Developing incremental capacity during seasonal demand for technologies beyond 90 NM in semiconductor fabrication industry
Mohd Azizi, Chik
MetadataShow full item record
This research is to develop a new approach to accelerate capacity increment from unexpected additional demand at semiconductor fabrication by 10%. The issue is from unpredictable demand at short notice to increase capacity by 10% in future period or as high as 35%historically. Current improvements practice for capacity increment such as increase in equipment throughput, availability and manufacturing efficiency will take up to six months to be effective are not able to solve this issues. This is due to the processing steps that range from 300 to 1000 with cycle time 30 to 90 days. Plus, the new improvements from equipment usually need data to verify electronically to understand the impact from the product quality and reliability. The traditional improvements from equipment are still relevant for medium or long-term solution. Another approach for capacity increment is through investing in new or refurbished equipment. Investing new equipment required Capital expenditure (CAPEX) range from USD 20 to 30 million for additional capacity of 1,000 wafers per month. This huge investment will takes up from 10 up 14 months for equipment ready to support production and only afford by big scale companies. Newly popular strategy is to do outsourcing to another fab, but again, initial stages needs six to nine months for product to qualify for production, however when the outsourcing fab is qualified and fully loaded, it is not guaranteed to absorb additional loading. Thus, make this research topic valuable and significant for the situation current economic trends. The new approach is based on the philosophy of borrowing the future capacity into current time when immediate capacity increment is needed. As the capacity usually determine by total number of processing steps that need to process at the constraint equipment, this research concluded that the higher re-entrance steps caused lower the capacity of a fab at exponential trend. Through similar approaches taken from specification from SEMI and others literature the baseline capacity successfully developed at 22,511 wafers per month. Therefore in this case to increase the capacity at a time, it needs lowered reentrance steps, by recalculating the capacity not based plan order demand but by using preferred WIP to reduce the impact from exponential impact. In order to this, the cycle time and the equipment output needs to be integrated in production systems for real time validation and the information then needs to store into databases and integrated with simulation model for the system to be able identify immediately for the WIP that able to help to increase the capacity with high accuracy planning. The data collection and the simulation models used commercial software from Applied Material which is Advance Productivity Family and AutoshedAP, that proven in this research at 97% accuracy for 30 days output WIP forecasting. The simulation results in the models successfully demonstrated accelerate capacity increment by 21%, and actual implementation of capacity increment by 33% and the research meets its original objective.