Ch-1 (1) Digital Filters.

pdf
Ch-1 (2) Impulse Invariant Method.pdf
Ch-1 (3) BLT Method.pdf
Ch-1 (4) Butterworth Filter Design.pdf
Ch-1 (5) Butterworth Filter Design by IIM.pdf
Ch-1 (6) Butterworth Filter Design by BLT.pdf
Ch-2 (1) FIR Filters-Concept of Linear Phase.pdf
Ch-2 (2) FIR Filters- POLEs ZEROs.pdf
Ch-2 (3) FIR Filter Design-Windowing Method.pdf
Ch-2 (4) FIR Filter Design-FSM 9.35.54 PM.pdf
Ch-3 (1) Multi Rate Signal Processing.pdf
Ch-4 (1) Adaptive-Filter.pdf
Introducton to Digital Filter
@
Sardar Patel Institute of
Technology, Andheri, Mumbai
Kiran Tulshiram Vasumati TALELE

Academic Year 2020-21

Bharatiya Vidya Bhavans’

Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015]
Digital Filter
H(Z) h(n) DE
RD PZ FR

* DT System *
x[n] y[n]
Ex. Digital Filter

Infinite Impulse Finite Impulse

Response(IIR) Filter Repsonse(FIR) Filter
Ex. h[n] = (0.5)n u[n] Ex. h[n] = { 1, 2, 3, 4 }
Infinite Length Finite Length
Causal Causal
STABLE STABLE
Kiran TALELE 9987030881 talelesir@gmail.com 33
 NOTE : To find Output of Digital Filter

(1) h[n]
x[n] (2) H[z] y[n]
(3) DE
(4) RD
(5) PZ
ZT IZT
(6) FR

X(z) Y(z)
H(z)

Kiran TALELE 9987030881 talelesir@gmail.com 4

4
NOTE : Relationship Diagram
1. Take ZT
2. Arrange in terms of
Y(z) & X(z) PZ
3. Write H(z) =Y(z) / X(z)
IZT

DE H(z)
ZT h[n]

Z=ejw
1. Write H(z) in –ve H(ejw) OR
Powers of Z RD H(w)
2. Let H(z) = Y(z)/X(z)
3. Cross multiply
4. Take Inverse ZT

5
 (*) What is Magnitude Response ?

Ans :

Magnitude = Magnitude of numerator

Response Magnitude of Deno min ator

Where,
Magnitude = (Re al ) + (Im aginary )
2 2

Kiran TALELE 9987030881 talelesir@gmail.com 6

 (*) What is Phase Response?
Ans :
Phase = Angle of – Angle of
Response Numerator Denominator

Where,  −1  Im aginary 
 tan  Re al  When Re al  0
  
Angle =
 −1  Im aginary 
180 + tan   When Re al  0
  Re al 

K.T.V. TALELE 9987030881 talelesir@gmail.com 7

7
 (*) What is stability condition for
causal and stable system?
Ans :
.
For causal and stable
system, all the POLES
must lie INSIDE the unit
circle

K.T.V. TALELE 9987030881 talelesir@gmail.com 8

8
(*) What is Minimum Phase System ?
For any system If  H() –  H(0) = 0
Then system is called a Minimum Phase System.
When All zeros are inside the unit circle, the
system is minimum phase.

(*) What is Maximum Phase System ?

For any system If  H() –  H(0) = M
Then system is called a Maximum Phase System.
When All zeros are outside the unit circle, the system is
Maximum phase.
NOTE :

If the system is Neither Minimum Phase

NOR Maximum Phase
Then System is Mixed Phase System.
 DSP System

x(t) x(n) y(n) Reconstruction y(t)

Sampling A D
Using S/H DDigital ProcessorA Using LPF
C C

K.T.V. TALELE 9987030881 talelesir@gmail.com 10

10
Digital filter is nothing but
a Mathematical Algorithm
implemented in
Hardware or Software.

Kiran TALELE  9987030881

 IMPORTANT

•An important step in the development of a

digital filter is the determination of a
realizable transfer function H(z)
approximating the given frequency
response specifications. The process of
deriving the transfer function H(z) is called
digital filter design.

Kiran TALELE 9987030881 talelesir@gmail.com

 • Position of the poles affects system stability
System Stability
• The position of zeroes does not
 System Stability. .. .

Kiran TALELE 9987030881 talelesir@gmail.com 14

 Ideal Filter Frequency Response

LPF HPF

PB SB SB PB

BPF BSF/BRF

SB PB SB PB SB PB

Kiran TALELE 9987030881 talelesir@gmail.com 15

 Practical Filter Frequency Response

Kiran TALELE 9987030881 talelesir@gmail.com 16

 Digital I I R FILTER DESIGN

Time Domain Transform Domain

Method Method
Impulse Invariant Method BLT Method

Kiran TALELE 99870 30 881 talelesir@gmail.com

 IIR Filters
Infinite impulse response (IIR) filter is described using :
(a) IIR filter difference equation

IIR filter output

depends also on
(b) IIR filter Transfer function the past outputs

Kiran TALELE 9987030881 talelesir@gmail.com 18

18
 Digital F I R FILTER DESIGN

Windowing Method Frequency Sampling Method

Kiran TALELE 99870 30 881 talelesir@gmail.com

Difference between between Analog Filter and Digital Filter

Parameter Analog filter Digital Filter

1 Digital(discrete time
Input/output signal Analog
sequences)
2 Lumped elements such as Software + digital
Composition
R.L. and C or analog /Cs hardware
3 In terms of system By difference
Filter representation
components equation
4 Flexibility Not flexible Highly flexible
5 Portability Not easily portable Portable
6 Design objective and Specifications to values of Specifications to
result R.L and C components difference equation
7 Environmental effects Negligible effect of Negligible effect of
environmental parameters environmental
parameters
8 Interference notes Minimum/negligible
Maximum effect
and other effects effect
9 Storage/maintenance Difficult storage and Easier storage and
failure maintenance and higher maintenance and
20
failure rate reduced failure rate
Difference between between IIR Filter and FIR Filter

FIR filter IIR filter

1 Length of h[n] is finite Length of h[n] is In-finite
2 Provides exact linear phase. Not linear phase.
3 Provides good stability. Stability is not guaranteed.
4 Order required is higher. Order required is lower.
5 Computationally not efficient. Computationally more efficient.
6 More memory required for the Less memory required fro storage of
storage of coefficients. coefficients.
7 Requires more processing time. Requires less processing time.
8 Requires M multiplications per Requires 2M + 1 multiplications per
output sample output sample.

Kiran TALELE 9987030881 talelesir@gmail.com 21

Q1 What are advantages of FIR Filters ?

1. They can easily be designed to be "linear

phase"
2. They are suited to multi-rate applications.
3. They have desirable numeric properties.
4. They can be implemented using fractional
arithmetic.
5. They are simple to implement.

Kiran TALELE 9987030881 talelesir@gmail.com 22

22
Q2 . What are the disadvantages of FIR Filters
(compared to IIR filters)?
Compared to IIR filters, FIR filters sometimes have the
disadvantage that they require more memory and/or
calculation to achieve a given filter response characteristic.

Q3. What are the advantages of IIR filters (compared

to FIR filters)?
IIR filters can achieve a given filtering characteristic using
less memory and calculations than a similar FIR filter

Kiran TALELE 9987030881 talelesir@gmail.com 23

23
Q4 . What are the disadvantages of IIR Filters
(compared to FIR filters)?

1) They are more susceptible to problems of finite-

length arithmetic, such as noise generated by
calculations, and limit cycles.

2) They are harder to implement using fixed-point

arithmetic.

3) They don't offer the computational advantages of

FIR filters for multirate (decimation and
interpolation) applications.

Kiran TALELE 9987030881 talelesir@gmail.com 24

24
Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 26

 Digital IIR Filter Design
using Impulse Invariant Method
@
Sardar Patel Institute of
Technology, Andheri, Mumbai

Kiran Tulshiram Vasumati TALELE

Academic Year 2020-21

Bharatiya Vidya Bhavans’

Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015]
 Digital I I R FILTER DESIGN

Time Domain Transform Domain

Method Method
Impulse Invariant Method BLT Method

Kiran TALELE 9987030881 talelesir@gmail.com 3

 An important step in the
development of a digital filter is the
determination of a realizable
Transfer Function H(z)
approximating the given frequency
response specifications.

Kiran TALELE 9987030881 talelesir@gmail.com 4

 Digital IIR filter Design
using Impulse Invariant
Method

Kiran Tulshiram Vasumati TALELE

 Concept of Digital IIR Filter Design

INPUT OUTPUT
Analog Filter
x(t) y(t)
Sampling Reconstruction

x[n] y[n]
Digital Filter

1. Programmable
2. Stable

Kiran TALELE 9987030881 talelesir@gmail.com 6

6
 Concept of Invariant Method

X(s) Analog Filter Y(s)

LT ILT
x(t) H(s) y(t)
Sampling Sampling

x[n] Digital Filter y[n]

ZT ZT
X(z) H(z) Y(z)

For Impulse Invariant Method, x(t) = (t)

For Step Invariant Method, x(t) = u(t)

Kiran TALELE 9987030881 talelesir@gmail.com 7

7
 Impulse Invariant Method

X(s)=1 Analog Filter Y(s)=H(s)

LT ILT
x(t)=(t) H(s) y(t)=h(t)
Sampling Sampling

x[n]= (n) Digital Filter y[n]=h[n]

ZT ZT
X(z)= 1 H(z) Y(z)=H(z)

Kiran TALELE 9987030881 talelesir@gmail.com 8

8
 IIR Filter Design by IIM Design Procedure
(I) Determine the Normalized Analog filter Transfer function H(s)
that satisfies the specification for the desired filter.

(II) Find X(s)

For Impulse Invariant Method ,
Take x(t) = (t)
Y(s)=H(s)
By LT, X(s) = 1 X(s)=1 Analog Filter
LT LT
H(s) y(t)=h(t)
x(t)=(t)

x[n]= (n) Digital Filter y[n]=h[n]

ZT ZT
(III) Find Y(s) H(z) Y(z)=H(z)
X(z)= 1
Now Y(s) = X(s) H(s)
Y(s) = H(s)

9
Kiran TALELE 9987030881 talelesir@gmail.com 9
 X(s)=1 Analog Filter Y(s)=H(s)
LT LT
H(s) y(t)=h(t)
x(t)=(t)

x[n]= (n) Digital Filter y[n]=h[n]

ZT ZT
H(z) Y(z)=H(z)
X(z)= 1

(IV) Find y(t) (VI) Find X(z) (VII) Find Y(z)

Now, Y(s) = H(s) Now x[n] = [n] Now y[n] = h[n]
By Inverse LT, By ZT, X(z) = 1 By ZT, Y(z) = H(z)
y(t) = h(t)

(V) Find x[n] (VII) Find y[n] (IX) Find H(z)

Now, x(t) = (t) Now, y(t) = h(t) H(z) = Y(z) / X(z)
Put t = nTs
Put t = nT
y[nT] = h[nT]
x(nT) = (nT) y[n] = h[n]
x[n] = [n]

10
Kiran TALELE 9987030881 talelesir@gmail.com 10
 •Impulse Invariant Method ALGO
H(s) → h(t) → h[n] → H(z)

Given H(s) To Find H(z)

Step -1 : Find h(t) by ILT
Step-2 : Find h[n] by Sampling
Step-3 : Find H(z) by ZT

Kiran TALELE 9987030881 talelesir@gmail.com 11

11
 Derive the relation between Analog filter POLE and
Digital filter POLE when Impulse Invariant Method
is used for filter design.
Solution : Consider Analog Filter,
Single
Order Filter

Eq-1
By Inverse Laplace Transform,

u(t)
-a
Put t = nT

Kiran TALELE 9987030881 talelesir@gmail.com 12

12
 u(n)
u(n)

By Z-Transform,

Digital POLE
Eq-2

By equating equation 1 and 2 we get,

Kiran TALELE 9987030881 talelesir@gmail.com 13

13
 Derive the relation between Analog filter
Frequency and Digital filter Frequency when
Impulse Invariant Method is used for filter design.

Solution :
In I I M,

Analog Filter POLE Digital Filter POLE

S1 = σ+j 

Kiran TALELE 9987030881 talelesir@gmail.com 14

14
 By substituting we get,

By equating we get,

Analog Filter
Frequency
(In radians/sec)

Digital Filter
Frequency
Frequency
(In radians)
(In Hz)

Kiran TALELE 9987030881 talelesir@gmail.com 15

15
A frequency mapping between analog frequency and digital freqiency.


IIM

Analog LPF

PB SB

− − c c +
0

Digital LPF PB SB

−  − c 0 c  

Kiran TALELE 9987030881 talelesir@gmail.com 16

16
Q1. A Digital filter is required to be designed with cutoff
frequency of 0.2π radian and sampling frequency of
1000 Hz using Impulse Invariant Method. What is
Analog Domain Cutoff frequency ?.

Given,
Digital Filter design by IIM
Cutoff Frequency = 0.2π
Sampling Frequency Fs = 1000 Hz

Find Analog Cutoff Frequency ?

i.e. c

Kiran TALELE 9987030881 talelesir@gmail.com 17

17
 The relation between AF freq. and DF Freq. is given by,

Where (1) Fs = 1000 Hz Now, c = wc Fs

(2) wc = 0.2π radian c = 0.2π 1000

c = 200 π r/s

c = 628 r/s

Kiran TALELE 9987030881 talelesir@gmail.com 18

18
Q2. A Digital filter is required to be designed with
cutoff frequency of 100 Hz and sampling frequency
of 1000 Hz using Impulse Invariant Method.
What is Analog Domain Cutoff frequency ?.

Given,
Digital Filter design by IIM
Cutoff Frequency = 100 Hz
Sampling Frequency Fs = 1000 Hz
Find Analog Cutoff Frequency ?

Kiran TALELE 9987030881 talelesir@gmail.com 19

19
 The relation between AF freq. and DF Freq. is given by,
Frequency
(In Hz)

Sampling
Frequency
Digital (In Hz)
Where (1) Fs = 1000 Hz Frequency
(NO Units)

Now, c = wc Fs

c = 0.2π 1000
c = 200 π r/s

c = 628 r/s

Kiran TALELE 9987030881 talelesir@gmail.com 20

20
 The relation between AF freq. and DF Freq. is given by,

Frequency
(In Hz)

Sampling
Frequency
Digital (In Hz)
Where (1) Fs = 1000 Hz Frequency
(NO Units)

Now, c = wc Fs

c = 0.2π 1000
c = 200 π r/s

c = 628 r/s

Kiran TALELE 9987030881 talelesir@gmail.com 21

21
 Explain the Mapping of POLES from Analog Filter
to Digital Filter in Impulse Invariant Method..

Analog Filter Digital Filter

S1 = σ+j 

σ =0
σ<0 σ>0

------------S-Plane---------------- ------------Z-Plane-----------------

Kiran TALELE 9987030881 talelesir@gmail.com 22

 Case-I When σ = 0 r = 1
Analog poles which lies on imaginary axis gets mapped onto
the unit circle in the z-plane.

