Small-signal and large-signal modeling of active devices using CAD-optimized neural networks

Abstract

Artificial neural networks (ANNs) are presented for the technologyindependent modeling of active devices in MMICs. ANNs trained with S-parameter and DC measurements over the entire bias and frequency operational band are used for the small-signal bias-dependent modeling of a low-noise GaAs HEMT device, without the need of the equivalent circuit parameter extraction. ANNs are also used within the large-signal model of a power MESFET device, modeling the drain-source current I ds and charges Q g and Q d obtained from integration of their partial derivatives. After training and testing, the ANN models have been implemented as two-port networks into a microwave circuit simulator. This enabled the ANN models to be used in the design, analysis, and optimization of microwave/mm-wave circuits. Improved techniques in network building to provide not only accurate but also fast simulation models have been applied.

A neural network approach for compact cryogenic modelling of HEMTs

“…An interesting ANN-based procedure implemented for CAD-oriented modelling of GaAs active devices has been reported in (Giannini et al 2002). In that work, the device performance was reproduced by multiple-cascaded ANNs leading to either a bias-dependent small-signal HEMT model and to a large signal MESFET model without involving the determination of an equivalent circuit.…”

Signal and Noise Modeling of Microwave Transistors Using Characteristic Support Vector-based Sparse Regression

“…Particularly the Back-Propagation Multi-Layer Perceptron (BPMLP)s have been employed for the nonlinear interpolation based on "learning" from the measured or simulated data, in the fast, accurate and reliable modeling of both active and passive microwave devices [1][2][3][4][5][6]. Today's fast, accurate and reliable Signal S-and Noise Nblack-box model of a microwave transistor can be achieved only by a single simple BPMLP with one hidden layer which is capable of the simultaneous generalization of 12 Scattering S-and Noise N-functions into the entire operation domain of the bias condition V DS /V CE , I DS /I C and the frequency f for all the configuration types [3].…”

“…On the other hand, the training time and memory of SVM are expensive and are strongly correlated to the number of training patterns  , as O( 3  ) and O( 2  ), respectively [17][18][19]. Thus approximately at least 50% of the training data is reduced that corresponds to 1/8 and 1/4 of the training time and the memory complexity respectively of the SVRM as the nonlinear interpolator in the S-and Nmodeling used in [3][4][5][6][7][8]. The proposed method is implemented on the modeling of a low-noise microwave transistor ATF-551M4 and VMMK-1225 as a case study.…”

Signal-noise support vector model of a microwave transistor

“…The neural networks are typical learning machines and used in the black-box modeling of nonlinear devices. Neural transistors [4][5][6][7] are typical black-box models of microwave transistor that generalizes the signal-and noise parameters from N-measured samples accurately over the whole operation domain of the device, which consists of bias conditions (V DS , I DS ) and the operation frequency f. These black-box models are employed successfully in designing microwave devices and circuits [8,9]. Neural networks are also utilized in characterization of the nonlinear properties of the device such as third order distortion characterization of a transistor in [10].…”