Cheb Yshev
Cheb Yshev
17
rractke Prohlcm S.3 Find the pole locatio ns of a 6th order Butterworth filte r with
nc- ::= I rad/sl'C .
IH(jO)I
IH(i~)I
1.0 l.O i - -- -- .-
I
1
Ji+ £2 [i+7
.
..
I
I
I
''I
I
I
.:'
• I
I
'
I
I
Ji +).l ------------------1---------
I
I
.
'I
----------- -------~-- -
... : ,:
'I
Os n n., n. n
I l'ypc I
IH(jn)1 2 =
1
- - -- - N=l,2 , ... (5.33)
1 + c2c2
N ( TI;
n)
where c: is a parame ter of the filter related to the ripple in-the passban d and CN(x)
is
the Nth order Chebyshev polynom ial defined as
(5.34)
where C 0 (x ) = 1 and C 1(x) = x .
The Chebyshev polynomials described by Eq. (5.34) or Eq. (5.33a) have the fol-
lowing properties.
5. 18 Digital Signnl Processing
IHGO)I
IH(iO)I
Nodd N even .
1.0
1.0
1
Jt +t2
I I
1
Ji + ,.,2 -- - ----------➔ ----
./t+).l ----- ----- ---r- - -
• I
I
I
I
I I
(a) (b)
Fig. 5.~o Lowpass Chebyshev filter magni tude response.
Taking logarithm for Ecf. (5.33) we get
0, = 10 log [ 1 + e 2c i ( ~:) ]
= l0log [1 + e2 {cosh(N cosh- 1
(O ,/ 0, )) } 2] [J3 (5 .3 7)
fo r e in Eq. (5 .3 7) , so .
Substituting Eq. (5 .3 6) .
.ext higher integer, w_ lving for N an d ro un
e ge t - - - - - di ng it to th e
- - ~ --- ,
. co sh - 1
N >
10 · a. _- r
1()0,la p - 1
cosh -1 -n. (5.38)
Substituting Eq. (5 .2 5)
and Eq . (5 .2 6) , in Eq. (5.38) w
e obtain
( · co sh -1 A
I~ N-~ ~ 1 .
h - (1 /k ) (5.39)
In F.q. (5.39) co sh - 1 (x
.
) can be evalwitea us in
g the-identity
co sh - 1 x = ln[x + ..jx 2 -1 }
e abOp (5.40)
ov e equation we g~t
We define
CN ( - ~ ) = ± j/ e = = -. lN ~ -~ ( t)]- -
f
~ow Eq. (5 .4 0a ) yields
co s- 1 (-~) = < l> -j 9 ✓ /~
Equating the re.al and imaginary parts of both sides of Eq.(5.41) result in
since cosh(N8) > Ofor 8 real, then in order to satisfy Eq.(5.42a), we have
(2k - l )1r .
<P--: 2
N k=l,2, . . . N (5.43)
Using this result and Eq.(5.42b) we can now solve fo~ 6, where sin N¢ = ±1.
Now we have ·
8=±..!_sinh-1
. N • e
(!') . (5.44)
' . . ~
Combining Eq.(5.44),. Eq.(5.43) and Eq.(5.40b), we obtain the left half plane
locations given by ' ·
or
µ. = e"inh-l(i-1) .- E-l + Vl + e-2
From Eq.(5.44) we can write
-
Infinite Impulse ke~ponse Filters , .2J
Now
. ( µ1 /N _ µ - l/N) ( µ1/N + 1,- 1/N ] ]
St ~ 0,, - Slll </J
[ 2 + J. cos ¢ (5 .49)
2 ,
µ1/N _
a=~ [ . 2
µ-1/N] (5.52)
'1ieat a and b are minor and major axes of the ellipse respecti;ely.
IUIPpol•
LHPpola
or14(-t)
ofH(■)
a,
Pole■ ror
nodd
Ch8 bySheV filter
/
f
• Fig. 5.11 Locus of the poles of .
J'
S.7.l Ch"bJ!he, l)ptt • 2 Filttr
OK-t,y~hcv rypc-2 filtc.r has both poles and zeros. The magnitude square resP<>nse
1
ghn by 1
IHLJ0)( =
1
1+£2
[C (!!a.)
2
Nu,, ]
(5.56)
ci(¥t)
v.-bcre CN(x) is the Nth-order Chebyshev polynomial, n.,
is the stopband frequency
and flp is lhe passband frequency.