ANALOG FILTER DIGITAL FILTER

=0  Marginally
Stable


------------S-Plane---------------- ------------Z-Plane-----------------

Kiran TALELE 9987030881 talelesir@gmail.com 23

 Case-II When σ < 0 r<1
Analog poles which lies on LEFT side of S-Plane gets mapped
INSIDE the unit circle in the z-plane.

ANALOG FILTER DIGITAL FILTER

r<1
<0 
 INSIDE
the Unit Circle
Stable
LEFTSIDE
Of S-Plane

Kiran TALELE 9987030881 talelesir@gmail.com 24

 Case-III When σ>0 r>1
Analog poles which lies on RIGHT side of S-Plane gets mapped
OUTSIDE the unit circle in the z-plane.

ANALOG FILTER DIGITAL FILTER

>0
r>1
 
Un-Stable

RIGHT SIDE OUTSIDE

Of S-Plane Of Unit Circle

Kiran TALELE 9987030881 talelesir@gmail.com 25

Examples :

Case –1, Let S1 = σ+j  2 

Case –2, Let S =  + j   + 
2  T 
S1T S 2T
Then Z1 = e Then Z2 = e
( + j ) T   2 
Z1 = e  + j 

 +  T

Z2 = e 
T

T j T
Z1 = e e Z 2 = e T e j  T e j 2
Put e j 2 = 1
T j T
Z2 = e e

Kiran
KiranTALELE 9987030881
TALELE 99 870talelesir@gmail.com
30 881 talelesir@gmail.com
2626
 Summary :-
Here, Analog POLESs S 1 ≠ S2
But the corresponding Digital POLEs Z1 = Z2
That means both S1 and S2 are mapped at the
same location in z-plane.

This overlapping of POLEs is called aliasing.

In IIM, the mapping of POLEs from AF to DF is

many-to-one.

Kiran TALELE 9987030881 talelesir@gmail.com 27 27

 S2

S1

z1

S-plane Z-plane

Kiran TALELE 9987030881 talelesir@gmail.com 28

28
 S3

S2

S1

z1

S-plane Z-plane

Kiran TALELE 9987030881 talelesir@gmail.com 29

29
 Freq Transformation from AF To DF in IIM

w


Digital W=T
Filter Linear
0 
0
Analog
Filter

0 

Kiran TALELE 9987030881 talelesir@gmail.com 30

 Summary :-

(i) In IIM, freq transformation from AF to DF is NOT

one-to-one.
 π
(ii) When AF freq exceeds the range 0,
 T

the frequency aliasing occurs.

(iii) IIM is suitable only when operating freq of

Input is in the range 0, π 

 T


(iv) IIM is NOT suitable to design HPF/BPF.

Kiran TALELE 9987030881 talelesir@gmail.com 31

31
 FAQ
What are the advantages of IIR Filters ?

Ans : IIR filters require Less memory

and calculations to achieve a
given filter response
characteristic

Kiran TALELE 9987030881 talelesir@gmail.com 32

32
 FAQ
What is the relation between Analog Filter
POLE and Digital Filter POLE when
Impulse Invariant Method is used for filter
design?.
Ans :
Z = eST
Digital
filter
POLE Analog
Sampling
filter
Time
POLE

Kiran TALELE 9987030881 talelesir@gmail.com 33

33
 FAQ
What is the relation between Analog
Filter Frequency and Digital Filter
Frequency when Impulse Invariant
Method is used for filter design.
Ans : Analog Filter
Frequency
(In radians/sec)

Digital Filter
Frequency
(In radians) Frequency
(In Hz)

Kiran TALELE 9987030881 talelesir@gmail.com 34

34
 FAQ

Why Impulse Invariant method is

not suitable for HPF / BPF design?

The mapping from the analog frequency 

to the frequency variable w in the digital
domain is many to one which reflects the
effect of aliasing due to sampling.
A one to one mapping is thus possible only
if frequency c lies in the principle range
of −     
T T

Kiran TALELE 9987030881 talelesir@gmail.com 35

35
Kiran TALELE 9987030881 talelesir@gmail.com 36
36
Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 38

 Digital IIR Filter Design
using BLT Method
@
Sardar Patel Institute of
Technology, Andheri, Mumbai
Kiran Tulshiram Vasumati TALELE
Academic Year 2020-21
Bharatiya Vidya Bhavans’
Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015]
Digital IIR filter Design
using BiLinear
Transformation (BLT)
Method

Kiran TALELE
 99870 30 881
 Derivation of BLT Equation
Consider Analog filter with transfer function
1
H (s) = Eq-1
s +1
• To find Differential Equation,
Y (s) 1 y’(t) + y(t) = x(t)
=
X (s) s +1 y’(t) = x(t) - y(t)
• Cross Multiply,
y’(nt) = x(nt) - y(nt)
S Y(s) + Y(s) = X(s)
• By Inverse LT, y’(n) = x(n) - y(n)

Kiran TALELE 9987030881 talelesir@gmail.com 4

4
 t
By integral calculus, y (t ) = y (t0 ) +  y ' (t ) dt
Let t = n T to
and t = (n-1)T
nT

y (nT ) = y (n − 1)T  +  (t ) dt

y '

( n −1) T

By Trapezoidal rule of Integration

y [nT ] = y [(n − 1) T ] +
T
2
 y [nT ] +
'
y ' [(n − 1) T ] 

y [n] = y [n − 1] +
T '
2

y [n] + y [n − 1]
'

Kiran TALELE 9987030881 talelesir@gmail.com 5
5
 y [n] = y [n − 1] +
T
2
 y [ n] +
'
y ' [n − 1] 
y [n] = y [n − 1] + 
T x[n] − y[n] x[n −1] − y[n −1]
+ 
2
−1
Y [ z] = z Y [ z ] +
T
 X [ z] − Y [ z ] + z −1 X [ z] − z −1 Y [ z] 
2
−1 T T T −1 T −1
Y [ z] = z Y [ z ] + X [ z] − Y [ z] + z X [ z] − z Y [ z]
2 2 2 2
 −1 T −1 T  T −1 T 
Y ( z ) 1 − z + +z  = X ( z)  + z 
 2 2  2 2

Kiran TALELE 9987030881 talelesir@gmail.com 6

6
  −1 T −1 T  T −1 T 
Y ( z ) 1 − z + + z  = X ( z)  + z 
 2 2 2 2
T
+ z −1
T T
2
(
1 + z −1 )
Y (z)
=
2 2 =
X (z) 1− z −1 +
T
+ z −1
T (
1−z −1
+)T
2
(
1 + z −1 )
2 2

1 1
H ( z) = =
(1 − z −1 ) 2 ( z − 1)
1 + 1 +
T
( 1 + z −1 ) T ( z + 1)
2

Kiran TALELE 9987030881 talelesir@gmail.com 7

7
 1
H ( z ) = 2 ( z −1) Eq-2
+ 1
T ( z +1) 1
H (s) =
s +1
By comparing Eq-1 and Eq-2 we get,

H(z) = H(s) s = 2 ( z −1)

T ( z +1) BLT
Equation
Digital Analog
Filter
Filter

Kiran TALELE 9987030881 talelesir@gmail.com 8

8
 Property of the Bilinear Transformation:
it preserves the stability regions.

s-plane z-plane

Kiran TALELE 9987030881 talelesir@gmail.com 9

 What is the relation between Analog filter POLE
and Digital filter POLE when BLT Method is used
for filter design.
Solution :
In BLT, H(z) = H(s) s = 2 ( z −1)
T ( z +1)

The relation between Analog filter POLE and Digital

filter POLE is given by,
Digital
Filter
POLE
Analog
Filter
POLE

Kiran TALELE 9987030881 talelesir@gmail.com 10

 What is the relation between Analog filter Frequency and
Digital filter Frequency when
BLT Method is used for filter design.

Solution :
In BLT, The relation between Analog filter POLE
and Digital filter POLE is given by,

Put S = σ+j 
Let σ = 0 r=1

S= j 

Kiran TALELE 9987030881 talelesir@gmail.com 11

11
 We get,

Kiran TALELE 9987030881 talelesir@gmail.com 12

12
 Analog Filter
Frequency in
radian/sec
Digital i.e. p, s, c
Filter
Frequency
Digital Filter
Analog Frequency in
Filter Thisequation
equationisiscalled
called
Frequency This radians
Prewarpingequation
Prewarping equation i.e. Wp, Ws, Wc

Kiran TALELE 9987030881 talelesir@gmail.com 13

A mapping between analog frequency and digital frequency.

 T 
w = 2 tan −1
 
 2  B LT

Analog LPF


− − c c +
0

Digital LPF

−  − c 0 c  
Kiran TALELE 9987030881 talelesir@gmail.com 14
 Q. A Digital filter is required to be designed with cutoff
frequency of 100Hz and sampling frequency of 1000
Hz using BLT Method.
What is Analog Domain Cutoff frequency ?.
Given,
Digital Filter BLT
Cutoff Frequency = 100 Hz
Sampling Frequency Fs = 1000 Hz

Find Analog Cutoff Frequency ?

Kiran TALELE 9987030881 talelesir@gmail.com 15

 The relation between AF freq. and DF Freq. is given by,

2  w
= tan  
T 2

Where (1)

Kiran TALELE 9987030881 talelesir@gmail.com 16

 Frequency warping in BLT
The relation between Analog Filter frequency
and Digital Filter Frequency is given by,
Solution :
2  w
 = tan  
T 2

−1  T 
w = 2 tan  
 2 

Kiran TALELE 9987030881 talelesir@gmail.com 17

 Case -1 For small value of w
For small
Now, 2  w value of 
= tan   tan () = 
T 2
2  w w
=   = w
T 2 T
w
=
T
w = ΩT 0


Kiran TALELE 9987030881 talelesir@gmail.com 18

Case -2 For Large value of w
 w
Now, =
2
T
tan  
2 w
T  w

= tan  
2 2
w  T 
= tan−1 
Non Linear

2  2 
−1  T 
w = 2 tan  
0
 2  

Kiran TALELE 9987030881 talelesir@gmail.com 19

19
 • In BLT, AF frequency in the range [ 0 to ) is
mapped to DF frequency in the range [ 0 to  ].
• This frequency Compression is called
Frequency Warping.
• The frequency warping equation is given by,

−1  T 
w = 2 tan  
 2 

Kiran TALELE 9987030881 talelesir@gmail.com 20

20
 Frequency Transformation from AF to DF in BLT
w w

Digital
Filter
Non-Linear

0
0 

Analog
Filter
0 

Kiran TALELE 9987030881 talelesir@gmail.com 21

21
 IIM
w


Digital W=T
Filter Linear
0 
0
Analog
Filter

0 

Kiran TALELE 9987030881 talelesir@gmail.com 22

22
Kiran TALELE 9987030881 talelesir@gmail.com 23
23
 Summary

(i) In BLT, freq transformation from AF to DF is

one-to-one and there is NO aliasing.

(ii) Freq bands of AF and DF do NOT match.

(iii) Freq bands of AF are compressed in DF.

Kiran TALELE 9987030881 talelesir@gmail.com 24

24


Difference between between IIM and BLT Method

Impulse Invariant Method (IIM) Bilinear Transformation
Method(BLT)
(1) In IIM Method, the relation between (1) In BLT Method, the relation
analog filter POLE and Digital filter POLE between analog filter POLE and
Digital filter POLE is given by
is given by Z=eST 2 ( z − 1)
S =
T ( z + 1)

(2) The relation between Analog freq. (2) The relation between Analog
and Digital freq. is given by,  freq. and Digital freq. is given by,
W=  T 
W = 2 tan −1  
Fs  2 

(3) Frequency relation doesn’t give one (3) Frequency relation gives one to
to one mapping between analog filter one mapping between analog
freq. and digital filter freq. W which filter freq. and digital filter freq.
reflects the effect of aliasing due to W
sampling.
(4). Due to aliasing, HPF or BPF with cut (4). BLT method is suitable for all
off frequency of analog filter c types of filter design.

greater than Tcan . not be designed using
impulse invariant method 25
Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 27

Butterworth Filter Design
@
Sardar Patel Institute of
Technology, Andheri, Mumbai

Kiran Tulshiram Vasumati TALELE

Academic Year 2020-21
Bharatiya Vidya Bhavans’
Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015]
 Analog Butterworth Filter

Kiran Tulshiram Vasumati TALELE

 Analog Butterworth filter
1. The magnitude response is
maximally flat in both Pass 1.0
band and Stop band. 0.707
PB SB
2. There are NO ripples in pass
band and stop band.
0 c 
3. The magnitude response of
Butterworth filter decreases
monotonically as the frequency
increases. 1.0 N=1
0.707
4. As N gets larger, Magnitude N=2
PB
Response approaches an Ideal
Low Pass Frequency
response. 0 c 

Kiran TALELE 9987030881 talelesir@gmail.com 4

4
(1) Magnitude Response of LPF

1
H (j) = 1.0
2N 0.707
 Ω 
1+   PB SB
Ω 
 c

Where N is order of filter. 0 c 

i.e. No of POLEs

At =0, | H(j)| = 1 for all N

At =Ωc, | H(j)| = 0.707 for all N

Kiran TALELE 9987030881 talelesir@gmail.com 5

 Example -1 : LPF c = 1 r/s N=1
ZEROS : Analog Butterworth LPF has NO ZEROS
POLES :
 N +1+ 2 k 
j  
Sk = c e  2N 

 2+ 2 k 
j   Now,
Sk = e  2 

1
H LP ( s) =
(s − s0 )
1
H (s) =
s +1

Kiran 9987030881
TALELE
Kiran TALELE talelesir@gmail.com
9987030881 66
talelesir@gmail.com
 Example -2 : LPF c = 1 r/s N=2
ZEROS : Analog Butterworth LPF has NO ZEROS
POLES :  N +1+ 2 k 
j  
Sk = c e  2N 
To find H(s)
 3+ 2 k 
j  
Sk = e  4 
H LP ( s ) =
1
(s − s0 )(s − s1 )
 3 
j  1
K = 0, S0 = e  4  Hˆ ( s) =
S 2 + 1.414 S + 1
 3 
− j 
K = 1, S1 = e  4 

