LED Automation Software: Understanding the Process Through Data is an Important First Step
Previously, I wrote about how LED manufacturers were striving to shorten the time between initiating the LED manufacturing process and measuring their performance in an effort to improve the product yields and possibly even boost LED performance.
Certainly one of the keys to making this possible is having rapid access to manufacturing data. There are many ways to gather data for analysis by the process engineers. It can come from the process tools—for LED, the EPI and Litho steps are quite critical to the performance of the final device. Data can also come from the measurement or metrology tools, which are critical for providing the correlation between process parameters and device performance. And manufacturing process routing or tracking information provides the context backbone that makes the rest of the data pertinent.
The true added value in having all of this data, comes from the ability to analyze it as a whole. It is a data mining activity in which multiple data sources need to be analyzed together, as shown in the figure above. Only through analysis and understanding can we turn raw data into information that is actionable.
In a recent presentation, Professor Kei May Lau of the Photonics Technology Center at Hong Kong University of Science and Technology, noted that when it comes to process improvement on MOCVD tools, the manufacturer must be very experienced in the tool and the process characteristics. “But a ’good grower’ must also possess the traits of a good detective,” she said. This struck me as a very elegant way to say that intuition is currently a big part of process success. Without a mechanism for capturing all of the process data and analyzing it, the process engineers will continue to rely chiefly on intuition. This likely means that tools and processes will be independently adjusted based on the instincts and experience of the individual engineers. The result may be some local optimization of certain tools, but the bigger productivity payoff comes from harnessing process information that validates the collective improvements of the local tool owners.
Going forward, for the LED manufacturer, being an effective detective requires more than just experience and intuition. It requires four key components:
1) Real time manufacturing system that traces lots, locations, times, etc.
2) Real time process data coming off the key tools that is captured in a database.
3) Real time measurement data from all of the metrology tools (in-situ, probe, defect tracking, etc.)
4) A solution that can pull the data together and easily drive real time analysis of the data.