Академический Документы
Профессиональный Документы
Культура Документы
C U ATU
FO E FE
S SU
I
Richard J. Cameron
U
ntil the early 1970s, nearly all filter synthesis
techniques were based on the extraction of
electrical elements (lumped capacitors and
inductors, transmission line lengths) from
the polynomials that represented the filter’s
electrical performance in mathematical terms. This was per-
fectly adequate for the technologies and applications that
were available at the time, and many important contribu-
tions were made to the art of advanced filter transfer and
reflection polynomial generation, and then their conversion
to electrical component values corresponding to the filter
technologies that were available at this time [1]–[13].
In the early 1970s a revolution in telecommunication sys-
tems and available technology was taking place. The first
satellite telecommunication systems were in operation and
demand for their services was growing enormously, mean-
ing that the ratio-frequency (RF) spectrum allocated to sat-
ellite communication systems had to be pushed to higher
October 2011 43
In the early 1970s a revolution the initial design of a filter network (the prototype
network) assumes frequency-independent coupling
in telecommunication systems elements and lossless and dispersionless resonators,
and available technology was these real-world effects may be accommodated when
taking place. The first satellite analyzing the matrix for filter performance predic-
tion. Different characteristics may be allocated to dif-
telecommunication systems were ferent elements if there is a mix of technologies within
in operation and demand for their the filter. Another advantage is the ability to recon-
services was growing enormously. figure the coupling matrix through similarity trans-
forms to arrive at a different coupling arrangement,
corresponding to the available coupling elements of
network, which forms the basis of the microwave the particular microwave structure selected for the
filter design, involves first generating the polynomi- application. This can be done without going right back
als, which represent the transfer and reflection char- to the beginning of the network synthesis process to
acteristics (S21 and S11), required from the network in start again on a different network synthesis route, as
order to satisfy the specifications. Then the prototype would have to be done if the classical element extrac-
elements (capacitors, inductors, lengths of transmis- tion method was being used. More recently, coupling
sion line) are extracted from the polynomials. As the matrix synthesis theory has been advanced to include
network is built up element by element, the polynomi- asymmetric filtering characteristics, which have
als decrease in degree until, when the network is fully become important for terrestrial telecommunication
synthesized, the coefficients of the polynomials are all systems, particularly mobile telephony systems.
zero apart from some constants. Different extraction Because of the prevalence of the coupling matrix in
strategies are required to obtain the specific topol- microwave filter design, this article will concentrate on
ogy of the prototype network that corresponds to the techniques for the synthesis and then the reconfigura-
electrical components available in the technology it is tion of the coupling matrix, ready for realization in a
intended to realize the microwave filter with. variety of microwave structures. Firstly the method for
The coupling matrix network synthesis method the generation of advanced polynomial filtering func-
provides an alternative to the extraction of electrical tions will be briefly outlined, followed by the synthe-
elements one by one. The coupling matrix is simply sis of one of the canonical networks—the transversal
a representation of the network, which may be used matrix. Then the reconfiguration of the transversal
for the initial design, then the tuning, modelling matrix into various forms for realization in a variety of
and analysis of microwave filter performance. One microwave structures will be discussed. Some exam-
important feature is the one-to-one correspondence ples are given to clarify certain aspects of the design
between the elements of the coupling matrix and the processes, and references cited if further information is
individual physical components of the filter. Although required by the reader.
M1,N
M1,N – 1
M1,j
M1,i Mi,N
Mi,N – 1
Mi,j
1F 1F 1F 1F 1F 1F
i1 i2 ii ij iN – 1 iN
1H ½H ½H ½H ½H ½H ½H ½H ½H 1H
(1) (2) (i ) (j ) (N – 1) (N )
M2,j
M2,N – 1
M2,N
44 October 2011
The Coupling Matrix The coupling matrix network synthesis
The basic circuit model that was used in [20] was a
bandpass prototype which is a generalized multicou- method provides an alternative to
pled network as shown in Figure 1. The circuit com- the extraction of electrical elements
prises a cascade of lumped element series resonators one by one.
intercoupled through transformers; each resonator
comprising a capacitor of 1 Farad in series with the self
inductances of the main-line transformers, which total 1 each loop, the capability of the circuit may be extended
Henry within each loop. This gives a centre frequency to include asymmetric cases (see Figure 2). These have
of 1 radian/second, and the couplings are normalized to been finding increasing application recently, as the RF
give a bandwidth of 1 rad/sec. In addition every loop is frequency spectrum becomes more crowded and rejec-
theoretically coupled to every other loop through cross- tion specifications more severe. The FIR (sometimes
mutual couplings between the main-line transformers. referred to as a self-coupling) represents a frequency
This network may be represented by an N 3N cou- offset of the resonator it is associated with, and its value
pling matrix where N is the number of resonators (the is entered along the diagonal of the coupling matrix.
degree or order) of the filter. The elements of the matrix Because the inverters are also frequency-invariant and
contain the values of the couplings between each of the there are no self-inductors, the network of Figure 2 may
resonators; between sequentially numbered resonator now be considered as a low-pass prototype, which sim-
nodes (main-line couplings), and nonadjacent nodes plifies the synthesis process somewhat.
(cross-couplings). Because the electrical elements of the The N12 short-circuit admittance matrix 3 yr 4 for
network are passive and reciprocal, the matrix is sym- the network of Figure 2 may be separated out into its
metrical about its principal diagonal. To more closely purely resistive and purely reactive parts:
3 yr 4 5 3 G 4 1 3 jM 1 U 4 5 3 G 4 1 3 y 4
represent a microwave circuit, the transformers may be
(1)
replaced by immittance inverters (90° lengths of trans-
mission line), which approximate the electrical character- where the purely real matrix 3 G 4 contains the conduc-
istics of many microwave coupling devices. By placing tive terminations GS and GL of the network and the
an inverter at each end of the network, the input and purely reactive admittance 3 y 4 5 3 jM 1 U 4 is the sum
output couplings of the filter may also be represented of the coupling matrix M and the diagonal matrix U
(Figure 2). With the extra inverters, the matrix increases which contains the frequency variable s 1 5 jv 2 , except
to N12 3 N12 in size—the so-called N12 coupling for USS and ULL which are zero.
matrix—and becomes the dual network of Figure 1. The N12 coupling matrix [M] contains the val-
The circuit in Figure 1 will only support symmetric ues of all the couplings in the network including
filtering characteristics, but with the addition of a series- the input/output couplings (which may connect to
connected frequency-invariant reactance (FIR) within internal resonators as well as the first and Nth). The
MS,L
MS,N
MS, j
MS,i Mi,L
Mi,N
C1 Ci Cj CN
GS MS1 M1i Mij Mj,N MN,L GL
(S ) jB1 (1) jBi (i ) jBj ( j ) jBN (N ) (L )
M1,j
M1,N
M1,L
Figure 2. Multicoupled network—equivalent low-pass prototype modified to include FIRs and immittance inverters.
