Artificial Neural Networks are emerging as a powerful technology for RF and microwave characterization, modeling, and design. Neural modeler helps us to immediately start developing neural models for RFhlicrowave components and circuits and helps to provide neural models for our simulators. In this study, a novel fuzzy neural network structure is used for behavior of an active microwave device. Here, the device is modeled by a black box whose small signal and noise parameters are evaluated through a fuzzy clustering neural network based upon the fitting of both of these parameters.
Signal and noise behaviours of microwave transistors are modeled through the neural network approach for the whole operating ranges including j-equency, bias and configuration types. Here, the device is modeled by a black box whose small-signal and noise parameters are evaluated through various neural network metho&, based upon the fitting of both of these parameters for multiple bias and confguration. Previous results are improved with a Conic Section Function Neural Network method in this work.
1.IntroductionAdequate representations of circuit elements for both signal and noise behaviours over their whole operation ranges are essential for the design of monolithic microwave integrated circuits. Even the characterisation of passive elements, which is relatively simple for low frequencies, can be difficult for microwave frequencies.In the case of semiconductor devices that are characterised by highly nonlinear models with a large set of parameters and complex relationships between them a proper selection of values for these parameters is a nontrivial task. On the other hand, since each circuit simulation involves a CPU-intensive procedure to solve the physics-based equations, such existing optimisation methods are more oriented towards off-line computations. They are not suitable for practical interactive design where designers may want to reoptimize the circuit after modifying the specifications, or even the circuit topology. To address these problems, two types of approximations have been previously used 1) Multidimensional polynomial (or its variants, such as splines or response d a c e ) model has been used to approximate and replace original simulations during optimisation. However, this approach can handle only mild nonlinearity in highdimensional space. It typically models building or updating during optimisation, consumes valuable on-line CPU time.2) The look-up table approach has been used 0-7803-5529-6/99/$10.00 01999 IEEE to approximate and replace accurately device or circuit simulations. However, the size of the table grows exponentially with the number of dimension and the table becomes too difficult to generate and manage when many parameters for a device or a circuit are involved On the other hand, neural networks have become a very important vehicle in the signal processing area for speech processing, vision, control systems, etc. Recently, they have been applied to microwave impedance matching, studying the effects of design factors on printed circuit board (PCB) assembly yield in modelling the properties of silicon dioxide films, and manufacturing process modelling. Neural networks enjoy the ability to learn from data, to generalise pattern in data, and to model nonlinear relationships. These appealing features make neural networks a good candidate to overcome some of the difficulties in traditional device and circuit modelling and optimisation. However, this potentially powmodelling approach has not been seriously addressed in the literature and bridging this gap is the objective of this paper...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.