Académique Documents
Professionnel Documents
Culture Documents
8.7
(8.87a)
I2 = i 1 i 2 ,
(8.87b)
I3 = i 3 i 4 ,
(8.87c)
I4 = i 3 + i 4 .
(8.87d)
First consider the line as being driven in the even mode by the i 1 current sources. If
the other ports are open-circuited, the impedance seen at port 1 or 2 is
e
= j Z 0e cot
.
Z in
(8.88)
(8.89)
cos (
z)
i1 .
sin
(8.90)
427
+V3
+V2
2
I2
I3
3
Z0e, Z0o
I1
1
+V1
I4
4
+V4
l
(a)
vb
i1
i1
i3
i2
Z0e, Z0o
i4
i3
4
va
l
(b)
O.C.
I3
Z0e, Z0o
I1
O.C.
(c)
FIGURE 8.42
Definitions pertaining to a coupled line filter section. (a) A parallel coupled line
section with port voltage and current definitions. (b) A parallel coupled line section with even- and odd-mode current sources. (c) A two-port coupled line section
having a bandpass response.
Similarly, the voltages due to current sources i 3 driving the line in the even mode are
va3 (z) = vb3 (z) = j Z 0e
cos z
i3 .
sin
(8.91)
Now consider the line as being driven in the odd mode by current i 2 . If the other ports
are open-circuited, the impedance seen at port 1 or 2 is
o
= j Z 0o cot
.
Z in
(8.92)
(8.93)
428
cos (
z)
i2 .
sin
(8.94)
Similarly, the voltages due to current i 4 driving the line in the odd mode are
va4 (z) = vb4 (z) = j Z 0o
cos z
i4 .
sin
(8.95)
(8.96)
where the results of (8.90), (8.91), (8.94), and (8.95) were used, and =
. Next, we
solve (8.87) for the i j in terms of the I s:
1
(I1 + I2 ),
2
1
i 2 = (I1 I2 ),
2
1
i 3 = (I3 + I4 ),
2
1
i 4 = (I4 I3 ),
2
i1 =
(8.97a)
(8.97b)
(8.97c)
(8.97d)
j
(Z 0e I1 + Z 0e I2 + Z 0o I1 Z 0o I2 ) cot
2
j
(Z 0e I3 + Z 0e I4 + Z 0o I4 Z 0o I3 ) csc .
2
(8.98)
This result yields the top row of the open-circuit impedance matrix [Z ] that describes the
coupled line section. From symmetry, all other matrix elements can be found once the first
row is known. The matrix elements are then
j
(Z 0e + Z 0o ) cot
2
j
=
(Z 0e Z 0o ) cot
2
j
=
(Z 0e Z 0o ) csc
2
j
=
(Z 0e + Z 0o ) csc
2
Z 11 = Z 22 = Z 33 = Z 44 =
(8.99a)
Z12 = Z 21 = Z 34 = Z 43
(8.99b)
Z13 = Z 31 = Z 24 = Z 42
Z14 = Z 41 = Z 23 = Z 32
(8.99c)
(8.99d)
A two-port network can be formed from a coupled line section by terminating two
of the four ports with either open or short circuits, or by connecting two ends; there are
10 possible combinations, as illustrated in Table 8.8. As indicated in the table, the various
circuits have different frequency responses, including low-pass, bandpass, all pass, and all
stop. For bandpass filters, we are most interested in the case shown in Figure 8.42c, as open
circuits are easier to fabricate in microstrip than are short circuits. In this case, I2 = I4 = 0,
Image Impedance
2Z 0e Z 0o cos
Zi1 =
Zi1
429
Zi2
Response
Re(Zi1)
Zi2 =
0
2 Low-pass
3
2
2 Bandpass
3
2
2 Bandpass
3
2
2 Bandpass
3
2
Re(Zi1)
Zi1
2Z 0e Z 0o sin
Zi1 =
Zi1
0
Re(Zi1)
Zi1
Zi1 =
2 sin
Zi1
Zi2
Zi2 =
Zi1
Zi1 =
Zi1
Z 0e Z 0o
Zi1 =
Re(Zi1)
(Z 0e + Z 0o) sin
Z 0eZ 0o
Zi1
Z 0e + Z 0o
2
All pass
Zi1
Zi1 =
2Z 0eZ 0o
Z 0e + Z 0o
All pass
Zi1
Zi1 =
Z 0eZ 0o
All pass
All stop
Zi1
2Z 0eZ 0o
cot
Z 0e + Z 0o
Z 0eZ 0o
Zi2 =
Zi1
Zi1
Zi1 = j
All stop
Zi1
Zi1
Zi1
Zi1 = j
Zi2
Z 0eZ 0o tan
Zi1
Zi1
Zi1
Zi1 = j
Z 0eZ 0o cot
All stop
430
Z 0e Z 0o
2
FIGURE 8.43
1
2
3
2
The real part of the image impedance of the bandpass network of Figure 8.42c.