(Reprinted with permission from [22].)
October 2011 45
and configure it such that its nonzero entries coincide
The filter design process begins
with the available coupling elements of the structure
with the generation of the rational it is intended to use for realizing the filter response.
polynomials embodying the transfer Finally, the dimensions of the coupling elements are
calculated from the coupling matrix values.
and reflection characteristics.
The procedure is illustrated in Figure 4 for a sixth-
degree characteristic with two TZs and realized in
diagonal contains the values of the frequency invari- coupled waveguide resonator technology. The direct
ant reactances which represent resonator frequency correspondence between the elements of the coupling
offsets (the negative values of the FIRs in Figure 2), matrix and the physical filter components is highlighted.
which are necessary for asymmetric characteristics.
Figure 3(a) shows a canonical coupling matrix which Generation of Transfer
has all couplings present, and Figure 3(b) an example and Reflection Polynomials
of a typical coupling and routing diagram, represent- In modern telecommunication, radar and broadcast
ing a possible inter-resonator coupling arrangement. systems, where the allocated RF frequency spectrum
has become very congested, the specifications on
Synthesis Procedure performance from the component microwave filters
The filter design process begins with the generation have become increasingly stringent. For these applica-
of the rational polynomials embodying the transfer tions the Chebyshev class of filtering characteristic is
and reflection characteristics S21 and S11, which satisfy very suitable on account of the inherent equiripple in-
the rejection and in-band specifications of the appli- band return loss level and the ability to build in TZs to
cation. Once these have been obtained, the next step provide high close-to-band rejection levels or in-band
in the process is to synthesize the coupling matrix group delay equalization, or both within the same
filtering function. Moreover, the TZs may be placed
asymmetrically to optimally comply with asymmetric
specifications. A method for generating the low-pass
S 1 2 3 4 L
prototype polynomials for the Chebyshev class of filter
S MSS MS1 MS2 MS3 MS4 MSL function is outlined below.
For any two-port lossless filter network composed
1 MS1 M11 M12 M13 M14 M1L of a series of N intercoupled resonators, the transfer
and reflection functions may be expressed as a ratio of
2 MS2 M12 M22 M23 M24 M2L
two polynomials [21]:
F 1 v 2 /eR P 1 v 2 /e
3 MS3 M13 M23 M33 M34 M3L
S11 1 v 2 5 S21 1 v 2 5
E1v 2 E1v 2
, (2)
4 MS4 M14 M24 M34 M44 M4L
46 October 2011
where lobes at 30 dB each on the upper side of the passband. The
polynomials and corresponding singularities of P 1 s 2 F 1 s 2
xn 1v2 5 and E 1 s 2 are given in Table 1, and plots of the transfer and
12vv n
,
v2v n
rejection characteristic are shown in Figure 5
and v n are the positions of the nfz finite-position TZs,
and the remaining N 2 nfz zeros at v 5 6 `. Coupling Matrix Generation
For a prescribed set of TZs that make up the poly- The second step in the synthesis procedure is to calculate
nomial P 1 v 2 and a given equiripple return loss level, the values of the coupling elements of a canonical cou-
the reflection numerator polynomial F 1 v 2 may be built pling matrix from the transfer and reflection polynomi-
up with an efficient recursive technique, and then als. Three forms of the canonical matrix are commonly
the polynomial E 1 v 2 found from the conservation of used—the folded [16], arrow [25] or transversal [24]. The
energy principle [21]–[23]. latter is particularly easy to synthesize, and the other two
An example of this synthesis method is given in [21] may be derived from it quite simply by applying a formal
for a fourth-degree prototype with 22 dB return loss series of analytically calculated similarity transforms.
level and two imaginary axis TZs at s 5 1j 1.3217 and The transversal coupling matrix comprises a series of
1j 1.8082 which are so positioned to give two rejection N individual first-degree low pass sections, connected
Prototype Polynomials S 1 2 3 4 5 6 L
S⎡ x ⎤
1 x
⎢ x x
⎥
S21(s ) =
P (s ) /ε ⎢ ⎥
E(s) 2⎢ x x x x ⎥
⎢ ⎥
3⎢ x x x ⎥
4⎢ ⎥
F (s ) /εR
S11(s ) = x x x
E(s) ⎢ ⎥
Folded 5⎢ x x x x ⎥
Coupling 6 ⎢ x x x⎥
Matrix ⎢ ⎥
Synthesis L ⎣⎢ x ⎦⎥
Reconfiguration of
Coupling Matrix
Resonant Cavity S 1 2 3 4 5 6 L
Tuning Offset S⎡ x ⎤
⎢
1 x x x
⎥
⎢ ⎥
2⎢ x x x ⎥
⎢ ⎥
3⎢ x x x x ⎥
Mainline Coupling
4⎢ x x x ⎥
⎢ ⎥
5⎢ x x x ⎥ Input/Output
6⎢ x x x x⎥ Coupling
Cross Coupling ⎢ ⎥
(Negative) L ⎣⎢ x ⎦⎥
Realization
2 3 6 L
S 1 4 5
Figure 4. Microwave filter design process—synthesis of the polynomials for the transfer and reflection function, synthesis
of canonical coupling matrix, reconfiguration of coupling matrix, realization in microwave coupled-resonator technology.
(Reprinted with permission from [22].)
October 2011 47
The approach that is employed
TABLE 1. 4–2 asymmetric Chebyshev filtering function with two
to synthesize the N12 transversal
prescribed transmission zeros.
coupling matrix is to construct the
Transfer and Reflection Function Polynomials. two-port short-circuit admittance
parameter matrix 3 YN 4 for the overall
si , i 5 E(s) F(s) P(s) network in two ways; the first from
0 20.1268 2 j2.0658 10.0208 2j2.3899
the coefficients of the rational poly-
nomials of the transfer and reflec-
tion scattering parameters S21 1 s 2 and
1 12.4874 2 j3.6255 2j0.5432 13.1299
2 13.6706 2 j2.1950 10.7869 j1.0 S11 1 s 2 which represent the character-
3 12.4015 2 j0.7591 2j0.7591 istics of the filter to be realized, and
4 11.0 11.0 the second from the circuit elements
of the transversal array network. By
equating the 3 YN 4 matrices as derived
Corresponding Singularities.
