Data processing: structural design – modeling – simulation – and em – Modeling by mathematical expression
Reexamination Certificate
2000-02-23
2004-08-10
Jones, Hugh (Department: 2128)
Data processing: structural design, modeling, simulation, and em
Modeling by mathematical expression
Reexamination Certificate
active
06775646
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to the determination of behavioral models for nonlinear devices, circuits, subsystems or systems. In particular the present invention is a method for creating excitation signals and a method for performing function fitting in the determination of behavioral models for nonlinear devices, circuits, subsystems, and systems from embeddings of time-domain measurements.
2. Description of the Related Art
Linear, time invariant (LTI) devices, circuits, subsystems, and systems are completely characterized by their transfer functions. To understand the performance of an LTI device, one need only determine the transfer function of the LTI device. Once the transfer function is known, the operation of the device in a system is known completely for all input conditions. The same is true for LTI circuits, subsystems and systems.
A transfer function is a complex frequency domain-function that describes the output of an LTI device in terms of its inputs and therefore, forms a complete description of the LTI device. The term complex function when used herein refers to a function that includes complex numbers having a real and an imaginary part. An equivalent form of the transfer function of an LTI device in the time-domain is called an impulse response of the LTI device. A one-to-one relationship exists between the transfer function in the frequency-domain and impulse response in the time-domain. In addition, the transfer function and the impulse response are not functions of and do not depend on the input signal that is applied to the LTI device.
The determination of the transfer function, especially if it involves measured data from the LTI device, is known as model development or model parameter extraction. Once a model of an LTI device is developed, or equivalently the transfer function is known, for a given device, the actual device may be replaced by a virtual device based on the model in any simulation of a system using the device. Often the development of the model involves extraction or determination of model parameters from a set of test data that represents the device of interest.
Transfer functions of LTI devices, circuits, subsystems, or systems can be extracted from measurements made with a vector spectrum or vector network analyzer. A swept or stepped frequency input signal is generated and the vector spectrum analyzer or network analyzer records the output of the LTI device. Then, a transfer function can be computed by comparing the input and output signals. Furthermore, models suitable for simulation of a given LTI device or circuit can extracted from transfer functions using, among other things, linear system identification techniques.
Time-domain measurements provide an alternate method of characterizing LTI devices or circuits. Pulse inputs that approximate an impulse are applied to a device and the outputs are measured and recorded. In one such well known, time-domain method, the poles and zeros of the Laplace transform of the governing differential equation of the device are estimated from the recorded output data. Once a suitable governing differential equation is determined, the device transfer function is calculated. In an alternative method, the measured data associated with the impulse response is transformed using a Fast Fourier Transform (FFT) to the frequency-domain where a linear system identification method is then used to extract the transfer function.
The characterization or modeling of nonlinear devices or circuits is much more difficult than that for LTI devices. Reference to a “nonlinear device” when used herein will be understood to include devices, circuits, subsystems or systems with a nonlinear input-output relationship. Unlike the linear case, the nonlinear device or circuit is not readily represented by a transfer function or impulse response, at least not one that is independent of the input signal or stimulus. However, there is still a need to model nonlinear devices so that their performance in systems can be evaluated efficiently. This is especially true when it is impractical or too expensive to use the actual device, such as when the device is still being designed.
It is desirable to have a method for characterizing and developing a model of nonlinear devices to avoid the need to have the actual device available whenever its performance in a system must be investigated. Furthermore it is advantageous to have such a modeling method utilize a finite set of measurements, either actual measurements or measurements of a simulation of the device. The model so generated must accurately predict the performance of the device over all expected operational conditions within a given level of accuracy and with an acceptable amount of computational cost.
The term “behavioral model” herein refers to a set of parameters that define the input-output behavior of a device or circuit. Generally, a behavioral model must be of a form suitable for rapid simulation. “Simulated measurements” refers to values of voltage, current or other physical variables obtained from device, circuit or system simulation software. The objective of building a behavioral model from actual or simulated measurements is to reduce simulation time by replacing a complex circuit description in the simulation with a simpler, easier to simulate, behavioral model.
In many cases, nonlinear devices are electronic in nature (e.g. transistors, diodes). In these cases the measurements used to produce a model of the device are typically measured voltages and currents in and out of the ports of the device or equivalently incident or reflected power waves present at the ports at various frequencies. The models extracted from the measurements generally need to reflect the dynamic relationships between the voltages and currents at the ports. The model can be used, for example, to compute the currents into the ports from recent values of the voltages across the ports. Often this is the essential computation that must be provided to electronic circuit simulators by a software module that represents a device.
Mechanical and hydraulic devices can also exhibit nonlinear behavior and, therefore, be modeled as nonlinear devices for which construction of a suitable behavioral model would be beneficial. For example, a vehicular system comprising driver inputs and vehicle response may be represented in terms of a nonlinear behavioral model. In the case of vehicular systems, the input measurements might be of variables such as steering wheel position, brake pressure, throttle position, gear selection and the response measurements might be of variables such as the vehicle speed, lateral and longitudinal acceleration, and yaw rate. The behavioral model extracted from the measurements needs to reflect the dynamic relationship between the driver inputs that are applied and the subsequent response of the vehicle. In other words, the model defines a “virtual car” that can be “driven” using previously recorded or real-time measured driver inputs.
A variety of methods have been developed to characterize and develop models of nonlinear devices. However, these methods generally have severe limitations associated with them. For instance, many of the techniques are limited to use with so called “weakly nonlinear devices”, those devices whose performance is nearly linear. Therefore, these techniques are not suitable for many nonlinear devices.
One such approach to characterization of weakly nonlinear devices is to simply assume that the device is linear, at least in the operational range of interest. Under this assumption, a variant of the time-domain impulse response method described hereinabove can be used. For devices that are, in fact, weakly nonlinear devices, this approach yields reasonably good results. However, the accuracy of such a model will degrade rapidly as the amount or degree of nonlinearity in the device increases.
Another class of methods for characterizing nonlinear devices is represented by the Volterra input-output maps method
Tufillaro Nicholas B.
Walker David M.
Agilent Technologie,s Inc.
Jones Hugh
LandOfFree
Excitation signal and radial basis function methods for use... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Excitation signal and radial basis function methods for use..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Excitation signal and radial basis function methods for use... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3342576