(8.100a)
V3 = Z 31 I1 + Z 33 I3 ,
(8.100b)
cos 1 = cos 2 =
Z 0e Z 0o
.
Z 0e + Z 0o
The propagation constant can also be calculated from the results of Table 8.1 as
Z 11 Z 33
Z 11
Z 0e + Z 0o
=
=
cos ,
(8.103)
cos =
2
Z
Z 0e Z 0o
Z 13
13
which shows is real for 1 < < 2 = 1 , where cos 1 = (Z 0e Z 0o )/(Z 0e +
Z 0o ).
Design of Coupled Line Bandpass Filters
Narrowband bandpass filters can be made with cascaded coupled line sections of the form
shown in Figure 8.42c. To derive the design equations for filters of this type, we first show
that a single coupled line section can be approximately modeled by the equivalent circuit
shown in Figure 8.44. We will do this by calculating the image impedance and propagation
constant of the equivalent circuit and showing that they are approximately equal to those
FIGURE 8.44
431
J
90
Z0
of the coupled line section for = /2, which will correspond to the center frequency of
the bandpass response.
The ABCD parameters of the equivalent circuit can be computed using the ABCD
matrices for transmission lines from Table 4.1:
cos
j Z 0 sin
cos
j Z 0 sin
0
j/J
A B
= j sin
j sin
j
J
0
C D
cos
cos
Z0
Z0
2
1
cos
J Z0 +
sin cos
j J Z 02 sin2
J
Z
J
0
=
(8.104)
.
1
1
2
2
j
sin J cos
J Z0 +
sin cos
J Z0
J Z 02
The ABCD parameters of the admittance inverter were obtained by considering it as a
quarter-wave length of transmission of characteristic impedance, 1/J . From (8.27) the
image impedance of the equivalent circuit is
J Z 2 sin2 (1/J ) cos2
B
= 0 2
,
(8.105)
Zi =
C
(1/J Z 0 ) sin2 J cos2
which reduces to the following value at the center frequency, = /2:
Z i = J Z 02 .
From (8.31) the propagation constant is
1
cos = A = J Z 0 +
sin cos .
J Z0
(8.106)
(8.107)
Equating the image impedances in (8.102) and (8.106), and the propagation constants of
(8.103) and (8.107), yields the following equations:
1
(Z 0e Z 0o ) = J Z 02 ,
2
Z 0e + Z 0o
1
= J Z0 +
,
Z 0e Z 0o
J Z0
where we have assumed sin 1 for near /2. These equations can be solved for the
even- and odd-mode line impedances to give
Z 0e = Z 0 [1 + J Z 0 + (J Z 0 )2 ],
(8.108a)
Z 0o = Z 0 [1 J Z 0 + (J Z 0 )2 ].
(8.108b)
432
FIGURE 8.45
Development of an equivalent circuit for derivation of design equations for a coupled line bandpass filter. (a) Layout of an (N + 1)-section coupled line bandpass
filter. (b) Using the equivalent circuit of Figure 8.44 for each coupled line section.