Reflection Zeros Transmission Transmission/Reflection
(Roots of F(s)) Zeros (Prescribed) Poles (Roots of E(s))
by these two methods, the elements
of the coupling matrix associated
1 2j0.8593 1j1.3217 20.7437 2 j1.4178 with the transversal array network
2 2j0.0365 1j1.8082 21.1031 1 j0.1267 may be related to the coefficients of
3 1j0.6845 j` 20.4571 1 j0.9526 the S21 1 s 2 and S11 1 s 2 polynomials.
1j0.9705 20.0977 1 j1.0976
In [24] it is explained how the
matrix 3 YN 4 is built up very simply
4 j`
3 YN 4 5 j c
N
d1a
the source and load terminations, respectively. The cir- 0 MSL 1
k51 1 sCk 1 jBk 2
cuit of the kth low-pass section is shown in Figure 6(b) MSL 0
M2Sk
3c d.
MSk MLk
(5)
MSk MLk M2Lk
Rejection/Return Loss (dB)
0
4-2 Asymmetric
Chebyshev Filter Construction of the N+2 Transversal Matrix
Now the two expressions for 3 YN 4 , the first in terms
10
S11
S21
20 of the residues r21k and r22k and the eigenvalues lk,
22 dB which have already been derived from the S21 and S22
30 polynomials of the desired filtering function, and the
second in terms of the circuit elements of the transver-
40
–3 –2 –1 0 1 2 3 4 sal array, may be equated. This leads to the following
Prototype Frequency (rad/s) relationships between the residues and the transversal
coupling matrix elements:
Bk 1 { Mkk 2 5 2lk
Figure 5. Low-pass prototype transfer and reflection
Ck 5 1 and
characteristics of the 4-2 asymmetric Chebyshev filter, with
MSL 5 K `
two prescribed transmission zeros at s1 = +j1.3217 and
2
s2 = +j1.8082. (Reprinted with permission from [22].) MLk 5 r22 k and MSk MLk 5 r21k .
48 October 2011
Therefore,
MLk 5 "r22k N
and N–1
October 2011 49
The eigenvalues of the matrix M1 after the trans-
S 1 2 3 4 L form are the same as those of the original matrix M0,
S sc m sx which means that an arbitrarily long series of trans-
1 m sc m sx ax forms with arbitrarily defined pivots and angles may
2 m sc m ax be applied, starting with M0 . Each transform in the
3 m sc m series takes the form
4 sx ax m sc m Mr 5 Rr # Mr21 # Rtr r 5 1, 2, 3, c, R, (8)
L sx ax m sc
and, under analysis, the matrix MR resultant at the end
(a) of the series of transforms will yield exactly the same
performance as the original matrix M0. With an analyti-
S 1 2
cally calculated angle u r and judiciously chosen pivot
positions, coupling elements may be zeroed (annihi-
lated) and others created to arrive at a coupling matrix
whose nonzero entries correspond to the available
interresonator coupling elements of the filter structure
L 4 3
it is intended to realize the required transfer and reflec-
Main Line Coupling tion characteristics with.
Cross Coupling By applying a series of rotations, the N12 trans-
Resonator Node versal matrix may be reduced to the folded form. The
Nonresonant Node (NRN)
pivots and a formula for calculating the angle of such a
(b) sequence is given below for a fourth-degree example,
annihilating the elements MS4, MS3, MS2, M2L , M3L ,
Figure 8. N+2 folded canonical network, fourth degree
and finally M13 in sequence (see Table 2). The resulting
example. (a) Coupling matrix: sc and ax couplings are
generally zero for symmetric characteristics. (b) Coupling folded configuration coupling and routing schematic
and routing schematic. Possible nonzero couplings: sc=self- is shown in Figure 8(b).
coupling, m=main-line coupling, ax=asymmetric cross- The final values and positions of the elements in the
coupling, and sx=symmetric cross-coupling. Couplings are cross diagonals of the folded coupling matrix (the cross
equivalued about the principal diagonal. couplings) are automatically determined—no specific
action to annihilate couplings within them needs to be
taken if they are not needed to realize the particular
M1 5 R1 # M0 # Rt1, (7)
transfer function under consideration.
50 October 2011
angles for degrees 6, 8, and 10 as shown
in Table 3. The sequences for degrees TABLE 2. Fourth-degree example: pivots and angles of the
similarity transform sequence for the reduction of the
12 and 14 may be found in refs. [29],
transversal (or any other) matrix to the folded configuration.
[30], and odd degrees may be accom- N21
Total number of transforms R 5 a n5 n 5 6.
modated by using the next-lowest even 1
degree, e.g., for a ninth-degree use the Element to u r 5 tan21 1 cMk l / Mmn 2
eighth-degree sequence. Transform Pivot be Annihilated
Although the in-line topology of Number r [i, j] in Matrix M0 k l m n c
the asymmetric in-line realization is 1 [3, 4] MS4 in row S S 4 S 3 –1
exactly the same as for the symmet-
2 [2, 3] MS3 “ S 3 S 2 –1
ric equivalent, the values of the in-
line coupling matrix, and therefore 3 [1, 2] MS2 “ S 2 S 1 –1
the dimensions of the corresponding 4 [2, 3] M2L in column L 2 L 3 L +1
physical coupling elements, are not 5 [3, 4] M3L “ 3 L 4 L +1
equal-valued about the physical cen- 6 [2, 3] M13 in row 1 1 3 1 2 –1
tre of the structure [29]. Although this
means more design effort to develop
and manufacture a working filter, Table 3. Pivot positions and rotation angles for the general
there is an advantage in that there asymmetric in-line realization, for degrees 6, 8, and 10.
