771 Views

Process Control in High Volume Semiconductor Manufacturing

February 14, 2017, anysilicon

Semiconductor manufacturing process starts with wafers having several dice. These dice are then probed by testers to verify quality of wafer that will be later on assembled and packaged into an integrated circuit commonly abbreviated as IC. These ICs are then placed inside large electrical systems and devices to be shipped to the customers for use. As the chip-making process concludes with electrical testing, the testing process is done with great due diligence to avoid any rejects or discards that can hamper yield and create operational inefficiencies. Further, the competition in the semiconductor manufacturing makes it highly disadvantageous for the business to lose the packaged wafer (IC) at the final test stage in the times when restriction of Fab’s wafer capacity, dwindling feature size and increase in the diameters of wafer is dominant. In order to overcome this issue and to stay ahead of increasing competition in the global marketplace, semiconductor manufacturing processes needs to be steady, efficient, and yielding high quality products capable of performing according to its specification after being shipped to the customer.

 

The man with a carrier of the silicone wafers

 

This means that all the stakeholders be it assembly floor operators, product or test engineers, or the senior management must constantly strive for operational improvement and reduce process variability. The reduction in variability of process can be achieved to a larger extent by devising strict control mechanism that is able to detect outliers and generate alarms in case the control limits are exceeded and process variation gets out of hand.

 

The companies are moving towards high volume production with the goal of achieving minimum defects. With this scale of manufacturing products it becomes impossible to test each unit, without incurring a very high cost and with the price of chips going down, it becomes even more critical to reduce the costs.  Thus, the ability to use statistical techniques to detect outliers and process anomalies become ever important. Semiconductor manufacturing is now relying all the time more on using statistical based methods to characterize, monitor and optimize processes. This involves collecting and analyzing large data sets and deducing trends and gaining insights about outliers that can hamper the smooth flow of manufacturing process.

 

This ability to sample a small number of units out of 1000s of products representative of the whole production makes statistical techniques all the more very valuable. Statistical Process Control for semiconductor manufacturing is one such widely used method in the industry for outlier detection and to monitor process variability. Statistical Process Control is a two-step process that seeks to distinct between systematic or normal variations from special or unusual variations. These techniques are widely applied to production yields as well as to wafer parametric test results at various process nodes in the high volume semiconductor manufacturing. The first step in Statistical Process Control is to calculate a Process Capability Index (Cpk) by measuring normal variations in the data collected at the various stages of the semiconductor manufacturing. The second step involves monitoring of production material continuously using control charts that calculates that there is no violation of rules. In case of violations, alarms are generated that give early warnings of issues, saving the company from packaging products having potential to not perform well.

 

Statistical Process Control in semiconductor manufacturing is not a new concept but it got into the limelight when computers became mainstream and manufacturers started using enterprise semiconductor SPC software. Since, then Statistical Process Control for semiconductor industry has become a new norm to successfully monitor the processes on the assembly floor to improve quality and to reduce scraps.  This timely control of process variation leads to enhance operational efficiency, reduced cycle times and increased financial savings as the variations are detected early and corrective measures are taken in due time.

 

 

Author Bio:

Irteza Ubaid is working as a senior strategy executive at yieldWerx, which is a semiconductor yield management software solution provider company that helps IC manufacturers carry out huge data extraction, its transformation and  loading lot genealogy and product data from MES and ATE systems. http://yieldwerx.com/statistical-process-control/