Filiz Güneş

KNOWLEDGE-BASED SUPPORT VECTOR SYNTHESIS OF THE MICROSTRIP LINES

In this paper, we proposed an efficient knowledge-based Support Vector Regression Machine (SVRM) method and applied it to the synthesis of the transmission lines for the microwave integrated circuits, with the highest possible accuracy using the fewest accurate data. The technique has integrated advanced concepts of SVM and knowledge-based modeling into a powerful and systematic framework. Thus, synthesis model as fast as the coarse models and at the same time as accurate as the fine models is obtained for the RF/Microwave planar transmission lines. The proposed knowledge-based support vector method is demonstrated by a typical worked example of microstrip line. Success of the method and performance of the resulted synthesis model is presented and compared with ANN results.

A knowledge-based support vector synthesis of the transmission lines for use in microwave integrated circuits

In this paper, we proposed an efficient knowledge-based support vector regression machine (SVRM) method to build synthesis models of the transmission lines for the microwave integrated circuits, with the highest possible accuracy using the fewest accurate data. This method is based comprehensively on the powerful generalization capability of support vector machine (SVM) over other classical optimization techniques; especially its working principle based on the small sample statistical learning theory is utilized in lessening the need for the accurate training and validation data together with the human time. Thus, synthesis models as fast as the coarse models and at the same time as accurate as the fine models are obtained for the RF/microwave planar transmission lines. Since the method employs the reverse relations between the analysis and synthesis processes, therefore firstly general definitions of analysis and synthesis processes are made for the RF/microwave planar transmission lines. Then the synthesis data are obtained by reversing the analysis data according to these definitions, where analysis process may be based on either the analytical formulation or empirical (coarse) formulas. Thereafter, generation process of the fine support vector (SV) expansion for synthesis from the coarse SVs is put forward in the form of block diagrams, depending on type of the analysis processes. Finally, the proposed knowledge-based support vector method are demonstrated by the two typical worked examples, representing the typical analysis processes which belong to the commonly used transmission lines, conductor backed coplanar waveguides with upper shielding and microstrip lines. Besides, artificial neural network (ANN)s are employed also in modeling as a competent regressor and it is also verified that only SVs would be sufficient to be used in training ANN models. Success of the method and performances of the resulted synthesis models are presented as compared to each other and the conventional ones.

THE MULTI-OBJECTIVE OPTIMIZATION OF NON-UNIFORM LINEAR PHASED ARRAYS USING THE GENETIC ALGORITHM

In this article, a linear phased antenna array for beam scanning is considered with a fixed narrow/broad interference out of the scanning region. This interference is aimed to be suppressed by optimizing the positions of array elements while avoiding the rise of maximum sidelobe level (MSLL) during the main beam is scanning within the prescribed region. These two objectives; suppressing the fixed interference and avoiding the rise of MSLL during scanning are in conflict with one another. In order to evaluate the effectiveness of such multi-objective approaches it is important to report Pareto optimal solutions which are the objective way of solving multi-objective optimization problems. Thus, in this work, the genetic algorithm (GA) is introduced for the purpose of obtaining the Pareto optimal fronts for the two conflicting objectives to show the effectiveness of the proposed method.

Neural network modeling of active devices for use in MMIC design

This work can be classified into three parts: The first part is a multidimensional signal‐noise neural network model for a microwave smallsignal transistor. Here the device is modeled by a black box, whose small signal and noise parameters are evaluated through a neural network, based upon the fitting of both these parameters for multiple bias and configuration with their target values. The second part is the computer simulation of the possible performance (F,Vi,Gtmax) triplets. In the final part, which is the combination of the first two parts, the performance curves are obtained using the relationships among operation conditions f, VCE, and ICE; the noise figure, input VSWR and maximum stable transducer gain. q 1999 Elsevier Science Ltd. All rights reserved.

Design of a broadband microwave amplifier using neural performance data sheets and very fast simulated reannealing

In this work, the neural performance data sheets of the transistor are employed to determine the feasible design target space in the optimization of a microwave amplifier. The basic amplifier with a single transistor between the input and output matching circuits (IMC and OMC) is also worked out. Very Fast Simulated Reannealing (VFSR) is utilized in the multi – objective optimization process for the global minimum of the objective function which is expressed as a function only gain of a matching circuit, in the negative exponential form to ensure the rapid convergence.