Kiran 9987030881
TALELE
Kiran TALELE talelesir@gmail.com
9987030881 77
talelesir@gmail.com
Example -3 : LPF c = 1 r/s N=3
ZEROS : Analog Butterworth LPF has NO ZEROS
POLES :
 N +1+ 2 k 
j  
To find H(s)
Sk = c e  2N 

 4+ 2 k 
j  
Sk = e  6 

 2 
j 
k = 0, S 0 = e  3 

k = 1, S1 = e − j = − 1
 2 
− j 
k = 2, S 0 = e  3 

Kiran TALELE 9987030881 talelesir@gmail.com 8

8
Example -4 : LPF c = 1 r/s N=4
ZEROS : Analog Butterworth LPF has NO ZEROS
POLES :
 N +1+ 2 k 
j  
Sk = c e  2N 

To find H(s)

Where
a0 = 1
a1 = 2.613
a2 = 3.414
a3 = 1

Kiran TALELE 9987030881 talelesir@gmail.com 9

9
Another method to find Normalized H(s) for LPF:

where a0 = aN == 1

  (k − 1)  
 cos 
ak =   2N  a
  k   k −1
 sin  
  2N  

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 10
10
Analog Butterworth LPF Design Parameters
Specifications:

Design Specifications :

Ap : Attenuation in PB
As : Attenuation in SB
p : Pass Band Freq.
s : Stop Band Freq.
pass transition stop
band band band

Kiran TALELE 9987030881 talelesir@gmail.com 11

 To Find Analog Butterworth LPF order N
dB
0 1
H (j) =
2N
 Ω 
1+  
Ω 
PB SB  c 

0 0
0.5p s
0.75 


Kiran TALELE 9987030881 talelesir@gmail.com 12

(1) To find Ap At Ω = Ωp
1
H (jp ) =
2N
Ω p
1+  
Ω 
 c 
 
 
Ap = 0 - 20 log 
( )
1
2N 
 1+
Ωp

 Ωc



Ap = 20 log 

1+
Ωp
Ωc
( ) 2N 



Kiran TALELE 9987030881 talelesir@gmail.com 13

 Ap = 20 log 

1+( )
Ωp
Ωc
2N 



  Ω p 2N 
Ap = 10 log  1 +  Ω  
  c 
Ap   Ω p 2N 
= log  1 +  Ω  
  c
10


 Ap  2N
10  10 
= 1+
 Ω p 
 Ωc 
2N
 Ω p 
Ap
10 10
−1 = I
 Ωc 
Kiran TALELE 9987030881 talelesir@gmail.com 14
 (2) To find As
At Ω = Ωs
1
H (js ) =
2N
Ωs
1+  
Ω 
 c
 
 
As = 0 - 20 log 
( )
1
2N 
 1+
Ωs

 Ωc



As = 20 log 

1+
Ωs
Ωc
( ) 2N 



Kiran TALELE 9987030881 talelesir@gmail.com 15

 As = 20 log 

1+
Ωs
Ωc
( ) 2N 


  Ω s 2 N 
As = 10 log  1 +  Ω  
  c 
  Ω s 2 N 
As = log  1 +  Ω  
  c 
10

(
As )  Ωs 
2N
10 10
= 1+ Ω 
 c
2N
As
−1 =  Ω s 
10 10
 Ωc 
II
Kiran TALELE 9987030881 talelesir@gmail.com 16
 Dividing equation II by I we get,

( )
As
10 10
−1
Ap
= Ωs 2N
Ωp
10 10
−1

10 10 − 1 
As
 Ωs 
log  Ap  = 2N log  
10 10 − 1  Ωp 

Kiran TALELE 9987030881 talelesir@gmail.com 17

  10 As 10 − 1 
log  Ap 

10 10 − 1 
for LPF N =
 Ωs 
2 log  
 Ωp 

 10 As 10 − 1 
log  Ap 

10 10 − 1 
for HPF N =
 Ωp 
2 log 
 Ωs 

Kiran TALELE 9987030881 talelesir@gmail.com 18

Prove that

Ωc =
p s
1 = 1
 Ap  2N
 As 
 10 − 1
2N
10 10 10 − 1
   
 
   

Kiran TALELE 9987030881 talelesir@gmail.com 19

 p
Step-1 : TO prove that Ωc = 1
 Ap  2N
 10 
10 − 1
 
 
The magnitude response of Analog
Butterworth LPF is given by,

1
H (j) =
2N
 Ω 
1+  
Ω 
 c 

Kiran TALELE 9987030881 talelesir@gmail.com 20

 At Ω = Ωp
1
H (jp ) =
2N
Ω p
1+  
Ω 
 c 
 
 
Ap = 0 - 20 log 
( )
1
2N 
 1+
Ωp

 Ωc


Ap = 20 log 

1+( )
Ωp
Ωc
2N 



Kiran TALELE 9987030881 talelesir@gmail.com 21

 
Ap = 20 log 

1+ ( )
Ωp
Ωc
2N 


  Ω p 2N 
Ap = 10 log 
  Ωc 
1+  

  

Ap  Ωp
2N

= log 
  Ωc 
1+  

10
  

Kiran TALELE 9987030881 talelesir@gmail.com 22

  Ap  2N
10

 10 
  Ωp
= 1+  Ω 
 c
2N
 Ω p 
Ap
10 10
−1 =
 Ωc 
2N
 Ω p 
Ap
= 10 10
−1
 Ωc 

Kiran TALELE 9987030881 talelesir@gmail.com 23

 1
 Ap 2N
Ωp  10 − 1
= 10 
Ωc  
 

Ωc 1
=
Ωp 1
 Ap  2N
 10 
10 − 1
 
 
Ωp
Ωc =
1
 Ap  2N
 10 − 1
 10 
 
 

Kiran TALELE 9987030881 talelesir@gmail.com 24

Step-2 : TO prove that s
Ωc = 1
 As  2N
10 10 − 1
 
 
The magnitude response of Analog
Butterworth LPF is given by,
1
H (j) =
2N
 Ω 
1+  
Ω 
 c 

Kiran TALELE 9987030881 talelesir@gmail.com 25

 At Ω = Ωs
1
H (js ) =
2N
Ωs
1+  
Ω 
 c
 
 
As = 0 - 20 log 
( )
1
2N 
 1+
Ωs

 Ωc



As = 20 log 

1+
Ωs
Ωc
( ) 2N 



Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 26
 
As = 20 log 

1+ ( )
Ωs
Ωc
2N 


  Ωs  2N

As = 10 log  1 +  Ω  
  c 
 2N

As = log  1 +  ΩΩ s  
  c
10


Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 27
 (As )  Ωs 
2N
10 10
= 1+  Ω 
 c
2N
10
As
10  Ωs 
−1 =  Ω 
 c
2N
 Ω s  = 10
As
10
−1
 Ωc 

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 28
 1

Ωs  As10  2N
= 10 − 1
Ωc  
 

Ωc 1
=
Ωp 1
 As  2N
10 10 − 1
 
 

Ωs
Ωc =
1
 As  2N
10 10 − 1
 
 

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 29
(1) Analog LPF to LPF Transformation
Normalized LPF De-normalized LPF

1.0 1.0

PB SB
PB SB

0 1 0 C

LPF

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 30
30
(2) Analog LPF to HPF Transformation
Normalized LPF De-normalized HPF

1.0 1.0

SB PB
PB SB

0
0 0.5
1
0.75 
0 C

LPF

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 31
 Ex-1. Given 1
H ( s) = 2
S + S +1
Find the transfer function of the following filters.
(a) A LPF with passband freq = 10 rad/sec
HLPF(s) = Ĥ(s) S
S=
10
1
H(s) = 2
S S
  +   +1
10  10 

100
H(s) = 2
S + 10S + 100

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 32
32
 (b) A HPF with a cutoff freq of 10 rad/sec
H HPF (S) = Ĥ LPF (S) 10
S=
S

1
H ( s) = 2
10  10 
  +  +1
S  S 

S2
H(s) = 2
S + 10S + 100

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 33
33
 © A BPF with a pass band of 10 rad/sec and a
center freq of 100 rad/sec
HBPF(s) = HLPF(s)
S2 + o2 where B = 10 and o = 100
S=
SB
1
HBPF(s) = 2
S + S +1 S2 +10000 S2 +104
S= =
10 S 10 S
1
= 2
 S 2 + 10 4   S 2 + 10 4 
 10S  +  10S  + 1
   

100 S2
HBPF(s) = 4
S +10 S3 + 20100S2 + 105 S + 108

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 34
34
 (d) A Band stop filter with a stopband of 2
rads/sec and a centre freq of 10 rads/sec
HBSF(s) = ĤLPF(s)
SB
S= where B = 10 and o = 100
S2 + o2

1
HBSF(s) = 2
S + S +1 S=
10 S
=
10 S
S +10000
2
S2 + 102

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 35
35
 NOTE :

(1) LT { (t) } == 1 1
S
1
(2) LT { u(t) } == S

(3) LT  e u(t)  ==
− at
1
S +a

 e − at cos( bt) u(t) 

S +a
(4) LT == ( S + a) 2 + b 2

(5) LT  e− at Sin ( bt ) u(t)  ==

b
( S + a)2 + b 2

• where b is Analog Resonant frequency.

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 36
36
 • ANALOG Butterworth Filter Design ALGORITHM

Given ➔ Ap As p s

(1) Calculate Filter Order N

(2) Calculate Normalized LPF

(3) Calculate De-normalized H(s)

Kiran TALELE 9987030881 talelesir@gmail.com 37

 • ANALOG Butterworth LPF/HPF Design

Given Ap As p and s

(2) Calculate Filter Order N

LPF HPF

38
Kiran TALELE 9987030881 talelesir@gmail.com 3838
(2) Calculate Normalized LPF Transfer Function

Method-1 to find Normalized H(s) for LPF:

LPF Order = N c = 1 rad/sec

(i) ZEROS : Analog Butterworth LPF has NO ZEROS
(ii) POLES : To find POLES
 N +1+ 2 k 
j  
Sk = c e  2N 
For k = 0, 1, 2, . . . . .N

(iii) To find Normalized H(s)

39
Kiran TALELE 9987030881 talelesir@gmail.com 3939
 Method-2 to find Normalized H(s) for LPF:

where a0 = aN == 1

  (k − 1)  
 cos 
ak =   2N  a
  k   k −1
 sin  
  2N  

Kiran TALELE 9987030881 talelesir@gmail.com 4040

(3) Calculate Denormalized LPF/HPF Transfer Function

Denomalized LPF Denomalized HPF

H HPF (s) = H LPF (s) S = c

S

Where Analog Cutoff Frequency c is given by,

p
c =
( 10 )
1
Ap / 10
−1 2N

41
Kiran TALELE 9987030881 talelesir@gmail.com 4141
 HW-Q1: Magnitude response of Analog LPF is given
below, determine Order and Cutoff Frequency of Filter.

Where Ap = 0.92 dB As = 40 dB
Ωp = 10 rad/sec Ωs = 20 rad/sec
42
Kiran TALELE 9987030881 talelesir@gmail.com 4242
 ANSWER:

N= 8

Ωc = 10.9528 rad/sec

43
Kiran TALELE 9987030881 talelesir@gmail.com 4343
HW-Q2 Given
Ap = 0.91 dB As = 13.97 dB
p = 2.00 rad/sec p = 4.828 rad/sec
Design Analog Butterworth Filter.

ANSWER:

8.2592
H ( s) =
s 2 + 4.0641 s + 8.2592

Kiran TALELE 9987030881 talelesir@gmail.com

21-11-2011 TALELE 44
44
HW-Q3 Determine the order and the poles of a
low pass Butterworth filter that has a – 3db
bandwidth of 500 Hz and an attenuation of 40 db
at 1000 Hz. Also draw the rough sketch of
frequency response characteristics of the
Butterworth filter.

ANSWER :

Order N = 7

Kiran TALELE 9987030881 talelesir@gmail.com

21-11-2011 TALELE 45
45
Digital Butterworth Filter

Kiran Tulshiram Vasumati TALELE

Practical Filter Frequency Response

Transition band
Peak ripple
value

Passband
edge Stopband edge
frequency frequency

Kiran TALELE 9987030881 talelesir@gmail.com 47

 Digital Butterworth LPF Specifications
dB
0 Transition
Ap
Band
Attenuation
Pass Stop As in
Stop Band
Attenuation Band Band
in
Pass Band

0 0.5 0.75 
Design Specifications :
Ap : Attenuation in PB
As : Attenuation in SB
Wp : Pass Band Freq.
Ws : Stop Band Freq.
Fs : Sampling Freq.

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 48
48
• DIGITAL Butterworth Filter Design using IIM

Ap As wp ws Fs p
wp =
s
ws =
Fs Fs
Ap As p s
Step-1 : Design
Analog Butterworth
Filter Order N
Filter
Normalized LPF
De-normalized H(s)

h(t)
Step-2 : Design
h(n)
DIgital Butterworth
Filter using IIM
H(z)
• DIGITAL Butterworth Filter Design using BLT

Ap As wp ws Fs
2  wp 
p = tan 
Ap As p s T 2 
Step-1 : Design  
Analog Butterworth 2 w 
Filter Order N
Filter s = tan s 
T  2 
Normalized LPF
De-normalized H(s)

Step-2 : Design
H ( z ) = H ( s ) s = 2 ( z −1)
DIgital Butterworth
Filter using BLT T ( z +1)
H(z)
 Laplace Transform TABLE

(1) LT { (t) } == 1
1
(2) LT { u(t) } == S

(3) LT  e u(t)  ==
− at
1
S +a

 e − at cos( bt) u(t)  ==

S +a
(4) LT ( S + a) 2 + b 2

(5) LT  e− at Sin ( bt ) u(t)  == b

( S + a)2 + b 2

• where b is Analog Resonant frequency.

