11

Modern
and
Filter Theory
Active RC
Filters

MODERN FILTER THEORY

The ideal low-pass
filter is
elements, approximations
passive elements,
ve two studied in this
and Cheh
Butterworth and Imations are chapter. Por realizing
apter.
are studied in it using
Dedures studied Chebyshev approximations.
studied in Choapproximatio
detail. These are the
the
procedures Chapter low-pass filtersUsing
9, low-pase network
network syn
modern filters based ynthesis
are realized. These
the based, LP.BP
as
they are, on modern network synthesis
thods.
methods. Using
Using LP-HP
LP-HP and
and LP-BP
are

be obtained from LP passive filters. transformations,

HP and BP iiters can

11.1 APPROXIMATION TO IDEAL LP FILTER

Eioure 11.1 shows the function |H(j»)| which represents an ideal LP flter.
The magnitude of |H(j») is cons
tant in the frequency range 0<w<Ku
the LP cut off is at w = o, and its
magnitude is zero for « >w Un- H(W
fortunately, such ideal
an fflter is not
realizable. The reason for this can
be understood only after a detailed
study of positive real functions and
their characterization. However, we 0
Bve an important relation which
g0es by the name of Paley-Wiener ldeal low pass
filter
Fig. 11.1
CTiterion for a function to be a causal
time function
(11.1)
should be nonzero throuoh
to be finite
|H(w}) becomes infinite the range
the.
For the integral In H(»)}
a In Fi
Fig. 11A
means
[H(w)=0
frequencies. For realizability, any
of
we notethat |Hw)0
tothe ideal LP filter
for
should
w> o.
have |H(»)|= for o ) . approximis ati
where e

and nonzero.
Anite

MAGNITUDE FUNCTION (MEA.

11.2
MANIMALLY-FLAT FMF)
that is close to being constantin
the region
region
We first seek out a function
0< a, but falls-off
beyond . Let a system function H(s) be
given by
H+Cs+Car +... Cms
Hs)= Ho(I+ ds + ds2 +.. . t d,s") (11.2)
The magnitude squared |H(j»)| can be obtained by putting s = jo, collec
terms, and oblaining the squared magnitudele.
ting real and imaginary
Hj)= H(jo)H(-jo)
=
|H(s)H(-s)| at s = jo. Thus |H(juP wilAlso,
of the form
1aqw + azo + ...+ a,mu
H(jo)= Ko
H(ju)P =

i+ bu? + bzu* ...+b,n (1.3)

We see that |H(ju)2 is a function of a, and is even. To represent a TD
filter, Hjo)F for > o should be zero, so that m<n. Equation (113
can be approximated to Ko at » =0 by a Maclaurin series. This wil!l ive
us a useful function H(s). This approximation for Eq. (11.3) is
T T4)
T HMjP=KoI+ + +.. (11.4
where the odd derivatives are absent and
d2T d'T
TO dw2"
T4 =
d 4 : etc.

For Eq. (11.4) to approximate K% at w

0, T(2), T«), TO, etc. at u=
=

should be zero, i.e. all the derivatives of the

magnitude square should e
Zero at 0. This idea will be used to obtain a
=

relation between d, and

b, in Eq. (11.3). The resulting function is called a
function. maximally-fAlat mogniraue
Since Eq. (11.3) can be approximated by a long division as
m u,
Hjo= Ko{l + (41 - bi)u + [(a2 - b) - b,(a1 - bi)lu" **"
Recalling the zero derivative requirements require
at w = 0 for MFM, Wo
a= bi, az= bi up to am= bm and
Thus the MFM b+=bn+2=b,-1 =0 and b, #0
squared function is

with
|H(jo)(MFM) Ko I+ajw2 +a2w4 +
=
.+a,w
(1.59
m << n. +am 1+aw2+azw*+. .o
+6,w"
Modern Filter
Theory ond Active
that in
in Eq. (11.5) the
that
the las
last term in RC Fatert
n o l e
M3
he only extra term
compared to the
denominator
We
the
whichi st h e .

numerator. is
ha
E r a n p l eI 1 1

tions among the

coefficients a, h and
p e t e r m i n et h e r e t

M F M ,g i v e n e
sch that T(
jo
TS) bs +
s+a

Soltion:

aJw
Tjo)(c -

w) +jb»
aw

eaeficients of like powers in the numerator and

Equa denominator,
acand b-2c ==1
Thus,for a=
1, l and b=V3. Hence
c=

T(s)= (s+ 1)(G*+ V3s+ 1) for a-

T(jo)= (1 +u(1+ u? +)
giVing

4 RUTTERWORTH FUNCTIONS AND SYNTHESIS

if all the a's are zero, we get the
Butterworth function B
In Ea. (11.5),
of MFM LP function.
which is a special case
B.(ju)= Ko(1 + ba)
= Ko'[1 + (w/un)"]
where wp= (1/b,)!/2n, i.e. we take
and magnitude,
we normalize the frequency
For simplicity,
w=1 and Ko=l
(I1.6)
S0 that
B.ju)P =
1//1 + w) )
are 1/(1 + ) , /+
values of B,(ju)|3
n=
1, 2 and 3 the becomes
or (11.6)
a / + w ' ) respectively.
For w = 1, Eq.

