2024 IEEE 10th International Symposium on Microwave, Antenna, Propagation and EMC Technologies for Wireless Communications (MAPE)

| 979-8-3315-3392-2/24/$31.00 ©2024 IEEE | DOI: 10.1109/MAPE62875.2024.10813787

Synthesis of Non-Classic Filter Topologies Based

on Orthogonal Matrix Decomposition
Guoqun Cao, Yi Zeng, and Ming Yu
Department of Eletctrical and Electronic Engineering
Southern University of Science and Technology
Shenzhen, China
[email protected], [email protected], [email protected]

Abstract—In this article, a semi-analytical synthesis method for However, all these methods are vulnerable to local minima and
filters with non-classic topologies is proposed based on orthogo- the success rate of the algorithms is not well established. In
nal matrix decomposition. The orthogonal transform matrix of comparison, the success rate of directly solving the orthogonal
filter topology reconfiguration is decomposed into a product of
rotation matrices with different pivots via the analytical Gaussian matrix will be higher. Recently, a synthesis approach using
elimination method. Hence, the path of rotation transform can interval Newton method has been proposed [4], [10], which
be obtained from these derived pivots. Once the rotation path is is capable of solving all feasible real solutions. Unfortunately,
available, the corresponding rotation angles can be easily solved the exhaustive synthesis method mentioned above is extremely
by optimization. Aided with the rotation path already figured computationally extensive and time-consuming, and the or-
out, the synthesis efficiency can be significantly improved. More
importantly, once the topology is determined, the rotation path is thogonal matrix should be recalculated once the parameters of
not affected by the position of transmission zeros and return loss. the filter are modified. In [11], all the rotation pivots should
Therefore, the obtained rotation path can be applied to accelerate be transversed to find the needed rotation path, which results
the synthesis of the same class of non-classic topologies. Finally, in a long computation time.
a sixth-order filter with the extended box topology is given as an
example to illustrate the effectiveness of the proposed method. In this paper, we propose for the first time a semi-analytical
Index Terms—Coupling matrix reconfiguration, non-classic synthesis method based on orthogonal matrix decomposition.
topology, rotation path synthesis The proposed method provides a more in-depth discussion
on orthogonal matrices. The rotation paths are derived to
I. I NTRODUCTION improve the synthesis efficiency of non-classic topologies. In
particular, the rotation paths are independent of the location of
Filter is an indispensable component in wireless communi- the transmission zeros and the return loss, which means that
cation systems, and its performance directly affects the overall the rotation paths are applicable to all configurations under
capability. Meanwhile, the increasing shortage of spectrum the same topology, greatly improving the performance of the
resources also puts forward more stringent filter requirements. algorithm. Finally, a sixth-order filter with the extended box
The coupling matrix (CM) plays a fundamental role in the topology is given, and the synthesis results demonstrates the
modern microwave filter design process, which is an abstract validity of the proposed method.
expression of the actual physical structures and provides a
guidance for the design and tuning process [1], [2]. To fulfill
different physical layout and response shapes, numerous inter- II. O RTHOGONAL M ATRIX D ECOMPOSITION
esting topologies such as arrow, cascaded triplet, and cascaded
quadruplet [3] have been proposed. Nevertheless, in addition A. Gaussian Elimination Method
to the classic topologies mentioned above, substantial non-
classic topologies are necessary to adapt to compact layout or The known classic CM of the filter is defined as M0 , which
accommodate parasitic couplings [4]. is a N+2 order real symmetric matrix, N being the order of
Usually, the CM with the folded or transversal topology the filter. The target CM with the non-classic topology is M1 ,
is synthesized first [5]. And then the CM is reconfigured to which can be obtained by orthogonal transformation:
the target topology via a series of CM rotations. Especially,
the CM rotations should follow a specific sequence, which is M1 = QM0 QT (1)
also named as the rotation path. However, the reconfiguration
path is unknown for most topologies, except a few classic
ones, which hinders the synthesis of many non-classic topolo- With only a slight modification of the Gaussian elimination
gies. Over the past few decades, the synthesis of non-classic method, the orthogonal matrix Q can be decomposed into a
topologies has received considerable attention. Gradient-based product of rotation matrices
CM synthesis methods were presented in [6], [7]. In [8] and
[9], CM synthesis is considered as an eigenvalue problem. Q1 = Rij Q (2)

Authorized licensed use limited to: Yuri Gagarin State Technical University of Saratov. Downloaded on January 05,2025 at 16:57:45 UTC from IEEE Xplore. Restrictions apply.
 where Rij is the rotation matrix and (i, j) is the rotation pivot. B. Cost Function
Rij takes the following form: Once the rotation paths for the non-classic topology are
  derived through the matrix decomposition, the next procedure
1 ... 0 ... 0 ... 0
... ... is to determine the corresponding rotation angles for various
... ... ... ... ...

