Skip to Main Content
This paper describes the yield-mining solutions in semiconductor manufacturing from the theoretical as well as application points of view. In specific, an innovative methodology is developed to enhance the correlation between CP yield data and electrical test (ET) data so that the engineers can diagnose problems and improve yield more effectively. Also, an interactive yield-mining system is developed to embed engineers' knowledge in the data mining process. To cope with the single-tool problem and the small sample size feature in a yield ramp-up or in a high product-mix, low volume production environment, a project-based multi-product analysis methodology is developed to enhance the diagnosis capability. The proposed solutions have been fine tuned and validated in a local foundry fab.