L. De Tommasi

Automatic model type selection with heterogeneous evolution: An application to RF circuit block modeling

Many complex, real world phenomena are difficult to study directly using controlled experiments. Instead, the use of computer simulations has become commonplace as a cost effective alternative. However, regardless of Moorepsilas law, performing high fidelity simulations still requires a great investment of time and money. Surrogate modeling (metamodeling) has become indispensable as an alternative solution for relieving this burden. Many surrogate model types exist (support vector machines, Kriging, RBF models, neural networks, ...) but no type is optimal in all circumstances. Nor is there any hard theory available that can help make this choice. The same is true for setting the surrogate model parameters (bias- variance trade-off). Traditionally, the solution to both problems has been a pragmatic one, guided by intuition, prior experience or simply available software packages. In this paper we present a more founded approach to these problems. We describe an adaptive surrogate modeling environment, driven by speciated evolution, to automatically determine the optimal model type and complexity. Its utility and performance is presented on a case study from electronics.

Low Order Transmission Line Modeling by Modal Decomposition and Minimum Phase Shift Fitting

Frequency dependent transmission line models of the traveling wave type are usually implemented via rational approximation of the characteristic admittance and the propagation function. For the propagation function, a delay needs first to be compensated before proceeding with the rational approximation. It has been shown that assumption of a lossless time delay can lead to significant loss of accuracy of the rational approximation itself. In this paper is shown a procedure suitable for approximating the magnitude of the propagation function with a minimum phase shift rational function. This enables a more efficient search for the optimal delay that gives best accuracy for a given order of the rational approximation, since the function does not need to be refitted in the delay optimization loop

Single-Input-Single-Output Passive Macromodeling via Positive Fractions Vector Fitting

This paper introduces a constrained vector fitting algorithm which can directly identify a passive driving point function (impedance or admittance) from frequency domain data. The proposed positive fractions vector fitting (PFVF) algorithm formulates the residue identification step as a convex programming problem, while the pole identification step follows the unaltered standard Vector Fitting procedure. A further extension to multi-input-multi-output functions is possible and is under investigation.

Identification of broadband passive macromodels for electromagnetic structures

Purpose – The paper aims to present an overview of techniques for the identification in the frequency domain of reduced order models for distributed passive electromagnetic structures.Design/methodology/approach – Most known approaches proposed in different application contexts are described within a unified framework.Findings – A passive reduced order model of an unshielded twisted pair is fully developed with the combination of vector fitting algorithm and the passivity enforcement via Hamiltonian perturbation.Originality/value – A state‐of‐the‐art picture of the frequency domain identification and passivity enforcement techniques is given, and a test case of actual interest fully analysed.

Accurate Macromodeling Based on Tabulated Magnitude Frequency Responses

The paper introduces an approach for rational macromodeling based on magnitude frequency data. This is achieved by fitting the magnitude square function using vector fitting (VF) with symmetrical basis functions. The procedure is demonstrated to be more robust than a direct application of VF to the magnitude square function.

Low-Order Identification of Interconnects With the Generalized Method of Characteristics

This paper deals with the identification of low-order accurate macromodels describing electrically long lossy multiconductor transmission lines. The formulation is based on the generalized method of characteristics, whose key feature is the extraction of delays from the propagation operators. We describe and compare different delay extraction approaches, in view of the identification of the macromodels. In particular, we propose a new identification procedure enhancing the results coming from a vector fitting algorithm by a further nonlinear identification step. This gives more general properties to the model implementation (e.g., a state space realization with no repeated pole), possibly improving the accuracy at the same time. Some reference case studies are analyzed to highlight the key features of the proposed approach.

A comparison of sequential design methods for RF circuit block modeling

When modeling complex systems, the locations of the data points are essential to the success of the algorithm. Sequential design methods are iterative algorithms that use data acquired from previous iterations to guide future sample selection. They are often used to improve an initial design such as a Latin hypercube or a simple grid, in order to focus on highly dynamic parts of the design space. In this paper, a comparison is made between different sequential design methods on a real-world electronics problem. Error-based and density-based methods are compared against a novel hybrid technique which incorporates both an error-based measure, using gradient estimations of the objective function, and a density-based measure, using a Voronoi tessellation approximation. The test results indicate that a considerable improvement of the average model accuracy can be achieved by using this new approach.