Qi J. Zhang

A generalized space-mapping tableau approach to device modeling

A comprehensive framework to engineering device modeling, which we call generalized space mapping (GSM) is introduced in this paper. GSM permits many different practical implementations. As a result, the accuracy of available empirical models of microwave devices can be significantly enhanced. We present three fundamental illustrations: a basic space-mapping super model (SMSM), frequency-space-mapping super model (FSMSM) and multiple space mapping (MSM). Two variations of MSM are presented: MSM for device responses and MSM for frequency intervals. We also present novel criteria to discriminate between coarse models of the same device. The SMSM, FSMSM, and MSM concepts have been verified on several modeling problems, typically utilizing a few relevant full-wave electromagnetic simulations. This paper presents four examples: a microstrip line, a microstrip right-angle bend, a microstrip step junction, and a microstrip shaped T-junction, yielding remarkable improvement within regions of interest.

A hierarchical neural network approach to the development of a library of neural models for microwave design

Neural networks recently gained attention as a fast and flexible vehicle to microwave modeling simulation and optimization. This paper addresses a new challenge in this area, i.e., development of libraries of microwave neutral models. A hierarchical neural network framework is presented utilizing the knowledge of basic relationships common to all library components. The proposed method improves the reliability of neural models, while significantly reducing the cost of library development through reduced need for data collection and shortened time of training.

A Neural Network Approach to the Modeling of Heterojunction Bipolar Transistors from S-Parameter Data

Artificial neural networks have gained attention as a fast, efficient, flexible and accurate tool in the areas of microwave modeling, simulation and optimization. In this paper, a novel neural network approach is proposed for the modeling of Heterojunction Bipolar Transistors (HBT) directly from their S-Parameter data. The neural network structure incorporates bias current and bias voltage as inputs. This enables us to use the same neural model under different bias conditions. The proposed technique provides reliable neural transistor models, while significantly reducing the cost effort and complexity involved in the modeling of HBT.

New directions in model development for RF/microwave components utilizing artificial neural networks and space mapping

This paper presents advances in model development for RF/microwave components exploiting two powerful technologies: artificial neural networks (ANN) and space mapping (SM). We survey the fundamental issues on classical neuromodeling. We review some state-of-the-art neuromodeling techniques, emphasizing SM based neuromodeling techniques. We show how SM based neuromodels decrease the cost of training, improve generalization ability and reduce the complexity of the ANN topology w.r.t. the classical neuromodeling approach. We illustrate these novel approaches through a practical microwave modeling problem. We conclude by proposing some possible exciting future applications of ANN and SM in microwave CAD.

On knowledge-based neural networks and neuro-space mapping

This article reviews the most significant milestones in CAD methodologies for intelligent EM-based modeling and design optimization using artificial neural networks and space mapping. We consider knowledge-based and automatic neural network model generation based on advanced data sampling algorithms. Computationally efficient neural space mapping methods for highly accurate EM-based modeling, statistical analysis and yield estimation are described. We briefly compare different strategies for developing suitable (input and output) neuro-mappings. Inverse modeling exploiting neural networks is addressed, including neural inverse space mapping optimization. Embedded passives, microstrip filters, active devices and waveguide structures illustrate the techniques.

Neural space mapping EM optimization of microwave structures

We propose, for the first time, Neural Space Mapping optimization for EM-based design. It exploits our Space Mapping-based neuromodeling techniques, avoiding troublesome parameter extraction. Simple neuromodels are trained, without testing points, during each optimization iteration. Coarse model sensitivities are exploited to select suitable fine model base points for the initial mapping.

Neural network structures for EM/microwave modeling

Neural networks have gained attention as a fast and flexible vehicle to EM/microwave modeling, simulation and optimization. Neural network device/circuit models can be used during EM/microwave design to provide instant answers. The structure of the neural network influences factors such as training data and training time required, and the accuracy that could be achieved. This paper discusses various microwave-oriented neural network structures.

Solutions of EM problems using finite element & complex frequency hopping techniques

The focus of this paper is to introduce the basic concepts and features of a new efficient technique for solution of time or frequency domain electromagnetic field problems. The method is based on Laplace domain finite element formulation of problems which are solved with the aid of Complex Frequency Hopping (CFH) technique. CFH is a moment matching technique used successfully in the circuit simulation area for solution of large set of ordinary differential equations. CPU measurements show that the proposed method is one to two orders of magnitude faster than the current solution methods associated with the finite element technique.

Yield-Driven EM Optimization using Space Mapping-Based Neuromodels

In this work, an efficient procedure to realize electromagnetics-based yield optimization and statistical analysis of microwave structures using space mapping-based neuromodels is proposed. A generalized relationship between the fine and coarse model sensitivities through the Jacobian of the neuromapping is proposed. Our technique is illustrated by the EM-based statistical analysis and yield optimization of an HTS microstrip filter.