GB2312100A - Method for designing E-plane bandpass filter with conductive strip - Google Patents

Abstract

A method of designing an E-plane bandpass filter having a conductive strip and a resonator is described. Firstly, design parameters of the E-plane bandpass filter are designed by way of network synthesis. Next the frequency response of the E-plane bandpass filter is optimized by evaluating an error function corresponding to the difference between the desired frequency response and a transfer function of the E-plane bandpass filter and varying the input parameters of the transfer function. Thus, the lengths of the conductive strip and the resonator which optimize the frequency response are determined.

Description

METHOD FOR DESIGNING E-PLANE BANDPASS FILTER WITH CONDUCTIVE STRIP Background of the Invention The present invention relates to a microwave communication system, and more particularly a method for designing Eplane bandpass filter with a conductive strip.

Microwave communication systems such as MDR (Microwave Digital Relay) can include small loss waveguide-type bandpass filters of extremely narrow bandwidth. Comprising a rectangular waveguide, a metal pole, an inductive window, and a diaphragm, the waveguide-type bandpass filter is very complicated in construction and difficult to set up in a mass production system. On the other hand, E-plane bandpass filters have small loss factors and are easy to manufacture and design precisely, so that fine tuning is not necessary.

The E-plane bandpass filter is constructed by inserting a conductive strip into a rectangular waveguide.

As shown in Fig. 1, an E-plane bandpass filter includes a waveguide 2 and a plurality of conductive strip E-plane circuits 4 ("conductive strips" for short). The conductive strips 4 may be precisely manufactured by means of photoetching and a photo-printing. The E-plane bandpass filter may be readily modified simply by changing the construction of the conductive strip 4, so that it may be manufactured economically and is suitable for mass production.

To secure the most preferable Q-value, dielectric-free purified conductive strips 4 are used, and the width of the slot pattern between the conductive strips 4 must be identical to the height of the waveguide 2. Therefore, the E-plane bandpass filter includes a plurality of resonators divided by the axial conductive strips 4. The width and length of the conductive strips 4 and the length of the resonator are the design parameters of the E-plane bandpass filter. For convenience, however, only the lengths of the conductive strips 4 and the resonator are generally used as design parameters, with the width of the conductive strip 4 being fixed. With reference to Fig. 1, the parameter 1 represents the length of the resonator and the parameter m represents the length of the conductive strip 4.

One typical proposal for designing E-plane bandpass filters is a design method based on network synthesis. It is possible to design a direct coupling resonance filter by means of this method, which is one of the circular filter theories. The E-plane bandpass filter is classified as a distributed constant-type step impedance filter, and is usually designed by a method, proposed by Levy, for designing direct coupling resonance filters. In this method for designing direct coupling resonance filters, a required filter is converted into an impedance circuit and then converted again into an equivalent inverted circuit. Then, the equivalent inverted circuit is converted into a circuit of reflection coefficients. Using this design method, it is possible to obtain the discontinuous reflection coefficients for a desired performance of the E-plane bandpass filter.

To design the E-plane bandpass filter, the reflection coefficients must be converted into the physical size of the actual E-plane construction. For this purpose, the reflection coefficient values of the conductive strip 4 are first expressed as a function of the length "m" of the conductive strip 4 and then compared with reflection coefficient values calculated by means of the network synthesis technique, so as to obtain the length parameters of the strip construction. A simple way to express the reflection coefficient as a function of the length "m" of the conductive strip 4 is to obtain a scattering matrix expressing the strip construction. For this purpose, mode matching and general scattering matrix techniques are widely used.

There follow network synthesis theories (1)-(5) used for designing an E-plane filter.

(1) calculating parameters by means of network synthesis n : number of resonators

(2) calculating a reflection coefficient S11 by means of network synthesis

TD(n,n) = eI2nD, n = 1, 2, ...

(3) scattering matrix of the conductive strip

(4) determining a function of the reflection coefficient S11 ST11 = Sa11 + Sa12 Sb11 TDSa24 ST12 = Sa12(I + Sb11TDSa22)Sb12 ST21 = Sb21 TD Sa21 ST22 = Sb22 + Sb21 TD Sa22 Sb12 Fig. 2 illustrates the construction of a branched discontinuous conjunction, taken by way of example for a better understanding of the scattering matrix of the conductive strip. Fig. 3 shows how a function of the reflection coefficient S1l is determined by means of the network synthesis technique.

