Ideal Filters

One of the reasons why we design a filter is to remove disturbances

SIGNAL s (n) x(n) Filter y ( n) s ( n )

v(n)
NOISE

We discriminate between signal and noise in terms of the frequency spectrum

S (F ) Y (F )

V (F )

F0 F0 F
F0 F0 F0 F
 Conditions for Non-Distortion
Problem: ideally we do not want the filter to distort the signal we want to recover.

x(t ) s (t ) y (t ) As (t T ) Same shape as s(t),

IDEAL just scaled and
2
FILTER 2
delayed.
1.5
1.5

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1

-1.5 -1.5

-2 -2
0 200 400 600 800 1000 0 200 400 600 800 1000

Consequence on the Frequency Response:

constant | H (F ) |
Ae j 2FT if F is in the passband
H (F )
F 0 otherwise
linear
H (F )

F
For real time implementation we also want the filter to be causal, ie.

h( n)

h(n) 0 for n 0

since

n
y (n) h(k )x(n k )
k 0
past values only

FACT (Bad News!): by the Paley-Wiener Theorem, if h(n) is causal and with finite energy,

ln H ( ) d

ie H ( ) cannot be zero on an interval, therefore it cannot be ideal.

1 2
log H ( ) log(0) log H ( ) d
1
Characteristics of Non Ideal Digital Filters

| H ( ) | IDEAL

p Positive freq. only

NON IDEAL
Two Classes of Digital Filters:
a) Finite Impulse Response (FIR), non recursive, of the form

y (n) h(0) x(n) h(1) x(n 1) ... h( N ) x(n N )

With N being the order of the filter.
Advantages: always stable, the phase can be made exactly linear, we can approximate any
filter we want;
Disadvantages: we need a lot of coefficients (N large) for good performance;

b) Infinite Impulse Response (IIR), recursive, of the form

y (n) a1 y (n 1) ... a N y (n N ) b0 x(n) b1 x(n 1) ... bN x(n N )

Advantages: very selective with a few coefficients;

Disadvantages: non necessarily stable, non linear phase.
 Finite Impulse Response (FIR) Filters

Definition: a filter whose impulse response has finite duration.

x ( n) y ( n)
h( n)

h( n)




n

h( n) 0 h( n) 0
Problem: Given a desired Frequency Response Hd ( ) of the filter, determine the impulse
response h(n) .

Recall: we relate the Frequency Response and the Impulse Response by the DTFT:

H d ( ) DTFT hd (n) h ( n)e d
j n

1
hd (n) IDTFT H d ( ) d
H ( ) e j n
d
2

Example: Ideal Low Pass Filter

sin ( c n )
c
1 j n
hd (n)
2 d A
Ae
c
n
A c sinc c

n

hd (n)
Hd ( )
DTFT
A

c
4
c c
n
Notice two facts:
the filter is not causal, i.e. the impulse response h(n) is non zero for n<0;

the impulse response has infinite duration.

This is not just a coincidence. In general the following can be shown:

If a filter is causal then

the frequency response cannot be zero on an interval;

h(n)
H( )
H( ) 0
h( n) 0

magnitude and phase are not independent, i.e. they cannot be specified arbitrarily

As a consequence: an ideal filter cannot be causal.

Problem: we want to determine a causal Finite Impulse Response (FIR) approximation of the
ideal filter.
rectangular window hw (n)
We do this by 2 0.25
1.8
0.2
a) Windowing 1.6

1.4 0.15

1.2
0.1

hd (n) 1

0.8 0.05

0.25 0.6
0
0.4
0.2
0.2 -0.05

0.15 0
-0.1

-100 -50 0 50 100

L
-100 -50 0 50 100
0.1
L L L
0.05

0
hamming window hw (n)
-0.05
2 0.25
-0.1
-100 -50 0 50 100 1.8
0.2
1.6

1.4 0.15

infinite impulse response 1.2

0.1

(ideal) 1

0.8 0.05

0.6
0
0.4

0.2 -0.05
0
-100 -50 0 50 100 -0.1
-100 -50 0 50 100

L L L L

finite impulse response

b) Shifting in time, to make it causal:

hw (n) h(n) hw (n L)
0.25
0.25

0.2
0.2

0.15 0.15

0.1 0.1

0.05 0.05

0 0

-0.05 -0.05

-0.1 -0.1
-100 -50 0 50 100 -100 -50 0 50 100
Effects of windowing and shifting on the frequency response of the filter:
1
a) Windowing: since hw (n) hd (n) w(n) then H w ( ) H d ( ) *W ( )
2
| W ( ) | rectangular window | H w ( ) |
40 20

20 0

0
-20

-20
-40

*
-40
-60
-60

H d ( ) -80
-80

-100
-100

-120
-120 0 0.5 1 1.5 2 2.5 3
0 0.5 1 1.5 2 2.5 3

c c 40
hamming window
20

20
0
0

-20

*
-20
-40
-40
-60
-60

-80
-80

-100 -100

-120 -120
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
For different windows we have different values of the transition region and the attenuation in the
stopband:
attenuation
20