are no restrictions on the pattern of
u r 5 tan21 1 cMk l / Mmn 2
TZs that the prototype may incorpo- Degree Rotation Pivot
rate (apart from the usual conditions, N No. r [i, j ]
k l m n c
i.e., the minimum path rule must be
obeyed, and symmetry of the pattern 6 1 [2, 4] 2 5 4 5 +1
of TZs about the imaginary axis (uni- 1 [4, 6] 3 6 3 4 –1
tary condition), and about the real 2 [2, 4] 2 7 4 7 +1
8
axis (symmetric characteristics) must 3 [3, 5] 2 5 2 3 –1
4 [5, 7] 4 7 4 5 –1
be preserved). Moreover the compu-
tations required to produce the asym- 1 [4, 6] 4 7 6 7 +1
metric in-line configuration are less 10 2 [6, 8] 3 8 3 6 –1
3 [7, 9] 6 9 6 7 –1
complex. Figure 10(a) and (b) shows
S 1 2 3 4 S 1 4 5 8 L
L 8 7 6 5 2 3 6 7
(a) (b)
1 4 5 8
M23 M67
M12 M34 M36 M56 M78
M8L
M14 M58
M45
MS1
2 3 6 7
Figure 10. Eighth-degree network: (a) cross-coupled folded configuration (b) after conversion to in-line topology. (c) Possible
realization in cylindrical dual-mode cavities. (Reprinted with permission from [22].)
October 2011 51
For most coupled-resonator minimum path rule, the Pfitzenmaier configuration is
able to realize N 2 2 TZs, the same as the folded struc-
technologies, the couplings of the ture. The coupling and routing diagram for a sixth-
transversal matrix are physically degree example is shown in Figure 11.
impractical or impossible to realize. The Pfitzenmaier configuration may be easily
obtained for any even degree symmetric characteris-
tic $ 6 by using a sequence of coupling matrix rota-
tions [28]. Unlike the asymmetric in-line realization,
L S the pivots and angles of the rotations in the sequence
S 1 2 3 2 may be defined with simple equations. Starting with
6 1 the folded matrix, a series of R 5 1 N 2 4 2 / 2 rotations is
applied according to (9) after which the Pfitzenmaier
configuration is obtained.
For the rth rotation, pivot 5 3 i, j 4 and associated
L 6 5 4 5 4 3
(a) (b) angle 5 u r , where:
i5r11
Figure 11. Pfitzenmaier configuration—6-4 symmetric
filtering characteristic. (a) Original folded configuration. j5N2i S r 5 1, 2, 3, c, R (9)
(b) After transformation to Pfitzenmaier configuration. u r 5 tan21 1 2Mi,N2r/ Mj,N2r 2
(Reprinted with permission from [22].)
and N is the degree of the filter (N 5 even integer $ 6 2 .
the reconfiguration of the coupling/routing diagram
for an eight–four characteristic in folded form to an in- Advanced Configurations
line form, and Figure 10(c) illustrates a possible realiza- In this section some advanced coupling matrix con-
tion in cylindrical dual-mode cavities. figurations will be considered. The first is the cul-de-
sac configuration, which is derived from the folded
Pfitzenmaier Configurations coupling matrix, and has the principal advantage that
Another configuration that is able to avoid the input/ it needs no diagonal cross couplings even if realizing
output isolation problem associated with the folded asymmetric characteristics. The second is the cascaded
configuration in a dual-mode structure was intro- trisection configuration which is derived from the
duced by Pfitzenmaier [31] for sixth-degree symmet- arrow canonical matrix. This has applications for the
ric filtering characteristics. In [31] it is shown that the generation of cascaded n-tuplets and box filters
synthesized sixth-degree circuit can be transformed
(not using coupling matrix methods) to a topology The Cul-de-Sac Configuration
where the input and output resonances (1 and 6) are The cul-de-sac configuration [24], [32] in its basic form
in adjacent cavities of the dual-mode structure, thereby is restricted to double-terminated networks and will
avoiding the isolation problem. Furthermore, because realize a maximum of N 2 3 TZs. Otherwise it will
it is possible to directly cross-couple resonances 1 and accommodate any even- or odd-degree symmetric or
6, the signal only has two resonances to pass through asymmetric prototype. Moreover its form lends itself to
between the input and output, and therefore by the a certain amount of flexibility in the physical layout of
its resonators.
A typical cul-de-sac configuration is shown in
5 4 3 2 10
Figure 12(a) for a tenth-degree prototype which will
L accommodate a maximum of seven TZs. There is a cen-
tral core of a quartet of resonators in a square forma-
S – tion [1, 2, 9, and 10 in Figure 12(a)], straight-coupled to
1 9 8 7 6 each other (i.e., no diagonal cross-couplings). One of
(a)
these couplings is always negative; the choice of which
3 2 7 one is arbitrary. The entry to and exit from the core
– L
quartet are from opposite corners of the square [1 and
10, respectively, in Figure 12(a)].
S Some or all of the rest of the resonators are strung
1 6 5 4
(b) out in cascade from the other two corners of the core
quartet in equal numbers (even-degree prototypes) or
Figure 12. Cul-de-Sac network configurations: one more than the other (odd degree prototypes). The
(a) tenth degree (7 TZs max) and (b) seventh degree last resonator in each of the two chains has no output
(4 TZs max) [24]. coupling, hence the nomenclature cul-de-sac for this
52 October 2011
configuration. An example of an odd degree character-
istic is shown in Figure 12(b) (seventh-degree).
To transform the folded coupling matrix to the cul- 1 M12 2 M23 3 M34 4
de-sac form, a series of rotations is applied according MS1 M26 M35
to the following procedure: M45
M36
i 5 1 N 1 2 2 / 2 2 r and j 5 N/ 2 1 r M7L
2 Mij
Angle = u r 5 tan21a b (cross-pivot
1 3 M23 2 7
2 1 Mjj 2 Mii 2 M27
rotation) (10a)
M12
For N odd and r 5 1, 2, 3, . . . , 1 N 2 3 2 /2 1 M16 6 M56 5 M45 4
For rotation # r: MS1
and j 5 1 N 1 1 2 / 2 1 r
Angle = ur 5 tan21 1 Mi, j21 / Mj21,j 2
Figure 13. 7-1-2 asymmetric filter example—coaxial cavity
(10b)
realizations: (a) folded network configuration and (b) cul-
de-sac configuration. (Reprinted with permission from [22].)
For example, for a seventh-degree filtering function
with three TZs: however that this topology will realize two fewer TZs
Number of rotations = 2 than the basic version.
