2
N
K
2
X(k) = Xe(k-N/2) - W N Xo(k-N/2)
N=8
x(2n) = xe(n) ; x(2n+1) = xo(n)
xe(0) = x(0) xo(0) = x(1)
xe(1) = x(2) xo(1) = x(3)
xe(2) = x(4) xo(2) = x(5)
xe(3) = x(6) xo(3) = x(7)
X(k) = Xe(k) + W8k Xo(k ) k = 0 to 3
= Xe(k-4) - W8k 4 Xo(k 4) k = 4 to 7
X(0) = Xe(0) + W80 Xo(0) ; X(4) = Xe(0) - W80 Xo(0)
X(1) = Xe(1) + W81 Xo(1) ; X(5) = Xe(1) - W81 Xo(1)
2 2
X(2) = Xe(2) + W8 Xo(2) ; X(6) = Xe(2) - W8 Xo(2)
X(3) = Xe(3) + W83 Xo(3) ; X(7) = Xe(3) - W83 Xo(3)
X(0) & X(4) having same i/ps with opposite signs
78
N N
This pt DFT can be expressed as combination of pt DFT.
2 4
N
Xe(k) = Xee(k) + W N2 k Xeo(k ) k = 0 to -1 (0 to 1)
4
N
2( k ) N
N W 4
Xeo(k ) N N
= Xee(k- )- N k= to -1 ( 2 to 3 )
4 4 4 2
Xo(k) = Xoe(k) + WN2 k Xoo(k ) k = 0 to N -1
4
N
2( k ) N N N
N
= Xoe(k- ) - WN 4
Xoo(k ) k= to -1
4 4 4 2
For N=8
0
Xe(0) = Xee(0) + W8 Xeo(0) ; xee(0) = xe(0) = x(0)
79
2
Xe(1) = Xee(1) + W8 Xeo(1) ; xee(1) = xe(1) = x(2)
0
Xe(2) = Xee(0) - W8 Xeo(0) ; xeo(2) = xe(2) = x(4)
2
Xe(3) = Xee(1) - W8 Xeo(1) ; xeo(3) = xe(3) = x(6)
Where Xee(k) is the 2 point DFT of even no. of xe(n) & Xeo(k) is the 2 point DFT of odd
no. of xe(n)
Similarly, the sequence xo(n) can be divided in to even & odd numbered sequences as
xoe(0) = xo(0) = x(1)
xoe(1) = xo(2) = x(5)
xoo(0) = xo(1) = x(3)
xoo(1) = xo(3) = x(7)
Xo(0) = Xoe(0) + W80 Xoo(0) ;
Xo(1) = Xoe(1) + W82 Xoo(1) ;
Xo(2) = Xoe(0) - W80 Xoo(0) ;
Xo(3) = Xoe(1) - W82 Xoo(1) ;
Xoe(k) is the 2-pt DFT of even-numbered of xo(n)
Xoo(k) is the 2-pt DFT of odd-numbered of xo(n)
Xee(0) = xee(0) + xee(1) = xe(0) + xe(2) = x(0) + x(4)
Xee(1) = xee(0) - xee(1) = xe(0) - xe(2) = x(0) - x(4)
80
Xee(0) = xee(0) + xee(1) = xe(0) + xe(2) = x(0) + x(4)
Xee(1) = xee(0) - xee(1) = xe(0) - xe(2) = x(0) - x(4)
81
No. of No. of Complex Speed
No. of points N Multiplications Improvement Factor:
Stages Direct N2 FFT N2
N
N Log 2 N
Log 2 N 2
2
2 4 16 4 4
3 8 64 12 5.33
4 16 256 32 8
5 32 1024 80 12.8
6 64 4096 192 21.33
For N=8
No of stages given by= Log2N = Log28 = 3.
No. of 2 i/p sets = 2( Log 2N -1 )
=4
Total No. of Complex additions using DITFFT is NLog2 N
82
= 8 * 3 =24
Each stage no. of butterflies in the stage= 2 m-q where q = stage no. and N=2m
Each butterfly operates on one pair of samples and involves two complex additions and
one complex multiplication. No. of butterflies in each stage N/2
DITFFT: ( different representation) (u can follow any one) ( both representations are
correct)
N
N
1 1
2
x(2n 1)W
2
x(2n)W 2 nk ( 2 n 1) k
X(k) = N + N
n 0 n 0
N N
1 1
2 2
= x (n)W
n 0
e
nk
N + W Nk x
n 0
o (n)W Nnk/ 2
2
K
4 pt DFT Xe(k) + W N Xo(k) k= 0 to N/2 -1 = 0 to 3
N
N (K ) N
Xe(k- ) - W N 2 Xo(k- ) k = N/2 to N-1 = 4 to 7
2 2
2 pt DFT Xe(k) = Xee(k) + W N2 K Xeo(k) k = 0 to N/4-1 = 0 to 1
N
2( k )
= Xee(k-N/4) - W N 4
Xeo(k-N/4) k = N/4 to N/2 -1 = 2 to 3
Xo(k) = Xoe(k) + W N2 K Xoo(k) k = 0 to N/4-1 = 0 to 1
N
2( k )
= Xoe(k-N/4) - W N 4
Xoo(k-N/4) k = N/4 to N/2 -1 = 2 to 3
W82 W41
N=8
X(0) = Xe(0) + W80 Xo(0) ; X(4) = Xe(0) - W80 Xo(0)
X(1) = Xe(1) + W81 Xo(1) ; X(5) = Xe(1) - W81 Xo(1)
X(2) = Xe(2) + W82 Xo(2) ; X(6) = Xe(2) - W82 Xo(2)
X(3) = Xe(3) + W83 Xo(3) ; X(7) = Xe(3) - W83 Xo(3)
Xe(0) = Xee(0) + W80 Xeo(0) ; Xe(2) = Xee(0) - W80 Xeo(0)
Xe(1) = Xee(1) + W82 Xeo(1) ; Xe(3) = Xee(1) - W82 Xeo(1)
Xo(0) = Xoe(0) + W80 Xoo(0) ; Xo(2) = Xoe(0) - W80 Xoo(0)
83
Xo(1) = Xoe(1) + W82 Xoo(1) ; Xo(3) = Xoe(1) - W82 Xoo(1)
N
1 1
=
4
Xee(k) = x ( 4n)W
4 nk 4 nk
N
x(4n)W N =x(0) + x(4) W84 k
0 n 0
Xee(0) = x(0)+x(4)
Xee(1) = x(0)-x(4)
x(0) x(0) x(0)
x(4) x(2) x(1)
x(2) x(4) x(2)
x(6) x(6) x(3)
x(1) x(1) x(4)
84
x(5) x(3) x(5)
x(3) x(5) x(6)
x(7) x(7) x(7)
Other way of representation
85
5.2 DIFFFT:
N
1
2 N 1
= x(n' )W
nk n 'k
X(k)
x(n)W N + N put n’ = n+N/2
n 0 n1 N / 2
N
1 N
1
2
= x ( n)W
2
x(n N / 2)W
nk ( n N / 2) k
N + N
n 0 n 0
N N
1 1
2 N 2
= x ( n)W x(n N / 2)W
nk k nk
2
N +W N N
n 0 n 0
N
1
2
= [ x(n) + (-1)
n 0
k
x(n+
N W nk
2
)] N
N
1
2
X(2k) = [
n 0
x ( n ) + x(n+
N W nk
2
)] N /2
N
1
2
X(2k+1) = {[ x ( n) - x(n+
N n nk
)] W N } W N / 2
n 0 2
Let f(n) = x(n) + x(n+N/2)
n
W
g(n) = { x(n) – x(n+N/2) } N
86
N=8
f(0) = x(0) + x(4)
f(1) = x(1) + x(5)
f(2) = x(2) + x(6)
f(3) = x(3) + x(7)
g(0) = [x(0) - x(4)] W80
g(1) = [x(1) - x(5)] W81
g(2) = [x(2) - x(6)] W82
g(3) = [x(3) - x(7)] W83
87
N
1
4
=
nk
X(4k)
[ f ( n) + f(n+ N )] W N / 4
n 0 4
N
1
4
X(4k+2) = [{ f (n) - f(n+ N
n 0 4
W n
W nk
)} N / 2 ] N / 4
N
1
4
X(4k+1) = [ g (
n 0
n ) + g(n+
N
4
W nk
)] N / 4
N
1
4 nk
X(4k+3) = [{ g ( n) - g(n+
nk
N / 2 ]WN / 4
N W
)}
n 0 4
X(4k) = f(0) + f(2) + [ f(1) + f(3) ] W84 k
2 4k
X(4k+2) = f(0) – f(2) + { [ f(1) – f(3) ] W8 } W8
X(0) = f(0) + f(2) + f(1) + f(3)
88
X(4) = f(0) + f(2) – [ f(1) + f(3) ]
X(2) = f(0) - f(2) + [ f(1) - f(3)] W82
X(6) = f(0) - f(2) - [ f(1) - f(3)] W82
Find the IDFT using DIFFFT
X(k) = { 4, 1-j 2.414, 0, 1-j 0.414, 0, 1+j 0.414, 0, 1+j 2.414 }
Out put 8x*(n) is in bit reversal order x(n) = { 1,1,1,1,0,0,0,0}
89
6.DIGITAL FILTER STRUCTURE
The difference equation
NP
a k x(n-k) + bk
M
y(n) = y(n-k)
k N F k 1
NP
a k z k Np NF
(1 C k ) Z 1
k N F
M
Z NF M
1 bk z (1 d
H(z) = k or = A
k 1 k ) Z 1
k 1 k 1
If bk= 0 non recursive or all zero filter.
6.1 Direct Form – I
1. Easily implemented using computer program.
2. Does not make most efficient use of memory = M+Np+N F delay elements.
6.2 Direct form-II
90
Smaller no. of delay elements = Max of (M, Np) + N F
Disadvantages of D-I & D-II
1. They lack hardware flexibility, in that, filters of different orders, having different no.
of multipliers and delay elements.
2. Sensitivity of co-efficient to quantization effects that occur when using finite-precision
arithmetic.
6.3 Cascade Combination of second-order section (CSOS)
y(n) = x(n) + a1 x(n-1) + a2 x(n-2) + b1 y(n-1) + b2 y(n-2)
1 a1Z 1 a2 Z 2
H(z) =
1 b1 Z 1 b2 Z 2
Ex:
91
z 5 5 1 Z 2 z 5 1 5Z 2 1
Z 1 Z Z 3
3 12 12 12 3 4 4 4
H(z) = Z 1
Z 2 = Z 1 Z 2
1 1
2 4 2 4
z 1 1
1 Z 1 Z 1 Z 2 1 2
3 4 1 1 1 1 Z Z
= Z 1 Z 2 = z 1 Z Z 1 Z 2
1 3 4 1
2 4 2 4
Ex:
Z Z 1
Z 2 Z
1 Z 2
Z 3
H(z) = = Z 1 Z 1 Z 1
Z 1 Z 1 Z 1
1 1 1 1 1 1
2 4 8 2 4 8
Z
0.65 0.45Z 1
Z 2 1.45 Z 1
= Z 1 Z 1 Z 1
1 1 1
2 4 8
1.45 Z 0.65 0.45Z
1 1
Z 2
=Z Z 1 Z 1 Z 2
1 1
2 4 32
92
6.4 Parallel Combination of Second Order Section (PSOS)
Ex:
z 5 5 1 Z 2 1 5 5 Z 3
Z Z Z 1 Z 2
3 12 12 12 3 12 12 12
H(z) = Z 1
Z 2 = Z 1 Z 2
1 1
2 4 2 4
Z 1 Z 2 Z 3 5 2 1 5 1 Z 1 7
1 Z Z
2 4 12 12 3 12 3 3
Z 3 Z 2 Z 1
12 6 3
___-____+____-_______
7 Z 2 Z 1 1
12 12 3
7 Z 2 7 Z 1 7
12 6 3
______________________
5 1 7
Z
4 3
7 5 1
1 Z
Z 3 4
H(z) = Z
2
3 1 Z 2
1 Z
2 4
93
Ex:
Z Z
Z 2 1
H(z) = obtain PSOS
Z 1 Z 1 Z 1
1 1 1
2 4 8
1 Z 1 2Z 1
A
1
B C
Z Z
1
Z = Z 1 Z 1
1 1
Z 1
1 1 1 1 1 1
2 4 8 2 4 8
A = 8/3 B = 10 C = -35/3
6.5 Jury – Stability Criterion
N ( z)
H(z) =
D( z )
N
D(z) = b Z
N i
i = bo ZN +b1 ZN-1 + b2 ZN-2 +….. bN-1 Z1 + bN
i 0
ROWS COEFFICIENTS
1 bo b1 ……. bN
2 bN bN-1 ……. bo
3 Co C1 ……. CN-1
4 CN-1 CN-2 ……. Co
5 do d1 ……. dN-2
6 dN-2 dN-3 ……. do
94
.
