Data processing: structural design – modeling – simulation – and em – Simulating electronic device or electrical system – Circuit simulation
Reexamination Certificate
1998-11-30
2001-11-06
Powell, Mark R. (Department: 2123)
Data processing: structural design, modeling, simulation, and em
Simulating electronic device or electrical system
Circuit simulation
C703S004000
Reexamination Certificate
active
06314390
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to semiconductor devices and, in particular, to a process employing a stochastic search algorithm to determine model parameters for MOSFETs.
2. Description of Related Art
In order to achieve increased performance metal oxide semiconductor field effect transistors (MOSFETs) at ever decreasing line width, for example, at one-quarter micron foundry technology now being used for manufacturing semiconductor devices, it is critical to employ accurate modeling while creating the initial designs of such MOSFETs. There are several commercial software packages available to solve the specific problem of semiconductor device modeling, for example, the UTMOST modeling software available from Silvaco Data Systems, Santa Clara, Calif. Additionally, proprietary software has also been developed by many companies to solve this problem. In all the known approaches, a great amount of human interaction is required in the fitting process.
The modeling software generally performs only a local optimization and, accordingly, may not find a global minimum unless reruns are made with several different initial guesses. Furthermore, the modeling software is not able to provide a measure of the fit of the overall model because of the wide range of magnitudes of currents in different regions of operation. Therefore, an operator must perform successive optimizations in different regions and iterate through the successive regions until overall convergence is obtained. Existing software does not evaluate reasonableness of values beyond the limits of the measured data until after fitting the data. If problems are found at that time, then additional iterations are required.
Although U.S. Pat. No. 5,136,686 teaches the application of genetic algorithms to model fitting, there is no disclosure or suggestion that it may be applied to the problem of fitting models for circuit simulation, particularly for modeling MOSFETs. Moreover, this patent defines the fitness function as the sum of the distances in the model range space between each pair of measured and simulated points. For semiconductor devices, the model output varies by many orders of magnitude such that only error regions of high output values contribute significantly to the sum. Other areas are not well modeled because the fitness function does not adequately reflect the fit in small valued regions.
Additionally, it has been found to be important that a fitness function also include a measure of the physical reasonableness of the model in regions where actual data cannot be measured, because the device will not operate there. Such constraints are important for models which will be used in complex computer simulations because violating them can lead to failure of the simulations to converge.
Bearing in mind the problems and deficiencies of the prior art, it is therefore an object of the present invention to provide an improved method of determining modeling parameters for MOSFETs.
It is another object of the present invention to provide a MOSFET modeling algorithm which may achieve both global and local optimization.
A further object of the invention is to provide a method of determining model parameters for MOSFETS which requires less human interaction is required in the fitting process.
It is yet another object of the present invention to provide a method of determining a set of parameters for modeling a MOSFET which provides a measure of the fit of the overall model over a wide range of magnitudes of currents in different regions of operation.
It is another object of the present invention to provide a method of calculating certain critical parameters which may limit the size of the search space.
Still other objects and advantages of the invention will in part be obvious and will in part be apparent from the specification.
SUMMARY OF THE INVENTION
The above and other objects and advantages, which will be apparent to one of skill in the art, are achieved in the present invention which is directed to, in a first aspect, a method of determining a set of parameters for modeling an active semiconductor device in which current flow through a channel or other area is regulated by voltage applied to the device terminals, for example, MOSFETs. The method comprises first providing a plurality of measured values for current as a function of voltage for a plurality of active semiconductor devices of differing geometries. There is then determined an initial population of vectors comprising individual values representing a plurality of desired active semiconductor device model parameters. Fitness is then evaluated for each of the vectors by comparing calculated values for current as a function of voltage from the population to the plurality of measured values for current as a function of voltage of the vectors, converting any current differences to voltage errors and adding any such voltage errors together to arrive at a fitness value for each vector. Vectors of best fitness are selected and at least one genetic operator is applied thereto to create a new population of the vectors. Vectors of best fitness are then selected. The steps of evaluating fitness and selecting vectors of best fitness are optionally repeated for such vectors of best fitness until a desired fitness is achieved to determine the desired active semiconductor device model parameters.
Preferably, the measured values for current as a function of voltage for a plurality of active semiconductor devices comprise values for drain current as a function of gate voltage, drain current as a function of drain voltage, and self gain.
In another aspect, the present invention provides a method of determining a set of parameters for modeling an active semiconductor device in which current flow through a channel or other area is regulated by voltage applied to the device terminals. A plurality of measured values for current as function of voltage are provided for a plurality of active semiconductor devices of differing geometries. An initial population of vectors comprising individual values representing a plurality of desired active semiconductor device model parameters is then determined. Values are calculated for a subset of parameters comprising fewer than all of the desired model parameters from each of the vectors. Fitness is evaluated for each of the vectors comparing vector individual values to the measured values using a fitness function. Vectors of best fitness are selected and at least one genetic operator is applied thereto to create a new population of the vectors. Diversity of the population of vectors is tested and optionally the parameter subset value calculation, fitness evaluation and best fitness vector selection are prepared until a sufficiently low vector population diversity is acheived.
Meta evolution parameters are added to each vector from the vectors having sufficiently low population diversity to control genetic operators and permit the rate and direction of change to evolve with each vector from generation to generation. At least one genetic operator is applied to the vectors having added meta evolution parameters to create a new population of the vectors. Values for the subset of parameters from each of the new population of vectors are calculated and fitness is evaluated for each of the vectors by comparing vector individual values to the measured values using a fitness function. Vectors of best fitness are selected and optionally the genetic operation, parameter subset value calculation and fitness evaluation are repeated for such vectors of best fitness until a desired fitness is achieved to determine the desired active semiconductor device model parameters.
Preferably, the measured values for current as a function of voltage for a plurality of active semiconductor devices comprise values for drain current as a function of gate voltage, drain current as a function of drain voltage, and self gain. Also, the subset of parameters is preferably selected from the group consisting of threshold voltage, mob
Bittner Calvin J.
Hoffmann James P.
Watts Josef S.
DeLio & Peterson LLC
International Business Machines - Corporation
Jones Hugh
Powell Mark R.
Reynolds Kelly M.
LandOfFree
Method of determining model parameters for a MOSFET compact... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method of determining model parameters for a MOSFET compact..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of determining model parameters for a MOSFET compact... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2597561