Kiran TALELE 9987030881 talelesir@gmail.com 51

51
 Z Transform TABLE

z 2 − r z cos(w)
z 2 − 2 r z cos(w) + r 2

rz sin( w)
z 2 − 2 r z cos(w) + r 2

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 52
52
Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 54

Digital Butterworth Filter Design
using Impulse Invariant Method
@
Sardar Patel Institute of
Technology, Andheri, Mumbai

Kiran Tulshiram Vasumati TALELE

Academic Year 2020-21
Bharatiya Vidya Bhavans’
Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015]
Digital Butterworth Filter
Design using Impulse
Invarient Method

Kiran Tulshiram Vasumati TALELE

 • DIGITAL Butterworth Filter Design using IIM ALGORITHM

Given ➔ Ap As wp ws Fs p s
wp = ws =
Fs Fs
Step-1 Ap As p s
Design
Analog Filter Order N
Butterworth
Filter H(s) Normalized LPF
De-normalized H(s)

Step-2 h(t)
Design
Digital h(n)
Butterworth
Filter H(z) H(z)

Kiran TALELE 9987030881 talelesir@gmail.com 4

 • DIGITAL Butterworth LPF/HPF Design using IIM
Given Ap As wp ws and Fs

Step-1 : Design Analog Butterworth Filter

(1) Calculate p and s p s
wp = ws =
Fs Fs
p = wp Fs s = ws Fs

(2) Calculate Filter Order N

LPF HPF

5
Kiran TALELE 9987030881 talelesir@gmail.com 55
(3) Calculate Normalized LPF Transfer Function

Method-1 to find Normalized H(s) for LPF:

LPF Order = N c = 1 rad/sec

(i) ZEROS : No ZEROS
(ii) POLES : To find POLES
 N +1+ 2 k 
j  
Sk = c e  2N 
For k = 0, 1, 2, . . . . .N

(iii) To find Normalized H(s)

6
Kiran TALELE 9987030881 talelesir@gmail.com 66
 Method-2 to find Normalized H(s) for LPF:

where a0 = aN == 1

  (k − 1)  
 cos 
ak =   2N  a
  k   k −1
 sin  
  2N  

Kiran TALELE 9987030881 talelesir@gmail.com 77

(4) Calculate Denormalized LPF/HPF Transfer Function

Denomalized LPF Denomalized HPF

H HPF (s) = H LPF (s) S = c

S

Where Analog Cutoff Frequency c is given by,

p
c =
( 10 )
1
Ap / 10
−1 2N

8
Kiran TALELE 9987030881 talelesir@gmail.com 88
 Step-2 : Design Digital Butterworth Filter

By Impulse Invariant Method

H(s) → h(t) → h[n] → H(z)

To Find H(z)
(1) Find h(t) by Inverse LT
(2) Find h[n] by Sampling
(3) Find H(z) by ZT

9
Kiran TALELE 9987030881 talelesir@gmail.com 99
 HW - Q1 : An analog domain filter has a
transfer function as given below. Analog
filter is to be converted into digital filter so
that its impulse response characteristics are
retained.
The sampling frequency is 100 Hz.
Find the transfer function H(z).
b
H ( s) =
( s + a) 2 + b 2

Hint : By Impulse Invariant Method

H(s) → h(t) → h[n] → H(z)

Kiran TALELE 9987030881 talelesir@gmail.com 10

 HW-Q2 : A Digital Butterworth is required to meet
the following specifications
• Pass band ripple  1 dB
• Stop band attenuation  40 dB
• Pass band edge = 4 KHz
• Stop band edge = 6 KHz
• Sampling rate = 24 KHz
Find the filter order and cutoff frequency of Digital
Filter if Impulse Invariant Method is used for filter
design.
ANSWER : N = 14
Wc = 1.09 rad
11
Kiran TALELE 9987030881 talelesir@gmail.com 11
 HW- Q3: Design second order Butterworth low pass
filter using impulse invariance technique.
The cut off frequency required to be 50 Hz and sampling
frequency 500 samples/second.
Show filter realization in appropriate form.

Kiran TALELE 9987030881 talelesir@gmail.com 12

HW-Q4 Given

Ap = 0.91 dB As = 13.97 dB
p = 2.00 rad/sec p = 4.828 rad/sec

Design Digital Butterworth LPF using

Impulse Invariant Method.

Kiran TALELE 9987030881 talelesir@gmail.com

21-11-2011 TALELE 13
13
 Laplace Transform TABLE

(1) LT { (t) } == 1
1
(2) LT { u(t) } == S

(3) LT  e u(t)  ==
− at
1
S +a

 e − at cos( bt) u(t)  ==

S +a
(4) LT ( S + a) 2 + b 2

(5) LT  e− at Sin ( bt ) u(t)  == b

( S + a)2 + b 2

• where b is Analog Resonant frequency.

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 14
14
 Z Transform TABLE

z 2 − r z cos(w)
z 2 − 2 r z cos(w) + r 2

rz sin( w)
z 2 − 2 r z cos(w) + r 2

Kiran TALELE 9987030881

9987030881talelesir@gmail.com
talelesir@gmail.com 15
15
Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 17

Digital Butterworth Filter Design
using BLT Method
@
Sardar Patel Institute of
Technology, Andheri, Mumbai
Kiran Tulshiram Vasumati TALELE
Academic Year 2020-21

Bharatiya Vidya Bhavans’

Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Incharge, Intellectual Property Rights Cell of SP-IT [2012]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015]
Digital IIR filter Design using
Bilinear Transformation (BLT)
Method

Kiran Tulshiram Vasumati TALELE

 • DIGITAL Butterworth Filter Design using BLT

Given ➔ Ap As wp ws Fs 2  wp 
p = tan 

Step-1 T  2 
Design
Ap As p s 2  ws 
s = 2 tan ws 
Analog s = T tan 2 
Butterworth Filter Order N T  2 
Filter H(s)
Normalized LPF
De-normalized H(s)
Step-2
Design
Digital H ( z ) = H ( s ) s = 2 ( z −1)
Butterworth T ( z +1)
Filter H(z) H(z)

Kiran TALELE 9987030881 talelesir@gmail.com 4

 • DIGITAL Butterworth LPF/HPF Design using IIM
Given Ap As wp ws and Fs

Step-1 : Design Analog Butterworth Filter

(1) Calculate p and s

2  wp  2  ws 
p = tan  s = tan 
T 2  T  2 
 
(2) Calculate Filter Order N

LPF HPF

5
Kiran TALELE 9987030881 talelesir@gmail.com 55
(3) Calculate Normalized LPF Transfer Function

Method-1 to find Normalized H(s) for LPF:

LPF Order = N c = 1 rad/sec

(i) ZEROS : No ZEROS
(ii) POLES : To find POLES
 N +1+ 2 k 
j  
Sk = c e  2N 
For k = 0, 1, 2, . . . . .N

(iii) To find Normalized H(s)

6
Kiran TALELE 9987030881 talelesir@gmail.com 66
 Method-2 to find Normalized H(s) for LPF:

1
H ( s) =
a0 + a1 S 1 + ........ + ..a N S N

where a0 = aN == 1

  (k − 1)  
 cos 
ak =   2N  a
  k   k −1
 sin  
  2N  

Kiran TALELE 9987030881 talelesir@gmail.com 77

(4) Calculate Denormalized LPF/HPF Transfer Function

Denomalized LPF Denomalized HPF

H HPF (s) = H LPF (s) S = c

S

Where Analog Cutoff Frequency c is given by,

p
c =
( 10 )
1
Ap / 10
−1 2N

8
Kiran TALELE 9987030881 talelesir@gmail.com 88
 Step-2 : Design Digital Butterworth Filter

•By BLT Transformation,

H ( z ) = H ( s)
2 ( z −1)
s=
T ( z +1)

9
Kiran TALELE 9987030881 talelesir@gmail.com 99
 Example-1 : Digital Butterworth is required
to meet the following specifications

• Pass band ripple  1 dB

• Stop band attenuation  40 dB
• Pass band edge = 4 KHz
• Stop band edge = 6 KHz
• Sampling rate = 24 KHz

Find the filter order and cutoff frequency if BLT

technique is used for filter design.

•
11
Kiran TALELE 9987030881 talelesir@gmail.com 11
 Solution :
• Ap = 1 dB As = 40 dB

w p = 2 ( 4 KHz
24 KHz ) == 
3 radian

ws = 2 ( 6 KHz
24 KHz ) == 
2 radian

(1) Calculate p, s

In BLT, 2  w
= tan 
T 2

12
Kiran TALELE 9987030881 talelesir@gmail.com 12
 (i) To find Ωp : (ii) To find Ωs :
2  wp  2  ws 
p = tan  s = tan 
T  2  T  2 
 
tan 3 
2
p =
1  2 
24000  

Ωp = 27712.8 rad/sec Ωs = 48000 rad/sec

13
Kiran TALELE 9987030881 talelesir@gmail.com 13
 (2) Calculate Filter Order N

N = 9.61

Let N = 10

14
Kiran TALELE 9987030881 talelesir@gmail.com 14
 (3) Calculate Analog Cutoff Freq. c

p
c =
( 10 )
1
Ap / 10
−1 2N

Put Ωp = 27712.8 rad/sec

N = 10 and
Ap = 1 dB
By solving we get,

Ωc = 29649.7 rad/sec

15
Kiran TALELE 9987030881 talelesir@gmail.com 15
 (4) Calculate Digital Cutoff Freq. wc

2  C 
C = tan 
T  2 
 
CT  C 
= tan 2


2  
−1  CT 
Wc = 2 tan  
 2 
Put Ωc = 29649.7 rad/sec
T = 1 / 24000 sec

Wc = 1.106 rad
16
Kiran TALELE 9987030881 talelesir@gmail.com 16
Example-2 : Design a first order High Pass DT
Butterworth filter whose cutoff frequency is
1 kHz at the sampling rate of 104 sample/sec.
Solution :
Given Order N = 1
HPF Butterworth filter
Cutoff frequency = 1 KHz
Sampling rate = 104 Hz.

Digital Cut-off frequency : C = 2  f C

wc = 2 [0.1] = 0.2  rad

Kiran TALELE 9987030881 talelesir@gmail.com 17

17
 STEP-1 Design Analog Butterworth HPF Filter

• (1) Calculate Normalized LPF

LPF c = 1 rad/sec N=1
POLES :
 N +1+ 2 k 
j  
Sk = c e  2 N 

 2+ 2 k 
j  
Sk = e  2 

K = 0, S0 = -1
Now, 1 1
H LP ( s) = H (s) =
(s − s0 ) s +1

Kiran TALELE 9987030881 talelesir@gmail.com 18

18
 (2) Calculate De-Normalized LPF

H HPF (s) = H LPF (s) S = c

S

2  C  2  0.2 
C = tan  =
 tan   = 6498.39 rad / sec
Ts  2  Ts  2 

1
H(s) =
6498.38
+1
S

s
H HP ( s) =
s + 6498.39

Kiran
KiranTALELE
TALELE9987030881 talelesir@gmail.com
9987030881 talelesir@gmail.com 19
19
 STEP-2 Design Digital Butterworth HPF Filter
•By BLT transformation,
H ( z ) = H ( s) s = 2( z −1)
T ( z +1)

Kiran TALELE 9987030881 talelesir@gmail.com 20

20
 Example-3 : Convert the analog filter with system
function s + 0.1
H ( s) =
( s + 0.1) 2 + 16

into a digital IIR filter by means of a bilinear

transformation.
The digital filter is to resonate at freq. wr = /2.

21
Kiran TALELE 9987030881 talelesir@gmail.com 21
 s+a
Solution : Let H ( s ) =
( s + a) 2 + b 2
where b is analog resonant frequency.

By comparing we get,
Analog Resonant Frequency b = 4.

Let r = b = 4

Digital resonant frequency wr = /2

22
Kiran TALELE 9987030881 talelesir@gmail.com 22
 2 w 
r = tan r 
T  2 

2  / 2
4= tan 
T  2 
By solving we get T = 0.5 sec

By BLT, Digital Filter is given by,

H ( z ) = H ( s)
2 ( z −1)
s=
T ( z +1)
Put T = 0.5 sec

23
Kiran TALELE 9987030881 talelesir@gmail.com 23
 s + 0.1
Now, H ( s) =
( s + 0.1) 2 + 16

s + 0.1
H ( z) =
( s + 0.1) 2 + 16
4 ( z −1)
s=
( z +1)

0.128 + 0.006 z −1 − 0.122 z −2

H ( z) = −1 −2
1 + 0.0006 z − 0.975 z

Kiran TALELE 9987030881 talelesir@gmail.com 24

 Example-4: Design a digital Butterworth filter that
satisfies the following constraint using bilinear
transformation. Assume T = 1s.

0.9  | H(ejw) |1 ; 0 w≤ 2

3
|H (ejw) |  0.2 for  w  .
4

21-11-2011 TALELE 25
Kiran TALELE 9987030881 talelesir@gmail.com 25
 0.9  | H(ejw) |  1 ; 0  w ≤ 0.5 
| H(ejw) |  0.2 ; 0.75   w ≤ 

1.0
0
0.9 20log(0.9)
0.2
20log(0.2)
0 0
0.50.5 0.75
0.75 
 0 0.5 0.75 

26
Kiran TALELE 9987030881 talelesir@gmail.com 26
 (i) Ap = 0 – 20 log ( 0.9)
= 0.9151 dB

(ii) As = 0 – 20 log ( 0.2)

= 13.97 dB

(iii) wp = 0.5  rad.

(iv) ws = 0.75  rad.

27
Kiran TALELE 9987030881 talelesir@gmail.com 27
 STEP-1 Design Analog Butterworth LPF Filter
(1) Calculate Ω p , Ω s
2  w
In BLT,  = tan  
T 2

2  wp  2 w 
(i )  p =tan   (ii)  s = tan  s 
T  2  T  2 
Assume T = 1 Sec

 0.5 
 p = 2 tan  
 2 
 p = 2 rad / sec

21-11-2011 TALELE 28
Kiran TALELE 9987030881 talelesir@gmail.com 28
 (2) Calculate Filter Order

N = 1.966  2

29
Kiran TALELE 9987030881 talelesir@gmail.com 29
 • (1) Calculate Normalized LPF
LPF c = 1 rad/sec N=2
ZEROS : NO ZEROs

POLES :  N +1+ 2 k 
j  
Sk = c e  2N 

 3+ 2 k 
j  
Sk = e  4 

 3 
j 
k = 0, S 0 = e  4 

 3 
− j 
k = 1, S1 = e  4 

30
Kiran TALELE 9987030881 talelesir@gmail.com 30
 1
H LP ( s) =
(s − s0 )(s − s1 )
1
Hˆ ( s) =
S2 + 2S +1

(4) Calculate De-normalized LPF

ˆ ( s)
H ( s) = H
S
where, S=
c

p
c = == 2.8478
(10 )
1
Ap / 10
−1 2N

Kiran TALELE 9987030881 talelesir@gmail.com 31

 1
H ( s) = 2
s + 2 s +1 S=
S
2.8738

1
H (s) = 2
 s   s 
 2.8738  + 2  2.8738  + 1

8.2592
H ( s) =
s 2 + 4.0641 s + 8.2592

21-11-2011 TALELE 32
Kiran TALELE 9987030881 talelesir@gmail.com 32
 STEP – 2 Design a Digital Butterworth LPF

H ( z ) = H ( s)
2 ( z −1)
s=
T ( z +1)

8.2592
H ( z) =
2
 2( z − 1)   2( z − 1) 
 ( z + 1)  + 4.0641  ( z + 1)  + 8.2592
   

8.2592( z + 1) 2
H ( z) =
2( z − 1) 2 + 8.182 ( z − 1) ( z + 1) + 8.2592 ( z + 1) 2

21-11-2011 TALELE 33
Kiran TALELE 9987030881 talelesir@gmail.com 33
 Example-5 Design and realize a Low Pass Filter using
the Bilinear Transformation Method to satisfy the
following characteristics.
(i) Monotonic Stop Band and Pass Band.
(ii) –3dB cutoff frequency of 0.5
(iii) Stop Band Attenuation of 15 dB at 0.75

Solution :
Monotonic response means Butterworth filter.