.
.
2N-3 r0 r1 r2
bo b N i
Ci = bN bi i = 0,1,…N-1
co c N 1i
di = c N 1 ci i = 0,1,…N-2
i. D(1) > 0
ii. (-1)N D(-1) > 0
iii. bo bN co c N 1 d o d N 2 ro r2
Ex:
Z4
D(z) = 4 Z 3Z 2Z Z 1
4 3 2
H(z) =
4Z 4 3Z 3 2Z 2 Z 1
1 4 3 2 1 1
2 1 1 2 3 4
3 15 11 6 1
4 1 6 11 15
5 224 159 79
D(1) = 4+3+2+1+1 = 11 > 0, (-1)4 D(-1) = 3 >0
bo b4 co c3 do d2 Stable.
Ex:
1 4Z 2
H(z) = = Ans: Unstable
7 1
1 Z 1 Z 2 4Z 2 7 Z 2
4 2
95
Non Recursive filters Recursive filters
Np M
y(n) =
k
ak x(n-k) y(n) = ak x(n-k) –
k 1
bk y(n-k)
k Nf
for causal system for causal system
Np M
=
k 0
ak x(n-k) y(n) = ak x(n-k) – bk y(n-k)
k0 k 1
For causal i/p sequence It gives IIR o/p but not always.
N
Ex: y(n) = x(n) – x(n-3) + y(n-1)
y(n) =
k 0
ak x(n-k)
NP
It gives FIR o/p. All zero filter. a
k N F
k z k
General TF : H(z) = M
Always stable. 1 bk z k
k 1
bk = 0 for Non Recursive
Nf= 0 for causal system
FIR filters IIR filters
1. Linear phase no phase distortion. Linear phase, phase distortion.
2. Used in speech processing, data Graphic equalizers for digital audio,
transmission & correlation processing tone generators filters for digital
telephone
3. Realized non recursively. Realized recursively.
4. stable Stable or unstable.
H(n) = an u(n) a<1 stable
= 0 a>1 unstable
5. filter order is more Less
6. more co-efficient storage Less storage
7. Quantization noise due to finite Quantization noise
precision arithmetic can be made
negligible
8. Co-efficient accuracy problem is More
96
less severe
9. used in multirate DSP (variable
sampling rate)
97
7. IIR FILTER DESIGN
Butterworth, chebyshev & elliptic techniques.
Impulse invariance and bilinear transformation methods are used for translating s-
plane singularities of analog filter to z-plane.
Frequency transformations are employed to convert LP digital filter design into HP,
BP and BR digital filters.
All pass filters are employed to alter only the phase response of IIR digital filter to
approximate a linear phase response over the pass band.
The system function = H(s)
The frequency transfer function = H(j ) = H(s) / s=j
The power transfer function = H ( j) = H(j ) H*(j ) = H(s) H(-s) / s=j
2
To obtain the stable system, the polse that lie in the left half of the s-plane are assigned to
H(s).
7.1 BUTTERWORTH FILTER DESIGN
1
The butterworth LP filter of order N is defined as HB(s) HB(-s) = 2N
s
1
j c
Where s = j c
1
H B ( j c ) H B ( j c ) db = -3dB ‘s
2
= or
2
It has 2N poles
2N
s
1 =0
c
j
2N
s
= -1
j c
S2N = -1 ( j c )2N
j
j
= e (e 2 c ) = c 2N e j e
2N j 2N
2
e j 2 m
1 N 2 m
2N j
S = c
2N e
98
1 N 2 m
j
2N
Sm = c e 0 m 2N 1
Ex: for N=3
( 4 2 m ) 2 4 5 2 7
j j j j j j
j
e 6
=e 3
,e ,e 3
,e 3
,e ,e 3
= 1200, 1800, 2400, 3000, 3600, 600
1
Vo( s ) CS 1
Vi( s ) 1 1 RCS
R
CS
1 1 1
c
RS s 2
1
c 1
c
Poles that are let half plane are belongs to desired system function.
1
H B ( j ) =
2
2N
1
c
For a large , magnitude response decreases as -N, indicating the LP nature of this
filter.
H B ( j ) = 10log10 H B ( j)
2
dB
2N
= -10 log10( 1 )
c
99
As
= -20 N log10
= -20 N dB/ Decade = -6 N dB/Octane
As N increases, the magnitude response approaches that of ideal LP filter.
The value of N is determined by Pass & stop band specifications.
Ex: Design Butterworth LPF for the following specifications.
Pass band:
-1< H ( j) dB 0
2
for 0 1404 ( p = 1404 )
Stop band:
H ( j) dB < -60
2
for 8268 ( s = 8268 )
If the c is given
2N
s
H ( js ) = [ 1
2
]-1 < 10-6 (-60dB)
c
log( 10 6 1)
=N> s
2 log( )
c
Since c is not given, a guess must be made.
The specifications call for a drop of -59dB, In the frequency range from the edge of the
pass band (1404 ) to the edge of stop band (8268 ). The frequency difference is equal to
log2 8268 = 2.56 octaves.
1404
100
1 oct ---- - 6N dB
2.56 ------ ?
=> 2.56 X - 6N dB = -59 dB’s
59
N= 3. 8
2.56 X 6
There fore: N =4
2N
s
Now H B ( js ) = [ 1
2
]-1 < 10-6
c
2N
s
1 > 106
c
s 2 N > 106 c 2 N
6
s 10 2 N > c => 1470.3 > c
c <1470.3
Let c =1470.3
At this c it should satisfy pass band specifications.
2N
p
H B ( jp) = [ 1
2
]-1 > 0.794 (= -1dB)
c
= 0.59
This result is below the pass band specifications. Hence N=4 is not sufficient.
Let N=5
6
c < s X 10 2 N = 2076.8
10
1404 -1
In the pass band H B ( jp) = [ 1
2
] = 0.98
2076
Since N=5
c = 2076
S1 = -2076
j144
S2, 3 = 2076 (cos (4 /5) j sin(4 /5)) = 2076 e
j108
S4, 5 = 2076 (cos (3 /5) j sin(3 /5)) = 2076 e
101
2076 5
HB(s) = s 2076 s 2 3359s (2076 ) 2 s 2 1283s (2076 ) 2
1. Magnitude response is smooth, and decreases monotonically as increases from 0 to
2. the magnitude response is maximally flat about =0, in that all its derivatives up to
order N are equal to zero at =0
Ex: c=1, N=1
H B ( j) = (1+ 2)-1
2
The first derivative
d 2
H B ( j ) =
2
=0 at =0
d
1 2
2
The second derivative
d2
H B ( j) = -2 at =0
2
d 2
N
3. The phase response curve approaches for large , where N is the no. of poles of
2
butterworth circle in the left side of s-plane.
Advantages:
1. easiest to design
2. used because of smoothness of magnitude response .
102
Disadvantage:
Relatively large transition range between the pass band and stop band.
Other procedure
Avo
When c = 1 Avs = 2N
w
1
wo
2 Avo
H B (s ) = 2N
s
1
j
If n is even S2N = 1 = e j ( 2 k 1)
j ( 2 k 1)
The 2N roots will be Sk= e 2N
k=1,2,….2N
Sk = Cos(2k 1) jSin(2k 1)
2N 2N
1
where
2
Therefore: H B (s ) = T(s) = N /2 k = (2k 1)
(s
2N
2
2Cos k s 1)
k 1
If N is odd
S2n =1 = e j 2 k
Sk = e j 2 k / N k=0,1,2….(2N-1)
1
T(s) = ( N 1) / 2
where k =k
N
k 1
( s 2 2Cos k s 1)
0 20 log H ( j) K1 for 1
20 log H ( j) K 2 for 2
103
k 1
1
2n
1 = K1
10 log 10 10 1
1 2 N c
1
c
k 2
2
2n
1 = K2
10 log 10 10
1
2 2 N c
1
c
k1
1
2n
10 1
Dividing k 2
10
2
10 10 1
10k1
10 1
log 10 k 2
10 10 1
n=
1
2 log 10
2
choosing this value for n, results in two different selections for c . If we wish to satisfy
our requirement at 1 exactly and do better than our req. at 2 , we use
1 2
c = 1
or c = 1
for better req at 2
k1
2n
k2
2n
10 10
1 10 10
1
End
7.2 CHEBYSHEV FILTER DESIGN
2 2 S
1
Defined as Hc(S) Hc(-S) = 1 CN
jp
= measure of allowable deviation in the pass band.
CN(x) = Cos(NCos-1(x)) is the Nth order polynomial.
Let x = Cos
CN(x) = Cos(N )
C0(x) = 1
C1(x) = Cos =x
C2(x) = Cos2 = 2 Cos2 -1 = 2x2-1
C3(x) = Cos3 = 4 Cos3 -3 Cos = 4x3-3x etc..
104
N CN(x)
0 1
1 x
2 2x2-1
3 4x3-3x
4 8x4- 8x2 +1
Two features of Chebyshev poly are important for the filter design
1. C N (x) 1 for x 1
1
2 1
H c ( j) 2 1 for 0 p
Transfer function lies in the range 1 2 1 H c ( j) 2 1 for 0 p
Whereas the frequency value important for the design of the Butterworth filter was the
c , the relevant frequency for the Chebyshev filter is the edge of pass band p .
2. x 1, CN (n) Increases as the Nth power of x. this indicates that for >> p , the
magnitude response decreases as -N, or -6N dB Octane. This is identical to Butterworth
filter.
Now the ellipse is defined by major & minor axis.
Define = 1 1 2
N1 1
N
Minor r = p
2
N1 1
N
Major R = p N = Order of filter.
2
SP = r Cos +j R Sin
105
Ex:
Pass band:
-1< H ( j) dB 0
2
for 0 1404
Stop band:
H ( j) dB < -60
2
for 8268
Value of is determined from the pass band
10 log 1 2
1
> -1dB -1dB = 0.794
< 10 0.1 12 = 0.508
1
= 0.508
Value of N is determined from stop band inequality
2 2 s
1
H c j s
2
= 1 C
N
<10-6
jp
1
s 10 6 1 2
Since 5.9 CN(5.9) > = 1969
p
2
Evaluating
C3(5.9) = 804 C4(5.9) = 9416 therefore N = 4 is sufficient.
Since this last inequality is easily satisfied with N=4 the value of can be reduced to as
small as 0.11, to decrease pass band ripple while satisfying the stop band. The value =0.4
provides a margin in both the pass band and stop band. We proceed with the design with
=0.508 to show the 1dB ripple in the pass band.
106
Axes of Ellipse:
=0.508-1 + (1+0.508-2)1/2 = 4.17
1404 1 1
R= 702 (1.43 0.67) 1494
4 4
4.17 4.17
2
1404 1 1
4.17 4.17 512
4 4
r=
2
7 5
poles locations : ,
8 8
7 7
S1,2 = 512Cos j1494Sin = 473 j572 742e j130
8 8
5 5
S3,4 = 512Cos j1494Sin = 196 j1380 1394e j 98
8 8
742 1394 2
Hc(S) =
[ S 2 S * 2 * 742Cos(130) (742 ) 2 ] [ S 2 S * 2 *1394Cos(98) 1394 2 ]
Chebyshev filter poles are closer to the j axis, therefore filter response exhibits a
ripple in the pass band. There is a peak in the pass band for each pole in the filter, located
approximately at the ordinate value of the pole.
Exhibits a smaller transition region to reach the desired attenuation in the stop band,
when compared to Butterworth filter.
Phase response is similar.
Because of proximity of Chebyshev filter poles to j axis, small errors in their
locations, caused by numerical round off in the computations, can results in significant
changes in the magnitude response. Choosing the smaller value of will provide some
margin for keeping the ripples within the pass band specification. However, too small a
value for may require an increase in the filter order.
It is reasonable to expect that if relevant zeros were included in the system function, a
lower order filter can be found to satisfy the specification. These relevant zeros could serve
to achieve additional attenuation in the stop band. The elliptic filter does exactly this.
7.3 IMPULSE INVARIANCE METHOD
107
H(z) = h( n) Z
n 0
n
e ST ) = h( n )e
STn
H(z) (at z =
re jw e j )T r = e T e jw e jT w T
Let S1 = j => Z1 = eT e jT
2
S2 = j ( ) => Z2 = eT e jT j 2 eT e jT
T
If the real part is same, imaginary part is differ by integral multiple of 2 , this is the
T
biggest disadvantage of Impulse Invariance method.
sa sa
Let HA(S) = =
s a b
2 2
s a jb s a jb
hA(t) = e at Cosbt for t 0 s1 = -a-jb
=0 otherwise s2 = -a+jb
h (nTs) = e anTsCos(bnTs) for n 0
1 e aTs Cos(bTs) Z 1 1 e aTs Cos(bTs) Z 1
H(z) = =
1 2e aTs Cos(bTs) Z 1 Z 2 (1 e ( a jb)Ts Z 1 )(1 e ( a jb)Ts Z 1 )
The pole located at s=p is transformed into a pole in the Z-plane at Z = e pTs , however, the
finite zero located in the s-plane at s= -a was not converted into a zero in the z-plane at Z =
e aTs , although the zero at s= was placed at z=0.