 0 ... cosθij ... −sinθij ... 0 
 responses under the same topology by an optimization method.
  The cost function is given as:
Rij = ... ... ... ... ... ... ... (3) X
 0 ... sinθij ... cosθij ... 0  cost = M1(m,n) (θ1 , θ2 , ...θl ) (10)
 
 . ... ... ... ... ... ... m,n
0 ... 0 ... 0 ... 1 where M1(m,n) is the element that should be zero in the target
where the cosθij and sinθij of the Rij are, respectively: topology. To minimize the cost function, the interior-point
algorithm is employed because of the accuracy and timeliness.
Qii Qji
cosθij = q , sinθij = q (4) III. S YNTHESIS E XAMPLES
2 2
Qii + Qji Qii + Q2ji
2
To demonstrate the validity of the proposed method, a sixth-
Note that Qii and Qji cannot be zero simultaneously, and order synthesis example with extended box topology is given.
after the Q1 = Rij Q transformation, the corresponding Q1(j,i) The specifications of the filter are as follows: the return loss
is modified as: RL = 20 dB and the transmission zeros (Tzs) wtz = [-1.3, 1.6].
The low-pass prototype response is depicted in Fig. 1.
Q1(j,i) = −Qii sinθij + Qji cosθij = 0 (5)
0
Repeating the formula (4), the non-zero elements of the first S21
column of the Q matrix except for Q11 can be eliminated: S11
-20
n
′ Y S-parameters(dB)
Q =( Ri1 )Q (6)
i=2 -40

Taking the same operation for the remaining columns, the

Q matrix can be reduced to a unitary matrix I: -60

n−1
Y n
Y
I=( Rij )Q (7) -80
j=1 i=j+1
-6 -4 -2 0 2 4 6
Both sides of the equation are multiplied by the inverses of Normalized frequency(rad/sec)
the rotation matrices:
Fig. 1. S-parameters of the sixth-order lowpass prototype.
n−1
Y n
Y
Q=( Ri,j )−1 (8) The target topology M1 is shown in Fig. 2, where M1(1,4)
j=1 i=j+1 and coupling M1(3,6) are cross-coupling. Actually, there are
The formula (9) can be rewritten as: multiple solutions to this topology, and one of the solutions is
presented, which is obtained by solving the orthogonal matrix
m
Y based on interval Newton method .
Q= Ri (9)
i=1

Where m represents the number of rotation transformations.

Note that when the rotation angle is nπ, the corresponding
pivot can be excluded. Hence, only a few rotation angles need
to be evaluated in the synthesis process, as shown in (9). In
this way, an orthogonal transform matrix is decomposed into
a series of rotation matrices analytically. And the remaining
rotations reveal the path of topology reconfiguration from M0
to M1 .
Fortunately, for filters with the same target topology, even Fig. 2. Sixth-order extended box topology filter. The black numbers define
if the return loss and transmission zeros are different, the the inter-coupling values and the red numbers denote the self-coupling values.
synthesis path remains unchanged. Therefore, the obtained
path can be reused to aid the synthesis of the filters with the The orthogonal matrix Q corresponding to Fig. 2 is on the
same topology but distinct specifications. top of the next page.

Authorized licensed use limited to: Yuri Gagarin State Technical University of Saratov. Downloaded on January 05,2025 at 16:57:45 UTC from IEEE Xplore. Restrictions apply.
  
1.0000 0 0 0 0 0 0 0
 0 1.0000 0 0 0 0 0 0 
 
 0
 0 −0.7213 −0.1109 0.6837 0 0 0 

 0 0 0 −0.1301 −0.0213 −0.9911 0 0 
Q=  (11)
 0
 0 −0.6927 0.1155 −0.7120 0 0 0 

 0
 0 0 0.9784 0.1587 −0.1328 0 0 

 0 0 0 0 0 0 1.0000 0 
0 0 0 0 0 0 0 1.0000

With the orthogonal matrix decomposition method outlined

0
above, the formula (11) can be decomposed as: S21
-10 S11
Q = R3,5 (θ1 )R4,5 (θ2 )R4,6 (θ3 )R5,6 (θ4 ) (12) -20
◦ ◦

S-parameters(dB)
where the rotation angle θ1 = 136.1593 , θ2 = 129.3168 , -30

θ3 = −78.0559◦ , and θ4 = −129.9328◦ . -40

With the rotation path known, for various responses under -50
the same topology only the corresponding rotation angles need
-60
to be solved as shown in formula (13).
-70
M1 = R3,5 (θ1 )R4,5 (θ2 )R4,6 (θ3 )R5,6 (θ4 ) · M0 ·
(13) -80
(R3,5 (θ1 )R4,5 (θ2 )R4,6 (θ3 )R5,6 (θ4 ))T
-90
-6 -4 -2 0 2 4 6
For example, the specification of the sixth-order extended Normalized frequency(rad/sec)
box topology filter is modified to: the return loss RL = 22 dB
and the transmission zeros (Tzs) wtz = [-1.8, -1.5]. Fig. 4. S-parameters of the modified sixth-order lowpass prototype.
The cost function is:
TABLE I
cost = M1(S,2) + M1(S,3) + M1(S,4) + M1(S,5) T HE C ORRESPONDING ROTATION A NGLES O F T HE F OUR S YNTHESIS
+ M1(S,6) + M1(1,3) + M1(1,5) + M1(1,6) S OLUTIONS

