Automation and Production Tools Deliver New Features for TCAD Calibration

The Automation and Production Tools of the VWF provide users with an experimental design system to create behavioral models (RSMs) of their processes. In the four years since its initial release, the VWF has grown to be the accepted model for TCAD tool automation. Our new release delivers many new features to support the integration of TCAD products and analysis techniques.

Enhanced Fitting Techniques For RSM Generation

  • Third order polynomials with cross terms added
  • Selection of transforms are supported for the experimental variables
  • Automatic fitting process to determine the best fit to the worksheet data
  • Denormalize RSM coeffs (without transforms only)
  • Warning in random designs if too few points for RSM generation

The majority of experimental designs used in literature can be fitted using quadratic functions. However we have found that for variables in TCAD often the standard experimental designs are often not enough. VWF already supports customized experimental designs. Now new features for the analysis of complex experiments include the use of transforms of the experimental variables together with third order polynomials. This is sufficient to fit many useful TCAD results such as threshold voltage versus Leff over a range of processing parameters. See Figure 2 for an example of a complex function.

Easier ATLAS Experimentation

  • An internal simulator has been developed for SIMMGR that allows users to define split variables for ATLAS in SET statements. It avoids the large time overhead of starting and stopping the simulator just to define the split points

Experiments using device simulation only were possible in earlier versions of VWF. However our latest release includes features to speedup and simplify experimental setup. An internal simulator has been developed to process the SET statements required in ATLAS experiments. Examples of device-only simulations are experiments on material and model parameters in Blaze and experiments on different lifetime or trap properties for power devices in Pisces. Figure 3 shows an example of results from a Blaze experiment.

Figure 1. GUI for fitting RSMs to simulated results

Support for SmartSpice Experimentation

  • Ability to define experiments in SmartSpice input files
  • Interface between a library of UTMOST Models and Smartspice to allow models created from process variations in ATHENA to be analyzed in terms of circuit performance

Figure 2. Response surface model fit for Vt versus length for a set of implant damage factors. A curve for any damage value factors can be created using an interactive slider in VWF.

Figure 3. Contour plot of Idsat in a MESFET as a function of saturation velocity and electron relaxation time for the energy balance model.

The new release of VWF will support experimentation using SmartSpice. Users can make experimental variations in smartspice input files in the same manner as with process or device simulators. More importantly SmartSpice can use the UTMOST models generated under VWF for process splits. Using the new version experiments can be run from MaskViews through ATHENA to ATLAS and then UTMOST to generate a library of SPICE models for both NMOS and PMOS transistors. The equivalent NMOS and PMOS models from the same process can then be combined into a single CMOS circuit simulation. This allows users to analyze the effect of process variables on circuit performance. Applications include examining capacitance versus drive current trade offs for process parameters and generating worst case circuit performance.

Improved GUI and System Administration

  • Remove data points from residual plots
  • Predictive error calculation
  • Worksheet paging for large data sets
  • Database pause while backup
  • Job queuing priority by user name

General improvements have been made in this release to the front end user interface for VWF. As VWF databases are used more frequently, backing up the database has become a problem. Many users use their VWF database continually. However it is necessary to stop accesses to the database during a backup to avoid database corruption. This is done using the VWFADMIN tool which can operate interactively or in batch mode. It will stop all jobs running under the VWF before the backup and restart them all afterwards.

Calibration Features Added

  • Measured data import and overlay onto simulation data
  • Automatic matching of response surface model to set of measured data point
  • Nominal case correction function
  • Improved yield synthesis

Calibration is an important issue for TCAD users. The combination of the VWF automation and production tools allows users to have a systematic approach to calibration. This is much more advanced than common look-and-see approach of rerunning the simulators with slight parameter variations. This approach can typically only operate on one or two tuning parameters at a time.

With VWF automation tools, users can define experiments to vary simulator tuning parameters in the same manner as `split lot experimentation' of process parameters. Behavioral models (RSMs) can be built defining the device performance in terms of the simulator tuning parameters. These RSMs can be analyzed in the Production Tools. A new feature of the Production Tools allows measured data to be imported along side the RSM. The CALIBRATION option then allows users to vary the simulator tuning parameters to tune the RSM to the measured data. Figure 5 shows an example of an RSM of oxide thickness as a function of time and temperature. The RSM has been made to match a set of measured data using tuning parameters a variables in an automated calibration scheme.


Figure 4. VWF experiment using UTMOST and SmartSpice.

Figure 5: Overlay of measured data for oxide thickness versus simulated results over a range of time and temperature.