108
Desing a Chebyshev LPF using Impulse-Invariance Method.
S1,2 = -473 j 572
S3,4 = -196 j 1380
[The freq response for analog filter we plotted over freq range 0 to 10000 . To set the
discrete-time freq range (0, ), therefore Ts = 10-4]
Ts
Z1,2 = e S1, 2Ts
= e 0.148 j 0.179 = 0.862 e j10.2
Z3,4 = e S 3, 4Ts
= e 0.061 j 0.433 = 0.94 e j 24.8
k
H(z) = 1
(1 2 * 0.862Cos10.2Z 0.862 Z )(1 2 * 0.94Cos24.8Z 1 0.94 2 Z 2 )
2 2
k
= 1 2
(1 1.69Z 0.743Z )(1 1.707 Z 1 0.88Z 2 )
Methods to convert analog filters into Digital filters:
1. By approximation of derivatives
dx x(nTs) x(nTs Ts)
/ t=nTs =
dt Ts
1 Z 1
S=
Ts
109
Or
Using forward-difference mapping based on first order approximation Z = e sTs 1+STs
Z 1
S=
Ts
Using backward- difference mapping is based on first order approximation
Z 1 e sTs 1 STs
Z 1 1 Z 1
S =
ZTs Ts
d 2x d dx
/t=nTs = / t nTs
dt 2
dt dt
x(nTs) x(nTs Ts) x(nTs Ts) x(nTs 2Ts)
= Ts Ts
Ts
x(nTs) 2 x(nTs Ts ) x(nTs 2Ts)
=
Ts 2
2
1 2Z 1 Z 2 1 Z 1
S 2
=
Ts 2 Ts
k
1 Z 1
k
Therefore S =
T
1 Z 1
Therefore H(z) = Ha(s) /s= using backward difference
T
1 0.5(1 STs)
Z= = 0.5 +
1 STs 1 STs
110
1 1 jTs
= =
1 jTs 1 Ts 1 2Ts 2
2 2
0.5(1 STs)
Z - 0.5 =
(1 STs)
z 0.5 0.5 is mapped into a circle of radius 0.5, centered at Z=0.5
Using Forward-difference
Z 1
S= Z=1+STs
Ts
u+jv = 1+ ( j) Ts
if =0 u=1 and j axis maps to Z=1
If >0, then u>1, the RHS-plane maps to right of z=1.
If <0, then u<1, the LHS-plane maps to left of z=1.
The stable analog filter may be unstable digital filter.
7.4 Bilinear Transformation
Provides a non linear one to one mapping of the frequency points on the jw axis in s-
plane to those on the unit circle in the z-plane.
111
This procedure also allows us to implement digital HP filters from their analog
counter parts.
2 Z 1 2 1 Z 1
S= =
Ts Z 1 Ts 1 Z 1
{Using trapezoidal rule y(n)=y(n-1)+0.5Ts[x(n)+x(n-1)]
H(Z)=2(Z-1) / [Ts(Z+1)] }
2 Z 1
To find H(z), each occurrence of S in HA(s) is replaced by
Ts Z 1
STs
1
And Z = 2
STs
1
2
1/ 2
2 Ts 2 j tan 1
Ts
Ts 1 e 2
j 1 2
e jw 2
Ts = 1/ 2
j 1 2 Ts
2 Ts
j tan 1
1 e
2
2
2
Ts
j 2 tan 1 Ts
e jw
e 2
w=2tan-1
2
The entire j axis in the s-plane - <j < maps exactly once onto the unit circle -
w such that there is a one to one correspondence between the continuous-time and
discrete time frequency points. It is this one to one mapping that allows analog HPF to be
implemented in digital filter form.
As in the impulse invariance method, the left half of s-plane maps on to the inside of the
unit circle in the z-plane and the right half of s-plane maps onto the outside.
112
2 w
In Inverse relationship is tan
Ts 2
w w3
Sin
w ....
For smaller value of frequency
2 2 = 2 2 8 w
2
Ts w Ts w Ts
Cos 1 ....
2 4
(B.W of higher freq pass band will tend to reduce disproportionately)
The mapping is linear for small and w. For larger freq values, the non linear
compression that occurs in the mapping of to w is more apparent. This compression
causes the transfer function at the high freq to be highly distorted when it is translated to
the w-domain.
Prewarping Procedure:
When the desired magnitude response is piece wise constant over frequency, this
compression can be compensated by introducing a suitable prescaling or prewarping to the
freq scale. scale is converted into * scale.
2 Ts
* = tan
Ts 2
We now derive the rule by which the poles are mapped from the s-plane to the z-plane.
1
Let HA(s) = S=Sp
S Sp
113
1 Ts(1 Z 1 )
H(z) = =
2 1 Z 1
Sp 2 SpTs 1 2 SpTs Z 1
Ts 1 Z 1 2 SpTs
2 SpTs
A pole at S=Sp in the s-plane gets mapped into a zero at z= -1 and a pole at Z =
2 SpTs
Ex:
Chebyshev LPF design using the Bilinear Transformation
Pass band:
-1< H ( j) dB 0 for 0 1404 =4411 rad
Stop band:
H ( j) dB < -60 for 8268 rad/sec =25975 rad/s
Let the Ts = 10-4 sec
Prewarping values are
2 Ts
p* = tan = 2*104 tan(0.0702 ) = 4484 rad/sec
Ts 2
2 Ts
And s* = tan = 2*104 tan(0.4134 ) = 71690 rad/sec
Ts 2
The modified specifications are
Pass band:
-1< H ( j*) dB 0 for 0 * 4484 rad/s
Stop band:
H ( j*) dB < -60 for * 71690rad/sec
10log 1 2
1
Value of : is determined from the pass band ripple 1dB
= 0.508
Value of N: is determined from
s * 1
H c js * = 1 2CN2
2
<10-6
p *
s *
Since 16
p *
114
CN2(16)
106 1
<
2
1
10 6 1 2
CN(16) < 2
= 1969
(0.508)
C3(16) = 16301
N = 3 is sufficient
Using Impulse Invariance method a value of N=4 was required.
=4.17
N1 1
N
4484
1
1
Major R = p * = 4.17 4.17 3 5001
3
2 2
4484
1 1
3
r= 4.17 3
4.17 2216
2
2
Since there are three poles, the angles are &
3
S1 = r cos + j Rsin = -2216
2 2
S2,3 = 2216 Cos j 5001 Sin = -1108 j 4331 = 4470 e j104.4
3 3
4.43 *1010
Hc(s) =
( s 2216)( S 2 2223s 4470 2 )
Pole Mapping
At S=S1
2 (2216 *10 4 )
In the Z-plane there is zero at Z = -1 and pole at Z = 0.801
2 (2216 *10 4 )
S2,3 = there are two zeros at Z=-1
2 (1108 j 4331) *10 4
Z= 4
0.801 j 0.373 0.9e j 24.5
2 (1108 j 4331) *10
-3 1 Z 1 1 2Z 1 Z 2
H(z) = 4.29 * 10
1 0.801Z 1 1 1.638Z 1 0.81Z 2
Pole Mapping Rules:
Hz(z) = 1-CZ-1 zero at Z=C and pole at Z = 0
115
1
Hp(z) = pole ar Z=d and zero at z=0
1 dZ 1
C and d can be complex-valued number.
Pole Mapping for Low-Pass to Low Pass Filters
Applying low pass to low pass transformation to Hz(z) we get
c 1
1 Z
Z
1
1 c
HLZ(Z) = 1-c = (1+c )
1 Z 1 1 Z 1
c
The low pass zero at z=c is transformed into a zero at z=C1 where C1 =
1 c
And pole at z=0 is Z=
Similarly,
1 Z 1
HLP(Z)=
1 d 1 d Z 1
1 d
d
Pole at z=d => Z=
1 d
Zero at z=0 => z =
1 Z 1 2Z 2Z
1 1 2
H(z) = K
1 0.622Z 1 1.07Z 0.674Z
1 1 2
K=
1 (1)(0.356)3 0.029
(1 0.801* 0.356)(1 (0.819 j 0.373)( 0.356))(1 (0.819 j 0.373)( 0.356))
116
8.FIR Filters
( )
Phase Delay: p
d ()
Group Delay: g
d
If p = g =constant and independent of frequency are called as constant time delay or
linear phase filters.
() o
o
p Changes with frequency
g = - =constant.
8.1 Type 1 Sequence
N 1
Center of Symmetry M= integer value
2
𝑁−3 𝑁−1
( 2 ) 𝑁−1 (𝑁−1)
H(𝑒 𝑗Ω𝑇
)= ∑𝑛=0 ℎ(𝑛)𝑒 −𝑗Ω𝑛𝑇 + h( ) 𝑒 −𝑗Ω𝑇( 2 ) +∑ 𝑁+1 ℎ(𝑛)𝑒 −𝑗Ω𝑛𝑇
2 𝑛= 2
Let N-1-n =n
𝑁−1 𝑁−3
(𝑁−3)/2 𝑁−1 −𝑗Ω𝑇( ) ( 2 )
∑𝑛=0 ℎ(𝑛)𝑒 −𝑗Ω𝑛𝑇 + h( )𝑒 2 + ∑𝑛=0 ℎ (𝑁 − 1 − 𝑛)𝑒 −𝑗Ω(𝑁−1−𝑛)𝑇
2
𝑁−1 𝑁−3
(𝑁−3)/2 𝑁−1 −𝑗Ω𝑇( ) ( 2 )
∑𝑛=0 ℎ(𝑛)𝑒 −𝑗Ω𝑛𝑇 + h( )𝑒 2 + ∑𝑛=0 ℎ(𝑛)𝑒 −𝑗Ω(𝑁−1−𝑛)𝑇
2
117
𝑁−1 𝑁−1 𝑁−3 𝑁−1
(𝑁−3)/2 𝑁−1 ( )
𝑒 −𝑗Ω 2 𝑇 [ ∑𝑛=0 ℎ (𝑛 )𝑒 𝑗Ω𝑇[ 2 −𝑛]
+ h( ) + ∑𝑛=0 2
ℎ(𝑛)𝑒 −𝑗Ω( 2 −𝑛)𝑇
2
𝑁−1
𝑗Ω𝑇 −𝑗Ω 𝑇 (𝑁−3)/2 𝑁−1 𝑁−1
H(𝑒 )= 𝑒 2 [ ∑𝑛=0 2ℎ(𝑛)𝑐𝑜𝑠Ω𝑇( − 𝑛) + h( ) ]
2 2
N 3
h( M ) 2 h(n )CosT ( n M )e jMT
2
H(w) =
n 0
N 1
T
2
Amplitude spectrum is even symmetric about w=0 & & both H(0) & H( ) can be non
zero.
8.2 Type 2 Sequence
h(n) = h(N-1-n)
N 1
Center of Symmetry M= half-integer value
2
N 1
) = h ( n )e
jnT
j T
H( e
n 0
N
1 N 1
h( n)e
2 jnT
= h( n)e
n 0
jnT
+ N
n
2
118
Let N-1-n = m
N
1
2 0
= h( n)e jnT
+ h ( N 1 m)e jT ( N 1 m )
n 0 m
N
1
2
But h(N-1-m) = h(m)
N N
1 1
2 2
== h( n)e
n 0
jnT
+ h ( m) e
m 0
jT ( N 1 m )
N
1 N 1 N
1
2 jT N 1
h ( n )e jnT 2 jT
+ h ( m)e
2 jT ( N 1 n ) 2
= e e
n 0 n 0
N
1 N 1 j ( nT 2T ( N 1) j ( T ( N 1 n )
T
( N 1)
2 jT
e e 2
= 2 h( n) e
n 0
2
2
N
1 N 1
2 jT N 1
= 2 h ( n )e cos T n
2
n 0 2
N
N 1 1
jT 2 N 1
=e
2
2h(n) cos T n
n 0
----Magnitude
2
N 1
T Linear Phase
2
N2 1
jMT
H(w) = 2 h(n )CosT (n M )e
n 0
The Amplitude spectrum is even symmetric about w=0 & odd symmetric about w= &
both H( ) is always zero for type 1 & 2 : Constant phase delay and group delay.