Rotation 1: i 5 3 j 5 5, therefore pivot 1 = [3, 5] and If the sequence is continued on for one further rota-
angle u 1 5 tan21 1 M34 / M45 2 tion than the basic sequence, the input and output cou-
Rotation 2: i 5 2 j 5 6, therefore pivot 2 = [2, 6] and plings will be included in the core quartet as shown
angle u 2 5 tan21 1 M25 / M56 2 in Figure 14(b) for the eighth-degree example. This
will realize two more TZs than the basic cul-de-sac
Figure 13 shows a realization in coaxial resonator tech- (i.e., seven for this eighth-degree prototype), but if it is
nology, firstly configured in folded form (a) and after convenient to include the source-load coupling MSL as
reconfiguration to the cul-de-sac form (b). In the cul- shown in Figure 14(b), then all eight TZs may be real-
de-sac form all the couplings will of the same sign except ized (fully canonical network). If the original prototype
for one in the central core quar-
tet—which one is arbitrary.
Also for this case where the 4 3 7 8 L
number of TZs is less than –
the maximum permissible, all
the couplings between resona- S
tors in the core quartet have the 1 2 6 5
same absolute value. (a)
4 3 2 1 L
Alternative Cul-de-Sac
Configurations –
In some cases it may give a
more convenient configura- S 8 7 6 5
tion and better input-output (b)
isolation if the final rotation 4 3 2 1 N2 1/ 2 L
in the sequence is omitted.
Even-Mode Branch
Such an example is shown 1/ 2 – 1/ 2
Odd-Mode Branch
in Figure 14(a) for an eighth-
degree example, which gives a S 1/ 2 N1 8 7 6 5
convenient rectangular topol- (c)
ogy and at least five resona-
tors between input and output Figure 14. Three alternative forms for the cul-de-sac configuration: (a) indirect-
as compared with the basic coupled, (b) fully canonical form, and (c) rat-race coupled even- and odd-mode networks.
cul-de-sac. It should be noted (Reprinted with permission from [22].)
October 2011 53
3-D) layout possibilities. However the simple topology
Coaxial tends to produce a rather sensitive device in practice.
Line ≈ λ /4
S L Bandstop Filters
MSL
To generate a bandstop characteristic from the regular
MS 1 M4L low-pass prototype polynomials it is only necessary to
1 4 1 M14 4 exchange the reflection and transfer functions (includ-
ing the constants) [22]:
2 3 M12 M34
P 1 s 2 /e F 1 s 2 / eR
S11 1 s 2 5 S21 1 s 2 5
E1s2 E1s2
(a) . (11)
2 M23 3
54 October 2011
the square-shaped core quartet
of couplings, with the source
and load terminals at adjacent S L MSL
a
corners at the input/output ≈ λg /4
end, while the other resonators MS 1 M4L
are strung out in two chains 2 1 4 3 2 M12 1 M14 4 M34 3
from the other two corners, in
equal numbers if N is even
(a) (b)
and one more than the other
0
if N is odd. There are no diago-
4-2 Asymmetric
October 2011 55
Ci + jBi
i 1 N 1
S MS 2 2 M23 3 M35 5 S MS 2
M23 M34
3
(e) (f)
Figure 19. Coupling and routing diagrams for trisections: (a) internal, (b) source-connected, (c) load-connected (d) canonical,
(e) nonconjoined cascaded, (f) conjoined-cascaded. (Reprinted with permission from [22].)
matrix, or to merge them to create quartet sections (two Figure 20(a) gives an example of the coupling and
trisections) or quintet sections (three trisections), etc. routing diagram for a fifth-degree fully canonical fil-
Following on below an efficient procedure for synthe- tering circuit, showing clearly why this configuration
sizing a cascade of trisections will be outlined [37]. is referred to as the wheel, with the main-line cou-
plings forming the (partially incomplete) rim and the
Synthesis of the Arrow cross-couplings and input/output coupling forming
Canonical Coupling Matrix the spokes. Figure 20(b) shows the corresponding cou-
The folded cross-coupled circuit and its corresponding pling matrix where the cross-coupling elements are all
coupling matrix was introduced above as one of the in the last row and column, and together with the main
basic canonical forms of the coupling matrix, capable line and self couplings on the main diagonals, give the
of realizing N TZs in an Nth-degree network. A second matrix the appearance of an arrow pointing downwards
form was introduced by Bell [25] in 1982, which later towards the lower right corner of the matrix. The arrow
become known as the wheel or arrow form. As with the matrix may be synthesized from the canonical transver-
folded form, all the main-line couplings are present, sal matrix with a formal sequence of rotations, similar to
and in addition the source terminal and each resonator that for the folded matrix [22].
node is cross-coupled to the load terminal. The basis of the trisection synthesis procedure relies
on the fact that the value of
the determinant of the self and
mutual couplings of the trisec-
2 S 1 2 3 4 5 L
tion evaluated at v 5 v 0, the
1
S MS1 MSL position of the TZ associated
with the trisection, is zero:
1 MS 1 M11 M12 M1L
det ` ` 50
3 Mk21, k Mk21, k11
2 M12 M22 M23 M2L
v 0 1 Mk, k Mk, k11
S L 3 M23 M33 M34 M3L (12)
where k is the number of the
4 M34 M44 M45 M4L
4 middle resonator of the tri-
5 M45 M55 M5L section.
Figure 21 gives the topology
5 L MSL M1L M2L M3L M4L M5L
and coupling matrix for the
(a) (b) fourth-degree filter with 22 dB
return loss and two TZs at
Figure 20. Fifth-degree wheel or arrow canonical circuit. (a) Coupling and routing diagram v 01 5 1.8082 and v 02 5 1.3217
(wheel). (b) N12 coupling matrix (arrow). (Reprinted with permission from [22].) that was used as an example
56 October 2011
Trisection S-1-2 (ω01) Trisection 3-4-L (ω02)
S 1 2 3 4 L
1 4
S ⎡ 0.0 0.9221 0.5921 0 0 0 ⎤
MS 1 ⎢ ⎥
M12 M34 M4L 1 ⎢ 0.9221 –0.8333 0.6259 0 0 0
⎥
2 ⎢ 0.5921 0.6259 0.6022 0.8634 0 0 ⎥
⎢ ⎥
S MS 2 2 M23 3 M3L L 3 ⎢ 0 0 0.8634 0.5812 0.2590 0.8471 ⎥
4 ⎢ 0 0 0 0.2590 –1.1092 0.6952 ⎥
⎢ ⎥
L ⎢⎣ 0 0 0 0.8471 0.6952 0.0 ⎥⎦
(a) (b)
Figure 21. Fourth-degree filter with two transmission zeros realized as trisections. (a) Coupling/routing diagram.