Kiran TALELE 9987030881 talelesir@gmail.com 34

34
 dB
0
Ap
-3
Pass Stop 15 dB As
Band Band
-15

wc = 0.5 ws= 0.75 

0 0.5 0.75 
0

Digital Butterworth LPF

Specifications :
Ap = 3 dB As = 15 dB
wp = 0.5 ws = 0.75

Kiran TALELE 9987030881 talelesir@gmail.com 35

35
Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 36

 Introducton to FIR Filter :
Concept of Linear Phase
@
Sardar Patel Institute of
Technology, Andheri, Mumbai
Kiran Tulshiram Vasumati TALELE

Academic Year 2020-21

Bharatiya Vidya Bhavans’
Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015] 2
Q1. Show that FIR filter is always STABLE.
ANS : Consider a Causal Digital FIR filter with impulse response


h [ n] = 1 2

3 4 
H(z) = 1 + 2 z −1 + 3 z −2 + 4 z −3

1 + 2 z −1 + 3 z −2 + 4 z −3
H(z) =
1

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

Kiran TALELE 9987030881 talelesir@gmail.com 3

 z 3 (1 + 2 z −1 + 3 z −2 + 4 z −3 )
H(z) = 3
z ( 1 )
z 3 + 2 z 2 + 3 z1 + 4
H(z) =
z3 In FIR Filter POLES
are always only at
POLES : P1 = P2 = P3 = 0 origin
For causal & stable
system, all the
All POLES are at ORIGIN
POLES must lie
INSIDE the unit
circle.
Therefore, FIR
Filter is always
STABLE

Kiran TALELE 9987030881 talelesir@gmail.com 4

 Q2. Plot Magnitude and Phase Spectrum of a Linear
Phase FIR filter with h[n]={ 3,2,1,2,3 }
ANS :
h [ n] = 3  
2 1 2 3 
H(z) = 3 + 2 z −1 + z −2 + 2 z −3 + 3 z −4
Put z = e jw
H(e jw ) = 3 + 2 e − jw + e − j 2 w + 2 e − j 3w + 3 e − j 4 w
 N −1 
− j   
H(e ) = e
jw  2 
 
 
Put N = 5

Kiran TALELE 9987030881 talelesir@gmail.com 5

 − jw − j 2w − j 3w − j 4w
H(e ) =
jw
3 + 2 e + e + 2 e + 3 e


H(e jw ) = e− j 2 3e j 2 + 2e j + 1 + 2e− j + 3e− j 2 
H(e jw ) = e− j 2  3(e j 2
+ e− j 2 ) + 2( e j + e− j ) + 1 
H(e jw ) = e− j 2  6 cos(2) + 4 cos() + 1 

Frequency
Response
Real Part of
H(w)
Frequency
Phase
Response
Response
Ф(w) Hr(w)

Kiran TALELE 9987030881 talelesir@gmail.com 6

 Summary
(1) Magnitude Response
H(w) = 6 cos( 2 ) + 4 cos( ) + 1

(2) Phase Response  (w) = e− j 2 = e j ( )

(3) Phase  (w) = −2 w

(4) Generalized Phase

− 2 w if Re{H ( w)}  0
 ( w) = 
− 2w +  if Re{H ( w)}  0

Kiran TALELE 9987030881 talelesir@gmail.com 7

 Sr. Freq. Hr(w) Phase 
NO W
1 0 11.00 0
2 0.1  9.66 – 0.2 
3 0.2  6.09 – 0.4 
4 0.3  1.50 – 0.6 
5 0.4  – 2.62 – 0.8  +  = 0.2 
6 0.5  – 5.00 – 1.0  +  = 0
7 0.6  – 5.09 – 1.2  +  = −0.2 
8 0.7  – 3.20 – 1.4  +  = −0.4 
9 0.8  – 0.38 – 1.6  +  = −0.6 
10 0.9  2.05 – 1.8  + (2 ) = 0.2 
11  3.00 – 2.0  + (2 ) = 0

Note : Range of freq. w is (-,  ]

Range of phase  is (-,  ]

Kiran TALELE 9987030881 talelesir@gmail.com 8

 Sr. Freq. Hr(w) Phase  Phase 
NO W
1 0 11.00 0 0
2 0.1  9.66 – 0.2  – 0.2 
3 0.2  6.09 – 0.4  – 0.4 
4 0.3  1.50 – 0.6  – 0.6 
5 0.4  – 2.62 – 0.8  +  = 0.2  0.2 
6 0.5  – 5.00 – 1.0  +  = 0 0
7 0.6  – 5.09 – 1.2  +  = −0.2  −0.2 
8 0.7  – 3.20 – 1.4  +  = −0.4  −0.4 
9 0.8  – 0.38 – 1.6  +  = −0.6  −0.6 
10 0.9  2.05 – 1.8  + (2 ) = 0.2  0.2 
11  3.00 – 2.0  + (2 ) = 0 0

Kiran TALELE 9987030881 talelesir@gmail.com 9

 Magnitude Spectrum
Sr. Freq. Hr(w)
NO W
10 1 0 11.00
9 2 0.1  9.66
8 3 0.2  6.09
7 4 0.3  1.50
6 5 0.4  – 2.62
5 6 0.5  – 5.00
4
7 0.6  – 5.09
3
2 8 0.7  – 3.20
1 9 0.8  – 0.38
10 0.9  2.05
0 0.1 0.2 0.4 0.6 0.8  w 11  3.00

Kiran TALELE 9987030881 talelesir@gmail.com 10

 Phase Spectrum


Sr. Freq. Phase 
 NO W
0.8
1 0 0
0.6 2 0.1  – 0.2 
0.4 3 0.2  – 0.4 
0.2
w 4 0.3  – 0.6 
0.2 0.4 0.6 0.8  5 0.4  0.2 
0 6 0.5  0
–0.2
7 0.6  −0.2 
8 0.7  −0.4 
–0.4 9 0.8  −0.6 
–0.6
10 0.9  0.2 
Linear 11  0
–0.8

-

Kiran TALELE 9987030881 talelesir@gmail.com 11

 Q3. What is Linear Phase Filter?

ANS : The Linear Phase Filter has phase response

linearly varying with frequency.

• i.e.  is proportional to frequency w.

Kiran TALELE 9987030881 talelesir@gmail.com 12

Q(4) What is the advantage of Linear Phase ?
If the phase response of the filter is Linear then
the output of the filter is same as original input
delayed by some constant, say .

There is NO distortion at the output.

Kiran TALELE 9987030881 talelesir@gmail.com 13

Q5. Consider a Linear Phase LPF with frequency response,
 e − j2w 0  w  0.5π
H(w) = 
 0 otherwise

Determine the output of the filter for any

1 given
input x[n].
Hence, Show that if the phase response of the filter is
Linear, then output of filter is delayed version of input
signal .

Kiran TALELE 9987030881 talelesir@gmail.com 14

A Linear Phase LPF with frequency response,

 e − j2w 0  w  0.5π | H(w) |

H(w) =  1

 0 otherwise PB SB
0 0.5 
| H(w) | = 1 for 0  w  0.5π

 = −2 w
  w (Hence, Linear Phase)

Kiran TALELE 9987030881 talelesir@gmail.com 15

 Consider x[n] = 3 Sin(0.1n) - 4 Cos(0.2n)
where w1 = 0.1 and w2 = 0.2
| H(w) |
1 W Mag. Phase  = − 2w
0.1 1 1 = - 0.2 
0 0.5 
0.2 1 2 = - 0.4 
To find output y[n] :
y[n]= 3 (1) Sin[0.1n +(-0.2 )] - 4(1)Cos[0.2n+(-0.4)]
y[n]= 3 Sin[0.1(n-2)] - 4 Cos[0.2(n-2)]

y[n] = x[n-2]

Here, output is delayed

16
version of input
 Q5. Show that if the phase response of the filter is NOT
Linear, then output of filter is distorted version of input
signal .
ANS : Consider a Linear Phase LPF with frequency response,
 2
 e − j 3w 0  w  0.5  | H(w) |
H(w) = 

 0 otherwise 1

PB SB
|H(w)|= 1 for 0  w  0.5
0 0.5 
2
 = −3 w (Non - Linear Phase )

Kiran TALELE 9987030881 talelesir@gmail.com 17

 Consider x[n] = 3 Sin(0.1n) - 4 Cos(0.2n)
where w1 = 0.1 and w2 = 0.2
2
W Mag. Phase  = −3 w
0.1 1 1- = -0.032
To find output y[n] : 0.2 1
2- = -0.12
2

y[n]= 3(1) Sin[0.1n+(-0.032)] - 4(1)Cos[0.2n+(-0.122)]

Here, output is distorted

version of input

Kiran TALELE 9987030881 talelesir@gmail.com 18

 Q5. What is Phase Delay and Group Delay ?
ANS :
• The phase delay (tp) and group delay (tg) of the filter
are given by,
− − d
tp = and tg =
w dw

Kiran TALELE 9987030881 talelesir@gmail.com 19

 Ex. Given y[n] = 0.25 x[n] + x[n − 1] + 0.25 x[n − 2]
Calculate Phase Delay and Group Delay.
Solution :
− − d
tp = and tg =
w dw

y[n] = 0.25 x[n] + x[n − 1] + 0.25 x[n − 2]

−1 −2
Y ( z) = 0.25 X ( z) + z X ( z) + 0.25 z X ( z)
(
Y ( z) = X ( z) 0.25 + z −1
+ 0.25 z −2
)
−1 −2
H ( z) = 0.25 + z + 0.25 z

Kiran TALELE 9987030881 talelesir@gmail.com 20

 H (e jw ) = 0.25 + e− jw + 0.25 e− j 2 w

H (e jw ) = e− jw 0.25 e jw + 1 + 0.25 e− jw 
H (e jw ) = e− jw  1 + 0.5 cos(w) 

−  ( w)
Phase Delay  g = == 1
w
− d ( w)
Group Delay  g = == 1
dw

Kiran TALELE 9987030881 talelesir@gmail.com 21

 NOTE : The necessary condition to have Linear
Phase is Group Delay of the filter must be
constant. ie independent of frequency.
• That is possible only when h[n] is either
Symmetric or Anti-symmetric.

Kiran TALELE 9987030881 talelesir@gmail.com 22

 For Symmetric h[n] :
 N −1 
• Phase : = − w
 2 
−  N −1 
• Phase Delay : ==  
w  2 
• Group Delay : − d  N −1 
==  
dw  2 

Example of Symmetric h[n]:

h[n] = { 1, 2, 3, 2, 1 } N=5 ODD
h[n] = { 1, 2, 2, 1 } N=4 EVEN

Kiran TALELE 9987030881 talelesir@gmail.com 23

 For Anti-Symmetric h[n] :
  N −1 
• Phase : = −  w
2  2 
− − 0.5   N −1 
• Phase Delay : = + 
w w  2 

• Group Delay : − d  N −1 
=  
dw  2 

Example of Antisymmetric h[n]:

h[n] = { 1, 2, 0, -2, -1 } N=5 ODD

h[n] = { 1, 2, -2, -1 } N=4 EVEN

Kiran TALELE 9987030881 talelesir@gmail.com 24

For a Linear PHASE FIR filter

h[n] must be either

Symmetric OR Antisymmetric

Kiran TALELE 9987030881 talelesir@gmail.com 25

 NOTE :
Linear Phase Filters are those Filters in which
the group delay is constant, ie independent of
frequency.
Linear Phase Filters are also called as constant
time delay Filters.

