The main aim of the IMPROD project is to insert enhanced statistical methods and processing control tools into the current IC manufacturing environment of two semiconductor manufacturing companies (SGS-Thomson and Bosch) in order to substantially increase process and product quality.
The first release of a test and parametric data management package was installed in an integrated circuit (IC) manufacturing line, including the ability to generate automatically Cpk values for process steps. An initial set of tools for engineering analysis has been implemented.
Examination of the process methods employed in microelectronic manufacturing has resulted in a proposed statistical methodology using multivariate analysis to detect relationships and dependencies in the production parameters, and a methodology has been defined which will automatically detected unwanted variation. Modelling of test structures has produced optimum new designs to be used on the semiconductor wafers so that essential process parameters can be extracted from the wafer.
The choice of data management methods, statistical analysis tools and selection of database format have been greatly assisted by the commercial software partner in the project. The tools developed in the project have been designed to be able to draw data from a variety of database systems.
The project will add the following to current process and device engineering capabilities:
- an expert system (using updated statistical methodology) to help improve the whole process of designing, manufacturing and testing ICs
- an information system to facilitate access to data belonging to different process steps (process measurement, wafer testing, final product output) with added special features to collect data by wafer and position on the wafer, validated and without redundancies: this will cover all the manufacturing sites that the product moves between during the maturity cycle (design, engineering production and volume production)
- application of the new tools in a real production scenario (multisite and multistep) in order to test the possibility of employing final quality control by variables rather than by attributes.
The improved system will be used for quality control in the production of automotive products at the two semiconductor manufacturers.
Bench-mark testing for the new system will be performed against the existing factory systems (which involve control by attributes rather than by variables) to demonstrate improved product quality. The goal is to improve this up to the level of equivalent Japanese products.
BT9 5AH Belfast
20041 Agrate Brianza Milano