Artificial neural networks approach to active inductor‐based filter design

Abstract

We present a novel technique based on artificial neural networks (ANNs), for design and optimization of active RF filters based on high‐ Q Active Inductors (AIs). ANNs are used to build the small‐signal S‐parameter model as well as the noise model of the transistor considered for the AI, by considering variable current and voltage biasing. With this approach, it is possible to rapidly identify the best design solution that lead to an optimal network configuration and filter order at the initial design steps. To demonstrate the advantage of this method, a bandpass filter was designed by means of a custom ANN model, made in MATLAB environment that is based on a commercial transistor, the BFP840ESD from Infineon. With this technique, we easily optimized the design to achieve an operating frequency of 5 GHz with a 3 dB bandwidth of about 30 MHz with a high Q‐factor and a minimum noise figure of 0.96 dB at the central frequency.

Design and Realization of Broadband Active Inductor Based Band Pass Filter

“…AIN can be simply designed by placing two end-toend connected transconductance amplifiers and one capacitor shown in Fig. 1(a) [7]- [10], which is also known as gyrator-C design. The inductance of gyrator-C AIN is proportional to the load capacitance C and inversely proportional to the product of trans-conductance parameters of the gyrator design as for ideal transconductance amplifier operation.…”

Novel neural network optimization approach for modeling scattering and noise parameters of microwave transistor

“…For power amplifiers, low‐noise amplifiers and other active microwave circuits including transistors, it is highly important to model the S and N parameters through all operational ranges of the transistors 6,12 . Many studies in the literature have used the Artificial Neural Network (ANN) approach in modeling S and N parameters which provides information on the linear behavior or microwave transistors 5,7,10,13–19 . In the literature, the ANN approach has not only been used in linear behavior modeling for microwave transistors.…”

“…Looking at the literature, it is understood that different ANN architectures have been used for linear behavior modeling of transistors. In studies, it is seen that parameters like the numbers of hidden layers, numbers of neurons and activation functions in the structure of the ANNs used have been determined manually with the trial and error method 4,5,7,10,13,15–20,22 . Gunes et al 13 used the General Regression Neural Network (GRNN) and Multi Layer Perceptron Neural Network (MLPNN) methods for modeling transistor S and N parameters.…”

“…6,12 Many studies in the literature have used the Artificial Neural Network (ANN) approach in modeling S and N parameters which provides information on the linear behavior or microwave transistors. 5,7,10,[13][14][15][16][17][18][19] In the literature, the ANN approach has not only been used in linear behavior modeling for microwave transistors. There are also studies where the ANN approach has been used for DC behavior modeling of transistors.…”

“…In studies, it is seen that parameters like the numbers of hidden layers, numbers of neurons and activation functions in the structure of the ANNs used have been determined manually with the trial and error method. 4,5,7,10,13,[15][16][17][18][19][20]22 Gunes et al 13 used the General Regression Neural Network (GRNN) and Multi Layer Perceptron Neural Network (MLPNN) methods for modeling transistor S and N parameters. They obtained the MLPNN number of hidden layers, number of neurons in each hidden layer and activation function parameters by the trial and error.…”