Kiran TALELE 9987030881 talelesir@gmail.com 26

 Show that if the Phase Response is
Linear the output of the Filter during
pass-band is delayed input.
Ans :
Consider a LPF with frequency
response H(e-jw) given by

− jw

 e | w |  wc
H (e ) = 
jw

 0 wc  w  

Kiran TALELE 9987030881 talelesir@gmail.com 27

 x[n] H(ejw ) y[n]
X (w) Y (w)

Let Y(w) = X(w) . H(w)

Y(w) = X(w) . e–jw

By iDTFT,
y[n] = x[n – ]  o/p of filter

Kiran TALELE 9987030881 talelesir@gmail.com 28

 NOTE : For Linear Phase filter h[n]
must be either Symmetric
or Antisymmetric
Examples of Examples of
Linear phase filters non Linear phase filters
h[n]= {3, 2, 1, 2, 3} h[n]={ 1, 2, 3, 1, 2, 3 }

h[n]= { 1, 2, 2, 1 } h[n]={3, 2, 1, 1, 2}

h[n]={1,-2, 0, 2,-1} h[n]={3, 2, 1,-2,-3}

h[n]= [n] + [n-3] h[n] = {3, -2, 2, 3}

Kiran TALELE 9987030881 talelesir@gmail.com 29

Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 31

31
 Introducton to FIR Filter :
POLES & ZEROS
@
Sardar Patel Institute of
Technology, Andheri, Mumbai
Kiran Tulshiram Vasumati TALELE

Academic Year 2020-21

Bharatiya Vidya Bhavans’
Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015] 2
 NOTE-1: When h[n] is either Symmetric
OR Antisymmetric, ZEROS of the
filter are always in Reciprocal
order.

i.e. If Z1 is ZERO of the filter,

Then 1/Z1 is also a ZERO of the filter

Kiran TALELE 9987030881 talelesir@gmail.com 3

 NOTE-2 : If ZEROS of the filter
are in reciprocal order,
Then filter is Linear Phase
FIR filter
 1 (z + 1) (z − 1)
 z −  (z − 2) H (z) =
H (z) = 
2
Z2
Z2

 1
(z + 1)  Z −  (z − 2)
H (z) =  2
3
Z

Kiran TALELE 9987030881 talelesir@gmail.com 4

 Position of Definite ZEROS
I. When h[n] is Symmetric
Type -2 filter : N EVEN Type -1 filter : N ODD

II. When h[n] is Anti-symmetric

Type -4 filter : N EVEN Type -3 filter : N ODD

Kiran TALELE 9987030881 talelesir@gmail.com 5

(Q1) Show three possible POLE ZERO pattern
of 4th order Linear Phase FIR filter with
(a) symmetric h[n] (b) Anti-symmetric h[n]

Solution :
• Linear Phase FIR filter Order = 4
Let No of POLES = 4
• Let No of ZEROS = 4
• N-1 = 4
• N= 5

Kiran TALELE 9987030881 talelesir@gmail.com 6

 (a) Symmetric h[n]
2 ej/4

-2 -0.5 0.5 2

4 POLES
0.5 2 2 e-j/4
4 POLES
4 POLES

Kiran TALELE 9987030881 talelesir@gmail.com 7

(b) Anti-symmetric h[n] with N odd has definite
zeros at z = 1 and z = -1

4 POLES 4 POLES

4 POLES
4 POLES

Kiran TALELE 9987030881 talelesir@gmail.com 8

(Q2) Anti-symmetric h[n] can-not be used for
LPF design.Justify.

| H(e jw ) |  1 0  w  wc
H(e ) = 
jw

 0 Otherwise

H(e )  0 for 0  w  w c
jw

PB SB

0 wc  H(e jw )  0 at w = 0

Kiran TALELE 9987030881 talelesir@gmail.com 9

| H(e jw ) | H(e jw )  0 at w = 0
jw
H(e )  0 for LPF
w=0

PB SB Put z = ejw
0 wc  At w = 0 z=1

H(z) z =1  0 for LPF

Kiran TALELE 9987030881 talelesir@gmail.com 10

 • For Anti-symmetric h[n], there exists definite
ZERO at z = 1.

H(z) z =1
= 0

Therefore, Anti-symmetric h[n] can-not be

used for LPF design.

Kiran TALELE 9987030881 talelesir@gmail.com 11

 Position of Definite ZEROS,…..
I. When h[n] is Symmetric
Type -2 filter : N EVEN Type -1 filter : N ODD
LPF LPF
HPF HPF
BPF

II. When h[n] is Anti-symmetric

Type -4 filter : N EVEN Type -3 filter : N ODD
LPF
LPF
HPF
HPF
BPF

Kiran TALELE 9987030881 talelesir@gmail.com 12

(Q3) One of the zeros of a third order causal linear phase
Low Pass FIR filter lies at z=0.5
• Find the location of the other zeros
• Find transfer function of filter.

Solution :
Linear Phase LPF
Order = 3

Kiran TALELE 9987030881 talelesir@gmail.com 13

• For Symmetric h[n] with N even there exists definite zero
at z = -1.

• ZEROS of a linear phase filter occurs at reciprocal location.

1
For ZERO at z1 = 0.5, There exists ZERO at == 2
z1

Kiran TALELE 9987030881 talelesir@gmail.com 14

  1
 z +  ( z + 2) ( z + 1)
H ( z) = 
2
z3
 2 5 
 z + z + 1 ( z + 1)
H ( z) =  
2
z3
7 7
z3 + z2 + z +1
H ( z) = 2 2
z3
7 −1 7 − 2
H ( z ) = 1 + z + z + z −3
2 2
7 7
h[n] = { 1, , , 1}
2 2

Kiran TALELE 9987030881 talelesir@gmail.com 15

• (Q4) One of the zeros of a third order
causal linear phase High pass FIR filter
lies at z=0.5
• Find the location of the other zeros
• Find the transfer function of filter.

Solution :

N –1 = 3 So N = 4 ( Even )

Kiran TALELE 9987030881 talelesir@gmail.com 16

• For Anti symmetric h[n] with N even there
exists definite zero at z = 1.

• ZEROS of a linear phase filter occur at

reciprocal location

1 1
For z0 = , == 2
2 z0

Kiran TALELE 9987030881 talelesir@gmail.com 17

  1
 z −  ( z − 2) ( z − 1)
H ( z) =  2
z3
 2 5 
 z − z + 1 ( z − 1)
H ( z) =  2 
z3
7 2 7
z − z + z −1
3

H ( z) = 2 2
3
z
7 −1 7 − 2 −3
H ( z) = 1 − z + z − z
2 2

Kiran TALELE 9987030881 talelesir@gmail.com 18

(Q5) Draw POLE-ZERO location of third order
Linear Phase LPF & HPF.
Solution :
Linear Phase LPF with Symmetric h[n] and
N EVEN (Definite ZERO at z = – 1)

Kiran TALELE 9987030881 talelesir@gmail.com 19

 Linear Phase HPF with Anti-Symmetric h[n]
and N EVEN. ( Definite ZERO at z = 1 )

Kiran TALELE 9987030881 talelesir@gmail.com 20

(Q6) An antisymmetric filter has one ZERO at
z=0.5 . What is the minimum order of this filter?

Solution :

• Zeros of a linear phase filter occur at

reciprocal location

1 1
z0 =  =2
2 z0

Kiran TALELE 9987030881 talelesir@gmail.com 21

 • Total No of zeros = 3
order = 3
i.e. N-1 = 3 so N = 4 ( Even )

Minimum order is 3.

Kiran TALELE 9987030881 talelesir@gmail.com 22

(Q7) Determine the coefficients of High Pass
linear phase FIR filter of length N = 4 which
has frequency response such that, and

H  ( 4) = 1
2

 3 
H  =1
 4 

Kiran TALELE 9987030881 talelesir@gmail.com 23

Solution :
HPF Linear Phase FIR filter
Length N = 4
Given frequency response :
H  ( 4) = 1
2
 3 
H
 4 
 =1

For High Pass Filter with N = 4 (even)

h[n] must be Anti-symmetric.

Kiran TALELE 9987030881 talelesir@gmail.com 24

Let h [n] = { h0, h1, –h1, –ho }

By ZT, H (z) = h0 + h1 z-1 – h1 z–2 – ho z–3

Put z = ejw ,

H (ejw) = h0 + h1 e–jw – h1 e–j2w – ho e–j3w

Kiran TALELE 9987030881 talelesir@gmail.com 25

 3
−j w  + j3w 1
j w − j
1
w − j
3
w
H (e jw ) = e 2
 0
h e 2 +he 2 −he 2 −h e 2
1 1 0 
 
 3   1 
3
−j w
H ( w) = e 2
2 j h0 sin  2 w  − 2 j h1 sin  2 w) 
    
3
−j w  3   1 
H ( w) = e 2
2 h0 sin  2 w  − 2 h1 sin  2 w 
    

 3   1 
| H ( w) |= 2 h0 sin  w  − 2 h1 sin  w 
 2   2 

Kiran TALELE 9987030881 talelesir@gmail.com 26

• (i) At w= 4
   3   
H   = 2 h0 sin   − 2 h1 sin  
4  8  8

½ = 1.85 h0 + 0.75 h1 __(I)

3
(ii) At w=
4
   9   3 
H   = 2 h 0 sin   − 2h1 sin  
4  8   8 

1 = – 0.765 ho + 1.85 h1 ------(II)

Kiran TALELE 9987030881 talelesir@gmail.com 27

 ½ = 1.85 h0 + 0.75 h1 __(I)
1 = – 0.765 ho + 1.85 h1 ------(II)

28
Kiran TALELE 9987030881 talelesir@gmail.com 28
Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 30

30
 FIR Filter Design using
Windowing Method
@
Sardar Patel Institute of
Technology, Andheri, Mumbai
Kiran Tulshiram Vasumati TALELE

Academic Year 2020-21

Bharatiya Vidya Bhavans’
Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
• Associate Professor, Electronics Engineering Department [1997]
• Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
• Convenor, SP-IT Innovation Council [2018]
@ Sardar Patel Technology Business Incubator(SP-TBI),
Funded by Department of Science & Technology(DST), Govt. of India
• Head, Academic Relations [2016]
• Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
• Treasurer [2020]
• Executive Committee Member [2015] 2
 Digital FIR Filter Design
using Windowing
Method

Kiran Tulshiram Vasumati TALELE

 ALGORITHM To Design Linear Phase F I R
Filter Using Window function.

Given ➔ W[n] N Wc Filter

Step-1 Desired Frequency

Hd(w) Response

Desired Impulse
Step-2 hd(n) Response

Step-3 h(n) Impulse Response

Kiran TALELE 9987030881 talelesir@gmail.com 4

 Examples of Window function :
(1) Rectangular Window

 1 0  n  N −1
w[n] = 
 0 otherwise
(2) Bartlet Window

 2n /( N − 1) 0  n  ( N − 1) / 2
w[n] = 2 − 2n /( N − 1) ( N − 1) / 2  N − 1
 0 otherwise
(3) Hanning Window

  2n 
 1 − cos    /2 0  n  N −1
w[n] =   N − 1 

 0 otherwise

Kiran TALELE 9987030881 talelesir@gmail.com 5

 (4) Hamming Window

  2n 
−   0  n  N −1
w[n] = 
0.54 0.46 cos
  N −1
 0 otherwise

(5) Blackman Window

  2n   4n 
0.42 − 0.5 cos   + 0.08 cos   0  n  N −1
w[n ] =   N −1  N −1
 0 otherwise

Kiran TALELE 9987030881 talelesir@gmail.com 6

 Characteristics of Window function :
Window Function Constant As (dB)
1 Rectangular C = 0.92 21
2 Bartlett C = 2.10 25
3 Hanning C = 3.21 44
4 Hamming C = 3.47 53
5 Blackman C = 5.71 74

Ref. “Digital Signal Processing : A Modern Introduction”

by Ashok Ambardar,
Cengage Learning India Edition 2007. Page 476

NOTE :
In FIR Filter, Order M = N-1
In IIR Filter, Order N = No of POLEs

Kiran TALELE 9987030881 talelesir@gmail.com 7

 Q(3) Given,
 1 0  F  800 Hz
H(w) = 
 0 otherwise
Assume Sampling frequency F = 5000 Hz.
Design FIR filter for length N = 5 using Hamming window.

Solution :

Now,

Kiran TALELE 9987030881 talelesir@gmail.com 8

 Step-1 : Find Hd(w)

Frequency
Response
Hd(w) = |Hd(w)| Ф(w)

Magnitude
Magnitude Response : Response

Where, wc =0.32 
PB SB
- -wc 0 wc 

Kiran TALELE 9987030881 talelesir@gmail.com 9

 Phase Response :
For a Linear phase FIR Filter with symmetric h(n),
.

N=5

By substituting in Hd(w) we get,

Kiran TALELE 9987030881 talelesir@gmail.com 10

 (II) Find hd[n]

1
hd [n] = 
jw jnw
Hd ( e ) e dw
2 −

1  wc

hd [n] =   e − j  w jnw
e dw
2 − wc 

1  wc

hd [n] =  e j ( n − ) w
dw
2 − wc 


1  e j ( n − ) w
 
wc

hd [n] =   
2  (n −  ) j − w 
Where wc =0.32  and  =2
 c 

Kiran TALELE 9987030881 talelesir@gmail.com 11

Step-3 : Find h[n]
h[n] = hd[n] w[n]
 wc sin ((n −  ) wc )
h[n] = 
  (n −  ) wc ) 

  2n 
 0.54 − 0.46 cos  
  4 

 n=0   n=0   n=0 

     
     
h[n] =     ==  
     
     
     

Kiran TALELE 9987030881 talelesir@gmail.com 12

Step-3 : Find h[n]
h[n] = hd[n] w[n]
 wc sin ((n −  ) wc )
h[n] =  
  ( n −  ) wc ) 

  2n 
 0.54 − 0.46 cos  
  4 

 n=0   n=0   n=0 

     
     
h[n] =     ==  
     
     
     

Kiran TALELE 9987030881 talelesir@gmail.com 13

Step-3 : Find h[n]
h[n] = hd[n] w[n]
 wc sin ((n −  ) wc )
h[n] = 
  (n −  ) wc ) 

  2n 
 0.54 − 0.46 cos  
  4 

 n=0   n=0   n=0 

     
     
h[n] =     ==  
     
     
     

Kiran TALELE 9987030881 talelesir@gmail.com 14

 Q(2) A filter is to be designed with the
following desired frequency response,
below-

 0 −
 w  4
H (e ) =  − j 2 w
jw 4

 e otherwise

Determine the filter coefficients using

Hamming window.

Kiran TALELE 9987030881 talelesir@gmail.com 15

 • Phase :  = −2w
For a Linear phase FIR Filter with symmetric h(n),
.

N=5
(I) Find Hd(w)

Where wc =0.25  and  =2

Kiran TALELE 9987030881 talelesir@gmail.com 16

 (II) Find hd[n]

− wc  
1  − j  w jnw − j  w jnw 
2   
hd [n] = e e dw + e e dw

 −  w c 

 j ( n − ) w − wc  j ( n − ) w  
1  e  e  
hd [n] =   +   
2  (n −  ) j −
  (n −  ) j 

 wc 

.......
.......
Kiran TALELE 9987030881 talelesir@gmail.com 17
  sin[( n −  ) ] wc sin ((n −  ) wc ) 
hd [n] =  − 
 [( n −  ) ]  [( n −  ) wc ] 

(III) Find h[n]

h[n] = hd[n] w[n]

 sin[( n −  ) ] wc sin ((n −  ) wc )

hd [n] =  − 
 [( n −  ) ]  [( n −  ) wc ] 

Kiran TALELE 9987030881 talelesir@gmail.com 18

 To Find h[n]
   
 sin[( n − 2) ] 1 sin  ( n − 2)    2n 

0.54 − 0.46 cos  
4
h[ n] =  − 
 [(n − 2) ] 4 [(n − 2) 4 ] 

  4 
 

 − 0.159 n = 0   0.08 n = 0   − 0.013 n=0 

 − 0.225   0.54   − 0.122 
     
h[n] =  0.75   1  ==  0.75 
     
 − 0.225   0.54   − 0.122 
 − 0.159   0.08   − 0.013 

Kiran TALELE 9987030881 talelesir@gmail.com 19

Q(3) Given Ap = 1 dB As = 40 dB
Wp = 0.2 
Ws = 0.8 
Design a digital filter using appropriate window
function.