8.3 Type 3 Sequence
119
N 1
M= integer value
2
N23
2 h(n) SinT ( M n) e jMT
H(w) = j
n 0
It shows generalized linear phase of MT and constant group delay of M. The
2
Amplitude spectrum is odd symmetric about w=0 & w= and H(0) & H( ) are always zero.
(Generalized means () may jump of at 0 if H(ejw) is imaginary.
8.4 Type 4 Sequence
N2 1
jMT
H(w) = j 2 h(n) SinT ( M n) e
n 0
Generalized linear phase and constant group delay of M. The Amplitude spectrum is odd
symmetric about w=0 & even symmetric about w= and H(0)=0 always.
120
8.5 Poles & Zeros of linear phase sequences:
The poles of any finite-length sequence must lie at z=0. The zeros of linear phase
sequence must occur in conjugate reciprocal pairs. Real zeros at z=1 or z=-1 need not be
paired (they form their own reciprocals), but all other real zeros must be paired with their
reciprocals. Complex zeros on the unit circle must be paired with their conjugate (that form
their reciprocals) and complex zeros anywhere else must occur in conjugate reciprocal
quadruples. To identify the type of sequence from its pole-zero plot, all we need to do is
check for the presence of zeros at z= and count their number. A type-2 seq must have an
odd number of zeros at z=-1, a type-3 seq must have an odd number of zeros at z=-1 and
z=1, and type-4 seq must have an odd number of zeros at z=1. The no. of other zeros if
present (at z=1 for type=1 and type-2 or z=-1 for type-1 or type-4) must be even.
121
Prove:
H(Z) = ∑𝑵−𝟏
𝒏=𝟎 𝒉(𝒏)𝒁
−𝒏
H(Z)at Z=Z0 = H(Z0)= ∑𝑵−𝟏 −𝒏
𝒏=𝟎 𝒉(𝒏) 𝒁𝟎
−(𝑵−𝟏)
=h(0) + h(1) 𝒁−𝟏 −𝟐
𝟎 ++ h(2) 𝒁𝟎 + …….. + h(N-1) 𝒁𝟎 =0
For linear phase h(N-1-n)=h(n)
−(𝑵−𝟏)
h(N-1) + h(N-2) 𝒁−𝟏 −𝟐
𝟎 ++ h(N-2) 𝒁𝟎 + …….. + h(0) 𝒁𝟎 =0
−(𝑵−𝟏) (𝑵−𝟏)
𝒁𝟎 [h(0) + h(1) 𝒁𝟏𝟎 ++ h(2) 𝒁𝟐𝟎 + …….. + h(N-1) 𝒁𝟎 ]=0
−(𝑵−𝟏)
𝒁𝟎 = ∑𝑵−𝟏 − -n
𝒏=𝟎 𝒉(𝒏)(𝒁𝟎 ) =0
Therefore H(Z0)= H(𝒁−𝟏
𝟎 )=0
If Z0 is a zero of H(Z), then 𝒁−𝟏
𝟎 is also a zero
1 If Z1 = -1 then 𝒁−𝟏
𝟏 = Z1 , then the zero lie at Z1=-1
This group contains only one zero on the unit circle
2) If Z2 is real zero with |Z2| <1 then 𝒁−𝟏
𝟐 is also a real zero and there are two zeros
in this group
3) If Z3 is a comple zero with |Z3| =1 then 𝒁−𝟏 ∗
𝟑 = 𝒁𝟑 and there are two zeros in this
group
4) If Z4 is a complex zero with |Z4| ≠1 then this group contain four zeros Z4, 𝒁−𝟏
𝟒 =
𝒁∗𝟒 , (𝒁∗𝟒 )-1
(𝒁∗𝟒 )-1
Z3 𝒁∗𝟒
Z1 Z2 𝒁−𝟏
𝟐
Z4
𝒁−𝟏 ∗
𝟑 = 𝒁𝟑
𝒁−𝟏
𝟒
122
FIR Filters
Fs Fs
8.6 Fourier series Method F
2 2
2Fs 2Fs
2F
2 2
s s
2 2
1. Frequency response of a discrete-time filter is a periodi function with period s
(sampling freq).
2. From the F.S analysis we know that any periodic function can be expressed as a linear
combination of complex exponentials.
Therefore desired freqency response of a discrete time filter can be represented by F.S as
H (e jT
) h( n )e
n
jnT
T = sampling period
The F.S co-efficient or impulse response samples of filter can be obtained using
1 s / 2
jT jnT
H ( e ) e d
h (n) =
s s / 2
clearly if we wish to realize this filter with impulse response h(n), then it must have finite
no. of co-efficient, which is equivalent to truncating the infinite expansion of H (e jT ) , which
leads to approximation of H (e jT ) , which is denoted by
M
H 1 (e jT
) h( n )e
n M
jnT
.
N 1
We choose M=
2 , in order to keep ‘N’ no of samples in h(n).
M
H1(z) = h( n) Z
n M
n
123
However, this filter can’t be physically realizable due to the presence of +ve powers of Z,
means that the filter must produce an output that is advanced in time with respect to the i/p.
N 1
This difficulty can be overcome by introducing a delay M= samples.
2
M
Therefore H(z) = Z-M H1(z) = Z-M h( n) Z
n M
n
H(z) = h(-M)Z0 + h(-M+1) Z-1 +…. +h(M) Z-2M
Let bi = h(i-M) i=0 to 2M
2M
H(z) = b Z
i 0
i
i
be the transfer function of discrete filter that is physically realizable.
Properties:
1. N=2M+1, impulse response co-eff, bi = 0 to 2M.
2. h(n) is symmetric about bM
Ex: M=4
3. The duration of impulse response is Ti = 2MT
4. Its magnitude and time delay function can be found in the following way
H (e jT ) e jMT H1 (e jT )
H (e jT ) H 1 (e jT )
This implies that magnitude response of the filter we have desired approximates the
desire magnitude response. The time delay of H(e jw) is a constant M. thus sinusoids of
different frequencies are delayed by the same amount as they are processed by the filter, we
124
have designed. Consequently, this is a linear phase filter, which means that it does not
introduce phase distortion.
Ex:
Design a LPF (FIR) filter with frequency response
H (e jT ) 1 for c
s
= 0 for c
2
1 c jnT
h(n) =
s c
e d
2 c
s 0
= Cos(nT )d
2 SincnT
=
s nT
2 1
SincnT SincnT s 2Fs 1
= 1 n w= T T
2Fsn. 2 2 Fs
Fs
bi = h(i-M)
2M
H(z) =
i
bZ i
i 0
Ex:
Design LPF that approximate following freq response.
H(F) =1 0 F 1000Hz
=0 else where 1000 F Fs/2
When the sampling frequency is 8000 SPS. The impulse response duration is to be
limited to 2.5ms
125
Ti = 2MT
2.5 *10 3
M= 10 N=21
1
2*
800
1 c jnT
s c
h(n) = 1.e d
1 Fc j 2FnT 1 Fc j 2FnT 2 Fc
2Fs Fc Fs Fc Fs 0
= 1.e 2dF = e dF Cos(2FnT )dF
1 1
= Sin 2FcnT Sin (0.25n )
n n
________________________________________________________________
OR
1
w = T = 2 *1000 *
8000 4
Hc(w) = 1 w
4
= 0 else where
1
1.e dw = 1 Sin (0.25n )
4 jwn
2
4 n
h(0) = 0.25 h(6) = -0.05305
h(1) = 0.22508 h(7) = -0.03215
h(2) = 0.15915 h(8) = 0
h(3) = 0.07503 h(9) = 0.02501
h(4) = 0 h(10) = 0.03183
h(5) = -0.04502
bi = h(i-10)
20
H(z) = b Z
i 0
i
i
126
FIR HPF
1 c
e jnT d
s / 2
jnT
h(n) = 1.e d
s s / 2 c
1 e jnT c e jnT s / 2
s jnT
= s / 2 c
jnT
jcnT s
j nT
s
j nT
1 e e 2
e 2 e jcnT
= s jnT
jcnT jcnT
s
j nT j nT
s
2 1 e e e 2
e 2
= s nT
2j 2j
2 s nT
=
2FsnT Sin c nT Sin 2
1
=
1
sin c nT Sinn = sin c nT
n n
FIR BPF
h(n) =
2 u
cos nT d =
1
sin u nT sin l nT
s l n
Ex:
Desing a BPF for H(f) = 1 160 F 200Hz
=0 else where
Fs = 800SPS
Ti = 20 ms
Ti 20 *10 3
M= 8 N = 17
2T 1
2*
800
127
sin 0.5n sin 0.4n
h(n) =
1
Sin 2F nT Sin 2F nT =
n
u l
n
h(0) = 0.1 h(4) = 0.07568
h(1) = 0.01558 h(5) = 0.06366
h(2) = -0.09355 h(6) = -0.05046
h(3) = -0.04374 h(7) = -0.07220 h(8) = 0.02338
16
H(z) = b Z
i 0
i
i
bi = h(i-8) h(-n) = h(n)
8.7 WINDOWING
Disadvantage of F.S is abrupt truncation of FS expansion of the freq response. This
truncation result in a poor convergence of the series.
The abrupt truncation of infinite series is equivalent to multiplying it with the rectangular
sequence.
WR(n) = 1 n M
=0 else where
h (n) h(n)WR (n)
H (e jw ) H (e jw ) *WR (e jw )
128
1
H (e
j
= )WR (e j ( w ) )d
2
WR(ejw) => FT of Rectangular Window
N 1 wN wN
2 NSa Sin
jw
WR(e ) =
1.e
N 1
jwn
dw
=
2
w = w
2
n Sin Sa
2 2 2
4
Main lobe width = & it can be reduced by increasing N, but area of side lobe will
N
be constant.
For larger value of N, transition region can be reduced, but we will find overshoots &
undershoots on pass band and non zero response in stop band because of larger side lobes.
So these overshoots and leakage will not change significantly when rectangular window is
used. This result is known as Gibbs Phenomenon.
The desined window chts are
1. Small width of main lobe of the frequency response of the window containing as
much as of the total energy as possible.
2. Side lobes of the frequency response that decrease in energy as w tends to .
3. even function about n=0
129
N 1
4. zero in the range n
2
Let us consider the effect of tapering the rectangular window sequence linearly from the
middle to the ends.
Triangular Window:
2n N 1
WT (n) 1 n
N 1 2
=0 else where
In this side lobe level is smaller than that of rectangular window, being reduced from -13
8
to -25dB to the maximum. However, the main lobe width is now . There is a trade off
N
between main lobe width and side lobe levels.
General raised cosine window is
2n N 1
W(n) = (1 )Cos for n
N 1 2
=0 else where
If =0.5 Hanning Window
If =0.54 Hamming Window
2n 4n
WB(n) = 0.42 + 0.5 Cos 0.08Cos Blackman Window
N 1 N 1
Kaiser Window
2n
2
Io 1
N 1
Wk (n) for n
N 1
Io( ) 2
=0 else where
is constant that specifies a freq response trade off between the peak height of the side
lobe ripples and the width or energy of main lobe and Io(x) is the zeroth order modified
Bessel function of the first kind. Io(x) can be computed from its power series expansion
given by
130
2
1 x k
Io(x) = 1+ k! 2
k 1
0.25 x 2 0.25 x 2 2 0.25 x 2
3
=1+
(1!) 2 + (2 !) 2 + (3 !) 2 +…..
Window Peak amplitude Transition width Minimum stop
of side lobe dB of main lobe band deviation dB
Rectangular -13 4 -21
k=1
N
Triangular -25 8 -25
k=2
N
Hanning -31 8 -44
k=2
N
Hamming -41 8 -53
k=2
N
BlackMan -57 12 -74
k=3
N
Kaiser variable variable -
If we let K1,W1 and K2,W2 represent cutoff (pass band) * stop band requirements for the
digital filter, we can use the following steps in design procedure.
1. Select the window type from table to be the one highest up one list such that the stop
band gain exceeds K2.
2. Select no. of points in the windows function to satisfy the transition width for the type
2
of window used. If Wt is the transition width, we must have Wt = W2-W1 k .
N
where K depends on type of window used.
131
K=1 for rectangular, k=2 triangular…..