(b) Coupling matrix. (Reprinted with permission from [22].)
earlier, now configured with two trisections to realize where the superfix (0) indicates that the coupling
the two TZs. The shaded areas in the matrix indicate values are taken from the original arrow matrix M102.
the couplings associated with each trisection. The trisection may then be pulled up the diagonal
Once the arrow coupling matrix has been formed, of the matrix with further rotations such as pivot
the procedure to create the first trisection realizing 3 N22, N21 4 and angle u 12 5 tan21 1 MN22, 112
N / MN21, N 2
112
the first TZ at v 5 v 01 begins with conditioning the until it is in its desired position. The procedure is illus-
matrix with the application of a rotation at pivot trated in Figure 22 for an asymmetric eighth degree
3 N21, N 4 and an angle u 01 to the original arrow example with four TZs.
matrix M102 . Now the process may be repeated for the second
The rotation angle u 01 is given by (13): trisection at v 5 v 02, and so on until a cascade of trisec-
102
tions is formed, one for each of the TZs in the original
MN21,
u 01 5 tan21 c d
N
(13) prototype, as shown in Figure 23(a). The trisections
102
v 01 1 MN, N may be realized directly if it is convenient to do so, e.g.,
1 3 5 7
M12 M34 M56 M78
MS1 M23 M45 M67
S 1 2 3 4 5 6 7 8 L
(a) S MS2 2 M24 4 M46 6 M68 8 M8L L
(a)
2 3 6 7
S 1 2 3 4 5 6 8 L
7
(b)
S 1 4 5 8 L
S 1 2 3 4 5 7 8 L (b)
6
(c)
2 M23 3 6 M67 7
M24
S 2 3 4 5 6 7 8 L
1 M12 M34 M56 M78
(d)
M
1 4 M45 5 57 8
MS1 M8L
Figure 22. 8-2-2 synthesis example. (a) Coupling and
routing diagram of initial arrow coupling matrix. (b) (c)
Conditioning rotation creates first trisection 6-7-8. (c)
Rotation 2 pulls the trisection to position 5-6-7 . . . etc. Figure 23. 8-4 asymmetric filter. (a) Trisection cascade.
(d) Rotation 7 finally creates trisection S-1-2. Note that (b) Merging of trisections. (c) Coaxial resonator realization.
when the trisection is in its final position, the outer cross Trisections S-1-2 and 2-3-4 merged to form quartet 1-2-3-4.
coupling of the arrow formation (M4L ) automatically Trisections 4-5-6 and 6-7-8 merged to form quartet 5-6-7-8.
disappears. (Reprinted with permission from [22].) (Reprinted with permission from [22].)
October 2011 57
the two trisections being merged are realizing TZ pairs
1 3 5 symmetrically located on the real or imaginary axes,
or in quartets with symmetry about both axes, then
Rest of
6 the diagonal couplings [M24 and M57 in Figure 23(b)]
Network
S 2 4 will be zero. It is essential that complex zeros are in
paraconjugate pairs, otherwise unrealizable complex
7 coupling values will result. Figure 23(c) shows a pos-
(a) sible coaxial-resonator realization for the two quartets.
3 This procedure may be extended to form even
higher order sections in cascade, for example three tri-
Rest of
7 sections may be merged to form a quintet section, as
2 4 Network
illustrated in Figure 24.
S 1 2 3 4 5 6 7 8 9 10 L
S 0 0.8388 0 0 0 0 0 0 0 0 0 0
1 0.8388 0 0.7144 0 –0.3894 0 0 0 0 0 0 0
2 0 0.7144 0 0.7832 0 0 0 0 0 0 0 0
3 0 0 0.7832 0 0.5052 0 0 0 0 0 0 0
4 0 –0.3894 0 0.5052 0 0.4423 0 0 0 0 0 0
5 0 0 0 0 0.4423 0 0.5174 0.4206 0 0 0 0
6 0 0 0 0 0 0.5174 0 0 0 0 0 0
7 0 0 0 0 0 0.4206 0 0 0.4574 0 –0.7156 0
8 0 0 0 0 0 0 0 0.4574 0 0.1590 0 0
9 0 0 0 0 0 0 0 0 0.1590 0 0.3871 0
10 0 0 0 0 0 0 0 –0.7156 0 0.3871 0 0.8388
L 0 0 0 0 0 0 0 0 0 0 0.8388 0
(a)
10-5 Dual-Band Filter
0
Rejection/Return Loss (dB)
S 1 4 5 7 10 L 10
20
2 3 6 8 9 30
(b)
40
50
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2
Frequency (rad/s)
(c)
Figure 25. Tenth-degree symmetric dual-band filter. (a) Coupling matrix. (b) Coupling/routing diagram. (c) Rejection/
return loss performance.
58 October 2011
2 M23 3 2 M24 4
1 M13 4 1 3
(a) (b)
Figure 26. 4-1 asymmetric filtering function. (a) Realized with conventional diagonal cross coupling (M13). (b) Realized
with the box configuration [33].
October 2011 59
[40] based on the use of the Groebner basis to solve
2 3 6 7 10 L nonlinear equations has become available through the
software package Dedale-HF, which is accessible on
– – the Internet [41].
S An interesting feature of extended box filters is that
1 4 5 8 9 multiple solutions for the coupling matrices exist for
(a) the same prototype filtering characteristic. This means
2 4 5 8 10 L that optimal coupling matrix values may be chosen for
– the RF technology it is intended to realize the extended
box filter with. The number of real solutions depends
S – on the degree and TZ pattern of the filtering function,
1 3 6 7 9 e.g., 16 for an eighth degree and 58 for a tenth degree
(b) characteristic. The multiple solution feature however
can cause a problem when trying to de-embed cou-
Figure 28. 10-2 asymmetric filter—coupling and routing pling values from a measured performance.
diagrams. (a) Synthesized with two trisections. (b) After
transformation of trisections to two box sections [33]. Conclusions
In this article, some of the more recent developments
Extended Box Sections in the art of filter synthesis have been outlined. These
The basic box section may be extended to enable a have been mainly based on the coupling matrix
greater number of TZs to be realized, but retaining representation of the filter’s coupling arrangements,
a convenient physical arrangement, as shown in Fig- because of the amenity of the coupling matrix to math-
ure 29 [33]. Here the basic fourth degree box section ematical manipulation, and the one-to-one correspon-
is shown and then the addition of pairs of resona- dence of the elements of the coupling matrix to the real
tors to form sixth, eighth and tenth degree networks. filter parameters.
