1,720,969 research outputs found
A comprehensive framework for training stable and passive multivariate behavioral models
Full text linkWe present a theoretical framework and related algorithms for the construction of behavioral models of linear or linearized devices. Unlike competing approaches, the proposed method is robust and guarantees theoretically the uniform stability and passivity of the models in a multivariate setting, where the model behavior depends not only on time or frequency but also on a number of design/stochastic parameters. Various examples demonstrate the high accuracy and reliability of proposed framework
Multiclass Sparse Centroids With Application to Fast Time Series Classification
Full text linkIn this article, we propose an efficient multiclass classification scheme based on sparse centroids classifiers. The proposed strategy exhibits linear complexity with respect to both the number of classes and the cardinality of the feature space. The classifier we introduce is based on binary space partitioning, performed by a decision tree where the assignation law at each node is defined via a sparse centroid classifier. We apply the presented strategy to the time series classification problem, showing by experimental evidence that it achieves performance comparable to that of state-of-the-art methods, but with a significantly lower classification time. The proposed technique can be an effective option in resource-constrained environments where the classification time and the computational cost are critical or, in scenarios, where real-time classification is necessary
Data-driven extraction of uniformly stable and passive parameterized macromodels
Full text linkA Robust algorithm for the extraction of reduced-order behavioral models from sampled frequency responses is proposed. The system under investigation can be any Linear and Time Invariant structure, although the main emphasis is on devices that are relevant for Signal and Power Integrity and RF design, such as electrical interconnects and integrated passive components. We assume that the device under modeling is parameterized by one or more design variables, which can be related to geometry or materials. Therefore, we seek for multivariate macromodels that reproduce the dynamic behavior over a predefined frequency band, with an explicit embedded dependence of the model equations on these external parameters. Such parameterized macromodels may be used to construct component libraries and prove very useful in fast system-level numerical simulations in time or frequency domain, including optimization, what-if, and sensitivity analysis. The main novel contribution is the formulation of a finite set of convex constraints that are applied during model identification, which provide sufficient conditions for uniform model stability and passivity throughout the parameter space. Such constraints are characterized by an explicit control allowing for a trade-off between model accuracy and runtime, thanks to some special properties of Bernstein polynomials. In summary, we solve the longstanding problem of multivariate stability and passivity enforcement in data-driven model order reduction, which insofar has been tackled only via either overconservative or heuristic and possibly unreliable methods
On stabilization of parameterized macromodeling
Full text linkWe propose an algorithm for the identification of guaranteed stable parameterized macromodels from sampled frequency responses. The proposed scheme is based on the standard Sanathanan-Koerner iteration in its parameterized form, which is regularized by adding a set of inequality constraints for enforcing the positiveness of the model denominator at suitable discrete points. We show that an ad hoc aggregation of such constraints is able to stabilize the iterative scheme by significantly improving its convergence properties, while guaranteeing uniformly stable model poles as the parameter(s) change within their design range
A Scalable Reduced-Order Modeling Algorithm for the Construction of Parameterized Interconnect Macromodels from Scattering Responses
Full text linkThis paper introduces an algorithm for the construction of reduced-order macromodels of electrical interconnects starting from their sampled scattering responses. The produced macromodels embed in a closed-form an approximate dependence of the model equations on external parameters such as geometrical dimensions or material characteristics. The resulting parameterized models are easily cast as parameter-dependent SPICE netlists, which can be used for system-level Signal and Power Integrity assessment via numerical simulation, including sensitivity and optimization tasks. The main novel contribution of this work is the formulation of the model fitting equations in a decoupled form, which allows for a very efficient implementation in case of interconnects with a large number of interface ports, as typically required in Signal and Power Integrity applications. The parameterized models are guaranteed stable and passive for any configuration of the external parameters, thus ensuring stable transient numerical simulations
Efficient EM-based variability analysis of passive microwave structures through parameterized reduced-order behavioral models
Full text linkIn this contribution we demonstrate how reduced-order behavioral models allow for extremely accurate and computationally efficient electromagnetic-based variability analysis of microwave passive structures. In particular, we report the MonteCarlo analysis of a wideband matching network at Ka-band, designed with a commercial foundry GaN-HEMT process PDK. As sources of variation we considered the thickness of the two dielectric layers available in the PDK to implement MIM capacitors of different order of magnitude, both exploited in the network. Based on a limited set of electromagnetic simulations, a parameterized behavioral model is extracted and then translated into a parameterized circuit equivalent (SPICE netlist) straightforward to be imported into RF CAD tools. The adopted model, implementing a rational approximation of the simulated S-parameters with rational dependence on the two parameters, provides excellent agreement with electromagnetic simulations, robustness against port impedance change and good extrapolation capabilities
Structured black-box parameterized macromodels of integrated passive components
Full text linkA novel black-box model representation and identification process is introduced, specifically designed to extract layout-scalable behavioral macromodels of passive integrated devices from sampled frequency-domain responses. An automated choice of structured frequency-domain basis functions enables extremely accurate approximations for responses characterized by high dynamic ranges over extended frequency bands, overcoming the main limitations of standard approaches. Numerical results confirm that the proposed structured approach provides robust and reliable scalable models, with guaranteed stability and passivity over the frequency band and parameter space of interest
An Adaptive Algorithm for Fully Automated Extraction of Passive Parameterized Macromodels
Full text linkWe present a general framework for the fully automated extraction of stable and passive parameterized macromodels from sampled frequency responses. The proposed iterative algorithm provides an automated selection of the optimal parameter configurations to be simulated by a field solver, based on a combination of data-driven and model-driven metrics. The resulting frequency responses are fitted by a parameterized rational macromodel, whose uniform stability and passivity are enforced. We demonstrate the effectiveness of this framework on a transmission-line network test case
A Framework for the Generation of Guaranteed Stable Small-Signal Bias-Dependent Behavioral Models
Full text linkWe present a numerical scheme for the identification of compact surrogate models of analog circuit blocks. The basic assumption is small signal operation, so that a local linearization can be applied around a given bias point, resulting in a bias-dependent linear state-space behavioral macromodel. The main novel contribution of this work is the ability to embed in the identification process a suitable set of constraints, that are able to guarantee the uniform stability of the model for any bias value within a prescribed design range
Enabling fast power integrity transient analysis through parameterized small-signal macromodels
Full text linkIn this paper, we present an automated strategy for extracting behavioral small-signal macromodels of biased nonlinear circuit blocks. We discuss in detail the case study of a Low DropOut (LDO) voltage regulator, which is an essential part of the power distribution network in electronic systems. We derive a compact yet accurate surrogate model of the LDO, which enables fast transient power integrity simulations, including all parasitics due to the specific layout of the LDO realization. The model is parameterized through its DC input voltage and its output current and is thus available as a SPICE netlist. Numerical experiments show that a speedup up to 700X is achieved when replacing the extracted post-layout netlist with the surrogate model, with practically no loss in accuracy
- …