2
Therefore N K
w2 w1
If analog freq are given, it must be converted in to Digital using w= T
Ex:
Apply the Hamming Window to improve the low pass filter magnitude response ontained
in ex1:
2n N 1
WH(n) = 0.54 + 0.46 Cos for n
N 1 2
=0 else where
N = 2M+1 = 21
WH(0) = 1 WH(6) = 0.39785
WH(1) = 0.97749 WH(7) = 0.26962
WH(2) = 0.91215 WH(8) = 0.16785
WH(3) = 0.81038 WH(9) = 0.10251
WH(4) = 0.68215 WH(10) = 0.08
WH(5) = 0.54
Next these window sequence values are multipled with coefficients h(n), obtained in ex1,
to obtain modified F.S Co eff h’(n).
h’(0) =0.25
h’(1) =0.22
h’(2) =0.14517
h’(3) =0.0608
h’(4) =0
h’(5) =0.02431
h’(6) =0.02111
h’(7) =-0.0086725
h’(8) =0
h’(9) =0.00256
h’(10) =0.00255
132
2M
H’(z) = b'
i 0
i Z i
bi' = h’(i-M) 0 i 20 h’(-n) = h’(n)
Ex:
Find a suitable window and calculate the required order the filter to design a LP digital
filter to be used A/D-H(Z)-D/A structure that will have a -3dB cutoff of at 30 rad/sec and
an attenuation of 50dB at 45 rad/sec. the system will use a sampling rate of 100 samples
/sec
Sol:
The desired equivalent digital specifications are obtained as
1
Digital ….. w1 wc cT 30 0.3 k1 3dB
100
1
w2 2T 45 0.45 k2 50dB
100
1. to obtain a stop band attenuation of -50dB or more a Hamming window is shosen
since it has the smallest transition band.
2. the approximate no. of points needed to satisfy the transition band requirement (or the
order of the filter ) can be found for w1 =0.3 rad &w2 = 0.45 rad, using Hamming
window (k=2), to be
2 2.2
N k =26.65
w2 w1 0.45 0.3
N = 27 is selected
133
Kaiser window
The attractive property of the Kaiser window is that the side lobe level and main lobe
width can be varied continuously by simple varying the parameter . Also as in other
window, the main lobe width can be adjusted by varying N.
we can find out the order of Kaiser window, N and the Kaiser parameters to design
FIR filter with a pass band ripple equal to or less that Ap, a minimum stop band attenuation
equal to or greater than As, and a transition width Wt, using the following steps:
Step 1 : Choose 𝛿 such that 𝛿 = 𝑀𝑖𝑛( 𝛿 p , 𝛿 s )
1 p
s 10 0.05 As , [ Prove As = 20 log 10 => As = 20 log s ]
s
10 0.05 Ap 1
p 0.05 Ap
10 1
1 p
[ Prove Ap = 20 log 10
1 p
1
100.05Ap =
1
1 100.05Ap = 1
Therefore: solving above eq for , we get
10 0.05Ap - 1
10 0.05Ap 1 ]
Step 2:
Calculate As using the shosen values
Aso= 20 log
134
Step 3:
Calculate the parameter as follows for
=0 for Aso 21 dB
= 0.5842(Aso -21)0.4 + 0.07886(Aso -21) for 21< Aso 50 dB
= 0.1102(Aso-8.7) for Aso>50 dB
Step 4:
Calculate D as follows
D = 0.9222 for Aso 21 dB
As 7.95
= for Aso >21 dB
14.36
Step 5:
Select the lowest odd value of N satisfying the inequality
samD
N 1
t
Wsam : Angular Sampling frequency
sam : Analog Freq
t = s- p for LPF
135
= p- s for HPF
= Min[( p1- s1), ( s2- p2)] for BPF
= Min[( s1 - p1), ( p2- s2)] for BSF
-3dB cutoff freq c can ve considered as follows
c =
1
p s for LPF & HPF
2
t t
c1 = p1 ; c2 p2 for BPF
2 2
t t
c1 = p1 ; c2 p2 for BSF
2 2
Ex:
Calculate the Kaiser parameter and the no. of points in Kaiser Window to satisfy the
following lowpass specifications.
Pass band ripple in the freq range 0 to 1.5 rad/sec 0.1 dB
Minimum stop band attenuation in 2.5 to 5.0 rad /s 40 dB
Sampling frequency: 10 rad/s
Sol:
The impulse response samples can be calculated using h(n) =
1
sin c nT
n
1
Where c = (1.5 2.5) =2rad/s
2
And the no. of points required in this sequence can be found as follows
Step1:
s 10 0.05( 40) 0.01
10 0.05( 0.1) 1
p 0.05( 0.1) 5.7564*10 3
10 1
Therefore we choose, 5.7564*10 3
Step 2:
3
Aso = -20 log( 5.7564*10 ) = 44.797 dB
Step 3 & 4:
136
= 0.5842 ( 44.797 -21)0.4 + 0.07886 ( 44.797 -21) = 3.9524
D = 2.566
Step 5:
10(2.566)
N 1 26.66 => N=27
1
2n
2
Io 1
N 1
Wk (n)
Io( )
Io( )
Wk (0) 1
Io( )
Io(3.94) 10.269
= 0.9899
Io(3.9524) 10.3729
137
9.OBJECTIVE PAPER-1
1)What is the parsval’s theorem expression in DTFT :
𝜋
∑n=-∞|x(n)|2=(1/2π)∫−𝜋 |𝑋(𝑤 )|2dw
Match the following:
2) E = , P = 0 a) power
3) E , P = 0 b) Neither energy nor power
4) E = , P 0, P c) Energy
Match the following
5) e-t u(t) a) power
6) u(t) b) Neither energy nor power
7) 1/t c) Energy
8) x (n) = 6e j 2 n / 4, what is the power of the signal
a) 36W b) 72W c) 18W d) none
Match the following: For a real valued sequence, the DTFT follow the properties as
9) Re [H (jw) ] a) Real valued function of w
10) Im[ H(jw) ] b) even function of w
11) F.T [even symmetric sequence] c) Imaginary valued function of w
12) F.T [odd symmetric sequence] d) odd function of w
13) x(n) = {4, 1, 3} h(n) = {2, 5, 0, 4} what is the output of the system.
a) {8, 22, 11, 31, 4, 12} b) {8, 22, 11, 31, 4, 12} c) {8, 22, 11, 31, 4, 12}
d) none
14) y(n) = x(n) * h(n) then y1 (n) = {0, 0, x(n), 0 } * { 0, h(n), 0 } is equal to
a) {0, 0, y(n), 0} b) {0, 0, 0, y(n), 0, 0} c) [0, 0, y(n), 0 } d) {0, y(n), 0, 0}
15)If x(n) and h(n) are having N values each, to obtain linear convolution using circular
convolution, the number of zeros to be appended to each sequence is
a) N – 1 b) 2N – 1 c) N d) N + 1
16)W49 = ?
a) – j b) + j c) + 1 d) -1
17) DFT [ x* (-n) ] = ?
a) X * (K) b) X * (-K) c) X * (N-K) d) none
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
b c a c a b a b d a c c b a a a
138
10.OBJECTIVE PAPER-2
1) The region of convergence of the Z-transform of a unit step function is
a)|Z| >1 b) |Z|<1 c) (real part of Z ) >0 d) (real part of Z ) <0
2) The Z T of the function f(nT) = anT is
a) Z/(Z-aT) b) Z/(Z+aT) c) Z/(Z-a-T) d) Z/(Z+a-T)
3) The Z T of the function (n k ) is
k 0
2
a) (Z-1)/Z b) Z/(Z-1) c) Z/( Z-1) d) (Z-1)2/Z
4) The Z T of a signal is given by X(Z)= Z -1(1-Z-4)/( 4(1-Z-1)2) its final value is
a) ¼ b) 0 c) 1 d) infinity
5) Consider the system shown in fig. The transfer function Y(Z) / X(Z) of the system is
x(n) y(n)
+ +
Z-1
-b a
-1 -1 -1 -1
a) (1+aZ )/ ( 1+bZ ) b) (1+bZ )/ ( 1+aZ )
c) (1+aZ-1)/ ( 1-bZ-1) d) (1-bZ-1)/ ( 1+aZ-1)
6) A linear discrete time system has the characteristic equation Z 3-0.8 Z=0, the system
a) is stable b) is marginally stable
c) is un stable d) stability cannot be assessed from the given information
7) The advantage of Canonic form realization is
a) smaller no of delay elements b) larger no of delay elements
c) hard ware flexibility d) none
3
akx(n k )
5
8) y(n) = - k 1
bky(n k ) the minimum no of delay elements
k 2
needed to realize the system is
a) 5 b) 10 c) 8 d) 11
9) Expand CSOS Ans: Cascaded form of second order section.
PSOS Ans: Parallel form of Second order section
10) To ensure a causal system, the total no of zeros must be less than or equal to the total
number of poles ( T / F )
139
1 2 3 4 5 6 7 8 9 10
a a c c a a a c T
11) The poles or zeros at the origin do effect the magnitude response ( T / F)
12) All poles and zeros of a minimum phase system lie inside the unit circle ( T / F)
13) To realize FIR filter
a) no feedback paths and forward path b) no feedback paths and no forward path
c) feedback paths and no forward path d) feedback paths and forward path
14) Find total no of complex multiplications using FFT for N=8: __________
15) Find total no of complex additions using FFT for N=8: __________
16) Find total no of real additions using direct DFT for N=8: __________
11 12 13 14 15 16
F T a 12 24 240
17) What is Z T of 2 (3n) u (-n-1): ____(-2)/(1-3z-1_)___ or (-2z)/(z-3)__________
18) (2M) Show the structure of Z
Direct form –II for 2nd order system
Z
x(n) y(n)
Z-1 a1
+ b1
b2 Z-1 a2
19) Show the structure of butterfly b3 Z-1 a3
Z-1
bnp
anp
Z-1
bm
140
11.OBJECTIVE PAPER-3
State TRUE or FALSE
1) u(n) = (n k )
K 0
2) x(n) = cos 0.5n is periodic sequence
3) Discrete-time sinusoidal signals with frequency that are separated by an integral
multiple of 2π are identical
4) y(n) =x(-n) is time invariant
Match the following
5) h(k )
1 Zero input response
6) Impulse response of difference equation is 2 linear
7) y(n) = |x(n)| 3 Stable
8) y(n) = x(n2) 4 Time invariant
CHOOSE THE CORRECT ANSWER
9) x(n) = Cos 0.125n , what is the period of the sequence
a) 8 b) 16 c) 125 / 2 d) none
10) y (n) = x (2n)
a) Causal b) Non-Causal c) Time invariant d) none
11) x(-n + 2) is obtained using following operartion
a) x (-n) is delayed by two samples b) x (-n) is advanced by two samples
c) x (n) is shifted left by two samples d) none
12) In situations where both interpolation and decimation are to be performed in
succession, it is therefore best to
a) Interpolate first, then decimate b) Decimate first and interpolate
c) Any order we can perform d) none
1 2 3 4 5 6 7 8 9 10 11 12
T F T F 3 1 4 2 B B A a
13) The output of anti causal LTI system is
n
a) y (n) = h( k ) x ( n k )
K 0
b) y (n) = h( k ) x ( n k )
K 0
1
c) y (n) = h( k ) x ( n k )
d) y (n) = h( k ) x ( n k )
141
14) (n-k) * x (n-k) is equal to
a) x(n-2k) b) x(n-k) c) x(k) d) none
15) Given x(n) the y(n) = x(2n – 6) is
a) x(n) is Compressed by 2 and shifted by 6 b) x(n) is Compressed by 2 and
shifted by 3
c) x(n) is Expanded by 2 and shifted by 3 d) none
16) Decimation by a factor N is equivalent to
a) Sampling x(t) at intervals ts / N b) Sampling x(t) at intervals tsN
c) N fold increase in sampling rate d) none
17) In fractional delay, x(n-M/N), specify the order of operation.
a) Decimation by N, shift by M, Interpolation by N
b) Shift by M, Decimation by N and Interpolation by N
c) Interpolation by N, Shift by M and Decimation by N
d) All are correct
18) Given g(n) = {1, 2, 3} , find x(n) = g (n / 2), using linear interpolation
a) 1, 0, 2, 0, 3 b) 1, 1, 2, 2, 3, 3 c) 1, 3/2, 2, 5/2, 3 d) none
19)
h1(n) + h3(n) y(n)
x(n)
h2(n)
In the figure shown, how do you replace whole system with single block
a) [ h1(n) + h2(n) ] * h3(n) b) h1(n)h3(n) * h2(n)h3(n)
c) [ h1(n) + h2(n) ] h3(n) d) none
20 The h(n) is periodic with period N, x(n) is non periodic with M samples, the output
y(n) is
a) Periodic with period N b) Periodic with period N+M
c) Periodic with period M d) none
13 14 15 16 17 18 19 20
C A B B C C A A
142
12.OBJECTIVE PAPER-4
1) If x(n) = {-1, 0, 1, 2, 1, 0, 1, 2, 1, 0, -1} What is X(0)
a) 6 b) 10 c) 0 d) none
2) If x(n) = 1, |n|≤2
0, other wise
Find DTFT
a) sin(5w)/sinw b) sin(4w)/sinw c) sin(2.5w)/sin(0.5w) d) none of the above
3) If x(n)=h(n)=u(n), then h(n) is equal to
a) (n+1)u(n) b) r(n) c) r(n-1) d) none
4) if x ~ (n) = { 1,0,1,1} and h ~(n) = { 1, 2, 3,1} find y ~(n)
a) {6 , 6, 5, 4} b) {1, 2, 4, 4} c) {5, 4, 1, 0} d) None
5) x(n) = {4, 1, 3} h(n) = {2, 5, 0, 4} what is the output of the system.
a) {8, 22, 11, 31, 4, 12} b) {8, 22, 11, 31, 4, 12} c) {8, 22, 11, 31, 4, 12} d)
none
6) y(n) = x(n) * h(n) then y1 (n) = {0, 0, x(n), 0 } * { 0, h(n), 0 } is equal to
a) {0, 0, y(n), 0} b) {0, 0, 0, y(n), 0, 0} c) [0, 0, y(n), 0 } d) {0, y(n), 0, 0}
7) If x(n) and h(n) are having N values each, to obtain linear convolution using circular
convolution, the number of zeros to be appended to each sequence is
a) N – 1 b) 2N – 1 c) N d) N + 1
8)W49 = ?
a) – j b) + j c) + 1 d) -1
9) DFT [ x* (-n) ] = ?
a) X * (K) b) X * (-K) c) X * (N-K) d) none
10) If x(n)X(K), then IDFT [ X (K), X(K) ] = ?
a) x (n / 2) b) 2x (n/2) c) ½ x (2n) d) none.
