Lecture

 5:  Analog  Filter  (III)  and  

Sampling
Topics:
– Chebyshev  Filter  Design
– Frequency  Transformations
– Bandpass  Filter  Design
– Highpass Filter  Design
– Bandstop Filter  Design
– Sampling
 Lecture  5:  Analog  Filter  (III)  and  
Sampling
Topics:
– Chebyshev  Filter  Design
– Frequency  Transformations
– Bandpass  Filter  Design
– Highpass Filter  Design
– Bandstop Filter  Design
– Sampling
Chebyshev filters
Normalised Chebyshev lowpass filter:

1
H ( jw) =
1+e 2 Cn (w)
2

| H ( jw ) | | H ( jw ) |

1 1
1 n=6 1 n=7
1+e 2 1+e 2

0 1 w 0 1 w
Chebyshev filters
Normalised Chebyshev lowpass filter:

1
H ( jw) =
1+e 2 Cn (w)
2

( )
Cn (w) = cos n cos-1w , for w <1,
or,
( )
Cn (w) = cosh n cosh -1w , for w >1.

Cn (w) = 2wCn-1(w)-Cn-2 (w), n > 2

( )
C0 (w) = cos 0 ´ cos-1w = cos(0)=1,
( )
C1(w) = cos 1 ´ cos-1w =w,
C2 (w) = 2wC1(w) -C0 (w)= 2w ´w -1= 2w 2 -1,
( )
C3(w) = 2wC2 (w) -C1 (w)= 2w ´ 2w 2 -1 -w = 4w 3 -3w,
etc. à Tables!
Chebyshev filters
Normalised Chebyshev lowpass filter:

1
H ( jw) =
1+e 2 Cn (w)
2

C0 (w) =1,
C1(w) =w,
C2 (w) = 2w 2 -1,
C3(w) = 4w 3 -3w,

2 ì0 when n is odd,

Þ Cn (0) = í
î1 when n is even.
DC gain:
ì 1 when n is odd,
ï
Þ H (0) = í 1 when n is even.
ï
î 1+e
2
Chebyshev filters
Normalised Chebyshev lowpass filter:

1
H ( jw) =
1+e 2 Cn (w)
2

| H ( jw ) | | H ( jw ) |

1 1
1 n=6 1 n=7
1+e 2 1+e 2

0 1 w 0 1 w

Ratio of maximum to
minimum gain in the
passband: r = 1+e 2

or, in dB: rˆ = 20 log10 1+e 2 =10 log10 1+e 2( ) Þ e 2 =10rˆ/10 -1

Chebyshev filters
Normalised Chebyshev lowpass filter:

1
H ( jw) =
1+e 2 Cn (w)
2

| H ( jw ) | | H ( jw ) |

1 1
1 n=6 1 n=7
1+e 2 1+e 2

0 1 w 0 1 w

| H ( jw ) |
1 1
At w =1 Þ H ( j1) = =
1+e 2 r 1
Gp

In dB: -Gˆ p « rˆ
Gs
0 w p w s w
Chebyshev filters
| H ( jw ) | | H ( jw ) |

1 1
1 n=6 1 n=7
1+e 2 1+e 2

0 1 w 0 1 w

Butterworth filters
H ( jw)
ideal (n = ¥)
1 n =1
0.707 n=2
n = 10 n=4

0 1 n=8 w
Chebyshev filters
Normalised Chebyshev lowpass filter:

Transfer function:

In general: Kn Kn
H (s) = = n
Cn¢ (s) s + an-1s n-1 +!+ a1s + a0
Cn¢ (s) à Tables!
ì a0 when n is odd,
ï
K n = í a0 = a0 when n is even.
ï
î 1+e
2
10rˆ / 20

Pole locations: Im

0 Re
Chebyshev filters

Frequency scaling: The normalised filter H (s) has wp =1.