(c) Equivalent circuit for transmission lines of length 2. (d) Equivalent circuit
of the admittance inverters. (e) Using results of (c) and (d) for the N = 2 case.
(f) Lumped-element circuit for a bandpass filter for N = 2.
433
right, but the filter can be reversed without affecting the response. Since each coupled line
section has an equivalent circuit of the form shown in Figure 8.44, the equivalent circuit of
the cascade is as shown in Figure 8.45b. Between any two consecutive inverters we have
a transmission line section that is effectively 2 in length. This line is approximately /2
long in the vicinity of the bandpass region of the filter, and has an approximate equivalent
circuit that consists of a shunt parallel LC resonator, as in Figure 8.45c.
The first step in establishing this equivalence is to find the parameters for the Tequivalent and ideal transformer circuit of Figure 8.45c (an exact equivalent). The ABCD
matrix for this circuit can be calculated using the results in Table 4.1 for a T-circuit and an
ideal transformer:
2 Z2
2 Z2
Z 11 Z 12
Z 11 Z 11
12
11
1 0
Z 12
Z 12
Z 12
Z 12
A B
. (8.109)
=
=
0
1
C D
1
Z 11
1
Z 11
Z 12
Z 12
Z 12
Z 12
Equating this result to the ABCD parameters for a transmission line of length 2 and characteristic impedance Z 0 gives the parameters of the equivalent circuit as
1
j Z0
=
,
C
sin 2
= Z 22 = Z 12 A = j Z 0 cot 2.
Z 12 =
(8.110a)
Z 11
(8.110b)
cos 2 + 1
= j Z 0 cot .
sin 2
(8.111)
The 1: 1 transformer provides a 180 phase shift, which cannot be obtained with the
T-network alone; since this does not affect the amplitude response of the filter, it can be
discarded. For /2 the series arm impedances of (8.111) are near zero and can also be
ignored. The shunt impedance Z 12 , however, looks like the impedance of a parallel resonant circuit for /2. If we let = 0 + , where = /2 at the center frequency
0 , then we have 2 =
=
/v p = (0 + )/0 = (1 + /0 ), so (8.110a)
can be written for small as
Z 12 =
j Z0
j Z 0 0
.
sin (1 + /0 )
( 0 )
(8.112)
j L02
,
2( 0 )
(8.113)
with 02 = 1/LC. Equating this to (8.112) gives the equivalent inductor and capacitor values as
L=
2Z 0
,
0
(8.114a)
C=
1
=
.
2
2Z
0 L
0 0
(8.114b)
The end sections of the circuit of Figure 8.45b require a different treatment. The lines
of length on either end of the filter are matched to Z 0 and so can be ignored. The end
inverters, J1 and J N +1 , can each be represented as a transformer followed by a /4 section
434
of line, as shown in Figure 8.45d. The ABCD matrix of a transformer with a turns ratio N
in cascade with a quarter-wave line is
j Z0
1
0
0
j
Z
0
0
A B
N .
j
(8.115)
= jN
= N
C D
0
0 N
0
Z0
Z0
Comparing this to the ABCD matrix of an admittance inverter [part of (8.104)] shows that
the necessary turns ratio is N = J Z 0 . The /4 line merely produces a phase shift and so
can be ignored.
Using these results for the interior and end sections allows the circuit of Figure 8.45b
to be transformed into the circuit of Figure 8.45e, which is specialized to the N = 2 case.
We see that each pair of coupled line sections leads to an equivalent shunt LC resonator,
and an admittance inverter occurs between each pair of LC resonators. Next, we show
that the admittance inverters have the effect of transforming a shunt LC resonator into
a series LC resonator, leading to the final equivalent circuit of Figure 8.45f (shown for
N = 2). This will then allow the admittance inverter constants, Jn , to be determined from
the element values of a low-pass prototype. We will demonstrate this for the N = 2 case.