11) Both discrete and periodic in one domain are also periodic and discrete in other
domain (T / F)
12) If h(n)= -h(-n) then H(K) is purely real (T / F)
143
13) Reversing the N point sequence in time is equivalent to reversing the DFT values (T /
F)
14) FT of non periodic discrete time sequence is non periodic (T / F)
Match the following: For a real valued sequence, the DTFT follow the properties as
15) Re [H (jw) ] a) Real valued function of w
16) Im[ H(jw) ] b) even function of w
17) F.T [even symmetric sequence] c) Imaginary valued function of w
18) F.T [odd symmetric sequence] d) odd function of w
n=N-1
19) Write DFF & IDFT formulas. X(k)=∑x(n)Wnnk
n=0 N-1
x(n)=(1/N)∑X(k)Wnnk
K=0
20) Total no of real multiplications in DFT is:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
A C A A C B A A A A T F T F B D A C 4n2
13.OBJECTIVE PAPER-5
Choose the Correct Answers
1. The Fourier transform of a finite energy discrete time signal, x(n) is defined as [ ]
a) X( )= x(n) e jn
b) X( )= x(n) en
n=- n=-
c) X( )= x(n) e -jn
d) X() = x(n) e-jn
n=- n=0
2. Inverse DFT (IDFT) of X(K) is x(n), where k=0,1,-----n-1. It is given as [ ]
N 1 j 2kn N 1 j 2kn
a) x(n) =
1
N
n 0
X(k) e N b) a) x(n) =
1
N
n 0
X(k) e N
j 2kn N j 2kn
c) x(n) =
1
N
n 0
X(k) e k d) a) x(n) =
1
N
n 0
X(k) e N
3. A N – periodic sequence x(n) and its DFT x(k) are known. Then the DFT of x(n) =
(n) will be
a) e-j2nk b) 1 c) e-j2nok/N d) e-j2nk /N [ ]
4. If the length of sequence x(n) is L and h(n) is M then the length of o/p sequence of the
circular convolution is [ ]
a) L+M b) L+M-1 c) L if L>M d) 2L if L=M
144
STATE TRUE OR FALSE
5. The DFT of a sequence is a continuous function of [ ]
6. The DFT of even sequence is purely imaginary and DFT of odd
sequence is purely real [ ]
7. The circular shift of an N point sequence is equivalent to linear shift of its periodic
extension [ ]
8. The multiplication of DFT of two sequences is equal to DFT of the linear convolution
of two sequences [ ]
Fill in the blanks
9. The 4-point DFT of a sequence x(n) is ________
10. DFT of a sequence x(n) = (n-n0) is __________
11. An N point sequence is called ________________ if it is antisymmetric about point
zero on the circle
12. The two methods of sectioned convolution are ________________ &
_______________
13. DFT of multiplication of two sequences DFT {x1 (n) x2(n) } =
_____________________
14. DFT of even sequence is X(k)= ________________________& DFT of odd sequence
is X(k) = _______________________
15. To get the result of linear convolution with circular convolution of sequence x(n) &
h(n), the sequences should extended to the length of __________________
16. Match the following
1 DFT [ x1(n) x2(n) ] a) X (N-K)
1
2. DFT [ x*(n) ] b) [ X1(k) X2(k)]
N
3. DFT [ x((-n))N ] c) X*(N-K)
4. X1(k) X2*(k) d) x1(n) x2(n)
e) x1(n) x2*(-n)
145
1= 2= 3= 4=
17. Show that the given sequence x(n) = { 1,-2,3,2,1,0} for the following conditions
using concentric circles.
a) x(-n) b) x(2-n) (2M)
18. Compute 4-point DFT of a sequence x(n) = {1,2,0,2} (2M)
14.OBJECTIVE PAPER-6
MULTIPLE CHOICES
1. In Impulse invariant transformation, the mapping of analog frequency to the digital
frequency is
a) one to one b) many to one c) one to many d) none
2. The digital frequency in bilinear transformation is
a) w = 2 tan-1(Ts/2) b) w = tan-1(Ts/2)
c) w = 2 tan-1(Ts) d) w = 2 tan-1(/2)
3. Which technique is useful for designing analog LPF
a) Butter worth filter b) Chebyshev filter
c) Both a and b d) none
4. Which filter is more stable?
a) Butter worth b) Chebyshev c) none
5. As increases , the magnitude response of LPF approaches with
a) –20Ndb/oct b) –6Ndb/oct c) –10Ndb/dec d) none
6. Using Impulse invariant technique the pole at S= SP is mapped to Z-plane as
a) Z=e-SPTs b) Z=e (SPTs) c) Z=eSP (Ts) d) None
TRUE or FALSE
7. The disadvantage of Chebyeshev filter is less transition region
8. The advantage of Butter worth filter is flat magnitude response.
9. for the given same specifications order of the Chebyshev filter is more than
Butterworth filter
10. Poles of Butterworth filter lies on circle.
1 2 3 4 5 6 7 8 9 10
B A C A B B F T F T
FILL IN THE BLANKS
11. The Butterworth LPF of order N is defined as: 1/(1+(s/jΩc)2N)
146
12. For N=3 what are the stable Butter worth angles :1200,1800,2400
13. –0.5db convert in to gain equivalent =0.994
14. Let S1,2 = 2076e±j144° Ha(S)= k/(s2-10552.7s+(2076π)2 (2M)
15. Given s = 2000; Ts = 10-4 ; *s = 2006
16. Using Bi-linear transformation, the pole at S = Sp is mapped into Z-plane using
(2M)
Z=1-(2+SpTS)/(2-SpTs)
17. Given allowable ripples in Pass band is –3 dB, the value of is 0.997 (2M)
15.OBJECTIVE PAPER-7
Choose the correct Answer
1. In impulse invariant transformation the mapping of analog frequency to digital
frequency is [ ]
a) one to one b) many to one c) one to many none
2. The digital frequency in Bi –linear transformation is [ ]
a) = 2 tan ( T /2)
-1
b) = tan ( T /2)
-1
c) = 2 tan ( T )
-1
d) = 2 tan-1( /2)
3. Using bilenear transformation for T = 1sec the pole p k is in S- Plane is mapped to Z –
plane using [ ]
1 z 1 1 z 1 1 z 1 Z 1
a) S = 2 b) S = c) S = 2 d) S=
1 z 1 1 z 1 1 z 1 Z 1
4. The normalized magnitude response of chebyshev type – I filter has a value of
____________ at cut off frequency are [ ]
1 1 1
a) b) c) d) 1 2
1 2
1 1 2
5. For high pass analog filter the transformation used is [ ]
a) SS/ b) S /S c) SS/c d) S c /S
6. The magnitude response of Type I – chebyshev LPF is given by [ ]
2 1 2 1
a) H a () = b) H a () =
1 C N ( / c )
2
1 C ( / c )
2 2
N
2 1 1
c) H a () = d) H a () =
1 C N ( / c )
2 2
1 C N ( / c )
2
147
7. The width of main lobe in rectangular window spectrum is [ ]
a) 2/N b) 4/N c) 8/N d) 16/N
8. The width of main lobe in Hamming window is [ ]
a) 4/N b) 2/N c) 8/N d) 16/N
9. The frequency response of rectangular window W R(w ) is [ ]
Sinwn / 2 Sinwn / 2 Sinwn / 2 Sinwn / 2
a) b) c) d)
Sinw / 2 Sinw Sinwn Sinwn / 2
10. In …………………. Window spectrum the width of main lobe is double that of
rectangular window for same value of N [ ]
a) Hamming window b) Kaiser window c) Blackman window d) none
State TRUE or FALSE
11. The disadvantage of chebyshev filter is less transition region [ ]
12. For chebyshev Type 2 filter ripples are present in pass band
and stop band [ ]
13. The advantage of Butter worth filter is flat magnitude response. [ ]
14. for cheby shev Type 1 filter equi–ripples are present only [ ]
in pass band.
15. For same specifications, the order N of chebyshev filter is less compared to Butter
worth filter. [ ]
16. FIR filter have non-linear phase characteristics. [ ]
17. FIR filters are non – recursive and stable filters. [ ]
18. The design of Digital transformation H (z) of IIR filter is direct and FIR is indirect
[ ]
19. Poles of chebyshev filter lies on circle [ ]
20. In FIR filter with constant phase delay the impulse response is symmetric
[ ]
148
16.OBJECTIVE PAPER-8
CHOOSE THE CORRECT ANSWER
1. The DTFT of a sequence x(n) is [ ]
a) x(n)e
n
jwn
b) x(n)e
n
jwn
c) x(n)e
jwn
dw d) x(n)e
jwn
dw
2. DTFT of ejwon x(n) is [ ]
j ( wwo )
a) x[ e j ( ww ) ] b) x[ e j ( w w ) ]
o o
c) x[ e ] d) x[ e j ( w wo )
]
3. DTFT of x1[n] * x2[n] is [ ]
1 1
a) X1[w] X2[w] b) X1[w] X2[w] c) X1[w] * X2[w] d) X1[w] * X2[w]
N N
4. The smallest value of N for which x(n +N) = x(n) holds is called [ ]
a) Fundamental period b) Fundamental frequency c) fundamental signal d) None
5. DFS of real part of periodic signal is [ ]
a) Xe(K) b) Xo (K) c) XR(K) d) XIm(K)
6. Expression for DFT is [ ]
N 1 N 1 N 1 N 1
a) x(n)W
n 0
N
Kn
b) x(n)W
n 0
Kn
N C) x(n)W
K 0
N
Kn
d) x(n)W
n 0
Kn
N
7. DFT of x1[n] x2[n] is [ ]
1 1
a) X1[K] * X2[K] b) X1[K] + X2[K] c) X1[K] * X2[K] d) X1[K] + X2[K]
N N
8. If M & N are the lengths of x(n) & h(n) then length of x(n) * h(n) is [ ]
a) M+ N –1 b) M + N +1 c) max (M,N) d) min (M,N)
9. Zero padding means [ ]
a) increasing length by adding zeros at the end of sequence
b) Decreasing length by removing zeros at the end
c) Inserting zeros in between the samples d) None of the above
II STATE TRUE OR FALSE
10. The F.T of discrete signal is a discrete function of [ ]
11. In a discrete signal x(n), if x(n) =x(-n) then it is called symmetric signal [ ]
12. The F.T of the product of two time domain sequence is equivalent to product
of their F.T [ ]
13. The DFT of a signal can be obtained by sampling one period of FT of the signal
[ ]
14. DFS is same as DTFS [ ]
149
17.OBJECTIVE PAPER-9
CHOOSE THE CORRECT ANSWER
1. Power signal is
a) Periodic b) aperiodic c) Continuous d) none [ ]
2. WN nK is
j 2 K j 2 Kn 2 Kn
j 2 nK
a) e N
b) e c) e N
d) e N
[ ]
3. When the sequence is circularly shifted in time domain by ‘m’ samples i.e. x((n-m))N
then on applying DFT, it is equivalent multiply sequence in frequency domain by
j 2 Km j 2 Km 2 Km
j 2 Km
a) e N
b) e N
c) e d) e N
[ ]
4. Multiplication of sequence in time domain, on apply DFT, it corresponds to circular
convolution in frequency domain and is given as
a) x1(n) x2(n) X1(K)
DFT
X2(K)
b) x1(n) x2(n) X1(K)X2(K)
DFT
c) x1(n) * x2(n) X1(K)
DFT
X2(K)
N 1
d) x1(n) x2(n) X1(K)X2(K)
DFT
K 0
5. Linear convolution of two sequences N1 and N2 produces an output sequence of length
a) N1 – N2 +1 b) N1 + N2 –1 c) N1 + N2 +1 d) 2N1 – N2 +1[ ]
FILL IN THE BLANKS
6. The basic signal flow graph for butterfly computation of DIT-FFT is
7. The Fourier transform of discrete time signal is called ………………………
8. FFT’s are based on the ………………………….. of an N-point DFT into
successively smaller DFT’s.
