- Buyers Guide
An IC Characterization and Analysis Program Update
Version 5.0 of the HP Integrated Circuit Characterization and Analysis Program (IC-CAP) has been developed and is available currently. This update to the popular device modeling program features a new statistics module to help process engineers and circuit designers improve yields and design more robust products by correlating device performance with process information. The new module was developed specifically for semiconductor manufacturers who need a self-contained, economical and easy-to-use statistical package. The statistical package offers a modeling database and a statistical analysis capability that provides parametric analysis features such as principal component analysis and factor analysis, as well as nonparametric analysis features. The module also includes boundary modeling, an efficient technique for worst-case modeling that models the actual probability density function and helps prevent circuit over designing.
IC-CAP is a UNIX-based program that automates the process of accurate device characterization. The program is used for numerous modeling tasks, including instrument control, data acquisition, graphical and statistical analyses, simulation and optimization. These processes are combined into a flexible and intuitive user environment for efficient and accurate extraction of active device and circuit model parameters, as shown in Figure 1 . IC-CAP also facilitates the building of model libraries for various simulators.
New Device Models
IC-CAP version 5.0 offers two new bipolar junction transistor (BJT) modeling packages for the silicon semiconductor industry: Philips Electronics’ most exquisite transistor model (MEXTRAM) and the vertical bipolar intercompany (VBIC) BJT model from the Bipolar/BiCMOS Circuits and Technology (BCTM) committee, a US industry consortium. These two BJT models are the most accurate in the industry and take into account the majority of the physical phenomena associated with modern BJT technologies.
The new software also provides the EEMOS1, a new metal-oxide semiconductor (MOS) model for discrete and power MOSFETs used in high frequency analog and RF circuits. An updated MOS (model 9) from Philips Electronics is included, which contains a newly added junction capacitance model. This new model requires fewer measurements and provides quick and accurate parameter extraction. The BSIM3v3.1, the latest release from U.C. Berkeley, contains new options for different noise models and new routines for avoiding discontinuities, and has also been added to IC-CAP 5.0.
An Enhanced User Interface
A number of usability and capability enhancements have been made available, including a redesigned user interface that provides more intuitive operation, enhanced graphical data presentations, improved data management and better links to other software tools. IC-CAP 5.0 also includes an update to the existing HP Root FET model, which is a driver for the model HP 8720D network analyzer. Additional new features include easier program installation and hard copy output.
Modeling WITH IC-CAP
Unlike other products, IC-CAP version 5.0 offers both accurate device models and advanced statistical analysis capabilities for building and maintaining accurate model libraries. The software contains two basic modeling functions: an independent statistical analysis package and a selection of models used for parameter extraction. The statistical analysis package is used in conjunction with any of the parameter extraction models, any American National Standard Code for Information Interchange (ASCII) format of measured data or even results from third-party process/device simulation tools.
A typical modeling procedure consists of selecting a model based on the application, extracting model parameters from device measurements using the program’s direct extraction algorithms and using the statistics module to build a statistical model for the devices. The statistical model then can be used with the IC-CAP simulators or exported to external simulators.
A number of high frequency circuit simulation models are contained in the IC-CAP software in addition to the new models mentioned previously. The HP Root models are a result of a highly automated data-acquisition system where DC and S-parameter measurements are taken over the device’s entire operating range.
The Statistics Package
The IC-CAP statistics package provides both parametric and nonparametric analysis features. Parametric analysis features include principal component analysis, principal factor analysis and multiple linear regression, making it easy to select the best model parameters to track in electrical test or to build models that predict SPICE parameters from dominant parameters or independent factors. These features help circuit designers and process engineers improve yields and design more robust products. The IC-CAP statistics package generates model files automatically from corner or Monte Carlo analysis. The new boundary modeling feature generates worst-case model candidates.
The IC-CAP nonparametric analysis tool uses a new patent-pending technique to handle arbitrary data distributions (Gaussian or non-Gaussian) and completes the nominal and boundary models. Nonparametric analysis works for all data from any arbitrary stichastic process. These processes need not be Gaussian or have any analytical probability density function.
The nonparametric statistics analysis begins when a nominal point is selected and boundary points from an arbitrary user-supplied data collection are chosen. The nominal point is the highest estimated local density and the boundary points are those that have an estimated local density greater than a particular threshold value, as shown in Figure 2 .
The IC-CAP software provides turnkey extraction modules for a wide range of popular device models. In addition, custom model equations and extraction techniques can be developed. The built-in extraction modules are easy to use and include the appropriate measurement setups, plot definitions, mathematical transformations, optimization routines and automation macros to help the user begin modeling quickly. IC-CAP covers a wide range of model types including MOS, BJT, MESFET/high electron mobility transistor (HEMT) and thin-film transistor (TFT). There are several models for each device type from which to choose. Figures 3, 4 and 5 show typical data plots obtained from the various models. The program supports the latest industry-standard models with highly effective and automated extraction algorithms, including several high frequency models with unique techniques for applications where the standard models will not fit. These unique models are also used in the company’s high frequency nonlinear circuit simulation tools.