æ s ö
For a different wp : H (s) = H ç ÷
ç w ÷
è p ø
1 1
H ( jw) = Þ H ( jw) =
1+e Cn (w)
2
2
æ w ö
1+e Cn ç ÷
2 2
ç w ÷
è p ø
Determination of the filter order n
| H ( jw ) |

1
Gp

Gs
0 w p w s w
é æ ws öù
Gˆ s (dB) = 20 log10 Gs Þ Gˆ s = 20 log10 H ( jws ) = -10 log10 ê1+e Cn ç ÷ú
2 2

êë ç w ÷ú
è p øû
Chebyshev filters

Determination of the filter order n

é æ öù
2 ws
Gs = 20 log10 H ( jws ) = -10 log10 ê1+e Cn ç ÷ú
ˆ 2

êë ç w ÷ú
è p øû
æ ö
2 ws ˆ
Þ e Cn ç
2 ÷ =10-Gs /10 -1
ç w ÷
è p ø Recall that: e 2 =10rˆ/10 -1
(
Cn (w) = cosh n cosh -1w )
1/ 2
æ æ ö ö é -Gˆ s /10 ù
ç -1ç ws ÷ ÷ 10 -1
Þ cosh n cosh = ê ú
ç ç w ÷ ÷ ê rˆ /10 ú
è è p ø ø ë10 -1û

æ é -Gˆ /10 ù1/ 2 ö

-1
ç 10 s -1 ÷
cosh ç ê ú ÷
ˆ
r /10
ç êë10 -1 úû ÷
Þ n = è ø
cosh -1(ws /w p )
 Lecture  5:  Analog  Filter  (III)  and  
Sampling
Topics:
– Chebyshev  Filter  Design
– Frequency  Transformations
– Bandpass  Filter  Design
– Highpass Filter  Design
– Bandstop Filter  Design
– Sampling
Frequency transformations

Bandpass filters

| H BP ( jw)| |H p ( jw)|

1 1
Gp Gp

Gs Gs
0 w s1 w p1 w p 2 ws 2 w 0 1 w s w
Bandpass filter Prototype  lowpass filter

ìï w p1w p2 -ws12 ws22 -w p1w p2 üï

ws = miní , ý
ïî p2 p1 s1 (w p2 -w p1)ws2 ïþ
(w -w ) w

s2 +w p1w p2
T (s) = ® H BP (s) =H p (T (s))
(w p2 -w p1 ) s
Frequency transformations

Bandpass filters

ìï w p1w p2 -ws12 ws22 -w p1w p2 üï

ws = miní , ý
ïî (w p2 -w p1)ws1 (w p2 -w p1)ws2 ïþ
s2 +w p1w p2 -w 2 +w p1w p2 w 2 -w p1w p2
T (s) =
(wp2 -wp1) s Þ T ( jw) = (w p2 -w p1) jw = j (w p2 -wp1)w
w ®0 Þ T ( jw) ® - j¥

ìï ws12 -w p1w p2 üï w p1w p2 -ws12

w =ws1 Þ Im{T ( jws1) }= Im í j ý = - £ -ws
(w -w
ïî p2 p1 s1 ïþ) w (w p2 -w p1 ) ws1

w p12 -w p1w p2

w =w p1 Þ T ( jw p1) = j
(w p2 -w p1)w p1 = - j1
w p22 -w p1w p2
w =w p2 Þ T ( jw p2 ) = j
(w p2 -w p1)w p2 = j1
ìï ws22 -w p1w p2 üï ws22 -w p1w p2
w =ws2 Þ Im{ T ( jws2 ) }= Im í j ý = ³ws
(w -w
ïî p2 p1 s2 ïþ) w (w p2 -w p1 ) ws2

w ®¥ Þ T ( jw) ® j¥

Frequency transformations

Bandpass filters
ìï w p1w p2 -ws12 ws22 -w p1w p2 üï
ws = miní , ý
ïî p2 p1 s1 (w p2 -w p1)ws2 ïþ
(w -w ) w
s2 +w p1w p2 -w 2 +w p1w p2 w 2 -w p1w p2
T (s) =
(wp2 -wp1) s Þ T ( jw) = (w p2 -w p1) jw = j (w p2 -wp1)w

ws
1
-1
-ws

w s1 w p1 w p 2 ws 2
 | H BP ( jw)|

w s1 w p1 w p 2 ws 2

|H p ( jw)|
w s
1
0
-1
-ws
Frequency transformations

Highpass filters

| H HP ( jw)| |H p ( jw )|

1 1
Gp Gp

Gs Gs
0 w s w p w 0 1 w p /w s w
Highpass filter Prototype  lowpass filter

w p æ w ö
T (s) = ® H HP (s) = H p çç p ÷÷
s è s ø
Frequency transformations

Bandstop filters

| H BS ( jw )| |H p ( jw )|

1 1
Gp Gp

Gs Gs
0 w p1 w s1 ws 2 w p 2 w 0 1 w s w
Bandstop filter Prototype  lowpass filter