The zeros arc located on the imaginary axis at the points
0
s" = i . :
Sill,yk
k = 1, 2, ... , N (5.57)
cosh-1 ( ~)
N =-- E (5.61&)
n
cosh-1-!
Op /
\- cosh-1A I
- cosh-11/k I
I
I (5.61b)
where ..J
.,\
A=-
E
k=~
n.
E = (100.lop _ 1)0.5 and
.,\ . ( 100,10•. _ 'l )0.5
Infinite Impulse Response Filters 5.23
Where
µ = e-1 + ✓e- 2• + 1
..
e = ✓ 1()0,lap - 1
Op = Passband frequency
Op =Maximum allowable attenuation in the passband
(·: For norrna1·
4. Cal 1zcd Chebyshev filter Op 1 rad/:sec)
= . .
t culate the poles of Chebyshev filter which lie on an elhpsc by US1ng
Ortnuta. the
1
5.2 ◄ Digital Signal Procr~~ing
(a) For N odd substitute s = 0 in the denominator polyriornial and find lite
value. This value is equal to th~ numerator of th~ transfer function.
(·: For Nodd the magnitude response IH(jn)I starts at 1.)
{b) Fo: N even substitute s = 0 in ~e denominator polynomial and divide
the resu It by J1 + E: 2• This value as equal to the numerator.
Solution
Step 1:
. cosh-•1 l0 · 0•-·1 ~
N;::: 1QO.lap - 1 - -1
cosh- 1 Oa - cosh . i 40001r
nP cosh- 2OO07r
= 1.91
Solution
8tep I: From the given data we can find that
1 1 1 .
--=• • -
./I+,~ ./2' vl + -,i:=::=R •
~:i
0.1,
11t(JU)I
1.0
0.707
I
Q.J - - - - - - - - - - - ,- - - - - I
= 0.596
- [µ1/N.+ µ-1/Nl -- 2 [(2.414)1/3 + {2.414tl/3]
b- nP
2 2
= 2.087
.
Step 4: To calculate the poles of Chebyshev filter
,1.,
'#'le
= ~2 + (2k2N
- l)7r k =1 2 3
' ,
<Pl = 120° (P2 = 180° cf,3 = 240°
I I
cos h- 1
10 · 1
O:s ~ 1 10 -1 I
----
:p - 1 >
cos h;
N '?. _ _;_ __ __0.lo
·lO
:~ ~,- --- •,.
100.1 - 1 =4.536
20007r
co sh - 1 -0s Cosh -1 --
, Op 100071"
i.c.,N=s
'-- --- -.
~ ·I
Q,!
.
.
•"
'th a
- -
l.s e 5.9 Design a Chebyshev filter wa maximum passoca,
' at 0, '= 20 rad/sec and the sto nua tion of 30 dB
--------~
, ~,,,. 11ttenuation of
8t
.. - 50 rad /~.
O, ;;..r:: Ma, '07)
pband ane · (AU ~
k = !1 = 0.4
n,
Now 31.607
c.osh- 1 - -
N -
> . 0.882 = 2.726•
1
cosh- 1 -
0.4
,.c.. N = 3
µ, = E-1 + fi+e -2 = 2.65
[µ,1/N _ µ,-1/N]
a= Op 2 = 6.6
µ1/N + µ,-l/N1
b -_ Hp,
n =-- --~
2 - = 21.06
[
Bk= acos</>k + jbsin¢1c_
; k = 1,2, 3
'Pk = ~ + ( 2k - 1 ) 1r; k = 1, 2, 3
2 2N
~l = 120° 1 ¢2 = 180°, (?3 = 240°
Bl = -3.3 + jl8.23
s2 = -6.6
B3 = -3.3 - jl8.23
. Denominator of H(,) = (, + 6.6)(,2 + 6.6s + 343.2)
Numerator of H(,) = (6.6)(343.2) = 2265.27
Tran,fer function H (,) = 2265 27
· ·
(, + 6.6)(,2 + 6.6, + 343.2) _____,._
sbt''I
Pndk:t Problem SA For die 1iven 1peclftcations find the order of the (hebY
filter
.
Practke Problem 5.5 Fmd the pole location, or a normalized CbebYshCY (llccC
•
f
Infinite Impulse Rcspom,e Filters .S.29
rrartkf Problem 5.6 De-sign a Chebyshev filter for the following ~pecifications
(5.62a)
- ~~ IH(j.0)1
I.O
1.0
::J._
~ --- ---- I
I •
Ji +£2 I
I
I
----------~--- - _L ----------:---- -
' I ./1 +Al • I I
---.J....._-- -1•~-- . nv n·, - a
°'-n.--n
'1s, (b)
(a)
IH(JU)I
10 1 - - - -.......