+ M1(2,4) + M1(2,5) + M1(2,6) R.A θ1 (◦ ) θ2 (◦ ) θ3 (◦ ) θ4 (◦ )

+ M1(3,5) + M1(4,6) + M1(1,L) Fig. 3(a) 54.6467 -78.1477 -54.5891 -104.4398
Fig. 3(b) -136.4566 -88.1588 175.4592 -67.9006
+ M1(2,L) + M1(3,L) + M1(4,L) + M1(5,L) Fig. 3(c) -98.3077 -164.3778 53.6227 12.6876
(14) Fig. 3(d) 112.0264 50.5371 31.1344 -147.8689
R.A: rotation angles.
By only optimizing the four rotation angles θ1 , θ2 , θ3 and
θ4 , the four synthesis solutions under this topology can be
computed, as shown in Fig. 3. The response of all the synthesis the same conditions, the algorithm runs 100 times in 1.944s,
less than the orthogonal matrix algorithm running time of
3.104s, which demonstrates the high efficiency of the proposed
method. The rotation angles of all the synthesis solutions in
Fig. 3 are listed in Table I.

IV. C ONCLUSION
This article introduces a novel rotation path synthesis
method, which can greatly improve the efficiency of the
algorithm. Furthermore, the rotation path is independent of
the location of the transmission zeros and the return loss, and
only depends on the topology. The interior-point algorithm
is exploited to evaluate the rotation angle. To validate the
proposed method, a sixth-order synthesis example is given.
Fig. 3. The four synthesis solutions of the sixth-order synthesis example. The
black numbers define the inter-coupling values and the red numbers denote R EFERENCES
the self-coupling values.
[1] A. Atia, A. Williams, and R. Newcomb, “Narrow-band multiple-coupled
cavity synthesis,” IEEE Trans. Circuits Syst., vol. CAS-21, no. 5, pp.
solutions in Fig. 3 is identical, as shown in Fig. 4. Under 649–655, Sep. 1974.

Authorized licensed use limited to: Yuri Gagarin State Technical University of Saratov. Downloaded on January 05,2025 at 16:57:45 UTC from IEEE Xplore. Restrictions apply.
 [2] R. J. Cameron, “General coupling matrix synthesis methods for Cheby-
shev filtering functions,” IEEE Trans. Microw. Theory Techn., vol. 47,
no. 4, pp. 433–442, Apr. 1999.
[3] S. Tamiazzo and G. Macchiarella, “An analytical technique for the
synthesis of cascaded N-tuplets cross-coupled resonators microwave
filters using matrix rotations,” IEEE Trans. Microw. Theory Techn., vol.
53, no. 5, pp. 1693–1698, May 2005.
[4] Y. Zeng and M. Yu, “On searching all solutions of microwave filter
synthesis based on interval arithmetic,” IEEE Microw. Wireless Techn.
Lett., vol. 33, no. 6, pp. 811–814, Jun. 2023.
[5] R. J. Cameron, “Advanced coupling matrix synthesis techniques for
microwave filters,” IEEE Trans. Microw. Theory Techn., vol. 51, no.
1, pp. 1–10, Jan. 2003.
[6] S. Amari, “Synthesis of cross-coupled resonator filters using an analyti-
cal gradient-based optimization technique,” IEEE Trans. Microw. Theory
Techn., vol. 48, no. 9, pp. 1559–1564, Sep. 2000.
[7] S. Amari, U. Rosenberg, and J. Bornemann, “Adaptive synthesis and
design of resonator filters with source/load-multiresonator coupling,”
IEEE Trans. Microw. Theory Techn., vol. 50, no. 8, pp. 1969–1978,
Aug. 2002.
[8] P. Kozakowski, A. Lamecki, P. Sypek, and M. Mrozowski, “Eigenvalue
approach to synthesis of prototype filters with source/load coupling,”
IEEE Microw. Wireless Compon. Lett., vol. 15, no. 2, pp. 98–100, Feb.
2005.
[9] L. Szydlowski, A. Lamecki, and M. Mrozowski, “Coupled-resonator
filters with frequency-dependent couplings: Coupling matrix synthesis,”
IEEE Microw. Wireless Compon. Lett., vol. 22, no. 6, pp. 312–314, Jun.
2012.
[10] Y. Zeng, P. Zhang, D. Lu and M. Yu, ”On Searching All Solutions of
Filter Topologies Based on Interval Arithmetic,” IEEE Trans. Microw.
Theory Techn., vol. 50, no. 8, pp. 659–669, Jan. 2024.
[11] G. Macchiarella, “A powerful tool for the synthesis of prototype filters
with arbitrary topology,” in IEEE MTT-S Int. Microw. Symp. Dig.,
Philadelphia, PA, USA, Jun. 2003, pp. 1467–1470.

Authorized licensed use limited to: Yuri Gagarin State Technical University of Saratov. Downloaded on January 05,2025 at 16:57:45 UTC from IEEE Xplore. Restrictions apply.