Olivera Pronic-Rancic

Analysis and validation of neural network approach for extraction of small-signal model parameters of microwave transistors

Abstract Extraction of parameters of a small-signal model is the first step in modeling transistors for advanced microwave applications. There are different extraction techniques, mostly based on optimizations or on direct analytical procedures. An alternative to the standard extraction methods are procedures based on the application of artificial neural networks. Namely, an artificial neural network is trained to determine equivalent circuit elements directly from the measured scattering parameters without the need for any additional tuning of the elements. In this study the results of a comprehensive analysis of the neural network based extraction procedures are presented. Stability of the extracted values with the choice of the input set of scattering parameters as well as accuracy of the final small-signal model were examined. Moreover, the influence of the number of measured data necessary for development of reliable neural models was investigated. The extraction procedure was examined for a HEMT transistor working under varying temperature conditions.

Efficient ANN based noise modeling of microwave FETs against temperature

An improved noise modeling technique for microwave MESFET/HEMT versus temperature is presented. It is based on an artificial neural network (ANN) that produces noise parameters as its outputs for device temperature, S parameters and frequency at its inputs. Once trained, the proposed model can be used for efficient prediction of transistor noise parameters over a wide temperature range. Since the model is based on ANN, all noise-generating mechanisms are included and therefore it is more accurate than empirical transistor models, as it is shown on a numerical example

Extraction of Pospieszalski's noise model parameters of microwave FETs based on ANNs

A new neural approach for extraction of the Pospieszalskis noise model parameters of microwave FETs is presented in this paper. This approach is based on the use of two artificial neural networks. The first network is aimed at calculating the intrinsic noise parameters from the given equivalent circuit parameters, transistor total noise parameters, frequency and ambient temperature. Since the gate noise temperature in the Pospieszalskis noise model is approximately equal to the ambient temperature, only the value of drain noise temperature is to be determined. Therefore, the second network is trained to determine drain noise temperature from the given extracted intrinsic noise parameters, equivalent intrinsic circuit parameters, frequency and ambient temperature. The proposed extracting approach enables avoiding time-consuming optimization procedures in microwave simulators, which are conventionally used for the determination of the noise model parameters. A detailed validation of the proposed approach was done by comparison of the measured transistor noise parameters with those obtained by using the extracted drain noise temperature.

ANN based inverse modeling of RF MEMS capacitive switches

RF MEMS switches have been efficiently applied in various applications in communication systems. Therefore, there is a need for reliable and accurate models of RF MEMS switches. Artificial neural networks (ANNs) have been appeared as very efficient alternative to time consuming full-wave and/or mechanical simulations of RF MEMS devices. However, to optimize the switch geometry it is usually necessary to perform certain optimization procedures. In this paper the development of ANN based procedures to be used as a feed-forward tool for determination of the switch geometrical parameters avoiding optimizations is proposed. The proposed procedure is developed for determination of the length of the bridge fingered part of a capacitive switch to achieve the desired electrical resonance frequency or the necessary actuation voltage.

ANN approach for analysis of actuation voltage behavior of RF MEMS capacitive switches

RF MEMS switches applications in communication systems have been increased in the recent time, creating a need for reliable and accurate switch models. This paper deals with the switch models based on the artificial neural networks. In particular, a model relating the switch actuation voltage and the switch geometry is considered. The model allows fast and efficient analysis of switch actuation voltage behavior, requiring significantly shorter time for the same analyses than standard mechanical simulators. The change of actuation voltage with the tolerances in the bridge dimensions is studied as well.

Artificial neural networks for efficient RF MEMS modeling

In this contribution we will show how artificial neural networks can be efficiently used to build models of RF MEMS components or to optimize them without enhanced numerical efforts. The method is especially interesting for designers and technologists, who want to modify or optimize switch parameters for a fixed technology without using heavy simulation tools. As examples the fast and accurate calculation of scattering parameters for an ohmic switch dependent on 4 different geometrical dimensions over frequency is shown. The second example is the derivation of some lateral dimensions for the resonant frequency of a capacitive switch without using optimization routines.

ANN based extraction of equivalent noise temperatures in microwave FET noise models

A procedure for extraction of equivalent noise temperatures of microwave FETs using artificial neural networks is presented in this paper. The Pospieszalskis noise model is considered. A neural network is trained to predict equivalent drain temperature for given equivalent intrinsic circuit elements, intrinsic circuit noise parameters, ambient temperature and frequency. The suggested procedure enables avoiding optimization procedures in microwave circuit simulators. The proposed approach is validated by comparison of the noise parameters calculated by using the extracted drain temperature with the reference ones obtained by Pospieszalskis approach.

Bias-dependent hybrid PKI empirical–neural model of microwave FETs

Empirical models of microwave transistors based on an equivalent circuit are valid for only one bias point. Bias-dependent analysis requires repeated extractions of the model parameters for each bias point. In order to make model bias-dependent, a new hybrid empirical–neural model of microwave field-effect transistors is proposed in this article. The model is a combination of an equivalent circuit model including noise developed for one bias point and two prior knowledge input artificial neural networks (PKI ANNs) aimed at introducing bias dependency of scattering (S) and noise parameters, respectively. The prior knowledge of the proposed ANNs involves the values of the S- and noise parameters obtained by the empirical model. The proposed hybrid model is valid in the whole range of bias conditions. Moreover, the proposed model provides better accuracy than the empirical model, which is illustrated by an appropriate modelling example of a pseudomorphic high-electron mobility transistor device.

An analytical approach to the HEMT noise wave model parameter determination

This paper presents an analytical approach to determination of the noise wave model parameters for a high electron-mobility transistor working under different temperature and frequency conditions. The presented approach is composed of two steps and provides more efficient determination of these parameters than in the case of optimization procedures commonly applied for that purpose in circuit simulators. The first step is extraction of the noise parameters of transistor intrinsic circuit from the measured noise parameters of whole transistor using an analytical noise de-embedding procedure. The second step is calculation of the noise wave model parameters from the de-embedded intrinsic noise parameters using existing formulas. The accuracy of the presented approach is validated in a wide frequency and temperature range by comparison of the transistor noise parameters simulated for the determined noise wave model parameters with the measured noise parameters.

COMPARATIVE ANALYSIS OF DIFFERENT CAD METHODS FOR EXTRACTION OF THE HEMT NOISE WAVE MODEL PARAMETERS

The noise wave model has appeared as a very appropriate model for the purpose of transistor noise modeling at microwave frequencies. The transistor noise wave model parameters are usually extracted from the measured transistor noise parameters by using time-consuming optimization procedures in microwave circuit simulators. Therefore, three different Computer-Aided Design methods that enable more efficient automatic determination of these parameters in the case of high electron-mobility transistors were developed. All of these extraction methods are based on different noise de-embedding procedures, which are described in detail within this paper. In order to validate the presented extraction methods, they were applied for the noise modeling of a specific GaAs high electron-mobility transistor. Finally, the obtained results were used for the comparative analysis of the presented extraction approaches in terms of accuracy, complexity and effectiveness.