I
l
j1 +£2
./1+~
I I
I
---- ---- --1---- 1
A+v
I
n·r '1c Q
(a) (b)
s2 +
S--+ ~ - - -
n,nu (5.63a)
s(f2u - n,)
0,. = min{IAI, IBI} (5.63b)
A= -n~ +n,nu (S.63C)
01(nu - n,)
B = n~ - n,n,. (S.63d>
02(0u - Oz)
"<i<ll I • IHOOM
1.0
I 1.0
li ♦ e2 ....J_
./i+J I
I
I
I
I
I
I
...'
I
I I ••-
•• •
----•--------·
(a)
(I,)
Fig. 5.15 lowpaaa to bandpass transformation.
Infinite Impu)se Response Filters SJ J
IH(iO)I
1.0,___ _
1.0
I
~ ---------
i '
I
I
I i
.,.--,
I
I
I I
t
(a) (b)
let Ille -~
~~ t....._ ign a normalized lowpass filter and. then use
suitable tranSformab.on
• of a bighpass filter.
· """llllfer function .
For IOWpass filter For higbpass filter
0c === 0,, = 500 rad/sec !le = Op = 1000 rad/sec
n. ::: 1000 rad/sec n. = 500 rad/sec
,
'. 5.32 Digital Signal Processing
IH(i!l)idB
,r,
o: n:
I
0 n. n
_
.,, . .n . .,,
►..S . ~
Fig. 5_17 Lowpass to highpass transformation
0c =n,. = 1ooondl•
. .
Substitute s - ~
s
H.(a) = H(s)I
.
---• 1000
1
= (a+ l)(a2 + • + l} __;~
I
-
- (a+ 1000)(,Z + 1000. + (1~) ~ '1/ 2
1 ,.
Infinite Impulse Response Filters 5.33
Fig. 5.18b
th .
tern function r,or the digital IIR filter
. obtained as aresuhof
Consequently e sys . . by finite difference. 1s .
. tions of the denvat1ves .
the approx1ma .
l ~- 1 _ 1 + jOT ·
z = 1- sT = 1- jnT - 1 + -n2~2
·1 · - jOT
(5.68)
= 1 + n2T2 + 1 + n2r2
=x+jy (5.69)
where X
1
= 1 + f!2T2 ~ . = 1 +nT
an y
. We can find that x and Y are related by
n2T2 ' . . .
( 1)
x--
2
+y2 =- 1 (S.111)
. 2 4 . . . f dill$ 1/2.
• of the 8 -plane 1s o ra
Thus the image in· the· z-plane of the Ju-axlS
•n
shown in Fig. 5.19, Which bas the following characteristics. - tf21
1. The left half s-plane maps inside a circle of radius· 1/2 centered at z - .
the z-plane. . . · ,: ' f radius 1/2'
2. The right half s-plane maps into the region outside the circle 0
the z-piane. · . i1J ~ ,.p,
3. The jO-axis maps on to the perimeter of the circle of radius 112 _.,,,,t
Th th
. us 1s . anal?~-to-di . • transformation
. g1ta1 a stab
. teehn'1que. d oes map on to the~ ~ tJ
1e 81"""'1
~ mto a stab!~ d1g1ta1 filter'. but the jll-axis docs not.~n map le 1.,call~1
cuele•.Thus 1h11 transformauon severely restricts the digital filter Po .5 ,.slfl r,
lie confined to relatively •mall frequencies. As a result, this mapp••~; ~
the design of lowpass filters and banc1pas. filters having relatively SDI olf"I"'"""
q~encies. I_t is not possible to transform a higbpass analog filter into a c
htghp- digital filter.
Infinite Impulse Response Filters 5.35
Im (Z)
jO
s-plone
I Re(z)
unit/,
circle
00
H(z)L=eAT = L h(n)e-aTn
n=O
Let us consider the mapping of points from the ·s-plane to the z-pl~e in\plied by ·
die relation · ·
z ~ •etfl'· · .. (5.12>
lfwe substitute 8
l:::: reiw
= <J + 'fl and express the complex .variable z in polar form as
3 . .
lb we get
.
re'w = e(e1+jU)T
. o'J' _;uT
=e ~ (5.73)
\ .
e11'
r == e (5.74a) .
w == UT (5.74b)
'I'
1\ . E (S.73), e" 'bas a magnitude of~ and Ill
1aa1e: first_term in the product anoni~ertn e;ur, has ~ity magnitude and an,anglc
•nr fQ _areal number. The sec is mapped to a place ID the %-plane of magnitude
· Therefore, our analog pale
· 5.36 Di~it.u Signal rro...-cssing
e~r anJ angk nr. The ~~1 p.u1 of the analog polo llctcrmines the rad·
. •usor!L
plM)e P'--'1e and the imaginary r ~1rt of the nnatog po1c d1ctatcs the angle of
pole~
l- the;.
