Filter Design

Methods of Filter Designing

FIR Filters • Windows method

• Frequency Sampling
(Non-recursive)

• Impulse-Invariant
IIR Filters Transformation
(recursive) • Bilinear Transformation
Ideal Filters
lowpass highpass

𝜔𝑐 𝜔𝑐

bandpass bandstop

𝜔𝑐 1 𝜔𝑐 2 𝜔𝑐 1 𝜔𝑐 2
4
Ideal Filters
Frequency response Desired Impulse response

{
𝑠𝑖𝑛 [ 𝜔𝑐 𝑛 ]
𝑛≠0
h 𝑑 [ 𝑛]= 𝜋𝑛
𝜔𝑐
𝑛=0
𝜔𝑐 𝜋
lowpass

{
𝜔𝑐
1− 𝑛=0
h 𝑑 [ 𝑛]= 𝜋
𝑠𝑖𝑛 [ 𝜔 𝑐 𝑛 ]
− 𝑛≠ 0
𝜋𝑛
𝜔𝑐
highpass 5
Ideal Filters
Frequency response Desired Impulse response

{
𝑠𝑖𝑛 [ 𝜔𝑐 2 𝑛 ] 𝑠𝑖𝑛 [ 𝜔 𝑐 1 𝑛 ]
− 𝑛≠0
h 𝑑 [ 𝑛]= 𝜋𝑛 𝜋𝑛
𝜔𝑐 2 − 𝜔 𝑐 1
𝑛= 0
𝜋
𝜔𝑐 1 𝜔𝑐 2
bandpass

{
𝑠𝑖𝑛 [ 𝜔𝑐 1 𝑛 ] 𝑠𝑖𝑛 [ 𝜔 𝑐 2 𝑛 ]
− 𝑛≠
h 𝑑 [ 𝑛]= 𝜋𝑛 𝜋𝑛
𝜔𝑐 2 − 𝜔 𝑐 1
1− 𝑛= 0
𝜋
𝜔𝑐 1 𝜔𝑐 2
bandstop 6
Ideal Filters
Type of Filter Response

LowPass Filter

HighPass Filter

BandPass

BandStop
 Time domain Frequency domain
S = 0.7*sin(2*pi*20*t) + 0.5*sin(2*pi*50*t)+
2*sin(2*pi*70*t) + sin(2*pi*100*t); 2.5
4

3 2

1
1.5
X(t)

1
-1

-2
0.5
-3

-4
0 5 10 15 20 25 30 35 40 45 50 0
t (milliseconds) 0 20 40 60 80 100 120

∗ ×
f (Hz)

¿
 Time domain Frequency domain
S = 0.7*sin(2*pi*20*t) + 0.5*sin(2*pi*50*t)+
2*sin(2*pi*70*t) + sin(2*pi*100*t); 2.5
4

3 2

1
1.5
X(t)

1
-1

-2
0.5
-3

-4
0 5 10 15 20 25 30 35 40 45 50 0
t (milliseconds) 0 20 40 60 80 100 120

∗
f (Hz)

×
1
From Fourier transform, we know

Time domain Frequency domain

 Impulse Response of Ideal Filters
Impulse Response Frequency Response

• Sink functions: infinite in extent

• Impulse response begins before t=0
• Therefore: Ideal filters are non-causal

Ideal filters can not be physically realized but can be closely approximated
11
Causality
A causal system is one whose output at any time depends only on values of the input at the
present time and in the past. It does not depend on future values

𝑦 [ 𝑛 ] = 𝑥 [𝑛 ] 𝑦 [ 𝑛 ] = 𝑥 [𝑛 −1 ] 𝑦 [ 𝑛 ] = 𝑥 [𝑛+ 1]
Causal Causal Non-Causal
 𝑥 (𝑡 ) 3

2
Past Future
1

−3 −2 −1 1 2 3 𝑡
Present
Present, Past and Future Values
3

 Present Past 2

1
Future

 Future
−3 −2 −1 1 2 3
 Past Present

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

𝑦 [ 0 ] = 𝑥 [ 0] 𝑦 [ 0 ] = 𝑥 [ − 1] 𝑦 [ 0 ] = 𝑥 [ +1]
This system uses present This system uses a past This system uses a future
value of the input to value of the input to value of the input to
calculate the output calculate the output calculate the output
 𝑦 [ 𝑛 ] = 𝑥 [𝑛 ] 𝑦 [ 𝑛 ] = 𝑥 [𝑛 −1 ] 𝑦 [ 𝑛 ] = 𝑥 [𝑛+ 1]
This system uses present This system uses a past This system uses a future
value of the input to value of the input to value of the input to
calculate the output calculate the output calculate the output