9. The Fourier transform of x(n)*h(n) is equal to …………………………..
10. Appending zeros to a sequence in order to increase the size or length of the sequence
is called ……………………..
11. In N-point DFT using radix 2 FFT, the decimation is performed …………… times.
12. In 8-point DFT by radix 2 FFT, there are …………… stages of computations with
…………………….. butterflies per stage.
13. If DFT of x(n) is X(K), then DFT of WN ln x(n) is …………………….
150
ANSWER THE FOLLOWING
14. What are the differences between linear and circular convolution?
15. How many multiplications and additions are required to compute N-point DFT using
radix 2 FFT
16. How many multiplications and additions are required to compute N-point DFT
17. What is the expression for N-point DFT of a sequence x(n) ?
18. What is the expression for N-point IDFT of a sequence X(K) ?
19. Define Aliasing error.
20. What is meant by Inplace computation.
151
18.OBJECTIVE PAPER-10
1. How we can calculate IDFT using FFT algorithm. (2M)
2. Draw the basic butterfly diagram for DIF algorithm.
3. Z[x(n)] = X(Z) then Z{x(n-m)} = …………………………………..
4. Define convolution property in Z-Transform.
5. Find the Z-Transform and ROC for the signal x(n) = an u(n).
6. Find the Z-Transform and ROC for the signal x(n) = - an u(-n-1).
7. Write the initial value theorem expression.
8. Z{(n)} = ……………………..
Z
9. Find inverse Z-Transform for X(z) = when ROC is Z<1
Z 1
152
10. What are the differences and similarities between DIT and DIF algorithms. (2M)
11. Give the Direct form II realization for second order system.
12. Give the Direct for I realization for second order system.
13. What is the relationship between Z-Transform and Fourier transform.
STATE TRUE OR FALSE:
14. ROC of a causal signal is the exterior of a circle of some radius r. [ ]
15. ROC of a anti causal signal is the exterior of a circle of some radius r. [ ]
16. ROC of a two sided finite duration frequency is entire Z-plane. [ ]
17. Direct form I required less no.of memory elements as compared to Canonic form.[ ]
18. A linear time invariant system with a system function H(Z) is BIBO stable if and only
if the ROC for H(Z) contains unit circle. [ ]
153
19.OBJECTIVE PAPER-11
ANSWER THE FOLLOWING
1. What are the advantages of digital filter over analog filter.
2. What is the relation between analog and digital radiant frequency in Impulse
Invariance design..
3. What is the relation between analog and digital radiant frequency in Bilinear
transformation design.
4. What are the drawbacks with Impulse Invariance method?
5. What is the disadvantage with Bilinear transformation technique.
6. What is the relation between S & Z in Bilinear transformation?
7. Mention any two techniques to design IIR Filter from analog filter.
8. What are the differences between Chebyshev type I and type II.
154
9. What are the differences between Butterworth & Chebyshev filter.
10. What is the expression for magnitude squared frequency response of Butterworth
analog filter?
11. What is the expression for magnitude squared frequency response of Chebyshev
analog filter?
TRUE OR FALSE
12. Poles of Butterworth filter lies on circle. [ ]
13. Poles of Chebyshev filter lies on circle. [ ]
14. Transition bandwidth for Chebyshev is more as compared to Butterworth filter.[ ]
15. Butterworth filters are all pole filters. [ ]
16. Chebyshev, type-II are all pole filters. [ ]
17. Chebyshev, type II filter exhibit equiripple behavior in the pass band and monotonic
characteristic in the stopband. [ ]
18. Chebyshev, type I filter exhibit equiripple behavior in the pass band and monotonic
characteristic in the stopband. [ ]
19. Butterworth filter exhibit monotonic behavior both in passband and stopband.[ ]
20. For the given specifications order of the Chebyshev filter is more as compared to
Butterworth filter. [ ]
155
20. OBJECTIVE PAPER-12
1. Define the following
a. Time variant system with an example(Equation)
b. Power signal with an example
c. Dynamic system
d. Recursive System
e. Non Recursive system
2. Give the example for FIR and IIR systems.
3. Give an example of Causal system
4. Write the condition to test the Linearity of the system
5. Plot y(n) = x(n-2) Give x(n) ={1,2,3,5,6}
6. Resolve the signal into impulse x(n)={4,5,4,4} ------ 2 Marks
156
7. Give the expression for Convolution sum y(n)=
8. Find the Convolution Sum Graphically with all the steps-------3 Marks
x(n)= 2 1 h(n)= 1 1
-1 0 0 1
9. Write the properties of Convolution Sum --------2 Marks
10. Write the expression for X(n) in terms of impulses
11. Write the necessary condition for the stability of the system
12. Write the general form of Difference equation
157
21.OBJECTIVE PAPER-13
State TRUE or FALSE
1. In direct –form II realization the number of memory locations required is more than
that of direct form –I realization [ ]
2. An LTI system having system function H(z) is stable if and only if all poles of H(z)
are out side the unit circle. [ ]
3. The inverse Z – transform of z/z-a is an u(n) [ ]
4. Digital filters are not realizable for ideal case. [ ]
5. As the order of Butter worth filter increases than the response is closer to ideal filter
response. [ ]
Answer the following
6. Find the transfer function H(z) of the given difference equation
Y(n) = 0.7 y(n-1) – 0.12y(n-2) + x(n-1) + x(n-2)
7. Indicate the poles and zeros of the given system and also check the stability of the
system
z ( z 1)
H(z) = (2M)
( z 0.2)( z 0.4)( z 0.5)
8. Realize the given system function H(z) using direct form –II
3 3.6 z 1 0.6 z 2
H(z) = (2M)
1 0.1z 1 0.2 z 2
158
9. Realize the given system function H(z) using cascade form (2M)
1
H(z) = 1
(1 0.5 z )(1 0.5 z 1 )
z
10. Find the inverse z-transform of x(z) = using partial fraction method.
( z 2)( z 3)
(2M)
11. Using cauchy residue method find the inverse z- transform of
z
x(z) = for ROC :z >2 (2M)
( z 1)( z 2)
12. Mention the two conditions to realize any digital filter
13. Draw the Magnitude response of Low Pass Butter Worth filter.
14. The order of the Butter Worth filter is obtained by using the formula N
______________________
15. The cut- off frequency c is obtained by using the formula
___________________________
159
22.OBJECTIVE PAPER-14
Fill in the Blanks
1. The expansion of FFT is _______________
2. The main advantage of FFT is _____________
3. The number of multiplications needed in the calculation of DFT using FFT with 32-
point sequence = ________________
4. __________________ number of additions are required to compute N – pt DFT using
radix –2 FFT.
5. What is decimation in time algorithm.
State TRUE or FALSE
6. For DIT –FFT algorithm the input is bit reversed and the output is in natural order
[ ]
7. By using radix –2 DIT –FFT algorithm it is possible to calculate 6-point DFT.[ ]
8. W NNK 1 [ ]
9. W NNK/ 2 1 [ ]
10. In DIT –FFT, the input sequence is divided into smaller subsequences [ ]
Answer the following
11. Calculate the DFT of the sequence x(n)={1,0,0,1} using DIT –FFT (2M)
12. Draw the Butterfly diagram for 8-point DFT using DIT –FFT algorithm (2M)
13. Find IDFT of the sequence X(k) = { 10, 0, 10, 0} (2M)
160
14. Write the steps for the calculation of IDFT using DIT –FFT (2M)
15.Write the values of the following
a) W80 b) W82 c) W83 d) W85
23.OBJECTIVE PAPER-15
CHOOSE THE CURRECT ANSWER
1. y(n)=x(2n) is a ____________ system [ ]
a) time invariant b) causal c) non causal d) none
2. y(n) = nx2(n) is a ____________system [ ]
a) Linear b) Non-linear c) time-invariant d) none
3. y(n)= x(n) +x(n-1) is a ____________ system [ ]
a) Dynamic b) Static c) time variant d) None
4. x(-n+2) is obtained by which of the following operations [ ]
a) x(-n) is shifted left by 2 samples b) x(-n) is shifted right by 2 samples
c) x(n) is shifted left by 2 samples d0 none
5. The necessary and sufficient condition for causality of an LTI system is [ ]
a) h(n) =0 for n=0 b) h(n) =0 for n>0 c) h(n) =0 for n<0
d) none
6. The convolution of two sequences x(n) =h(n) = {1, 2, -1} [ ]
a) { 1,4,2,-4,1} b) {1,-4,1,2,4} c) { 1,1,2,-4,4}
d) { 4,-4,2,1,1}
II STATE TRUE OR FALSE
7. An IIR system exhibits an impulse response for finite interval [ T/F ]
8. If the energy of a signal is infinite then it is called energy signal [ T/F ]
9. Static systems does not require memory [ T/F ]
10. A linear system is stable if its impulse response is absolutely summable[T/F ]
161
III Answer the following:
11. The average power of a discrete time signal with period N is given by ___________
12. The convolution sum of causal system with causal sequence is ____________
13. Give the graphical representation of the following discrete signals.
i) x(n) = (5-x) [ 4(x) – 4(x-3)}
ii) x(n) = -0.5(n+1) + 0.5(n) – 0.75 (n-2)
14. x(n) = {3, -2, 1,0,-1} show for x(-n) (1M)
15. If x(n) = {1,2,-2,-1} show for x(n-2) & X( -n+2) (2M)
16. Find the convolution of u(n) * u(n-2) (1M)
162
17. If the impulse response h(n) = 2n u( -n) then determine the corresponding system is
causal or stable. (1M)
18. Test the given discrete system for linearity , causality and time invariance
h(n) = n ex(n) ( 2M)
ASSIGNMENT
1 (a) Draw the frequency response of N-point rectangular window.
(b) Design a fifth order band pass linear phase filter for the following specifications.
i. Lower cut-off frequency = 0.4 πrad/sec
ii. Upper cut-off frequency = 0.6 πrad/sec
iii. Window type = Hamming
Draw the filter structure. [4+12]
2) Design a band pass filter to pass frequencies in the range 1-2 radians/second using
Hanning window N=5. Draw the filter structure and plot its spectrum. [16]
3) (a) Compare the performances of rectangular window, hamming window and Keiser
window
(b) The desired response of a low pass filter is
Hd(ej!) = _ e−j3!, −3π _ ω _ 3π/4
0 , 3π/4 _ |ω| _ π
Determine H(ej!) for M=7 using a Hamming window. [6+10]
4) (a) Design a linear phase low pass filter with a cut-off frequency of π/2
radians/seconds. Take N=7
(b) Derive the magnitude and phase functions of Finite Impulse Response filter when
i. impulse response is symmetric & N is odd
ii. impulse response is symmetric & N is even. [8+8]
5) (a) Design a low pass filter by the Fourier series method for a seven stage with cut-off
frequency at 300 Hz if ts = 1msec. Use hanning window.
(b) Explain in detail, the linear phase response and frequency response properties of
Finite Impulse Response filters. [8+8]
6) (a) Outline the steps involved in the design of FIR filter using windows.
(b) Determine the frequency response of FIR filter defined by y(n) = 0.25x(n)+ x(n-1)+
0.25x(n-2). Calculate the phase delay and group delay. [8+8]
7) (a) Define Infinite Impulse Response & Finite Impulse Response filters and com-pare.
163
(b) Design a low pass Finite Impulse Response filter with a rectangular window for a
five stage filter given: Sampling time 1 msec; fc = 200Hz.Draw the filter structure with
minimum number of multipliers. [6+10]
ASSIGNMENT
1) a) What are the advantages of Multirate signal processing?
b) Differentiate between Decimator and Interpolator?
2) Prove that spectrum of down sampler is sum of M uniformly shifted and stretched
version of X(ejw) scaled by a factor 1/M and also discuss the aliasing effect?
3) State and prove any one identity property in down sampler and any one identity
property in up sampler?
4) Let x(n)={1,3,2,5,-1,-2,2,3,2,1},find
a) Up sample by 2 times and down sample by 4 times
b) Down sample by 4 times and up sample by 2 times c) Justify why these outputs are
not equal.