Fig. 4 shows a flow chart for designing an E-plane bandpass filter based on the network synthesis technique. In the following, the overall flow for designing the filter based on the network synthesis technique will be described with reference to Fig. 4. First, the parameters n, a, and h are calculated at step 100 and consecutively, a reflection coefficient S1l is calculated at step 102. Then, the scattering matrix of the conductive strip is expressed at step 104, and a function of the reflection coefficient S11 is determined at step 106. If the parameter n (number of resonators) is higher than or equal to 1 at step 108, the process proceeds to step 110 to decrease the parameter n by 1. Then, the length of the conductive strip 4 is changed at step 112, and the function of the reflection coefficient S1l is calculated at step 114. The calculated function value of the reflection coefficient Sll is compared with the function value of the reflection coefficient Sll at step 116. If the two values are not identical to each other, the process returns to the step 112. Otherwise, if the S1l calculated value is identical to the S1l function value, the length m of the conductive strip 4 is determined at step 118, and then the process returns to the step 108 to repeat the succeeding steps. In this way, the length m of the conductive strip 4 required in the waveguide 2 may be determined.

If the parameter n is smaller than 1 in the step 108, the process proceeds to step 120 to determine the length 1 of the resonator. In Fig. 1, the number of resonators n is 3.

This conventional filter design method based on the network synthesis theory takes the only dominant mode into consideration. Accordingly, the accuracy of the filter at the higher frequency range is undesirably low, although the filter may be precisely designed at the lower frequency range. It is natural that the high frequency components are easily attenuated at the lower frequency range and, however, are not attenuated at the higher frequency range, being propagated for a relatively longer distance. Thus, in the higher frequency range, not only the dominant mode but also the higher modes are transmitted, and the higher modes are combined with each other, causing an influence on the features of the dominant mode.

Fig. 5 shows convergence of the reflection coefficient according to the number of modes in a bandpass filter designed based on the network synthesis technique. The drawing illustrates the features of the above mentioned conventional method for designing bandpass filters.

Accordingly, when designing a bandpass filter having a pass band at the higher frequency range based on the network synthesis technique, the central frequency of the designed filter is shifted due to the influence of the higher modes, resulting in a reduction of the pass bandwidth.

It is therefore an object of the present invention to provide a method of designing an E-plane bandpass filter, in which the design accuracy is improved in the higher frequency range.

Summary of the Invention Accordingly, the present invention provides a method of designing an E-plane bandpass filter having a conductive strip and a resonator comprising: determining design parameters of the E-plane bandpass filter by way of network synthesis; optimizing the frequency response of the E-plane bandpass filter by evaluating an error function corresponding to the difference between the desired frequency response and a transfer function of the E-plane bandpass filter and varying the input parameters of the transfer function; determining the lengths of the conductive strip and the resonator which optimize the frequency response.

Preferably, the transfer function is obtained from the scattering matrix of the filter.

Preferably, the method further comprises correcting the design specifications of the filter according to the determined design parameters, to bring the input parameters of the transfer function close to the optimum, so as to determine the initial values of the input parameters for the optimization of the frequency response.

The initial values may be used as design parameters of the network synthesis and only one parameter is changed while the other parameters are fixed, so as to reduce the optimization time.

Preferably, the smaller the error function, the closer the transfer function is to the desired frequency response.

The present invention also provides a method of manufacturing an E-plane bandpass filter comprising designing the filter by a method according to any preceding claim and manufacturing the filter to that design.