Solution: Given Ap = 1 dB As = 40 dB
Wp = 0.2  Ws = 0.8 
window function ➔ Linear Phase FIR Filter

Kiran TALELE 9987030881 talelesir@gmail.com 20

 ALGORITHM To Design Linear Phase F I R
Filter Using Window function.

Given ➔ W[n] N Wc Filter

Step-1 Desired Frequency

Hd(w) Response

Desired Impulse
Step-2 hd(n) Response

Step-3 h(n) Impulse Response

Kiran TALELE 9987030881 talelesir@gmail.com 21

 ALGORITHM To Design Linear Phase F I R
Filter Using Window function.
Given ➔ Ap As wp ws

W[n] N Wc Filter

Desired Frequency
Step-1 Hd(w) Response

Desired Impulse
Step-2 hd(n) Response

Step-3 h(n) Impulse Response

Kiran TALELE 9987030881 talelesir@gmail.com 22

Solution :
(1) Select Window Function (2) Calculate N

Given As = 40 dB
For,
f2 = 0.4 and f1 = 0.1
Hanning window = 44 dB
For Hanning window,
Hamming window = 53 dB C = 3.21
Blackman window = 74 dB

Select,
Hanning window = 44 dB

Kiran TALELE 9987030881 talelesir@gmail.com 23

 (3) Calculate Wc (4) Identify Filter

| H(e jw ) |

PB SB

0 0.2 0.8 
wp ws

Filter is LPF

Kiran TALELE 9987030881 talelesir@gmail.com 24

 Step-1 : Find Hd(w)

Frequency
Response
Hd(w) = |Hd(w)| Ф(w)

Magnitude Response :

| H(e jw ) |
 1 - wc  w  wc
H(w) = 
 0 otherwise
PB SB
Where, wc = 0.5 
- -wc 0 wc 

Kiran TALELE 9987030881 talelesir@gmail.com 25

 Phase Response :

For a Linear phase FIR Filter with symmetric h(n),

.

By substituting in Hd(w) we get,

Where wc =0.5  and  =5

Kiran TALELE 9987030881 talelesir@gmail.com 26

(II) Find hd[n]

1   e j ( n − ) w

wc

hd [n] =   
2  (n −  ) j − w 
 c 

 wc sin ((n −  ) wc ) 
wc
 h[n] =  
1  − 
hd [n] =  e − j  w jnw
e dw  ( n ) wc ) 
2 − wc 
Where wc =0.5  and  =5

1  wc

hd [n] =  e j ( n − ) w
dw
2 − wc 

Kiran TALELE 9987030881 talelesir@gmail.com 27

Step-3 : Find h[n]
h[n] = hd[n] w[n]
 wc sin ((n −  ) wc )
h[n] = 
  (n −  ) wc ) 

 n=0   n=0   n=0 

     
     
h[n] =     ==  
     
     
     

Kiran TALELE 9987030881 talelesir@gmail.com 28

Step-3 : Find h[n]
h[n] = hd[n] w[n]
 wc sin ((n −  ) wc )
h[n] = 
  (n −  ) wc ) 

 n=0   n=0   n=0 

     
     
h[n] =     ==  
     
     
     

Kiran TALELE 9987030881 talelesir@gmail.com 29

Step-3 : Find h[n]
h[n] = hd[n] w[n]
 wc sin ((n −  ) wc )
h[n] = 
  (n −  ) wc ) 

 n=0   n=0   n=0 

     
     
h[n] =     ==  
     
     
     

Kiran TALELE 9987030881 talelesir@gmail.com 30

Q(4) Design a causal digital High Pass filter using
windowing technique to meet the following
specifications :

Passband edge : = 9.5 KHz

Stopband edge : = 2 KHz
Stopband Attenuation :  50 dB
Sampling frequency : = 25 KHz

Solution: Given As = 40 dB
Wp = 0.76  Ws = 0.167  HPF
window function ➔ Linear Phase FIR Filter

Kiran TALELE 9987030881 talelesir@gmail.com 31

Q(5) Design sixth order Linear Phase Band Pass FIR filter with
pass-band frequencies w1 = 0.25  and w2 = 0.65
using Rectangular window function.

Solution :
Order M = N-1 ==6
Therefore, N=7 and so,  = 3

• Band Pass Filter, with w1 = 0.25 and w2 = 0.65

• Rectangular window function.

Kiran TALELE 9987030881 talelesir@gmail.com 32

 Step-1 : Find Hd(w)

Frequency
Response
Hd(w) = |Hd(w)| Ф(w)

Magnitude Response : Band Pass Filter,

with pass band
|Hd(w)|
frequencies :

SB PB SB
• w1 = 0.25 and
• w2 = 0.6
- -w2 -w1 0 w1 w2  w

Kiran TALELE 9987030881 talelesir@gmail.com 33

 Magnitude Response :
|Hd(w)|

SB PB SB

- -w2 -w1 0 w1 w 2  w

Kiran TALELE 9987030881 talelesir@gmail.com 34

 Phase Response :
For a Linear phase FIR Filter with symmetric h(n),
.

By substituting in Hd(w) we get,

Kiran TALELE 9987030881 talelesir@gmail.com 35

 (II) Find hd[n]

.......
.......
Kiran TALELE 9987030881 talelesir@gmail.com 36
Kiran TALELE 9987030881 talelesir@gmail.com 37
Step-3 : Find h[n]

h[n] = hd[n] w[n] Where w[n] is Rectangular Window

is such that,

 1 0  n  N −1
w[n] = 
 0 otherwise

Kiran TALELE 9987030881 talelesir@gmail.com 38

Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 40

40
 FIR Filter Design using FSM
@
Sardar Patel Institute of Technology,
Andheri, Mumbai
Kiran Tulshiram Vasumati TALELE

Academic Year 2020-21

Bharatiya Vidya Bhavans’

Sardar Patel Institute of Technology Andheri(w) Mumbai
K. T. V. Talele
@ Bharatiya Vidya Bhavans’ Sardar Patel Institute of Technology
Andheri(w) Mumbai
•  Associate Professor, Electronics Engineering Department [1997]
•  Incharge, Innovation & Entrepreneurship Development
Centre(IEDC) Developed by DST-Govt. of India [2010]
•  Convenor, SP-IT Innovation Council [2018]

@ Sardar Patel Technology Business Incubator(SP-TBI),

Funded by Department of Science & Technology(DST), Govt. of India
•  Head, Academic Relations [2016]
•  Chief Finance Officer (CFO) [2017]
@ IEEE Bombay Section
•  Treasurer [2020]
•  Executive Committee Member [2015] 2
 ALGORITHM To Design Linear Phase F I R
Filter Using Frequency Sampling Method

N Wc Filter

Hd(w)

H[k]

h(n)
3
Q1 The frequency response of Low Pass Filter is
given by,
jw ⎧ e − j 3w 0 ≤ w ≤ 0.5π
H (e ) = ⎨
⎩ 0 0.5π ≤ w ≤ π
Design the filter using Frequency Sampling Method.
Solution :
Given, ⎧ e − j 3w 0 ≤ w ≤ 0.5π
H (e jw ) = ⎨
⎩ 0 0.5π ≤ w ≤ π

Kiran TALELE 9987030881 talelesir@gmail.com 4

 •  Phase : φ = −3w
For a Linear phase FIR Filter with symmetric h(n),
.
⎛ N −1 ⎞
φ = −⎜ ⎟ w = −3w
⎝ 2 ⎠

N=7

(I) Find H[k]

jw ⎧ e − j 3w 0 ≤ w ≤ 0.5π
H (e ) = ⎨
⎩ 0 0.5π ≤ w ≤ π

Kiran TALELE 9987030881 talelesir@gmail.com 5

 2π k 2π k
Put w = =
N 7
⎛ 2π k ⎞
⎧ −j 3⎜ ⎟
0 ≤ w ≤ wc = 0.5π
⎪ e ⎝ 7 ⎠
H [k ] = ⎨
⎪⎩ 0 otherwise

H [k ] = ⎡ 1 k =0 w=0 ⎤
6π
⎢ −j ⎥
⎢ e 7
k =1 w = 0.28π ⎥
⎢ 0 k =2 w = 0.56π ⎥
⎢ ⎥
⎢ 0 k =3 w = 0.84π ⎥
⎢ 0 k =4 ⎥
⎢ ⎥
⎢ 0 k =5 ⎥
6π
⎢ j ⎥
⎢⎣ e 7
k =6 ⎥⎦
Kiran TALELE 9987030881 talelesir@gmail.com 6
 (II) Find h[n]
By Inverse DFT,
N −1
1
h[n] = ∑ H[k ] WN
−nk

N n=0

Kiran TALELE 9987030881 talelesir@gmail.com 7

 Q2. Design sixth order Linear Phase HPF with cutoff
frequency Wc = 0.65 π using Frequency
Sampling Method.

Solution :

Linear Phase HPF Design

Order N-1 = 6
N=7

Cutoff frequency Wc = 0.65 π

Kiran TALELE 9987030881 talelesir@gmail.com 8

 Step-1 : Find Hd(w)
Frequency
Response
Hd(w) = |Hd(w)| Ф(w)

Magnitude Response :
|Hd(w)|

⎧⎪ 0 0 ≤ w < 0.65π ⎫⎪
SB PB | H d (w) |= ⎨ ⎬
⎪⎩ 1 0.65π ≤ w ≤ π ⎪⎭
0 0.65π π

Kiran TALELE 9987030881 talelesir@gmail.com 9

Phase Response :
For a Linear phase FIR Filter with symmetric h(n),
⎛ N −1 ⎞
φ = −⎜ ⎟ w = −3w .

⎝ 2 ⎠
jφ − j 3w
φ (w) = e =e
By substituting in Hd(w) we get,

⎧⎪ 0 0 ≤ w < 0.65π ⎪⎫
H d (w) = ⎨ − j 3w ⎬
⎪⎩ e 0.65π ≤ w ≤ π ⎪⎭
 (II) Find H[k]

By Frequency Sampling,

2π k 2π k
Put w = =
N 7
⎧
⎪ 0 ⎫
0 ≤ w < 0.65π ⎪
H[k]= ⎨ -j3 2πk ⎬
( 7 )
⎩ e
⎪ 0.65π ≤ w ≤ π ⎪
⎭
2π k
w =
7
= 0.285π k ( )
Kiran TALELE 9987030881 talelesir@gmail.com 11
 ⎧
⎪ 0 ⎫
0 ≤ w < 0.65π ⎪
H[k]= ⎨ -j3 2πk ⎬
( 7 )
⎩ e
⎪ 0.65π ≤ w ≤ π ⎪
⎭
H[k] = ⎡ ⎤
0 k=0 w=0
⎢ ⎥
⎢ 0 k =1 w = 0.285π ⎥
⎢ 0 k =2 w = 0.571π ⎥
⎢ ⎛ 18π ⎞ ⎥
− j⎜ ⎟
⎢ ⎝ 7 ⎠ ⎥
⎢ e k =3 w = 0.857π ⎥
⎛ 18π ⎞
⎢ j⎜ 7 ⎟ ⎥
⎢ e ⎝ ⎠
k =4 ⎥
⎢ 0 k =5 ⎥
⎢ ⎥
⎢⎣ 0 k =6 ⎥⎦
 (III) Find h[n]
By Inverse DFT,

N −1
1
h[n] = ∑ H[k ] WN
−nk

N n=0

Kiran TALELE 9987030881 talelesir@gmail.com 13

Q5 The frequency response of Low Pass Filter is
given by,
jw ⎧ e − j 3w 0 ≤ w ≤ 0.5π
H (e ) = ⎨
⎩ 0 0.5π ≤ w ≤ π
Realize the filter using Frequency Sampling Method.
Solution :
⎧ e − j 3w 0 ≤ w ≤ 0.5π
H (e jw ) = ⎨
⎩ 0 0.5π ≤ w ≤ π

Kiran TALELE 9987030881 talelesir@gmail.com 14

 •  Phase : φ = −3w
For a Linear phase FIR Filter with symmetric h(n),
.
⎛ N −1 ⎞
φ = −⎜ ⎟ w = −3w
⎝ 2 ⎠

N=7

(I) Find H[k]

jw ⎧ e − j 3w 0 ≤ w ≤ 0.5π
H (e ) = ⎨
⎩ 0 0.5π ≤ w ≤ π

Kiran TALELE 9987030881 talelesir@gmail.com 15

 2π k 2π k
Put w = =
N 7
⎛ 2π k ⎞
⎧ −j 3⎜ ⎟
0 ≤ w ≤ wc = 0.5π
⎪ e ⎝ 7 ⎠
H [k ] = ⎨
⎪⎩ 0 otherwise

H [k ] = ⎡ 1 k =0 w=0 ⎤
6π
⎢ −j ⎥
⎢ e 7
k =1 w = 0.28π ⎥
⎢ 0 k =2 w = 0.56π ⎥
⎢ ⎥
⎢ 0 k =3 w = 0.84π ⎥
⎢ 0 k =4 ⎥
⎢ ⎥
⎢ 0 k =5 ⎥
6π
⎢ j ⎥
⎢⎣ e 7
k =6 ⎥⎦
Kiran TALELE 9987030881 talelesir@gmail.com 16
 (II) Frequency Sampling Realization
•  By Freq Sampling, H ( z) = N1 H 1 ( z) H 2 ( z)

Where (1) N = 7
(2) H1 (z) = 1 − z − N == 1 − z −7
N −1
H[k]
(3) H 2 (z) = ∑ j2 π k
k =0
1− e N
z −1
H [0] H [1] H [6]
H 2 ( z) = + +
1 − z −1 j
2π
j
12π
1− e 7 z −1 1− e 7 z −1
6π 6π
−j j
1 e 7
e 7
H 2 ( z) = −1
+ 2π
+ 2π
1− z j −j
1− e 7
z −1 1 − e 7
z −1
Kiran TALELE 9987030881 talelesir@gmail.com 17
 6π 6π
−j j
1 e 7
e 7
H 2 ( z) = −1
+ 2π
+ 2π
1− z j −j
1− e 7
z −1 1 − e 7
z −1

Kiran TALELE 9987030881 talelesir@gmail.com 18

 1 ⎡ − 1.8019 + 1.8019 z −1 ⎤
H 2 ( z) = −1
+ ⎢ −1 −2 ⎥
1− z ⎣ 1 − 1.246 z + z ⎦
X [n] 1/7 y [n]

z–1 z–1

z–1

z–1 -1.8019
z–1
z–1
z–1 1.246
z–1
z–1 1.8019
z–1
-1
-1

Kiran TALELE 9987030881 talelesir@gmail.com 19

Stay Connected……..