Application of the minimum path rule indicates The methods described in this article probably do
that a maximum of 1, 2, 3, 4, . . . , 1 N22 2 / 2 TZs may not cover all those available today for filter network
be realized by the 4th , 6th, 8th, 10th, . . . , Nth-degree synthesis. It is known that some advanced research
networks respectively. The resonators are arranged work is ongoing into the synthesis of lossy filters [42]–
in two parallel rows with half the total number of [47] which are used to compensate for a low resona-
resonators in each row, input is at the corner at one tor Q and give very linear in-band performance but
end and output from the diagonally opposite corner at the expense of a high-ish insertion loss (not a real
at the other end. Even though asymmetric character- problem in low-power circuits). Also, some work is
istics may be prescribed, there are no diagonal cross- ongoing into the synthesis of coupling matrices for
couplings. wideband devices, where the coupling elements have
There appears to be no regular pattern for deter- a frequency dependency [48]. Some novel synthe-
mining the sequence of rotations to synthesize the sis techniques have recently come available for the
coupling matrix for the extended box sections from the design of circuits incorporating the nonresonant node
folded network or any other canonical network. The (NRN) element, which are useful in high power appli-
networks may be synthesized using optimization tech- cations and for easing the design of dielectric and pla-
niques [22], [38], [39], but more recently a procedure nar circuits [49]–[51].
The Dedale-HF CAD pack-
age mentioned above for cre-
2 4 L 2 3 6 L ating extended box solutions
may also be used to solve
other topological cases which
are not amenable to a series
S 1 3 S 1 4 5
of analytical transforma-
(a) (b)
tions, and which can only be
2 3 6 8 L 2 3 6 7 10 L
solved with an optimization
approach. Another CAD opti-
mization procedure known
S 1 4 5 7 S 1 4 5 8 9 as space mapping has also
(c) (d) become available recently, and
has been widely used for the
Figure 29. Coupling and routing diagrams for extended box section networks. (a) fourth- design of complex filters and
degree (basic box section). (b) sixth-degree. (c) eighth-degree. (d) tenth-degree. multiplexers [52].
60 October 2011
References [31] G. Pfitzenmaier, “An exact solution for a six-cavity dual-mode el-
[1] O. Brune, “Synthesis of a finite 2-terminal network whose driving liptic bandpass filter,” in IEEE MTT-S Int. Microwave Symp. Dig.,
point impedance is a prescribed function of frequency,” J. Math. San Diego, CA, 1977, pp. 400–403.
Phys., vol. 10, no. 3, pp. 191–236, 1931. [32] W. M. Fathelbab, “Synthesis of Cul-de-Sac filter networks utiliz-
[2] S. Darlington, “Synthesis of reactance 4-poles which produce ing hybrid couplers,” IEEE Microwave Wireless Comp. Lett., vol. 17,
insertion loss characteristics,” J. Math. Phys., vol. 18, pp. 257–353, no. 5, pp. 334–336, May 2007.
1939. [33] R. J. Cameron, A. R. Harish, and C. J. Radcliffe, “Synthesis of ad-
[3] H. W. Bode, Network Analysis and Feedback Amplifier Design. Princ- vanced microwave filters without diagonal cross-couplings,” IEEE
eton, NJ: Van Nostrand, 1945. Trans. Microwave Theory Tech., vol. 50, pp. 2862–2872, Dec. 2002.
[4] M. E. van Valkenburg, Network Analysis. Englewood Cliffs, NJ: [34] R. J. Cameron, “General prototype network synthesis methods
Prentice-Hall, 1955. for microwave filters,” ESA J., vol. 6, pp. 193–206, 1982.
[5] H. J. Carlin, “The scattering matrix in network theory,” IRE Trans. [35] R. Levy, “Filters with single transmission zeros at real or imagi-
Circuit Theory, vol. CT-3, pp. 88–96, June 1956. nary frequencies,” IEEE Trans. Microwave Theory Tech., vol. 24, pp.
[6] R. F. Baum, “Design of unsymmetrical band-pass filters,” IRE 172–181, Apr. 1976.
Trans. Circuit Theory, vol. CT-4, pp. 33–40, June 1957. [36] R. Levy and P. Petre, “Design of CT and CQ filters using approxi-
[7] S. B. Cohn, “Direct coupled cavity filters,” Proc. IRE, vol. 45, mation and optimization,” IEEE Trans. Microwave Theory Tech.,
pp. 187–196, Feb. 1957. vol. 49, pp. 2350–2356, Dec. 2001.
[8] E. A. Guillemin, Synthesis of Passive Networks. New York: Wiley, 1957. [37] S. Tamiazzo and G. Macchiarella, “An analytical technique for
[9] W. Cauer, Synthesis of Linear Communication Networks. New York: the synthesis of cascaded N-tuplets cross-coupled resonators
McGraw-Hill, 1958. microwave filters using matrix rotations,” IEEE Trans. Microwave
[10] R Saal and E. Ulbrich, “On the design of filters by synthesis,” IRE Theory Tech., vol. 53, pp. 1693–1698, May 2005.
Trans., vol. CT-5, pp. 284–327, Dec. 1958. [38] S. Amari, “Synthesis of cross-coupled resonator filters using an
[11] G. Matthaei, L. Young, and E. M. T. Jones, Microwave Filters, Im- analytical gradient-based optimization technique,” IEEE Trans.
pedance Matching Networks and Coupling Structures. Norwood, MA: Microwave Theory Tech., vol. 48, no. 9, pp. 1559–1564, Sept. 2000.
Artech House, 1980. [39] W. A. Atia, K. A. Zaki, and A. E. Atia, “Synthesis of general topol-
[12] R. Levy, “Theory of direct coupled cavity filters,” IEEE Trans. Mi- ogy multiple-coupled resonator filters by optimization,” in IEEE
crowave Theory Tech., vol. 11, pp. 162–178, May 1963. MTT-S Int. Microwave Symp. Dig., 1998, vol. 2, pp. 821–824.
[13] H. J. Orchard and G. C. Temes, “Filter design using transformed vari- [40] R. J. Cameron, J. C. Faugere, F. Roullier, and F. Seyfert, “Exhaus-
ables,” IEEE Trans. Circuit Theory, vol. CT-15, pp. 385–408, Dec. 1968. tive approach to the coupling matrix synthesis problem and the
[14] J. D. Rhodes, Theory of Electrical Filters. New York: Wiley, 1976. application to the design of high degree asymmetric filters,” Int. J.