Design of an Ultra-Wideband, Low-Noise Amplifier Using a Single Transistor: a Typical Application Example

In this work, a design method of an Ultra-Wideband (UWB), low-noise amplifler (LNA) is proposed exerting the perfor- mance limitations of a single high-quality discrete transistor. For this purpose, the compatible (Noise F, Input VSWR Vi, Gain GT) triplets and their (ZS, ZL) terminations of a microwave transistor are exploited for the feasible design target space with the minimum noise Fmin(f), maximum gain GT max(f) and a low input VSWR Vi over the available bandwidth B. This multi-objective design procedure is reduced into syntheses of the Darlington equivalences of the ZSopt(f), ZLmax(f) terminations with the Unit-elements and short-circuited stubs in the T-, L-, ƒ-conflgurations and Particle Swarm Intelligence is successfully implemented as a comparatively simple and e-cient optimization tool into both veriflcation of the design target space and the design process of the input and output matching circuits. A typical design exam- ple is given with its challenging performance in the simple ƒ- and ƒ-conflgurations realizable by the microstrip line technology. Further- more the performances of the synthesized ampliflers are compared us- ing an analysis programme in MATLAB code and a microwave system simulator and verifled to agree with each other.

A GENERALIZED DESIGN PROCEDURE FOR A MICROWAVE AMPLIFIER: A TYPICAL APPLICATION EXAMPLE

In this work, a generalized procedure is carried out for the design of a microwave amplifier. First of all, the Performance Data Sheets (PDS) resulted from the active device characterization are used as Feasible Design Target Space (FDTS). Employing the PDS, the compatible (Noise F , Input VSWR Vi, Gain GT ) is determined over the predetermined bandwidth B between fmin and fmax operation frequencies with the source ZS and load ZL terminations as the design target. In the design stage, the Simplified Real Frequency Technique (SRFT) is utilized in the scattering-parameter formulation of the front- and back-end matching two-ports to provide the source and load terminations to the transistor, respectively. As an application example, a novel high technology transistor is chosen and the design targets are determined using the PDSs of the device and its front- and back-end matching two-ports are characterized by the scattering- parameters using the novel SRFT for each design target. Furthermore, the performances of the resulted amplifier circuits are analyzed and compared with the simulated results.

Efficient scattering parameter modeling of a microwave transistor using Generalized Regression Neural Network

In this paper, a simple, accurate, fast and reliable black-box modeling is presented for the Scattering (S-) parameters of a microwave transistor from the reduced amount of the discrete data using General Regression Neural Network (GRNN). GRNN is a probability- based Neural Network and has been used in the generalization applications in the cases of the existence of the poor data bases. In this work, the GRNN-based modeling is implemented to the microwave transistor BFP640 with the separate interpolation and extrapolation applications and the comparative results are given. It can be concluded that the superior extrapolation ability of a GRNN can be used in generalization of the reduced amount of scattering parameter data accurately to the entire operation domain of device, thus in S- parameter modeling of a microwave transistor can be achieved.

A competitive approach to neural device modeling: support vector machines

Support Vector Machines (SVM) are a system for efficiently training linear learning machines in the kernel induced feature spaces, while respecting the insights provided by the generalization theory and exploiting the optimization theory. In this work, Support Vector Machines are employed for the nonlinear regression. The nonlinear regression ability of the Support Vector Machines has been demonstrated by forming the SVM model of a microwave transistor and it has been compared with its neural model.

Performance characterization of a microwave transistor subject to the noise and matching requirements

In this paper, the gain GT of a microwave transistor is expressed analytically in terms of the mismatchings Vini¾ź1, Vouti¾ź1 at the ports, noise figure Fi¾źFmin and the [z]-parameter and noise parameters. Firstly, because the input termination ZS determines the noise Fi¾źFmin, thus the input termination ZS is pre-determined to lie on the tangent constant noise and available gain circles so that the maximum power delivery is ensured for the given noise. Then, a design configuration is constructed in the input impedance Zin- plane covering the gain and the required input and output mismatch circles within the Unconditionally Stable Working Area for the predetermined input termination ZS. Finally, the compatible Fi¾źFmin, GT, Vini¾ź1, Vouti¾ź1 quadrates for either required or optimum Vini¾ź1, Vouti¾ź1 couples are obtained with their ZS, ZL couples from the analysis of the design configuration. Furthermore, a case study is also presented for the full flexible performance characterization of a selected microwave transistor. It can be concluded that the near future microwave transistor is expected to be identified by performance data base built by its compatible Fi¾źFmin, GT, Vini¾ź1, Vouti¾ź1 quadrates and the ZS, ZL terminations within the device operation domain to overview all the possible low-noise amplifier designs using the full device capacity. Copyright