IC-CAP MOS models include the Philips Electronics MOS model 9 with quick extraction and junction capacitance; the Berkeley BSIM3v3.1 MOS model; the Berkeley BSIM1, 2 MOS models; the UCB MOS level 2, 3 models; the HP EEsof high frequency MOS level 3 model; the HP EEMOS1 MOSFET model; and the HP Root MOSFET model. IC-CAP BJT models include the BCTM VBIC BJT model, the Philips MEXTRAM BJT model, the Gummel-Poon BJT model, the HP EEsof high frequency Gummel-Poon BJT model and the EEBJT2 BJT model. IC-CAP MESFET models are the Curtice, Statz MESFET models; the HP EEFET3/EEHEMT1; and the HP Root MESFET/HEMT models. The a-Si TFT and p-Si TFT are the IC-CAP TFT models.
The User Interface
The new IC-CAP user interface makes it easy to open or create models, set up hardware drivers for measurement, extract parameters, and simulate and optimize model parameters. New icons and menu bars make IC-CAP version 5.0 look and feel like many of the latest Windows-style software products. New dialog boxes have been created for browse capabilities and editing. In addition, the program has a new status window for all IC-CAP messages, including log file capabilities. For all operations, the number of windows is minimal; a simple click on a tab, icon or menu button is all that is required.
Stimulus/response device measurements are configured by setting DC voltages and currents, RF frequencies and other user-defined parameters. Multiple measurement setups can be defined, including fixed- or swept-stimulus values in linear, log, exponential or list modes.
IC-CAP’s built-in instrument drivers make automated testing easy. All of the necessary measurements are defined and implemented from the software, allowing remote testing. Drivers for a wide variety of test instruments are included. All data for a collection of device geometries and temperatures can be stored in a single ASCII-formatted file. Subsets of this file can be imported into a given setup, providing for separate measurements and extractions and allowing multiple models for only one set of measured data.
Model parameters are extracted by applying mathematical transforms to the measured data. Simple extractions can be created by writing an equation directly in the IC-CAP transform window. A built-in parameter extraction language (PEL), similar to BASIC, can be used to create more intricate transforms and develop new extraction routines that can be tested immediately on actual data.
The function library provides convenient access to routines for trigonometric, complex, statistical, two-port and matrix computations. For example, the two-port function can be used to convert between two-port data types such as S- and Y-parameters. The function library also can be augmented with user-defined functions. The PEL, function library and user-defined functions provide the power to develop exactly the device and circuit models required.
After simulated data are compared to measured data, model parameters can be adjusted (optimized) in an iterative manner so that the simulated data more closely match the measured data. The IC-CAP software also includes several optimization algorithms.
Tasks within an extraction, or even entire extraction routines, can be automated using macros. This capability allows extraction techniques developed in the laboratory to be automated and leveraged to the production floor where minimal user interaction and high productivity are desired.
After a preliminary list of model parameters is extracted, a simulation can be performed. Simulation results can be plotted together with measured data to determine how well device performance predicted by the model compares to measured data. IC-CAP includes three SPICE simulators and provides direct links to several external simulators.
The software provides a color display in linear, log, real/imaginary, polar, Smith chart and tabular formats for viewing measured data. Statistical distribution of the data also can be displayed. Simulated and transformed data such as derivatives can be shown on the same plot.
IC-CAP is not limited to modeling discrete devices. The modeling of circuits, such as logic gates and operational amplifiers, is a natural extension of single-device modeling. The program’s flexible structure allows easy measurement of circuit characteristics, extraction and optimization of model parameters, and simulation of the circuit’s performance.
IC-CAP contains five components: the user environment, an analysis module, instrument drivers, a statistical package and extraction modules. It is available as a complete modeling suite or as individual modules to provide exactly the capability needed. The software is supported on HP 9000 series 700 workstations and Sun SPARC workstations. The operating systems are HP-UX 9.0x, HP-UX 10.2, SunOS V.4.1.X or Solaris 2.5. At least 64 MB of random access memory and 150 MB of disk space are recommended. Typically, an HP-IB card is required for instrumentation control.
Each IC-CAP module is available in two forms: a node-locked version that allows the software to execute on a single workstation only, and a network-licensed version for execution on multiple workstations, which allows various workgroups to share the software. More information can be obtained from the company’s Web site at http://www.hp.com/go/hpeesof.
Hewlett-Packard Co., HP EEsof Division, Westlake Village, CA (800) 477-6111, ext. 4909 or 9731.