Bsim3 学习笔记12

Model Parameter Extraction 提取

 

There are two different optimization strategies which can be used for parameter extraction: global optimization and local optimization.

Global optimization lets the computer find one set of parameters which best fit all the available experimental data. Global optimization may minimize the error between the simulation results and the available experimental data. However, any particular parameter extracted by global optimization may not have a close resemblance to its actual physical value. In local optimization each parameter is extracted in a certain operation region where its corresponding device’s behavior is dominant. Parameters optimized locally may not perfectly fit the experimental data in all the operating regions, but they are closely related to the physical processes at work.

 

Parameter Extraction for BSIM3v3

 

Extraction routines

1. 1.      Devices and measurements needed for parameter extraction.

For scalable model parameter extraction without using a binning approach. three sets of different size devices are needed to extract the model parameters.

One set of devices has a fixed channel width and different channel lengths.

One set of devices has a fixed long channel length and different channel widths.

One set of devices have a fixed shortest channel length and different channel widths.

 

Five sets of data are recommended to be measured for each device and saved into different files for DC parameter extraction.

(a) Ids vs.Vgs at different Vbs and Vds=0.05V (linear region measurement).

(b) Ids vs. Vds at different Vgs and Vbs=0V (linear & saturation region measurements).

(c) Ids vs. Vgs at different Vbs and Vds = Maximum Vds (saturation region measurement).

(d) Ids vs. Vgs at different Vgs and Vbs = Vbb (linear & saturation region measurements,

and |Vbb|is the maximum body bias).

(e) I sub vs. Vgs at different VdS and Vbs = 0 and Vbb (substrate current measurements)

 

1. Optimization method

The optimization process recommended for BSIM3v3 is a combination of Newton-Raphson iteration and a linear-least-square fit with either one, two, or three variables.

1. Parameter extraction procedures
2. Some process parameters have to be provided by the users before starting the parameter extraction.

Input Parameter Names	Physical Meaning of the input Parameter
TOX	Gate oxide thickness
NCH	Doping concentration
T	Temperature at which the data is taken
Ldrawn	Designated channel length
Wdrawn	Designated channel width
Xj	Junction Depth

1. Extract DC model parameters
2. Extract the model parameters for the overlap and intrinsic capacitances
3. Extract the model parameters for the temperature dependencies

 

 

 

Binning Methodology

A different set of model parameters is used in each bin. The numbers of bins are selected according to the required model accuracy. Generally, more bins are assigned to the region of short channel lengths and narrow channel widths, and less bins are needed for the region of large channel lengths and wide channel widths.

但是大部分或者几乎所有MOS技术可以被很好的用偏压和几何建模，而不是使用binning approach。

所以不建议使用binning approach。

The task of parameter extraction can be greatly simplified with the help of an automated software tool. A high quality tool provides the local and global optimization routines and supports the single device and group device approaches, as well as the single-bin and multi-bin methodologies. Such a tool can usually extract the parameter for many popular MOSFET compact models as well as bipolar transistor models. Convenient user interfaces including curve plotting functions are among the features that can be expected from these tools.

All the example model files used in this book are extracted using the parameter extraction software – BSIMPro 用这个！！！