ìï (w p2 -w p1)ws1 (w p2 -w p1)ws2 üï

ws = miní , ý
-
ïî s1w 2
+ w w
p1 p 2 ws22 -w p1w p2 ïþ

(w -w p1) s

T (s) = ® H BS (s) =H p (T (s))
p2

s 2 +w p1w p2
 Lecture  5:  Analog  Filter  (III)  and  
Sampling
Topics:
– Chebyshev  Filter  Design
– Frequency  Transformations
– Bandpass  Filter  Design
– Highpass Filter  Design
– Bandstop Filter  Design
– Sampling
Sampling

Brief review of Fourier series

Periodic function: f (t) = f (t +T0 ), for all t.

T0 = period

Exponential Fourier series: ¥

f (t) = å Dn e jnw0 t
n=-¥

2p
w0 = = fundamental frequency
T0

1
Dn = ò f (t) e- jnw0 t dt
T0 T0
Sampling

Brief review of Fourier series

¥
f (t) = å Dn e jnw0 t
n=-¥

Recap: Exercise 2: Show that e jw0t Û 2pd (w -w0 )

1 ¥ 2p
F [2pd (w -w0 )]= ò-¥[2pd (w -w0 )] e jwt dw = e jw0t = e jw0t
-1

2p 2p

¥
F (w) = 2p å Dn d (w - nw0 )
n=-¥
Sampling

Example: Impulse train (‘Dirac comb’)

dT (t )

T -T 0 T 2T t

¥
2p
dT (t) = å d (t - nT ) , ws = = 2pFs
n=-¥ T
1 T /2 1 T /2 1 - jnw ´0 1
Dn = ò dT (t) e - jnws t
dt = ò d (t) e- jnws t dt = e s =
T -T / 2 T -T / 2 T T

1 ¥ jnws t 2p
dT (t) = å e , ws =
T n=-¥ T
Sampling

Example: Impulse train (‘Dirac comb’)

dT (t )

T -T 0 T 2T t

1 ¥ jnws t 2p
dT (t) = å e , ws =
T n=-¥ T

2p ¥ ¥
F [dT (t)]= å d (w - nws ) = ws å d (w - nws )
T n=-¥ n=-¥
Sampling

The sampling theorem

F (w )
f (t ) A

t - 2pB 2pB w
- B 0 B F (Hz)
dT (t )
¥
dT (t) = å d (t - nT )
t n=-¥
T

f (t )
¥
f (t) = f (t) dT (t) = å f (nT ) d (t - nT )
t n=-¥
Sampling

The sampling theorem

¥
f (t) = f (t) dT (t) = å f (nT ) d (t - nT )
n=-¥

Recap: Frequency-convolution property of Fourier transform

1
f1(t) f2 (t) Û F1(w) * F2 (w)
2p

1
f (t) = f (t) dT (t) Û F (w) * F [dT (t)]
2p
2p ¥
F [dT (t)]= å d (w - nws )
T n=-¥

1 2p ¥
F (w) = F [ f (t)] = F (w) * å d (w - nws )
2p T n=-¥
Sampling

The sampling theorem

1 2p ¥
F (w) = F [ f (t)] = F (w) * å d (w - nws )
2p T n=-¥

Recap: Properties of convolution

4. Shift property: if f1(t) * f2(t) = c(t), then f1(t) * f2(t -T ) =c(t -T )
¥
5. Convolution with an impulse: f (t)*d (t) = ò f (t )d (t -t )dt = f (t)
-¥