Brief Description of the Drawings The present invention will now be described by way of example with reference to the accompanying drawings in which: Fig. 1 illustrates an E-plane bandpass filter with a waveguide; Fig. 2 illustrates a double branched discontinuous conjunction; Fig. 3 illustrates the way in which a function of the reflection coefficient S11 is determined based on the network synthesis technique; Fig. 4 is a flow chart of a method for designing an Eplane bandpass filter based on the network synthesis technique; Fig. 5 is a curve illustrating the convergence for the reflection coefficient according to the number of modes in an E-plane bandpass filter designed by the network synthesis technique; Fig. 6 is a flow chart of an optimal method for designing an E-plane bandpass filter according to the present invention; Fig. 7 is a curve illustrating the simulated result of the optimal method according to the present invention in comparison with the known method based on the network synthesis technique (in which the two methods are implemented under the conditions that fo=18GHz, BW=60MHz, ripple characteristic is 0.5dB, and 40dB is attenuated at a position 120MHz from the central frequency); and Fig. 8 is a curve illustrating the simulated result of the optimal method according to the present invention in comparison with the prior art method based on the network synthesis technique (in which the two methods are implemented under the conditions that fo=34GHz, BW=1.2GHz, ripple characteristic is 0.05dB, and 50dB is attenuated at a position 2GHz from the central frequency).

Detailed Description of the Preferred Embodiment An optimal method for designing an E-plane bandpass filter according to the present invention considers not only the dominant mode but also the higher modes which do not attenuate, so that it is possible to design an E-plane bandpass filter operating accurately even at the higher frequency range. In this method, the whole E-plane bandpass filter is expressed with the lengths of the conductive strip and the resonators and a scattering matrix which is a function of frequency. Then, an error function is determined using transmission coefficients and reflection coefficients of the scattering matrix. The error function is used in an optimization process. Here, the optimization process is based on a direct evaluation strategy which changes the input values to minimize the error function.

To reduce optimization time, the values determined by the network synthesis technique are employed as input parameters in the optimization process. In a process for obtaining a transfer function having information on the frequency characteristics of the filter, a mode matching technique which is a full wave analysis taking the higher modes into full consideration is used, so that the transfer function may be used as a simulation function for checking the frequency characteristics of the filter.

The optimization time is influenced by the initial values of the input parameters. In the circular filter design theory, the length of and distance between the conductive strips are shorter than a wavelength calculated at the central frequency, so that it may be possible to determine a range of the input parameters. A method for changing the input parameters, by a critical value, within this range cannot be used when designing a multistage strip and resonator filter, since the time taken to change the input parameters increases in a geometrically whenever the number of input parameters increases by one. If a wide pass band is required of the filter, a discontinuous multistage construction becomes necessary. Thus, when designing particularly a wide bandpass filter, reducing the optimization time is most important.

Fig. 6 is a flow chart of an optimal method for designing an E-plane bandpass filter according to one embodiment of the present invention. In the drawing, the design values, i.e. the length m of the conductive strip and the length 1 of the resonator, evaluated by the network synthesis design are simulated at step 200. Then, design specifications are corrected, at step 202, according to the simulated result of step 200, so as to make the input parameters become more similar to the actually optimized result. The input parameters are then determined as initial values of the optimization process, at step 204.

Describing the steps 200 through 204 in more detail, the transfer function (i.e. simulation function) for accurately predicting the frequency response of the designed filter can be obtained from the scattering matrix of the filter.

The simulation function is employed in the optimization process, to obtain the design parameters. If parameters for the filter size are arbitrarily fixed, the frequency response of the filter can be obtained accordingly.

Particularly, the step 202 for correcting the design specifications minimizes an input range of the input parameters and allows the input parameters to be quickly optimized from the initial value, thereby markedly reducing the optimization time for designing of an E-plane filter.

Then, the process proceeds to the step 206 through step 204, to change the lengths m and 1 of the conductive strip and the resonator, respectively. Here, the changes of the lengths m and 1 are set to become a critical value that the designer can manage. It is preferable that the changes become about 10~4-10~5mm.

The process then proceeds to step 208, to check whether changing the lengths is finished or not. Namely, it is checked whether or not the lengths have been fully changed by the preferred values as stated above. If changing the lengths is not fully finished, the error function is calculated at step 210. Then, it is checked whether or not the calculated error function has a decreasing value, at step 212. If the calculated error function has a decreasing value, the lengths m and 1 of the conductive strip and the resonator are designated at step 214.

The process then returns to the step 206 to repeat the succeeding steps. The above stated steps 206 through 214 will be now be described in more detail. To optimize the frequency response, the present invention employs an evaluation strategy which changes the input parameters continuously to judge whether or not the input parameters are optimized. Further, an error function is used as a standard of judging the optimization. The error function is a function corresponding to the difference between a desired frequency response and the simulated frequency response. It should be understood that the simulated result gets closer to the desired frequency response as the error function decreases.