Rectangular 4 / N -13dB 0

Bartlett 8 / N -27dB -20

8 / N attenuation
Hanning -32dB -40

Hamming 8 / N -43dB -60

16 / N
-80
Blackman -58dB
-100

-120
with N 2L 1 0 0.5 1 1.5 2 2.5 3

hw (n) 0.25

0.2

transition
0.15

0.1
region
0.05

-0.05

-0.1
-100 -50 0 50 100
n
L L
Effect of windowing and shifting on the frequency response:
b) shifting: since h(n) hw (n L) then H ( ) H w ( )e jL
Therefore
H ( ) H w ( ) no effect on the magnitude,
H( ) H w ( ) L shift in phase.

See what is H w ( ).

For a Low Pass Filter we can verify the symmetry hw (n) hw ( n). Then

H w ( ) h
n
w ( n ) e j n
hw (0) 2 hw (n) cos( n)
n 1

real for all . Then

0 in the passband;
H w ( )
don' t care, otherwise
The phase of FIR low pass filter:

H ( ) L in the passband;
Which shows that it is a Linear Phase Filter. 20

-20

H ( ) -40

-60
dB
-80

-100

-120
0 0.5 1 1.5 2 2.5 3

H ( )
degrees

dont care
Example of Design of an FIR filter using Windows:
Specs: Pass Band 0 - 4 kHz
Stop Band > 5kHz with attenuation of at least 40dB
Sampling Frequency 20kHz

Step 1: translate specifications into digital frequency 40dB

Pass Band 0 2 4 / 20 2 / 5 rad
4 5 10F kHz
2 5 / 20 / 2 rad 2
Stop Band
5 2

Step 2: from pass band, determine ideal filter impulse

10
response

c c 2 2n
hd (n) sinc n sinc
5 5
Step 3: from desired attenuation choose the window. In this case we can choose the hamming
window;

Step 4: from the transition region choose the length N of the impulse response. Choose an odd
number N such that:
8

N 10

So choose N=81 which yields the shift L=40.

Finally the impulse response of the filter

2 2(n - 40) 2 n
sinc 0.54 0.46 cos , if 0 n 80,
h ( n) 5 5 80
0 otherwise

The Frequency Response of the Filter:

H( )

dB

H( )

rad

 A Parametrized Window: the Kaiser Window

The Kaiser window has two parameters:

N Window Length

To control attenuation in the Stop Band

1.5

w[n] 0
1

0.5 5
10
0
0 20 40 60 80 100 120
n
There are some empirical formulas:

Attenuation in dB A N

Transition Region in rad

Example:

2

Sampling Freq. 20 kHz
P , S
Pass Band 4 kHz 5 2
Stop Band 5kHz, with 40dB Attenuation

N 45
S P rad
10 3.3953
A 40dB
Then we determine the Kaiser window

w kaiser ( N , )

w[n]

n
Then the impulse response of the FIR filter becomes

sin ( c (n L) )
h[n] w[n]
(n L)

ideal impulse
response

with c 1
2
( P S ) 9
20

N 45 2 L 1 L 22
h[n]

Impulse Response

| H ( ) |dB

Frequency Response

(rad)
Example: design a digital filter which approximates a differentiator.
Specifications:
Desired Frequency Response:
j 2 F if 4kHz F 4kHz
H d (F )
0 if F 5kHz

Sampling Frequency Fs 20kHz

Attenuation in the stopband at least 50dB.
Solution.
Step 1. Convert to digital frequency

2 2
jFs j 20,000 if - 5 5
H d ( ) H d ( F ) F F / 2
s
0
if | |
2
Step 2: determine ideal impulse response
2
5
1
hd (n) IDTFT H d ( ) j 20,000 e j n d
2 2

5
ax
e 1

ax
From integration tables or integrating by parts we obtain xe dx x
a a
Therefore
2n 2n
cos sin
20,000
4 5 2 5 if n 0
hd (n) 5 n n 2



0 if n 0
Step 3. From the given attenuation we use the Blackman window. This window has a transition
region region of 12 / N . From the given transition region we solve for the complexity N
as follows
2 12
0.1
2 5 N

which yields N 120. Choose it odd as, for example, N=121, ie. L=60.

Step 4. Finally the result

2 (n 60) 2 (n 60)
cos sin
4 5 2 5 0.42 0.5 cos 2n 0.08 cos 4n
h(n) 20,000
5 n 60 (n 60) 2 120 120

for 0 n 120
 Frequency Response
5
x10
H ( )
2

1.8

1.6

Impulse response h(n) 1.4

4
1.2
x10
3 1

0.8
2
0.6

1 0.4

0.2
0 0

0 0.5 1 1.5 2 2.5 3

-1

-2

-3
0 20 40 60 80 100 120 140 150

H ( ) dB 100

50

-50

-100

-150

-200

-250

0 0.5 1 1.5 2 2.5 3 3.5