1 2p ¥ 1 ¥
F (w) = F (w) * å d (w - nws ) = å F (w - nws )
2p T n=-¥ T n=-¥

1 ¥
F (w) = å F (w - nws )
T n=-¥
Sampling

The sampling theorem 1 ¥

F (w) = å F (w - nws )
T n=-¥

F (w )
f (t ) A

t - 2pB 2pB w
- B 0 B F (Hz)
dT (t )

T t F (w )
Lowpass
filter

A/T
f (t )

-w s - 2pB 2pB ws w

t
- Fs - B 0 B Fs F (Hz)
Sampling

The sampling theorem

1 ¥
F (w) = å F (w - nws )
T n=-¥

F (w )
A
ì
ï Fs ³ 2B
w Nyquist - Shannon ï
- 2pB 2pB íor, equivalently,
- B 0 B F (Hz)
sampling theorem : ï 1
ïî T £
2B
Lowpass
F (w ) filter

A/T Fs = 2B is the Nyquist rate

1
-w s - 2pB 2pB ws w T= is the Nyquist interval
2B
- Fs - B 0 B Fs F (Hz)
Sampling

Signal reconstruction: The interpolation formula F (w ) Lowpass

filter

A/T

-w s - 2pB 2pB ws w

Ideal lowpass filter:
- Fs - B 0 B Fs F (Hz)
H (w )
T æ w ö
H(w)=T rect ç ÷
è 4pB ø
- 2pB 0 2pB w

W æ w ö
Recap: h(t) = sinc (Wt ) Û H(w)= rect ç ÷
p è 2W ø
2pB
Thus: h(t) =T sinc (2pBt ) = 2!
BT sinc (2pBt ) = sinc (2pBt )
p =1
1
Nyquist rate: T = Û 2BT =1
2B
Sampling

Signal reconstruction: The interpolation formula

h(t) = sinc (2pBt ) h (t )

1
-1 1
2B 2B

0 t
¥
f (t) = å f (nT ) d (t - nT ) Filter         f (t )
n=-¥

¥ ¥
f (t) = å f (nT ) h(t - nT ) = å f (nT ) sinc [2p B (t - nT )] Interpolation formula
n=-¥ n=-¥

Sampled  signal Reconstructed  signal

f (t ) f (t )

t
Sampling

Practical difficulties in signal reconstruction

1- Need for ideal filters

F (w )

Sampling at
Nyquist rate:

0 2pB w s w

F (w )

Sampling above
Nyquist rate:
0 2pB w s w
Sampling

Practical difficulties in signal reconstruction

2- Aliasing

F (w )
Recovered  spectrum

-w s 0 w s w s w
Lost  tail  gets  
folded  back 2 Lost  tail.

0 Fs Fs F
2
Sampling

Practical sampling

F (w )
f (t )

t - 2pB 2pB w
- B 0 B F (Hz)
pT (t)
¥
2p
pT (t) = å Dn e jnws t , ws =
t n=-¥ T
T

f (t )

f (t) = f (t) pT (t)

t
Sampling

Practical sampling

f (t )
f (t) = f (t) pT (t)
t

Recap: Frequency-convolution property of Fourier transform:

1
f1(t) f2 (t) Û F1(w) * F2 (w)
2p

Þ F (w) = F [ f (t)] =
1
F (w) * F [ pT (t)]
2p
Sampling

Practical sampling

1
F (w) = F [ f (t)] = F (w) * F [ pT (t)]
2p
¥
2p
pT (t) = å Dn e jnws t , ws =
n=-¥ T

Recap: Exercise 2: Show that e jw0t Û 2pd (w -w0 )

¥
F [ pT (t)]= 2p å Dn d (w - nws )
n=-¥

¥
1
F (w) = F [ f (t)] = F (w) * 2p å Dn d (w - nws )
2p n=-¥
¥
= å Dn F (w - nws )
n=-¥
Sampling
¥
Practical sampling F (w) = å Dn F (w - nws )
n=-¥

F (w )
f (t )

t - 2pB 2pB w
- B 0 B F (Hz)
pT (t)

T t F (w )
Lowpass
filter

f (t )

2pB ws w
t
0 B Fs F (Hz)