or 0.707
=
1/W2= function
B,(ju) is
B,MJ1) values ofn,
the
for B,(s).
From
increasing expressions
Figure 11.2 shows thatfor proceed
to find
latter at the We now
(11.7a)
orig 1/D,
bq.(11.6), since » = S) 1/{1+(-I)'s)
=
B A S ) B , ( - 5 ) =
374 Circuit Theory

1.0

3 0.7 -

n:1
n 2
n4

0 1.0

Fig. 11.2 Butterworth filters of order n.

Thus D=1-s2, D2 =1+ s', D3 = 1 -s6, etc.

To obtain B,(s), we factorize D, and take
only those factors with roots
in LHP so that B,(s) is a
stable, realizable function.
Thus,
B0) =1/(0+5)
Ba(s)= 1/(1 + V2s +s) (11.76)

B(o=1/((s + 1(2 + s+ 1))

and so on.
The poles of B,(s) are roots of 1
+(-1ys= 0.
Modern Filter Theory and Aetive RC Filers 375

Butterworth Functions
Synthesis o f Br
Hurwitz polynomial. Recalling that (Chapter
= 1/D(5) where Ds) is
a
be realized as an LC two-
function Z,= 1/(M2 + N2) can
fer when M2+ N2 is Hurwitz, we have a
9 ) a t r a n s f e r

1 2 resistor,
ed by a V2(s)/I(s) of such
transfer impedance Z, =
p ó r tt e r m i n a t e

as a
realizing B(s)
m e t
od of
h o d

a two-port.

Thus, 1
B(s)=Z= M2 + N2

1/N2
1M2/N2
so that Z21= 1/N2

and Z22=M2/N2 zeroes at

(transmission
in Cauer
I form
is
connected
connected to
to
ladder resistor
is realized as a n LC a l
1ao
in and then
infinity) from rom port 2 to port 1,
Dor
port 2.

Example 11.3 nce

1.4 CHEBYSHEV FILTERS

the ideal low-pass filter is

toLented briety
to presen.
approximation
Another
realization, recourse can be
emphasizing the
concepts. For ken to exten
literature.
available in the
sive tables
low-pass filter 18 one realizable approximati
The Butterworth from it increa th
exact at w =0 and departs
is
ideal LP filter. It dB down) of its maximum an
increases and at w
=1 it is 0.707 (3 value. The
0 p < l is the 3 dB band. In
the stop-band where
where >1, the
the
range
asymptotic slope of 6n dB/oeta.
magnitude |B.(»)| falls off with
an
ere
filter. The higher the value of n, the is the
n is the order of theto and more complex is the network realia tion
filter response (close w=0)
to the ideal filter, the ded
In order to have a better approximation epat
ture from the ideal can be
distributed throughout the pass-band so tha
between the maximum value of 1 (0da.
the magnitude |H(jw)| oscillates
for a certain number o
and a value 1/(1 + )2 (which is the minimum),
to be defined. We introduce
times n. n and e are the parameters that have
the Chebyshev approximation to the LP filter as
1
(119
7 T+¢C{o)
where Cu) is the Chebyshev polynomial of order n in w, explained presen
ly. is a parameter specifying thc ripple in the pass band and n s i
order of the filter as well as the Chebyshev polynomial. n also gives n
and
number of ripples in the pass band. It is the total number of maxima
cally
minima in the range 0<w < 1. For w 1,T.(w) decreases monotond
While B,(u) has a slope 6n dB per octave, T(w) has a dB loss dBper
In
octave46fn 1) dB.
1), i.c. additional constant value of 6(4
-

an
Eq. (11.9),
KIDS
r(s)=K/D,(s), herc
where
|T{»}f T(syT(- Theory and
ct Ki
K Active RC
ireplaces in l 4
co).
»

2C»).
plane. Thu T(s) can he Th
be
is a
constant.
The roots Füters
and C elements
Tealized as aa.A8)
realized of
DAs)Dsy-1
D) as
(o)
Bascd on C,(»)
transfer are in e
and the
ve D,($) as a
whCral Values of n).polyno
value of funct
olvnomipoleal
,
ionthe left
of, Hand-hook
Z,As)
half
haif of
uing the
¢C.where
n (integrak

is s)
The and in
the Hand- hook
R
the

locationsfactored tatles
manner.