Kiran TALELE
9987030881
ktvtalele@spit.ac.in

www.facebook.com/Kiran-Talele-1711929555720263

Kiran Tulshiram Vasumati TALELE 21
21
 Multi Rate
Signal
Processing
Kiran TALELE
talelesir@gmail.com
MSD By TALELE 1
Down-sampling using Decimator
Down sampling reduces sampling
frequency of input signal.

Decimator consists of Antialiazing

LPF with cuoff frequency wc=π/D
followed by Down-sampler.

Kiran TALELE 99870 30 881 talelesir@gmail.com

 Down-sampler
x[n] Antializing a[n] y[n]
D
LPF H(w)
Fs=Fsx Fs=Fsx Fs=Fsy=Fsx/D

Example :
a[n] = { 1, 2, 3, 4, 5, 6, 7, 8 }
For D=2, y[n] = { 1, 3, 5, 7 }= a[nD]
For D=3, y[n] = { 1, 4, 7 }
Kiran TALELE 99870 30 881 talelesir@gmail.com
INPUT OUTPUT

Now, W=2π f Now, W=2π f

W = 2π F/Fs W = 2π F/Fs

For input x[n] For output y[n]

W = Wx W = Wy
Fs = Fsx Fs = Fsy = Fsx/D

Wx = 2π F/Fsx Wy = 2π (F/Fsx) D

Wy = Wx D
 Frequency Domain Representation for D=2

| X(w) |

-π π
| H(w) |

-π -π/2 π/2 π a[n] = x[n] *h[n]

| A(w) |
A(w)=X(w) H(w)
-π/2 π/2
-π π
| Y(w) | Wy = 2 Wx

-π π

Kiran TALELE 99870 30 881 talelesir@gmail.com

 Example :

xa [t] x[n] w[n] v[n] y[n]

ADC X H(w) 10

Cos(0.8πn)

Given Fs = 2500

(a) Obtain the spectrum of x[n] and y[n]

 Solution

xa (t) x[n]
ADC
Fs =2500

|Xa(w)|

-1100 Hz -900Hz 900 Hz 1100Hz

|X(w)|

− w2 − w1 w1 w2

Kiran TALELE 99870 30 881 talelesir@gmail.com

 ⎛ 900 ⎞ ⎛ 1100 ⎞
w1 = 2π ⎜ ⎟ = 0.72π w2 = 2π ⎜ ⎟ = 0.88π
⎝ 2500 ⎠ ⎝ 2500 ⎠

|X(w)|

− 0.88π − 0.72π 0.72π 0.88π

Kiran TALELE 99870 30 881 talelesir@gmail.com

 xa [n] x[n] w[n]
ADC X

Cos(0.8πn)

W[n]=x[n] Cos(0.8πn)
⎛ e j 0.8πn + e − j 0.8πn ⎞
w[n] = ⎜ ⎟ x[n]
⎜ 2 ⎟
⎝ ⎠
1⎛ j 0.8πn − j 0.8πn ⎞
w[ n] = ⎜ x[ n] e + x[ n] e ⎟
2⎝ ⎠
1
ByDTFT , W ( w) = ( X ( w + 0.8π ) + X ( w − 0.8π ) )
2

Kiran TALELE 99870 30 881 talelesir@gmail.com

 W (w)

−π − 0.48π − 0.32π − 0.08π 0.08π 0.32π 0.48π π

Kiran TALELE 99870 30 881 talelesir@gmail.com

 xa [n] x[n] w[n] v[n]
ADC X H(w)

Cos(0.8πn)
|W(w)|

−π − 0.48π − 0.32π − 0.08π 0.08π 0.32π 0.48π π

|H(w)|

−π 0.1π π
− 0.1π
|V(w)|

−π − 0.08π 0.08π π
xa [n] x[n] w[n] v[n] y[n]
ADC X H(w) 10

Cos(0.8πn)

|V(w)|

− 0.08π 0.08π π
-π

| Y(w) |

-π − 0.8π 0.8π π

Kiran TALELE 99870 30 881 talelesir@gmail.com

 (ii)  Show that it is possible to obtain
y[n] by sampling Xa(t) with period
T = 4 mili second

Solution :
Ts = 4 msec
Fs = 250 Hz

Kiran TALELE 99870 30 881 talelesir@gmail.com

 xa (t) Y[n]
Sampling

|Xa(w)|

-1100 Hz -900Hz 900 Hz 1100Hz

|Y(w)|

− w2 − w1 w1 w2

Kiran TALELE 99870 30 881 talelesir@gmail.com

 ⎛ 900 ⎞
w1 = 2π ⎜ ⎟ = 7.2π w1 == − 0.8π
⎝ 250 ⎠
⎛ 1100 ⎞
w2 = 2π ⎜ ⎟ = 8.8π w2 == 0.8π
⎝ 250 ⎠

|Y(w)|

− 0.8π 0.8π
Kiran TALELE 99870 30 881 talelesir@gmail.com
 Up-sampling using Interpolator

Up sampling increases sampling

frequency of input signal.

Interpolator consists of Upsampler

followed by Anti-Imaging LPF with
cuoff frequency wc=π/L

Kiran TALELE 99870 30 881 talelesir@gmail.com

 x[n] a[n] Anti-Imaging y[n]
L
LPF
Fs=Fsx Fs=L Fsx
Fs=Fsy=L Fsx

Example :
a[n] = { 1, 2, 3, 4 }
For L=2, y[n] = { 1, 0, 2, 0, 3, 0, 4, 0, }
For L=3, y[n] = { 1, 0,0, 2, 0, 0, 3, 0,0, 4, 0,0, }

Kiran TALELE 99870 30 881 talelesir@gmail.com

INPUT OUTPUT

Now, W=2π f Now, W=2π f

W = 2π F/Fs W = 2π F/Fs

For input x[n] For output y[n]

W = Wx W = Wy
Fs = Fsx Fs = Fsy = L Fsx

Wx = 2π F/Fsx Wy = 2π (F/L Fsx)

Wy = Wx / L
 Frequency Domain Representation for L=2

| X(w) |

-3π -π π 3π
| A(w) |
-π -π/2 π/2 π
| H(w) |

-π -π/2 π/2 π
| Y(w) | Y(w)=X(w) H(w)

-π -π/2 π/2 π

Kiran TALELE 99870 30 881 talelesir@gmail.com

 Multi stage approach to sampling rate
conversion.

(1) Multi stage approach for Interpolator :

N
L = ∏ Li == L1 L2
i =1

X(n) á L LPF á L2 LPF y(n)

1

H1(z) H2(z)

Fs = Fx Fs = L1 Fx Fs = L1 L2 Fx
 (2) Multi stage approach for Decimator
N
D = ∏ D i == D1D 2
i =1

X[n] LPF âD1

LPF âD2
H1(z) H2(z)
Fs=Fsx
Fs=Fsx/D1 Fs=F
sx /D1D
2
 Sampling Rate Conversion by Non-
integer factor

X(n) á L LPF LPF y(n)

H2(z) âD
H1(z)

Fs = Fx Fs=L Fsx Fs=(L/D) Fsx

Fs=(441/480) Fsx
 Explain Subband Coding
Sub-band coding is a method
where the speech signal is
subdivided into several
frequency bands and each band
is digitally encoded separately.

Kiran TALELE 99870 30 881 talelesir@gmail.com

 Sub-band coding filter bank consists of
Analysis filter bank at encoder side and
Synthesis filter bank at decoding side.

The analysis filter bank consists of M

filters each of them having the same
bandwidth.

Kiran TALELE 99870 30 881 talelesir@gmail.com

ENCODER
X(n) LPF M=4
Coding A1

BPF M=4 Coding A2

BPF M=4 Coding A3

BPF M=4 Coding A4

Kiran TALELE 99870 30 881 talelesir@gmail.com

LPF |H1(w)|

0 π/4 π
BPF |H2(w)|
0 π/4 2π/4 π
BPF |H3(w)|

0 π/2 3π/4 π
HPF |H4(w)|
0 3π/4 π
Kiran TALELE 99870 30 881 talelesir@gmail.com
DECODER

A1 Decoding M=4 LPF X(n)

A2 Decoding M=4 BPF

A3 Decoding M=4
BPF

A4 Decoding M=4 HPF

Kiran TALELE 99870 30 881 talelesir@gmail.com

Q(5)Explain Quadrature Mirror Filter

Quadrature Mirror Filter Bank is

a two channel subband filter
bank with complementary
frequency responses.

Kiran TALELE 99870 30 881 talelesir@gmail.com

 It consists of two sections.
1.  Analysis Section and
2.  Synthesis Section

The analysis filter bank consists of two

filters : LPF and HPF with cutoff
frequency

Kiran TALELE 99870 30 881 talelesir@gmail.com

Analysis Filter Bank

X(n) LPF 2
a(n)

0 π/2 π

HPF 2 b(n)

LPF : h0[n]
HPF : h1[n]=(-1)n h0[n] 0 π/2 π

Kiran TALELE 99870 30 881 talelesir@gmail.com

Synthesis Filter Bank

a(n) 2
LPF X(n)

0 π/2 π

b(n) 2 HPF

LPF : g0[n]
HPF : g1[n]=(-1)n g0[n] 0 π/2 π

Kiran TALELE 99870 30 881 talelesir@gmail.com

 Poly Phase Decomposition
Poly Phase structure of a Decimator
for 3 branches :

The Poly Phase decomposition

results in reduction of computation
complexity in filter realization.

Kiran TALELE 99870 30 881 talelesir@gmail.com

 Consider a Digital FIR filter with N=9
point h[n] ,
By ZT,
-1 −2 −3 −4
H(z) = h[0] + z h[1] + z h[2] + z h[3] + z h[4] + +

H(z) = [ h[0] + z −3
h[3] + z −6
h[6] ]
+ [ h(1) z -1 −4
+ z h[4] + z −7
h(7) ]
+ [ h(2) z -2
+z −5
h[5] + z −8
h(8) ]
Kiran TALELE 99870 30 881 talelesir@gmail.com
 H(z) = [ h[0] + z −3
h[3] + z −6
h[6] ]
+z -1
[ h(1) + z −3
h[4] + z −6
h ( 7) ]
+z −2
[ h(2) + z −3
h[5] + z −6
h(8) ]

H(z) = [ E1( z ) ]
3

+z -1
[ E 2( z ) ]
3

+ z −2
[ E3( z ) ]
3

Kiran TALELE 99870 30 881 talelesir@gmail.com

 x[n] y[n]
E1(z3) D=3
z-1
E2(z3 )
z-1
E3(z3 )

Kiran TALELE 99870 30 881 talelesir@gmail.com

 x[n] y[n]
D=3 E1(z)

z-1

D=3 E2(z)
z-1

D=3 E3(z)

Kiran TALELE 99870 30 881 talelesir@gmail.com

Adaptive Filter

By
Kiran Kumar
TALELE
talelesir@gmail.com
Adaptive Filter as a Noise Canceller
dk=sk+nk

Digital yk ek
Xk +
Noise Filter - Noise

Adaptive
Algorithm
 Adaptive Digital Filter

• An Adaptive Filter is a Digital

Filter which adapts automatically
to changes in its input signals.
• Two signals dk and xk are applied
simultaneously to the adaptive filter.

• The signal dk is contaminated signal

containing both the desired signal Sk
and additive noise nk.
• Xk is processed by the Digital filter to
produce an estimate of nk.

• An estimate of the desired signal is

obtained by subtracting the digital
filter output yk from the
contaminated signal dk.
 ek = dk – yk

ek = sk + nk – yk

The objective in noise cancelling is to produce

an estimate of the noise in the contaminated
signals
. i.e. an optimum estimate of the desired signal.
The output signal serves ek
ŝ k
serves tw o purposes
(i) as an estimate of the
desired signal
(ii) as an error signal w hich
is used to adjust the filter
coef f s.
Weiner Filter
Desired
Response:
Input: x(n) Filter d(n)
Output: + Estimation
H(z) y(n) Error: e(n)
Linear Filter _ 

The basic concept behind “Wiener Filter”

theory is to minimize the difference
between the filtered output and some
desired output.
• This minimization is based on the
least mean square approach which
adjusts the filter coefficients to
reduce the square of the difference
between the desired and actual
waveform after filtering.
 Desired
Response:
Based on the basic Input: x(n) Filter d(n)
configuration of the Wiener Output: + Estimation
H(z) y(n) Error: e(n)
Filter: Linear Filter _ 

• The impulse response of winner filter is given

by, W[n] = { w[0], w[1], w[2], w[3], . . . . .w[m] }

Where M is the length of the FIR filter.

In fact, it is the sum of e(n)2 which is
minimized:
• The estimation error e[n]= d[n]-y[n]

• Where, y[n]  W T X

The purpose of the filter is to produce an

estimate of the dersired response d[n].

To optimize the filter design, minimize the

MSE value of e[n],

Now, e[n]= d[n] - y[n]

e[n]  d[n] - W T Xn
  xn   w[0]
x   w[1] 
x[n]   n 1 
w 
 xn  2   w[ 2]
   
     

• The square of the error is given by,

e 2 [n]  d[n] - 2 d[n] X T W  W T X X T W

J  E(e[n])
d[n]
2
- 2 E ( d[n] X T W )  E ( W T X X T W)

 2  E(d 2 [n]) is variance of d[n]

P  E(d[n] X ) is N pt cross correlatio n vector
T T

R  E( X X T ) is the NxN autocorrelation matrix.

• By substituting we get,

J  2
2P T
W W T
R W
• The gradient is given by,

dJ
  2 P  2 R W  0
dW

 2 P  2 R W  0
2 R W  2 P
R W P
R W  P

WR -1
P
Woptimum  R P -1
 Drawback of wiener filter
1). It requires autocorrelation matrix R and
Cross correlation matrix P
2) It involves Matrix inversion which is time
consuming.
3) If the signals are non stationary, then
both R and P will change with time,
so Wopt will have to be computed
repeatedly.
Wiener Filter as a Noise canceller.

+
Σ
 
y[k] = s[k] +n[k] - ek  y[k ] n[k ]  s[k ]
(signal + noise)
(output = signal estimate)

Wiener Filter
 N 1
n[k ]   w[i ]x[ k  i]
x[k] (noise)

i 0

(noise estimate)