164
24.Reference
1 Digital Signal Processing by John G Proakis
2 Discrete time signal processing by Alan V Oppenheim Ronald W schafer
3 Digital signal processing by MITHRA
4 Digital signal processing by Tharun kumar rawat
5 Analog and Digital Signal Processing by Ashok Ambardar
165
You might also like
- Wiley - Interscience.introduction - To.digital - Signal.processing - And.filter - Design.oct.2005.ebook LinG88% (8)Wiley - Interscience.introduction - To.digital - Signal.processing - And.filter - Design.oct.2005.ebook LinG440 pages
- Multidimensional Signal, Image, and Video Processing and CodingNo ratings yetMultidimensional Signal, Image, and Video Processing and Coding8 pages
- Ece-V-Digital Signal Processing (10ec52) - Notes PDFNo ratings yetEce-V-Digital Signal Processing (10ec52) - Notes PDF161 pages
- Signal Processing: Understanding DigitalNo ratings yetSignal Processing: Understanding Digital7 pages
- Digital Signal Processing 10EC52: - A Unit - 1100% (1)Digital Signal Processing 10EC52: - A Unit - 1160 pages
- EC6502 Principles of Digital Signal Processing 1- By EasyEngineering.netNo ratings yetEC6502 Principles of Digital Signal Processing 1- By EasyEngineering.net116 pages
- Frequency Synthesizers From Concept to Product Artech House Microwave Library 1st Edition Alexander Chenakin instant download100% (1)Frequency Synthesizers From Concept to Product Artech House Microwave Library 1st Edition Alexander Chenakin instant download50 pages
- Digital Signal Processing With SelectedNo ratings yetDigital Signal Processing With Selected591 pages
- (Domine M.W. Leenaerts) Circuit Design For RF Tran (BookFi)No ratings yet(Domine M.W. Leenaerts) Circuit Design For RF Tran (BookFi)350 pages
- (Signal Processing and Data Analysis) ContentsNo ratings yet(Signal Processing and Data Analysis) Contents10 pages
- Handbook of Real Time Fast Fourier Transform100% (4)Handbook of Real Time Fast Fourier Transform484 pages
- Fourier Optics in Image Processing - 24!09!28!16!48 - 07No ratings yetFourier Optics in Image Processing - 24!09!28!16!48 - 07195 pages
- 2004 - Static and Dynamic Performance LimitationsNo ratings yet2004 - Static and Dynamic Performance Limitations229 pages
- Contents 50 Pic Microcontroller ProjectsNo ratings yetContents 50 Pic Microcontroller Projects3 pages
- Digital Signal Processing System Analysis and Design 2nd Edition Paulo S. R. Diniz instant downloadNo ratings yetDigital Signal Processing System Analysis and Design 2nd Edition Paulo S. R. Diniz instant download62 pages
- Feed Forward Amplifiers For Wideband Communication SystemsNo ratings yetFeed Forward Amplifiers For Wideband Communication Systems187 pages
- Download ebooks file Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu all chapters100% (3)Download ebooks file Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu all chapters61 pages
- B.-P.-Lathi-Zhi-Ding-Modern-Digital-and-Analog-Communication-Systems-Oxford-University-Press-2009 [7-12]No ratings yetB.-P.-Lathi-Zhi-Ding-Modern-Digital-and-Analog-Communication-Systems-Oxford-University-Press-2009 [7-12]6 pages
- Instant ebooks textbook Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu download all chapters100% (6)Instant ebooks textbook Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu download all chapters40 pages
- Instant download Digital Signal Processing and Applications Second Edition Dag Stranneby pdf all chapterNo ratings yetInstant download Digital Signal Processing and Applications Second Edition Dag Stranneby pdf all chapter67 pages
- Digital Signal Processing 2nd Edition PauloS.R.Diniz EduardoA.B.da Silva SergioL.Netto ISBN 0521887755 978-0521887755 download100% (3)Digital Signal Processing 2nd Edition PauloS.R.Diniz EduardoA.B.da Silva SergioL.Netto ISBN 0521887755 978-0521887755 download55 pages
- Frequency Agile Antennas for Wireless Communications 1st Edition Aldo Petosa instant download100% (1)Frequency Agile Antennas for Wireless Communications 1st Edition Aldo Petosa instant download77 pages
- Multimedia Signal Processing: Theory and Applications in Speech, Music and CommunicationsFrom EverandMultimedia Signal Processing: Theory and Applications in Speech, Music and CommunicationsNo ratings yet
- Multiresolution signal decomposition transforms subbands and wavelets 2nd ed Edition Ali N Akansu - The ebook is now available, just one click to start reading100% (1)Multiresolution signal decomposition transforms subbands and wavelets 2nd ed Edition Ali N Akansu - The ebook is now available, just one click to start reading47 pages
- Multi-GHz Frequency Synthesis & Division Frequency Synthesizer Design For 5 GHZ Wireless LAN SystemsNo ratings yetMulti-GHz Frequency Synthesis & Division Frequency Synthesizer Design For 5 GHZ Wireless LAN Systems157 pages
- Digital Signal Processing System Analysis and Design 1st Edition Paulo S. R. Diniz - The ebook in PDF and DOCX formats is ready for download now100% (1)Digital Signal Processing System Analysis and Design 1st Edition Paulo S. R. Diniz - The ebook in PDF and DOCX formats is ready for download now86 pages
- Chassis: Cp-830F: Model: Dth-29U7 DTH-29U8 DTH-2881 DTH-28W8 DTH-28W9No ratings yetChassis: Cp-830F: Model: Dth-29U7 DTH-29U8 DTH-2881 DTH-28W8 DTH-28W998 pages
- Digital Signal Processing Techniques and Applications in Radar Image ProcessingFrom EverandDigital Signal Processing Techniques and Applications in Radar Image ProcessingNo ratings yet
- Pcn Time of Flight Diffraction (Tofd) MaterialNo ratings yetPcn Time of Flight Diffraction (Tofd) Material151 pages
- Get Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu PDF ebook with Full Chapters NowNo ratings yetGet Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu PDF ebook with Full Chapters Now40 pages
- Comparison of Different Filtering/Smoothing Filters in Digital Image ProcessingNo ratings yetComparison of Different Filtering/Smoothing Filters in Digital Image Processing5 pages
- Brinkmann: Manual For Integrated AmplifierNo ratings yetBrinkmann: Manual For Integrated Amplifier4 pages
- Robot Builder's Cookbook Build and Design Your Own Robots 1st edition by Owen Bishop ISBN 0750665564 9780750665568 - The ebook in PDF format is ready for immediate accessNo ratings yetRobot Builder's Cookbook Build and Design Your Own Robots 1st edition by Owen Bishop ISBN 0750665564 9780750665568 - The ebook in PDF format is ready for immediate access42 pages
- 20 XR-1 X-Ray Apparatus Stationary Type Not Less Than 500ma (Analog) With CRNo ratings yet20 XR-1 X-Ray Apparatus Stationary Type Not Less Than 500ma (Analog) With CR6 pages
- UPF Interactions: 5G Standalone Access Registration: 1:PFCP Session Modification RequestNo ratings yetUPF Interactions: 5G Standalone Access Registration: 1:PFCP Session Modification Request2 pages
- Wiley - Interscience.introduction - To.digital - Signal.processing - And.filter - Design.oct.2005.ebook LinGWiley - Interscience.introduction - To.digital - Signal.processing - And.filter - Design.oct.2005.ebook LinG
- Introduction to Digital Signal Processing and Filter DesignFrom EverandIntroduction to Digital Signal Processing and Filter Design
- Multidimensional Signal, Image, and Video Processing and CodingMultidimensional Signal, Image, and Video Processing and Coding
- Ece-V-Digital Signal Processing (10ec52) - Notes PDFEce-V-Digital Signal Processing (10ec52) - Notes PDF
- EC6502 Principles of Digital Signal Processing 1- By EasyEngineering.netEC6502 Principles of Digital Signal Processing 1- By EasyEngineering.net
- Frequency Synthesizers From Concept to Product Artech House Microwave Library 1st Edition Alexander Chenakin instant downloadFrequency Synthesizers From Concept to Product Artech House Microwave Library 1st Edition Alexander Chenakin instant download
- Design of Integrated Circuits for Optical CommunicationsFrom EverandDesign of Integrated Circuits for Optical Communications
- (Domine M.W. Leenaerts) Circuit Design For RF Tran (BookFi)(Domine M.W. Leenaerts) Circuit Design For RF Tran (BookFi)
- Fourier Optics in Image Processing - 24!09!28!16!48 - 07Fourier Optics in Image Processing - 24!09!28!16!48 - 07
- Digital Signal Processing System Analysis and Design 2nd Edition Paulo S. R. Diniz instant downloadDigital Signal Processing System Analysis and Design 2nd Edition Paulo S. R. Diniz instant download
- Feed Forward Amplifiers For Wideband Communication SystemsFeed Forward Amplifiers For Wideband Communication Systems
- Download ebooks file Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu all chaptersDownload ebooks file Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu all chapters
- B.-P.-Lathi-Zhi-Ding-Modern-Digital-and-Analog-Communication-Systems-Oxford-University-Press-2009 [7-12]B.-P.-Lathi-Zhi-Ding-Modern-Digital-and-Analog-Communication-Systems-Oxford-University-Press-2009 [7-12]
- Instant ebooks textbook Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu download all chaptersInstant ebooks textbook Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu download all chapters
- Instant download Digital Signal Processing and Applications Second Edition Dag Stranneby pdf all chapterInstant download Digital Signal Processing and Applications Second Edition Dag Stranneby pdf all chapter
- Digital Signal Processing 2nd Edition PauloS.R.Diniz EduardoA.B.da Silva SergioL.Netto ISBN 0521887755 978-0521887755 downloadDigital Signal Processing 2nd Edition PauloS.R.Diniz EduardoA.B.da Silva SergioL.Netto ISBN 0521887755 978-0521887755 download
- Frequency Agile Antennas for Wireless Communications 1st Edition Aldo Petosa instant downloadFrequency Agile Antennas for Wireless Communications 1st Edition Aldo Petosa instant download
- Multimedia Signal Processing: Theory and Applications in Speech, Music and CommunicationsFrom EverandMultimedia Signal Processing: Theory and Applications in Speech, Music and Communications
- Multiresolution signal decomposition transforms subbands and wavelets 2nd ed Edition Ali N Akansu - The ebook is now available, just one click to start readingMultiresolution signal decomposition transforms subbands and wavelets 2nd ed Edition Ali N Akansu - The ebook is now available, just one click to start reading
- Multi-GHz Frequency Synthesis & Division Frequency Synthesizer Design For 5 GHZ Wireless LAN SystemsMulti-GHz Frequency Synthesis & Division Frequency Synthesizer Design For 5 GHZ Wireless LAN Systems
- Digital Signal Processing System Analysis and Design 1st Edition Paulo S. R. Diniz - The ebook in PDF and DOCX formats is ready for download nowDigital Signal Processing System Analysis and Design 1st Edition Paulo S. R. Diniz - The ebook in PDF and DOCX formats is ready for download now
- Chassis: Cp-830F: Model: Dth-29U7 DTH-29U8 DTH-2881 DTH-28W8 DTH-28W9Chassis: Cp-830F: Model: Dth-29U7 DTH-29U8 DTH-2881 DTH-28W8 DTH-28W9
- Digital Signal Processing Techniques and Applications in Radar Image ProcessingFrom EverandDigital Signal Processing Techniques and Applications in Radar Image Processing
- Ultra Wideband Signals and Systems in Communication EngineeringFrom EverandUltra Wideband Signals and Systems in Communication Engineering
- Get Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu PDF ebook with Full Chapters NowGet Multiresolution Signal Decomposition Transforms Subbands and Wavelets 2nd Edition Akansu PDF ebook with Full Chapters Now
- Comparison of Different Filtering/Smoothing Filters in Digital Image ProcessingComparison of Different Filtering/Smoothing Filters in Digital Image Processing
- Robot Builder's Cookbook Build and Design Your Own Robots 1st edition by Owen Bishop ISBN 0750665564 9780750665568 - The ebook in PDF format is ready for immediate accessRobot Builder's Cookbook Build and Design Your Own Robots 1st edition by Owen Bishop ISBN 0750665564 9780750665568 - The ebook in PDF format is ready for immediate access
- 20 XR-1 X-Ray Apparatus Stationary Type Not Less Than 500ma (Analog) With CR20 XR-1 X-Ray Apparatus Stationary Type Not Less Than 500ma (Analog) With CR
- UPF Interactions: 5G Standalone Access Registration: 1:PFCP Session Modification RequestUPF Interactions: 5G Standalone Access Registration: 1:PFCP Session Modification Request