Low Noise Amplifier Synthesis Using Multidimensional MLP Neural Network
- 23 August 2017
- journal article
- research article
- Published by Taylor & Francis in IETE Journal of Research
- Vol. 64 (3) , 374-386
Abstract
In this paper, a new method is proposed for determining the design parameters of radio frequency (RF) low noise amplifier (LNA) to obtain a set of desired design specifications. The proposed method is a simulation-based approach in which the multidimensional multilayer perceptron (M-MLP) neural network (NN) is utilized as the synthesis tool. The outputs of the M-MLP NN constitute the design variables, i.e., the numerical values of the passive components, transistors’ sizing, and the biasing conditions. The algorithm is implemented in MATLAB and the HSPICE simulator is utilized to verify the synthesis results. Strictly speaking, in each feed-forward process of the NN, a set of design parameters are produced at the output of the M-MLP NN, corresponding to which a set of design specifications are derived using HSPICE. These specifications include the input and output return loss, power gain, reverse isolation, noise figure, K stability factor, 3-dB bandwidth, and centre frequency. Considering the differences between the derived specifications and the desired ones, the connection weights of the M-MLP NN are altered in a way that it improves the overall behaviour of the algorithm toward satisfying the competing design measures. In other words, having the desired specifications, our method yields a completely automated design procedure. The formulation of the approach is described in detail. The dynamic learning rate is utilized to further improve the performance of the algorithm to search the solution space efficiently. The TSMC model for the 0.18 µm CMOS process is used in the LNA design.Keywords
This publication has 23 references indexed in Scilit:
- On Adaptive Learning Rate That Guarantees Convergence in Feedforward NetworksIEEE Transactions on Neural Networks, 2006
- Reference Point Based Multi-Objective Optimization Using Evolutionary AlgorithmsInternational Journal of Computational Intelligence Research, 2006
- CMOS Low-Noise Amplifier Design Optimization TechniquesIEEE Transactions on Microwave Theory and Techniques, 2004
- Regularising neural networks using flexible multivariate activation functionNeural Networks, 2004
- EM-Based Optimization of Microwave Circuits Using Artificial Neural Networks: The State-of-the-ArtIEEE Transactions on Microwave Theory and Techniques, 2004
- Artificial neural networks for rf and microwave design-from theory to practiceIEEE Transactions on Microwave Theory and Techniques, 2003
- Extraction and use of neural network models in automated synthesis of operational amplifiersIEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2003
- Multilayer perceptron for nonlinear programmingComputers & Operations Research, 2002
- Optimal design of a CMOS op-amp via geometric programmingIEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2001
- Utilizing feedforward neural networks for acceleration of global optimization procedures [SMES problems]IEEE Transactions on Magnetics, 1998