A database and data analysis strategy is proposed for multi-designer test chips that involve a wide array of different test
structures aimed at process characterization. The database described in this paper has been custom built for the multi-student
FLCC testchip that has six contributing students and over 15,000 individually probably transistors/test
structures. It has an interface with the Parametric Yield Simulator (PYS) that drives simulation parameters and
automatically populates the database with simulation results of each test structure at the specified process conditions. A
well-designed database forces structure into measurement and design related data, but includes enough flexibility as to
adapt to different types of test structures and experiments. The database is split into four separate sections that store
description of test structures, simulated results, experiment results, and process conditions. All data is centrally located
and web accessible for easy access from any computer with Internet access. Simulation results can be uploaded from the
server running the PYS, experimental results can be uploaded directly from the lab and data can be compared and
queried by all users of the database. Data analysis strategies can be compared and reused as queries and data analysis
results can be shared among users through the website. Queries can be saved, loaded, rated, and reused, so even novice
SQL users can utilize advanced queries. Advanced queries form the basis of a strategy that first identifies good process
monitors based on simulation results and then uses them to extract process conditions from electrical measurements via
an iterative process. This paper describes strategies that can be used to help facilitate collaboration and hence leverage
the benefits of combining multiple sets of test structures from different designers on one chip.
|