𝑦 [ 𝑛 ] = 𝛿 [𝑛 ] 𝑦 [ 𝑛 ] =𝛿 [𝑛 − 1] 𝑦 [ 𝑛 ] = 𝛿 [𝑛 +1]
𝛿 [𝑛 ] 𝛿 [𝑛 − 1 ] 𝛿 [𝑛 + 1 ]

0 1 -1

Causal Causal Non-Causal

1 0.8

0.8 0.6

0.6 0.4

0.4 0.2

0.2 0

0 -0.2

-0.2 -0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10

-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10

Non-Causal Causal
• When the output of a system depends on the past and/or future
input, the system is said to have a memory. For example

Present past future

• A memoryless system is one in which the current output depends

only on the current input; it does not depend on the past or future
inputs. e.g.
Causality
A causal system is one whose output at any time depends only on values of the input at the
present time and in the past. It does not depend on future values

𝑦 [𝑛 ]=𝑥 [ 𝑛] − 𝑥 [ 𝑛− 1]+ 𝑥 [𝑛 +1]

Present Delayed Advanced /

/ past future
FIR Filter Design
Using Windows Method
 Windows Method
Problem-1: Sinc signal is infinite 1

Solution: truncate it 0.8

(multiply sinc with a windows of limited length) 0.6

0.4

Problem-2: Sinc signal is non-causal 0.2

(exists on the left side as well)

0

-0.2

Solution: shift it to right side

-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10
 1

0.8

0.6

Explanation 0.4
Desired Impulse Response
0.2

Problem-1: Sinc signal is infinite -0.2

Solution: truncate it
-0.4
-10 Windows (Rectangular)
-8 -6 -4 -2 0 2 4 6 8 10

(multiply sinc with a windows of

limited length to make it finite)

0.8

0.6

0.4
Truncated Impulse Response
0.2

-0.2

-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 1
 1

Explanation Truncated Impulse Response

0.8

0.6

0.4

0.2

Problem-2: Sinc signal is non- 0

causal -0.2

(exists on the left side as well) -0.4

-10 -8 -6 -4 -2 0 2 4 6 8 10

To compute any sample of the

output, we need to know all Delayed & Truncated Impulse Response
samples of the input, both in
past and the future.
1

0.8

0.6

Solution: shift it to right side 0.4

0.2

-0.2

-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10

After Truncation and shifting, we can now implement this system

 1

Explanation Truncated Impulse Response

0.8

0.6

0.4

0.2

Problem-2: Sinc signal is non- 0

causal -0.2
0

(exists on the left side as well) -0.4

-10 -8 -6 -4 -2 0 2 4 6 8 10

To compute any sample of the

output, we need to know all Delayed & Truncated Impulse Response
samples of the input, both in
past and the future.
1

0.8

0.6

Solution: shift it to right side 0.4

0.2

-0.2
0
-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10

After Truncation and shifting, we can now implement this system

 1

Questions? Truncated Impulse Response

0.8

0.6

0.4

0.2

How many samples to delay? 0

-10 0 +10
¿ 𝑁 /2 -0.2

-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10

Delayed & Truncated Impulse Response

1

0.8
𝐷𝑒𝑙𝑎𝑦 =10
0.6

0.4

0.2

-0.2
0 21
-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10

After Truncation and shifting, we can now implement this system

Frequency response of what we designed
10 1

10 0

10 -1

10 -2

10 -3

10 -4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
 Time domain Frequency domain
S = 0.7*sin(2*pi*20*t) + 0.5*sin(2*pi*50*t)+
2*sin(2*pi*70*t) + sin(2*pi*100*t); 2.5
4

3
2

2
1.5
1
X(t)

0 1
-1

-2
0.5

-3
0

∗
-4 0 20 40 60 80 100 120

×
0 5 10 15 20 25 30 35 40 45 50
t (milliseconds) f (Hz)

10 1

10 0

0.8

0.6 10 -1

0.4

10 -2
0.2

0
10 -3

-0.2

10 -4
-0.4 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-10 -8 -6 -4 -2 0 2 4 6 8 10
 Analysis
Multiplying the sinc function with a rectangular windows
1

0.8
Desired Impulse
0.6 Response
0.4

0.2

∗
-0.2

×
-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10

Windows (Rectangular)

Time domain Frequency domain

Analysis
1

Multiplying the sinc function with a rectangular

0.8

windows in time domain means convolving sinc and

0.6

rectangular in frequency domain 0.4

0.2

• The side-lobes are responsible for

0

-0.2

the ripples (in passband and -0.4

-10 -8 -6 -4 -2 0 2 4 6

stopband)
• Width of the main-lobe is equal to
the transition width (in frequency
response)
What if we change the length/order

As the width of main-lobe decreases by increasing M, we can expect lesser transition width
for higher orders
Thank you

30