Herein, the optimizing process of designing the filter involves changing the input parameters until the desired frequency response is obtained and then halting the changes to the input parameters. By way of this optimization process, it is possible to design a filter which meets the desired design specifications. It may also be possible to employ the following error function out of various error functions.

In this way, if changing the length is fully finished at step 208, the process proceeds to step 216 to evaluate the lengths m and 1 of the conductive strip and the resonator.

It should be understood from the above descriptions that the initial values are used as the design parameters of the network synthesis. Further, only one parameter is changed with the other parameters being fixed, instead of sequentially changing all the parameters. Accordingly, the optimization time may be reduced drastically.

A simulation function which accurately predicts the characteristics of the filter has been used to compare the optimal design method with the existing method using network synthesis. To that end, two different filters have been designed according to the respective design methods, and their characteristics simulated. The filters were designed under the conditions that the central frequency is 18GHz, the pass bandwidth is 60MHz, a ripple is 0.5dB, and 40dB is attenuated at a position 120MHz from the central frequency. Fig. 7 illustrates the frequency responses of the filters designed by the respective design methods. In the drawing, the Tchebychev function is also shown to express the desired frequency characteristics of the filter.

It can be appreciated from Fig. 7 that the central frequency of the filter designed by the network synthesis method is shifted by 14MHz with respect to the desired frequency characteristic. On the other hand, the frequency response of the filter designed by the optimal method is almost coincident with the desired frequency characteristic. Even though the central frequency is located in the relatively lower frequency range, the frequency response of the filter designed by network synthesis has been changed because the designed filter has a narrow bandwidth. Namely, as the attenuation characteristics are higher for higher frequency band of the central frequency, the central frequency is shifted more.

This is because as the frequency increases, a combination of the higher modes has a significant effect on the dominant mode resonance; on the other hand, with a frequency of a narrower band, the frequency response of the filter designed by the network synthesis is much closer to the desired frequency characteristics.

To compare frequency response at the higher frequency range, two different filters have been designed respectively according to the present method and the conventional method, under the condition that the central frequency is 34GHz, the pass bandwidth is 1.2GHz, the ripple is 0.05dB, and 50dB is attenuated at a position 2GHz from the central frequency. Fig. 8 illustrates a simulation result.

It can be appreciated from Fig. 8 that the frequency response of the filter designed by network synthesis is shifted, at the central frequency, with respect to the desired frequency response, although the pass bandwidth is relatively wide. This is because the wider the pass bandwidth of the frequency response becomes, the longer is conductive strip. Therefore, the design formulas of the network synthesis cannot accurately predict the bandwidth any more. On the other hand, the frequency response of the filter designed by the optimal design method coincides with the desired frequency response, even though the central frequency is at the higher frequency range.

As described, the optimal method for designing the filter according to the present invention takes not only the dominant mode but also the higher modes into consideration, so that it is possible to design a filter which operates accurately even at the higher frequency range.

Claims (7)

CLAIMS:

1. A method of designing an E-plane bandpass filter having a conductive strip and a resonator comprising: determining design parameters of the E-plane bandpass filter by way of network synthesis; optimizing the frequency response of the E-plane bandpass filter by evaluating an error function corresponding to the difference between the desired frequency response and a transfer function of the E-plane bandpass filter and varying the input parameters of the transfer function; determining the lengths of the conductive strip and the resonator which optimize the frequency response.

2. A method according to claim 1 in which the transfer function is obtained from the scattering matrix of the filter.

3. A method according to claim 1 or claim 2 further comprising correcting the design specifications of the filter according to the determined design parameters, to bring the input parameters of the transfer function close to the optimum, so as to determine the initial values of the input parameters for the optimization of the frequency response.

4. A method according to claim 3 in which the initial values are used as design parameters of the network synthesis and only one parameter is changed while the other parameters are fixed, so as to reduce the optimization time.

5. A method according to any preceding claim in which the smaller the error function, the closer the transfer function is to the desired frequency response.

6. A method of designing an E-plane bandpass filter substantially as described with reference to and/or as illustrated in FIG. 6 of the accompanying drawings.

7. A method of manufacturing an E-plane bandpass filter comprising designing the filter by a method according to any preceding claim and manufacturing the filter to that design.