[15] J. D. Rhodes, “Filters approximating ideal amplitude and arbi- RF Microwave Comput.-Aided Eng., vol. 17, no. 1, pp. 4–12, Jan. 2007.
trary phase characteristics,” IEEE Trans. Circuit Theory, vol. 20, pp.
[41] Dedale-HF page. [Online]. Available: http://www.sop-inria.fr/
150–153, Mar. 1973.
apics/Dedale
[16] J. D. Rhodes, “The generalized direct-coupled cavity linear phase
[42] M. Oldini, G. Macchierella, G. Gentili, and C. Ernst, “A new ap-
filter,” IEEE Trans. Microwave Theory Tech., vol. 18, pp. 308–313, June
proach to the synthesis of microwave lossy filters,” IEEE Trans. Mi-
1970.
crowave Theory Tech., vol. 58, no. 5, pp. 1222–1229, May 2010.
[17] A. E. Atia and A. E. Williams, “New types of bandpass filters for
[43] V. Miraftab and M. Yu, “Advanced coupling matrix and admittance
satellite transponders,” COMSAT Tech. Rev., vol. 1, pp. 21–43, Fall 1971.
function synthesis techniques for dissipative microwave filters,” IEEE
[18] A. E. Atia and A. E. Williams, “Narrow-bandpass waveguide filters,”
Trans. Microwave Theory Tech., vol. 57, no. 10, pp. 2429–2438, Oct. 2009.
IEEE Trans. Microwave Theory Tech., vol. 20, pp. 258–265, Apr. 1972.
[44] J. D. Rhodes and I. C. Hunter, “Synthesis of reflection-mode pro-
[19] A. E. Atia and A. E. Williams, “Nonminimum-phase optimum-
totype networks with dissipative circuit elements,” Proc. Inst. Elect.
amplitude bandpass waveguide filters,” IEEE Trans. Microwave
Eng.—Microwave, Antennas, Propagat., vol. 144, no. 6, pp. 437–442,
Theory Tech., vol. 22, pp. 425–431, Apr. 1974.
Dec. 1997.
[20] A. E. Atia, A. E. Williams, and R. W. Newcomb, “Narrow-band
[45] W. Fathelbab, I. C. Hunter, and J. D. Rhodes, “Synthesis of lossy
multiple-coupled cavity synthesis,” IEEE Trans. Circuits Syst., vol.
reflection mode prototype networks with symmetrical and asym-
21, pp. 649–655, Sept. 1974.
metrical characteristics,” Proc. Inst. Elect. Eng.—Microwave, Anten-
[21] R. J. Cameron, “General coupling matrix synthesis methods for
nas, Propagat., vol. 146, no. 2, pp. 97–104, Apr. 1999.
Chebyshev filtering functions,” IEEE Trans. Microwave Theory
Tech., vol. 47, pp. 433–442, Apr. 1999. [46] A. Guyette, I. Hunter, and R. Pollard, “The design of microwave
[22] R. J. Cameron, C. M. Kudsia, and R. R. Mansour, Microwave Filters bandpass filters using resonators with nonuniform Q,” IEEE Trans.
for Communication Systems, Fundamentals, Design and Applications. Microwave Theory Tech., vol. 54, no. 11, pp. 3914–3922, Nov. 2006.
New York: Wiley, 2007. [47] I. C. Hunter, A. Guyette, and R. Pollard, “Passive microwave
[23] J. D. Rhodes and S. A. Alseyab, “The generalized Chebyshev low-pass receive filter networks using low-Q resonators,” IEEE Microwave
prototype filter,” IEEE Trans. Circuit Theory, vol. 8, pp. 113–125, 1980. Mag., vol. 6, no. 3, pp. 1959–1962, Sept. 2005.
[24] R. J. Cameron, “Advanced coupling matrix synthesis techniques [48] S. Amari, F. Seyfert, and M. Bekheit, “Theory of coupled resonator
for microwave filters,” IEEE Trans. Microwave Theory Tech., vol. 51, microwave bandpass filters of arbitrary bandwidth,” IEEE Trans.
pp. 1–10, Jan. 2003. Microwave Theory Tech., vol. 58, no. 8, pp. 2188–2203, Aug. 2010.
[25] H. C. Bell, “Canonical asymmetric coupled-resonator filters,” IEEE [49] S. Amari and U. Rosenberg, “New building blocks for modular
Trans. Microwave Theory Tech., vol. 30, pp. 1335–1340, Sept. 1982. deisgn of elliptic and self-equalized filters,” IEEE Trans. Microwave
[26] F. R. Gantmacher, The Theory of Matrices, vol. 1. New York: Chel- Theory Tech., vol. 52, pp. 721–736, Feb. 2004.
sea, 1959. [50] S. Amari and U. Rosenberg, “A universal building block for
[27] J. D. Rhodes, “A low-pass prototype network for microwave lin- advanced modular design of microwave filters,” IEEE Microwave
ear phase filters,” IEEE Trans. Microwave Theory Tech., vol. 18, pp. Wireless Comp. Lett., vol. 13, pp. 541–543, Dec. 2003.
290–300, June 1970. [51] D. C. Rebenaque, F. Q. Pereira, J. P. Garcia, A. A. Melcon, and M.
[28] R. J. Cameron, “A novel realization for microwave bandpass filters,” Guglielmi, “Two compact configurations for implementing trans-
ESA J., vol. 3, pp. 281–287, 1979. mission zeros in microstrip filters,” IEEE Microwave Wireless Comp.
[29] R. J. Cameron and J. D. Rhodes, “Asymmetric realizations for Lett., vol. 14, pp. 475–477, Oct. 2004.
dual-mode bandpass filters,” IEEE Trans. Microwave Theory Tech., [52] J. W. Bandler, Q. S. Cheng, S. A. Dakroury, A. S. Mohamed, M. H.
vol. 29, pp. 51–58, Jan. 1981. Bakr, K. Madsen, and J. Sondergaard, “Space mapping: The state-
[30] I. C. Hunter, Theory and Design of Microwave Filters (Electromag- of-the-art,” IEEE Trans. Microwave Theory Tech., vol. 52, pp. 337–361,
netic Waves Series 48). London: IEE, 2001. Jan. 2004.
October 2011 61