Automated Testing and Statistical Process Control - From Wafer to Product

In supporting both low-noise and power measurement requirements, many economies of both scale and scope can be utilized. With the increased emphasis today on low-cost products with highly repeatable performance, an integrated approach to on-wafer, packaged and module level characterization is essential. Use of a distributed network of controllers feeding a common database across a network allows the standardization of statistical calculations across products, and rapid response to performance variations in new products. At the wafer level, DC parametric data taken from the HP4142 parametric analyzer is sorted on the fly, and then stored on disk, using a systematic naming system based on the wafer and run numbers. Similarly, noise and s-parameter data taken from the noise figure Summit AutoProber, along with the site location identifiers is also stored on disk using similar naming conventions. Power on-wafer will also soon be available, and pulse power data will be added after that. At the module test level, when the devices are inserted into modules or amplifiers, each product has its data stored in a separate directory in ASCII format, for retrieval later. The module data can then be imported into a relational database such as Paradox, where multiple tables are appropriately keyed in order to link them, and to maintain maximum database integrity by automatically updating the data for retested modules. A wide range of statistics or correlation tests can then be run, comparing the wafer level performance to the module level for particular wafers, based on highly flexible criteria.