Noise wave modeling of microwave transistors based on neural networks

Abstract

that is, over-wavelength focusing is achieved if the time interval during which the source is switched on is larger than the travel time of the light between the source and the focus, along the corresponding ray of light. This is a rather intuitive and obvious condition from the standpoint of geometrical optics. CONCLUSIONThe focusing properties of physical LHM slabs have been found to be mainly limited by the slab thickness, the material losses, and the dispersive behavior of the LHM, rather than by the free-space wavelength of the incident field. The limitations for sub-and over-wavelength focusing, imposed by material losses and dispersion, have been quantitatively expressed, and numerical examples which illustrate these expressions have been provided. A very strong (logarithmic) dependence of the sub-wavelength focusing properties on both the loss tangent and the dispersive characteristics of the LHM has been found. This strong dependence may constitute a severe limitation from a practical standpoint, but does not constitute a fundamental limit for sub-wavelength focusing. In addition, it has been shown that the condition for over-wavelength focusing of realistic quasimonochromatic sources has a simple and rather intuitive interpretation in terms of geometrical optics. ACKNOWLEDGMENTSThis work has been supported by the Spanish Ministry of Science and Technology and FEDER funds (project no. TIC2001-3163).

Improved Noise Wave Model of Microwave FETs based on Artificial Neural Networks

“…The proposed model is more efficient and more accurate than the noise wave model where ANNs are applied to model frequency dependence of noise wave temperatures, [7], while the number of the measured data needed for model development remains the same. Namely, in order to model frequency dependence of the noise wave temperatures, after the values of equivalent circuit elements are extracted, development of the model whose noise wave temperatures vary with frequency requires extraction of the noise wave temperatures at several frequencies.…”

“…Comparing to the existing device noise wave model [7], the proposed model is more efficient and more accurate, while the number of the measured data needed for model development remains the same.…”

“…These temperatures, assumed to be constant over the whole frequency range, are obtained on the basis of certain experimental noise data by applying standard optimization procedures. Modeling accuracy is improved further assuming that noise wave temperatures vary with frequency, [6], [7]. Those frequency dependences can be modeled by polynomials, [6], or by artificial neural networks (ANN), [7].…”

Comparative Analysis of Different Cad Methods for Extraction of the Hemt Noise Wave Model Parameters

“…Most of the proposed transistor noise models represent the noise generated in a component using equivalent voltage and/or current sources [3,[5][6]. However, at microwave frequencies, a wave representation of noise looks attractive since it allows the use of scattering matrices for the noise computations, leading to advantages in Computer-Aided Design (CAD) of microwave networks [2,[7][8][9][10][11][12][13][14][15][16][17][18]. The power of the wave representation of noise lies in its ability to treat the network noise in terms of incident and reflected waves, which is compatible with the scattering matrix description of microwave networks.…”

Artificial neural network applications in improved noise wave modeling of microwave FETs

“…Those frequency dependences can be modeled by polynomials or by artificial neural networks (ANNs) [9].…”