You are on page 1of 4

Proceedings of the 36th European Microwave Conference

A Novel Method for the Design and Optimization of Microstrip Multi-section Bandpass Combline Filters
Homayoon Oraizi and Nima Azadi-Tinat Iran University of Science and Technology, Tehran, 16846-13114, Iran e-mail: h_oraizi@iust.ac.ir and nima.azaditinat@gmail.com

Abstract A Novel least square based method is developed for the microstrip multi-section bandpass combline filter. First, the overall transmission matrix of the filter is obtained in terms of its various geometrical dimensions. The effects of couplings among adjacent and nonadjacent coupled conducting strips and also the dispersion relations of coupled lines on microstrips are considered. Then, the filter scattering parameters and its insertion loss is determined. Subsequently, an error function is constructed based on the optimization of specified insertion losses in the passband, transition bands and stopbands. The performance of the proposed least square method for the combline filter design and optimization is verified by the available electromagnetic software packages, which highlights its several advantages ( such as simultaneous input and output impedance matching) compared with the available design procedures. Index terms Combline filters, CAD, filter design, impedance matching, optimization

I. INTRODUCTION The design of combline filters for inhomogeneous structures such as microstrips results in discrepancies between the predicted behavior from the theoretical models and the actual performance of fabricated filters, because of neglecting couplings among nonadjacent coupled conducting strips. This situation also adversely affects the performance of MIC and MMIC filters in practice, which are based on the conventional design procedures [1]. The classic filter synthesis methods [2],[3] give acceptable designs, only in such situations where the simplifying assumptions (which result in reducing the complexities of design procedures) are more or less correct in practice for the particular filter structure under considerations, having N parallel coupled lines in inhomogeneous media (such as combline and interdigital filters). Simplifying assumptions such as ignoring nonadjacent couplings among strips, a single phase velocity on all strips and dispersionless media, lead to inaccurate and at least unoptimum designs. In this paper, the considerable couplings among nonadjacent coupled conducting strips, and the effect of different phase velocities, which arise in the N coupled microstrip lines, and the dispersion relations for the inhomogeneous microstrip lines are taken into account in the development of the filter design procedure. First, the proposed numerical procedure for the design of N-section combline bandpass filters is developed. Then, the computer simulation and the numerical examples of the proposed method of filter design are described.
Fig. 1. Combline filter with coupled lines.

II. NUMERICAL PROCEDURE We start by considering the differential equations of N coupled transmission lines in the following matrix forms [4] (see Fig. 1) d [V] = ( j[L] + [R ] )[I] dz d [I] = ( j[C] + [G ] )[V] dz (1) (2)

The entries of matrices [L] and [C] are given in references [5],[6], in which the dispersion effects may be included [7]. Now, the wave equation for the voltage vector [V] of N lines may be obtained from (1) and (2), for which the characteristic impedances are equal to the eigenvalues of the following equation: (3) ( j[L] + [R ])( j[C] + [G]) = 2 [U]NN and the corresponding eigenvectors are obtained by:

( j[L] + [R ])( j[C] + [G])[M V ]i

= i2 [M V ]i

(4)

where [U]NN is a unit matrix of dimension N and [MV]i is the i'th column of the eigenvector matrix [MV], with entries xij.

2-9600551-6-0 2006 EuMA

1217

September 2006, Manchester UK

The k'th frequency in the band is denoted by subscript k. The voltage vector and current vector on the N lines are, respectively [8], a 1e j1,k z + a 1 e j1,k z [V] = [M V ]k M a e j N , k z + a e j N ,k z N N a 1e j1, k z a 1 e j1, k z [I] = ( [Ya ]k .[M V ]k ) M a e j N ,k z a e j N ,k z N N (5)

I1 V1 I = [Yf ]V N N Then its transmission matrix may be derived as [ 10]: V1 A B V2 I = 1 C D I 2 1 A B 1 Yf , 22 T= = det (Y ) Y Yf ,21 f f ,11 C D

(15)

(16)

(17)

(6)

where Yf,ij is the i'th entry of the filter admittance matrix, from which the scattering parameter may be written as [10]: s 21, k = 2 A k + B k Yl,k + CZ s, k + D k Z s ,k Yl,k (18)

where [Ya] is the characteristic admittance matrix with entries yij, which denotes the characteristic admittance of the j'th mode on the i'th line. The voltages and currents at the two ends (x=0 x=L) of the i'th line are respectively: Vi,k (0) = Vi ,k (L ) =
N

where Zs and Yl are the source impedance and load admittance of filter. The insertion loss is defined as: IL = 1 s 21
2

x (a
ij,k j=1 j

+ a j

)
j j,k L

(7)

(19)

x (a e
ij,k j=1

j j,,k L

+ a je

) )

(8)

I i,k (0) = I i,k (L ) =

y
j=1

ij,k x ij,k

(a j a j )
a je
j j, k L

(9)

Finally, different error functions may be constructed with reference to Fig. 2. We may consider, the following error function, based on the specification of the insertion loss in the passband (ILPB) and those in the upper and lower stopbands (ILSB). e = wt 1

y
j=1

ij,k x ij,k

(a e
j

j j,k L

(10)

(IL k ILSBk )2 + wt 2
k =1

NSL

N PU

k = N PL

(IL
2

ILPB k )

The complete admittance matrix for the filter maybe obtained by removing coefficients aj and a'j among equations (7) to (10) whereby the relations among the input and output currents and voltages are obtained: [I] 0 [V] 0 I = [Y] V [ ] L [ ] L (11)

+ wt 3

k = n SU

(IL

ILSBk )

(20)

where wt's denote weighting functions, which may be functions of frequency.