With reference to Figure 8.45e, the admittance just to the right of the J2 inverter is
C2
1
0
2
+ Z 0 J3 = j
+ Z 0 J32 ,
jC2 +
jL 2
L 2 0
since the transformer scales the load admittance by the square of the turns ratio. Then the
admittance seen at the input of the filter is
J22
1
1
+
Y = 2 2 jC1 +
jL 1
J1 Z 0
j C2 /L 2 [(/0 ) (0 /)] + Z 0 J32
J22
1
C1
0
= 2 2 j
+
. (8.116)
L 1 0
J1 Z 0
j C2 /L 2 [(/0 ) (0 /)] + Z 0 J32
These results also use the fact, from (8.114), that L n Cn = 1/02 for all LC resonators.
Now the admittance seen looking into the circuit of Figure 8.45f is
1
1
+
jL
1
jL
2 + 1/jC2
+ Z 0
C1
0
1
= j
,
+
L 1 0
j L
2 /C2
[(/0 ) (0 /)] + Z 0
Y = jC1
+
(8.117)
which is identical in form to (8.116). Thus, the two circuits will be equivalent if the following conditions are met:
C1
C1
1
=
,
(8.118a)
L
1
J12 Z 02 L 1
J12 Z 02 C2
L
2
=
,
(8.118b)
L2
C2
J22
J12 Z 03 J32
J22
= Z0.
(8.118c)
435
(8.119a)
(8.119b)
L
2 =
g2 Z 0
,
0
(8.119c)
C2
=
,
0 g2 Z 0
(8.119d)
where = (2 1 )/0 is the fractional bandwidth of the filter. Then (8.118) can be
solved for the inverter constants with the following results (for N = 2):
J1 Z 0 =
C1 L
1
L 1 C1
1/4
=
C2 C2
L 2 L
2
J2
J3 Z 0 =
=
.
J1
2g2
J2 Z 0 = J1 Z 02
1/4
,
2g1
(8.120a)
=
,
2 g1 g2
(8.120b)
(8.120c)
After the Jn are found, Z 0e and Z 0o for each coupled line section can be calculated from
(8.108).
The above results were derived for the special case of N = 2 (three coupled line sections), but more general results can be derived for any number of sections, and for the case
where Z L = Z 0 (or g N +1 = 1, as in the case of an equal-ripple response with N even).
Thus, the design equations for a bandpass filter with N + 1 coupled line sections are
,
(8.121a)
Z 0 J1 =
2g1
for n = 2, 3, . . . , N ,
Z 0 Jn =
2 gn1 gn
.
Z 0 J N +1 =
2g N g N +1
(8.121b)
(8.121c)
The even- and odd-mode characteristic impedances for each section are found from (8.108).
Design a coupled line bandpass filter with N = 3 and a 0.5 dB equal-ripple response. The center frequency is 2.0 GHz, the bandwidth is 10%, and Z 0 = 50 .
What is the attenuation at 1.8 GHz?
Solution
The fractional bandwidth is = 0.1. We can use Figure 8.27a to obtain the attenuation at 1.8 GHz, but first we must use (8.71) to convert this frequency to the
normalized low-pass form (c = 1):
0
1 1.8 2.0
1
= 2.11.
gn
Z 0 Jn
Z 0e ( )
Z 0o ( )
1
2
3
4
1.5963
1.0967
1.5963
1.0000
0.3137
0.1187
0.1187
0.3137
70.61
56.64
56.64
70.61
39.24
44.77
44.77
39.24
Note that the filter sections are symmetric about the midpoint. The calculated
response of this filter is shown in Figure 8.46; passbands also occur at 6 GHz,
10 GHz, etc.
Many other types of filters can be constructed using coupled line sections; most of
these are of the bandpass or bandstop variety. One particularly compact design is the interdigitated filter, which can be obtained from a coupled line filter by folding the lines at their
midpoints; see references [1] and [3] for details.
0
10
Attenuation (dB)
436
20
30
40
50
1.0
1.5
2.0
2.5
3.0
Frequency (GHz)
FIGURE 8.46