denominator form are

iion Chebyshev statements presented givenavaithe lsame

what follows, some able
of T are fora
In what
of the also
a brief presentation of above order in
of
ed presentat
values ofand Chebyshev
Weinber Pol
filters
n, ynomi al
and fors. For
will be
elaborated after an
an
should be
T) is usually plottedi in dB
excellent
extensive tahles her and detail
consulted. detail-
through values of 20
for
varinen
and
Tmaximum) =
1=0 dB log Tw). Thes,
T.(minimum)= (+a-10 10 log
The ripple in dB is (1+a
N 10 log (1 + )
For a typical
value ofe (11.10)
1/2, the ripple is
=
1 dB
Figure 11.3 shows |7T,(«)| for n 2 and n =
3 with I dB
==

rippie.
(a (b)

dB dB

1
0
0
n = 2,3
dB ripple
Chebyshev
filters of I
Fig. 11.3 New York,
Book Co.,
M c G r a w - H i l l

L. Weinh alusis
qnd
Synthesis,
9Anate)
Pole at s D.s)
s+2.0
-2.0
s+1.1s+ 1.1
2 -5.5+j0.90
-0.50, -0.25 +0.97 +0.99s2+1.2 +0.49
3 (s+0.5Ns* +
0.49s + 1
j0.98, -0.34 t /0.41
(s +0.28s +0.99M
-0.14 +0.7.
+92
.

Chebyshev Polynomials C,()

is defined as follows:
C,u)
CA) = cos n(cos- ) for0 <I
= cosh n(cosh-l ) for w>I (11.la
Let
w = COs 6
for 0 w<l (1L.0b
and
w =cosh 4
for w>1
Then,
Ca)= cos no for 0 1
and
CAw)= cosh nß for >
Since
eine= cos nô +jsin nt
=
(cos + jsin 8)
cos ne=Real part of (cos 0 + j sin 6)"
Also,
Colco)=cos 0 = 1
C() = cos
In obtaining
C(») from the binomial expansion of (cos 0 + j sin 8, the
real part will have even
powers of sin 0 as factors like sin2 8, sin' 6.
which can be ...

expressed in terms of cos2 0 since sin2 0= We, I- cos* 8.

therefore, get C,(») as a
polynomial in cos 0= w as listed below.

Chebyshev Polynomials
Co1
C= cos 6 -= co
C2=cos 0- sin? 0 =
2w2- 1
C cos 6- 3 cos 6 sin2 0 =
These expressions show 4w- 3w, etc.
that C S1 for
0<w<1.
Recurrence Relation for Modern Filter
Theory and Active RC Filters 379
Using the identity C
cos (A + B)
+ cos
r+ (A- B) 2 cos A cos B
=

C,-1= cos (n + 1)0 +

cos (n- 1)"
giving the
recurrence relation=2 cos nf cos 0 2wC,

For A B the
above Cral2wC, -C,-
identityb (11.12)
comes
For A= nô, cos 24
+ 1 = 2 cos2 A
cos (2n8) +1=
or 2 cos2 ne
cos ne }[cos
giving another useful relation (2n0) + 1)

hich can be used to avoid C-C2+ 1) (11.13)

shoud be
be drawn.
drawn. În the squaring C. Plots of Ca as a function of
range 0 w<1, value of C<1, and for
the
1,C increases. Our interest is in
C and e2C?, and finally in

Poles of Ts)
T.(6)T(-9) = 1/[1 + 2CAslj)]

Poles of T($) are the left s-plane roots of

1+)-
-4
roots of C, occur at
It is shown in Weinberg that the
(2n- 1)
k E jw for k =0, 1,..
.,
Sk=
left half of the
given by those sk which are in the
of
he poles T(s) are
plane. These poles are at

=(-sinh asD
(11.14)
where fork =0, 1,....(n -) (11.15)
na

a-sinh(
or
sinh
a Can of e and n.
be determined for tiven values
380 Cireuit Theoryv

shev
Remarks. The ripple bandwidth of the Chebyshev flte.
filter is up to
where 7,1) = 1/(1 + ) . Beyond w - 1, T,(») falls ofT Eup
falls off. For
to
can be approximated by ,TA
T.)C2-
The last expression is obtained by taking only the highest
CThe 3 dB point occurs in the stop band at w=cosh a,
highest powe
There is a useful expression for C,(u) when w > 1. This
This isis ohs.
CAlw)= cosh (n cosh- w). Let cosh- ø = x, so that coshx= obtained f
help of the identities With he
cosh x=(e* + e-*)/2
sinh x =(e* - e-*)/2
e =cOSsh x + sinh x
and
sinh2x= cosh2 x -1
Cw) can be expanded as

Cfw)-(9 +(S
( + Vu?-1 +, (w + Vu- 1)-"
2 2
and be
can
approximated by
C.lo) +Vu-1 for o1
(l1.1

Example 11.4