Jerzy Julian Michalski

Filter Tuning Based on Linear Decomposition of Scattering Characteristics

This paper proposes a microwave fllter post-production tuning based on an optimization process which flnds the vector of deviations of tuning elements that should be applied to tune the fllter. To build the system, the coarse set of scattering parameters is collected in such a way that every tuning element is detuned while other elements remain in their proper positions. In the concept, it is assumed that the relation between the positions of tuning elements and fllter scattering characteristics can be modelled by the sum of one argument polynomial functions. Each polynomial function depends on the value of only one tuning element. Therefore, the measured fllter characteristics can be linearly decomposed to characteristics from the collected coarse set and corresponding tuning element deviations can be found. This is done by way of optimization process. The presented numerical and physical experiments on the 7th order cross-coupled, bandpass fllter have verifled our approach.

Principal component analysis in application for filter tuning algorithm

This elaboration presents the improvement of an algorithm based on artificial neural network (ANN) for microwave filter tuning. In the applied algorithm, which is based on direct mapping of the detuned filter characteristics to the tuning element error, the sets of ANN learning vectors containing scattering filter characteristics and corresponding tuning element deviations are used. In the concept presented here filter characteristics are converted to “principal component” representation before ANN training. Such representation allows us to truncate less effective data components and thus significantly reduce the number of neurons in the ANN input layer. Experimental results of ANN training have shown that, when the presented approach is used, the ANN input vector dimension can be reduced even 32 times without losing ANN generalization ability.

Microwave filter tuning for different center frequencies based on Artificial Neural Network and phase compensation

This paper proposes a new approach for tuning microwave filters, which are tunable in a certain frequency range. An Artificial Neural Network (ANN) is used to build a filter inverse model, which maps scattering parameters into tuning elements deviations. Once the model is built, it may be applied for tuning filters of the same type on different center frequencies provided that the filter design allows to shift its band. Described considerations showed that a transformation of phase has to be performed on working signal. Five different types of phase compensation methods are proposed based on experiments performed on 4th order filter. Such compensation modifies scattering parameters in a way that the filter can be tuned for different frequencies using the same inverse model. Experiments performed on 7th order filter verified this approach.

The Analytic Extraction of the Complex-Valued Coupling Matrix and its Application in the Microwave Filter Modeling

The idea behind the coupling matrix identiflcation is to flnd the coupling matrix corresponding to the measured or designed scattering characteristics of the microwave fllter. The typical attitude towards coupling matrix parameter extraction is to use some optimization methods to minimize the appropriate cost function. In this paper, we concentrate on the analytic solutions | how they may be found and their application in further optimization processes. In general case, the suggested method generates complex-valued coupling matrix. For a special case of the fllter without cross-couplings we give fast and simple recursive method of flnding such complex- valued coupling matrix. The method is based on Laplaces formula for expanding the determinant. The complex-valued coupling matrix is used as a good starting point for the optimization methods to flnd the regular coupling matrix. The examples are presented showing that the optimization arrives to global minimum starting from real parts of complex-valued entries considerably more often than when the starting point is selected randomly.

Filter tuning and coupling matrix synthesis by optimization with cost function based on zeros, poles and hausdorff distance

This elaboration proposes and investigates a new cost function used in filter tuning algorithm and coupling matrix synthesis. The cost function is defined based on the Hausdorff distance between the template sets (the sets of zeros and poles of template filter reflection and transmission characteristics) and the sets of zeros and poles described from optimized characteristics of a filter. The experiments showed the performance of the proposed cost function definition.

Wireless Sensor Network analysis and optimization by 3D electromagnetic simulations for research rocket application

In this work we propose a method based on 3D electromagnetic simulations to design and optimize Wireless Sensor Network (WSN). We investigate the performance of WSN based on ZigBee modules for the application of collecting the data obtained from sensors distributed inside a research rocket. The propagation conditions inside the rocket are simulated and optimized using commercial 3D electromagnetic simulator CST MWS. The experiments performed in a real research rocket based on the measurement of lost packets shows a good network performance thus validated proposed approach.

Coupling matrix synthesis for lossy filters by optimization using Fréchet distance

A new cost function definition based on the Fréchet distance used in the process of coupling matrix synthesis by optimization is proposed. Such a definition finds a kind of similarity by direct comparison of an ideal - tuned filter response (template) and filter response obtained from optimized coupling matrix. The numerical experiments of coupling matrix optimizations using proposed cost function for two different sets of scattering parameters, one from ideal simulated filter response and second one as measurement of physical lossy filter, show very good results in sense of optimization time, convergence and accuracy.

Systematic solution of nonlinear equations with close or double roots in complex domain: Application of SCVS algorithm for crosswise complex dispersion characteristics

This paper shows the application of Simplex Search Vertices Searching (SCVS) algorithm in solving nonlinear equations of the form F(f, γ) = 0, where f ϵ ℝ, γ ϵ C and F is complex function, for the case where we face the problem of close or double complex roots. This problem is solved with the use of standard SCVS procedure, recently reported. In this case 2- and 3-Simplexes are used for root tracking routine. Analyzing the sign of function F at the vertices of considered simplexes we can infer whether there is one or two roots in its volume. Numerical results of calculating dispersion characteristics, for selected ferrite line, are shown to prove the performance of the proposed method.

Approximation of filter characteristic to tuning elementpositions using coarse set

This paper proposes a new approach In building a multidimensional approximator which maps filter scattering characteristic to the vector containing deviations of tuning element positions. In the concept, it is assumed that relation between positions of tuning elements and filter scattering characteristic can be modeled by the sum of polynomial functions. Each polynomial function only depends on the value of one tuning element. A coarse set of filter characteristics and tuning element positions is used in building the approximator which models this relation. This set is collected by detuning only one tuning element from its proper position at a time, while other elements remain in their proper positions. The presented numerical and physical experiments verified our approach.

The isolator tuning using sequential method with applied Artificial Neural Network

This paper presents the approach and results of utilizing sequential method, reported recently, with applied Artificial Neural Network (ANN) in postproduction ferrite isolator tuning. For isolator with R tuning elements, based on physically measured scattering characteristics, ANNs are used as a multidimensional approximators realizing inverse models for all R sub-devices. The sub-isolators (sub-devices) are obtained by successive detuning and removing tuning screws. For each subisolator, the input and output vectors are defined as physical scattering characteristics and the corresponding positions of the tuning element, detuned, in controlled way. Throughout the tuning process, these inverse models are used for calculating the tuning element increments needed for adjusting the tuning element in the proper position. Earlier that method was successfully used in filter tuning process so it encourage authors to adopt and verify that method in other microwave devices tuning. The obtained and presented results of our investigations prove that mentioned above method may be successfully used for other devices and systems that require tuning.