11141
CoosiJe-r any pole (1n th<' j O-axis, where u = 0 ns shown in Fig. 5_
20
I'l'ks map to the i-plane at a radius r = e0·T = 1. Therefore. the impulse .' ~
mawing map poks from the s-pl~e·s jn-nxis_to the z-plane's unit circle. 'llVari-
lm(z)
' ljO
s~plane
n
a
Now consider the poles in the left half of s-plane where u < 0. These pol~ 1111P
inside the unit circle as shown in Fig: 5.21, because r ~ e"T < 1 for u < O.
1
· lm(z) jll
z-plane ---
.a-plane
· O
Rc(z)
Fig. 5-:21 Stable poles mapping Inside the unit circle ...:i
Therefore all
the . . • a-plane poles with negative real parts· map to z-plaDC paleS~
· iD""
. ~1t arcle - stable analog poles are ""..__. to tabl diaital poles,
'Jbe
IDvanant mappin . ---ri-.a s c o- ...a
. g ~reserves the stability of the filter. . ~~
• JAIi ~les m the right half of the s-plane map to digital p0les outside
• CU'C e. .
,f •
r = e"r > 1 for u > O
The mapping is shown in Fig. 5_22_
.I
Infinite Impulse R;sponse Fitters 5.37
Im (z) jO
z-planc
a-plane
Rc(z) a
circle
Fig. 5.22 Unstable poles mapping, outside the unit
le, it is not one-to-one map ping
Although the jfl-a xis is map ped into the unit circ
poin ts in s-pl ane are map ped to a
rather it is many-to-one map ping , whe re ·man y
~in this is to consider two poles in
single point in the z-pl ane. The easi est way to expl .
inar y components differing by 2;
Gt a-plane with iden tical real parts, but ~ith imag
•shown in Fig. 5.23.
let the poles be ·
s1 = u + jO .
s2 = u + ,j ( n + ~ )
(5.7S)
t3~
lm(z ) ,-plane
z.plane
These poles map to z-planc poles z1 and z2, via impulse invariant mapPin1,
From Eq.(5.76a) and Eq.(5. 76b) we find that these poles map to the same location
in the z-plane. There are an infinite number of s-plane poles that map to the same
location in the z-plane. They must have the same real parts and imaginary ~
that differ by some integer multiple of ¥.
This is the main disadvantage of impoJse
invariant mapping. The s-plane poles having imaginary parts greater than f or lea
than -f cause aliasing, when sampling analog signals. The analog poles will a«
be aliased by the impulse invariant mapping if they are. confined to the a-plane's
"Primary strip" (within 1r/T of the real axis).
Let Ha(s) is the system function of an analog filter. This can be expressed ii
partial fraction fonn as
N
Ha(s) = L- Ck (S.7T)
.k=l 8 - Pk
wh~ {pA:} are .the poles oftbe analog filter and {c1c} are die coefficients in the partial
· &action expansion. · ·
The inverse Laplace transform of Eq.(S.77) is
N
. h"(t) = L c1ce"•' t ·> 0 (5.78)
k=l
h(n) • ho(nT)
N
• L c1,t!•nT
"•1
Weknow
00
H(i) = L h(n)i-n
n•O
Infinite Impul se Rcsponse Filters 539
. N . N
•
'r'.Jtisif Ha(s) =L Ck then H(z) =~
· ~I-
Ck
ePtT z- 1
•
s - Pk k=l
k=l
is high. Therefore,
f-or high sampling rate s (for sma ll T), the digital filter gain
m!CadofEq. (5.81a) we can use · "'
N .
· · Tck (5.81b)
H(z ) .= L
k=l
1 - ePtT z-1
opriate for
Due to the presence of aliasing, the impulse invariant method is appr
ce method is
~ design of lowpass and ban dpa ss filters only. ~e impulse invarian
~lter.
Dnsuctcssful for imp lem enti ng digital filters such as a bighpass