Virtual Wafer Fab

VWF™ enables users to perform advanced analysis tasks like design of experiments (DOE) or optimization, using any of the Silvaco simulators.

Introduction

Silvaco’s Virtual Wafer Fab (VWF) software suite addresses this challenge through its use of statistical design of experiments (DoE), a powerful tool for quickly optimizing processes. The modern alternative to single-variable experimentation,  DOE overcomes the information limitations of successive approximation experiments and quickly provides the kind of understanding and results that are needed. Optimal designs reduce the cost of experimentation by allowing statistical models to be estimated with fewer experimental runs and facilitating reduced time to market.

Silvaco’s Virtual Wafer Fab (VWF) software suite for automated simulations helps users to cut development costs dramatically and reduce silicon learning cycles and the risk of missing a market window.  As an alternative to DOE, the user can utilize an optimization framework consisting of several local and global optimization strategies to achieve a predefined target. This will allow users to automate tasks such as optimization of process parameters or calibration of model parameters. The VWF software is not limited to running only a single simulator but can combine any of the Silvaco simulation tools available allowing the user to optimize full-flow simulations. To support massive parallel operation, grid computing software like Oracle grid-engine (OGE) and LSF can be utilized. There is virtually no limit as to how many simulations can be carried out in parallel.

Features

  • Graphical user interface (GUI)
  • Support for: SQL-92 based database engine, filemode to keep data in XML files, RSM plots of simulated data
  • Import/export features to copy data to/from the database
  • Massive parallel execution of simulations through Oracle Grid Engine and LSF supported
  • Results exportable into Excel or OpenOffice spreadsheet formats

Benefits

  • VWF can represent an entire flow from process simulation to spice circuit performance, or just part of the flow
  • Powerful scripting interface allows users to customize DOEs, run experiments in the background, and compute complex optimization target functions
  • Process characteristics (i.e. oxide thickness), device characteristics (i.e. threshold voltage), and circuit characteristics (i.e. rise time) can be measured from entire flow
  • Each experimental variation can be run on a network of hosts to shorten simulation time
  • Response models can be generated over experimental spread and then input parameters can be rubber banded to see the effect on output responses
  • Full worksheet of input and measured output responses can be exported to SPAYN for additional statistical analysis or to TonyPlot for viewing purposes
  • Advanced security features allowing fine-grained access control

Applications

VWF is shipped with many examples demonstrating how the user can best utilize a cluster of available computing hardware and how you can automate many types of applications, such as:

  • Parameterize important geometrical characteristics of a test structure, such as thicknesses of different layers, spacer width, gate length, etc., and investigate effects on important device characteristics
  • Allow optimization algorithm to vary process parameters, such as gate oxide time and temperature or Vt adjust implant dose and energy, to achieve targeted device characteristic, such as given threshold voltage
  • Start with a gds file of an inverter and simulate the 3D backend process to extract and analyze corresponding capacitances
  • Run fully automated calibrations of simulation model parameters – script allows user to combine several simulation runs into single target, enabling maximum flexibility

 

Access experiments through an explorer or tree view graphic.
Drain Current vs. Gate Voltage shown for 4 different process parameter combinations. Curves correspond to the selected cells 8.2, 8.5, 8.6, and 8.9 (red, green, dark blue, light blue).

 

 

RSM generation done in SPAYN and then visualized in TonyPlot. The ring oscillator frequency is shown as a function of process parameter (gate oxidation time and Vt implant dose)

Technical Specifications

Supports:

  • Standard DOE experiment types
  • Defining own DOE types by means of script
  • DOE importing from spreadsheet (excel)
  • Various local and global optimization strategies
  • Oracle grid engine  (OGE) and LSF cluster computing software