Data Mining and Statistics


Once a specific topology or rule is captured, iDRM can scan through the layout of a block or complete design and find all instances that use that topology and take all the relevant measurements of the parameters (variables) that were used in the rule definition. The result is a complete and dynamically sort-able list of all such instances, each with complete information of the relevant measurements, orientations, location, etc.

iDRM can present the results in tables, various graph formats (e.g. 1d histograms, 2d occurrence graphs, Pareto charts, etc.), or export them to a spreadsheet. The user can dynamically switch between different views, filter data and zoom in to areas of interest. An integrated layout viewer provides a one-click hop from each table or graph entry to the layout locations where this specific set of values is found.

Using layout data mining, users can get valuable complete and accurate quantitative information on all layout features, patterns and devices, looks for value distributions or focus on specific values such as unexpected outliers.

Use case examples:

  • For all MOS devices: measure the distances from the gate to the diffusion contacts: this provides a quick overview of SD resistance for all devices.
  • Measure all distances from the gate to the well edge and calculate the WPE impact for each
  • Get LOD (length of diffusion) measurements for all devices (a factor in device stress)

The WPE and LOD statistics can quickly show if all devices are in the expected range and outliers can be immediately identified and further investigated.

  • Classify all PMOS and NMOS devices, by diffusion shape, poly shape and measure all Ws and Ls
  • Get all distances from via cuts to an external metal edge – these affect reliability caused by time-dependent dielectric breakdown (TDDB).
  • Measure distances on critical rules, e.g. all line-end enclosures for each via
  • Check for patterns and rule values which may hinder process migration or a rule change, and check distances if there is sufficient slack to accommodate such changes and many more…

Сheck out the following application-note example of data mining:
Profiling_layout_dependent_effects_on_variability: LOD and WPE