The boundary conditions on the voltages at the terminals of conducting strips are: Vi (0) = 0 , i = 2,..., N 1 1 Vi (L ) = I i (L ), i = 2,..., N 1 jC s,i 1 (12) (13)

After imposing the above boundary conditions, the filter output port currents may be written in terms of the voltages at ports 1 and N:

[I1 (0) L = [Ytot ][V1

I N (0) I1 (L ) L I N (L )]T 0 L 0 VN

0 L 0]T

(14)
Fig. 2. Speciefied frequency response of the filter, showing the stop, transition and pass bands.

from which we may extract the admittance matrix of the two port network between the input and output ports.

1218

III. COMPUTER SIMULATION For the minimization of the error function the algorithm FMINCON from the MATLAB optimization toolbox has been used. The program stops after a specified criteria is satisfied such as the value of error function. The input data are: microstrip substrate characteristics, center frequency, bandwidth, order of filter N (with N+2 strips), source and load impedances, initial values of strip widths (w) and spacings (s), length of strips (L), capacitance of lumped capacitors (C), desired insertion losses in the stop and pass bands (ILSB, ILPB), and the error criteria. The initial values of w and s may be obtained by a random generator RAND. The length is initially chosen as L=5/18 and the values of lumped capacitors are obtained from reference [3]. Two examples of combline filter designs are considered. For the first example, the order of filter is 4 (with 6 strips) and the center frequency is 4 GHz. The frequency responses of the filter before and after optimizations are drawn in Fig. 3. The performance of the filter is compared with the simulation results by Microwave Office.

Fig. 4. Comparison of the frequency response of example 1 with Microwave Office.

For the second example, the filter is of order 4 with center frequency 2 GHz and bandwidth of 20%. Its performance is drawn in Fig. 5 and compared with the results from Microwave Office in Fig 6.

Fig. 3. The combline filter frequency response showing the insertion loss and return loss for example 1.

Fig. 5. The combline filter frequency response showing the insertion loss and return loss for example 2.

1219

softwares with favorable results, which indicates the efficiency of the proposed combline filter design and optimization method. REFERENCES
[1] T. A. Milligan, "Dimensions of microstrip coupled lines and interdigital structure," IEEE Trans. Microwave Theory Tech., Vol.25, No.5, pp. 405-410, pp 405-410, May 1977 [2] G. Matthaei, "Comb-line band-pass filter of narrow or moderate bandwidth," Microwave J., Vol. 6, No. 8, pp 82-96, Aug 1963 [3] Ian Hunter, Theory and Design of Microwave Filter , The Institution of Electrical Engineers, London, 2001 [4] G. G. Gentili, and M. Salazar-Palma, "The definition and computation of modal characteristic impedance in quasi-TEM coupled transmission lines," IEEE Trans. Microwave Theory Tech, Vol. 43, No. 2, pp 338-343, Feb 1995 [5] M. Khalaj Amirhossenini, "Determination of capacitances and conductance matrices of lossy shielded coupled microstrip transmission lines," Progress In Electromagnetics Research, PIER 50, pp. 267-278, 2005 [6] S. S. Bedair, "Characteristics of some asymmetrical coupled transmission Lines," IEEE Trans. Microwave Theory Tech., Vol. MTT-32, No.1, Jan 1984 [7] V. K. Tripathi, "A dispersion model for coupled microstrips," IEEE Trans. Microwave Theory Tech., Vol. MTT-34, No. 1, pp 66-71, Jan 1986 [8] V. K. Tripathi, "Asymmetric coupled transmission lines in an inhomogeneous medium," IEEE Trans. Microwave Theory Tech., Vol. MTT-23, No. 9, Sep 1975 [9] V. K. Tripathi, "On the analysis of symmetrical three-line microstrip circuits," IEEE Trans. Microwave Theory Tech., Vol. MTT-23, No. 9, Sep 1975 [10] D. M. Pozar, Microwave Engineering, 3rd Edition, NY: John Wiley and Sons, 2004 [11] Microwave Office, Applied Wave Research, Inc, Ver 5.00, 2002 [12] R. Levy, R. V. Snyder and G. Matthaei, "Design of microwave filters," IEEE Trans. Microwave Theory Tech., Vol. 50, No. 3, pp 783-793, Mar 2002 [13] I. Shapir, V. A. Sharir and D. G. Swanson, "TEM modeling of parasitic bandwidth expansions in combline filters," IEEE Trans. Microwave Theory Tech., Vol. 47, No.9, pp 1664-1669, Sep 1999

Fig. 6. Comparison of the frequency response of example 2 with Microwave Office.

IV. CONCLUSION In this paper, a least square based procedure is developed for the design and optimization of microstrip combline filters, which provides for the specification and realization of arbitrary center frequency, bandwidth, insertion loss, source and load impedances. The full-wave dispersion relations of the microstrips are also incorporated in the design algorithm. The proposed procedure considers the adjacent and nonadjacent couplings among the coupled conducting strips. Evanescent modes are also excited in the microstrip combline filter [12]. It is pointed out in reference [13], that their effect may be accounted for by connecting a capacitor inside the filter configuration, which may be readily incorporated in the proposed design procedure. The performance of the designed and optimized filters is compared with the available full-wave electromagnetic

1220

You might also like