如何做到分析效率提升30%？2大半導體製程優化關鍵

[#1]

When talking about the semiconductor industry, people often think of the massive data and fast-paced product optimization process. Such an industrial ecosystem has a high degree of adhesion to JMP, and its fast and accurate interactive statistical analysis features are widely loved by semiconductor customers. In the working environment of the semiconductor industry, it is often necessary to identify the source of the problem and quickly introduce improvement measures for the adverse conditions faced. The most difficult stage is how to quickly find the most relevant influencing factors from hundreds or thousands of parameters. In the past, engineers could only rely on their past experience to slowly try and error out suspected parameters to find the key factors. This approach is not only time-consuming, but may also overlook important indicators. How to quickly find directions for process optimization is definitely a major challenge faced by every engineering team.

Case 1: How to find the key parameters in test data with more than 600 parameters

Taking the EWS test data of semiconductors as an example, there are more than 600 parameter indicators. To compare and rank their impact on special Bins, it may take a lot of time if the traditional method is used. Using JMP's Predictor screening and Response screening, you can quickly compare the impact of multiple parameters on the results, complete the task in milliseconds, and quickly find the parameters with the greatest impact on Bin 10. Then, using the interactive graphic analysis tool Graph builder, you can quickly draw a scatter plot of key parameters and Bin 10, and color them according to different Lots, quickly clarifying that most of the problematic data comes from Lots 16 to 29, and analyzing the source of the problem only takes a moment.

Figure 1: Using JMP’s Predictor screening and Response screening, you can quickly compare the impact of multiple parameters on your results.

Figure 2: Using JMP’s scatter plot to quickly understand the source of the problem

Case 2: How to improve the process and quickly find the best factor combination

Another common problem in semiconductors is how to improve the process and find the best formula combination. Most of the previous statistical analysis software cannot provide suggestions for effective experimental combinations, or the designed experiments have many experimental limitations, such as the types of groups and factor settings. Such problems make data analysis full of inaccuracies and cause additional costs. JMP's custom experiment platform provides optimized experimental design, helping users to build accurate models using the most cost-effective experimental combinations.

Taking the wire bond case of packaging and testing as an example, if we want to obtain the maximum value of ball shear and IMC%, we can use JMP's custom experiment platform to design experiments based on the number of groups we can afford and the appropriate types of factors. With the data collected through such experimental design combinations, users can follow the analysis path guided by JMP to obtain a good narrative model. What's more, using JMP's Profiler platform, we can quickly find the optimal formula combination that satisfies multiple conditions on the multiple models obtained. This function can greatly help users in developing new processes or improving existing processes.

Figure 3: Using JMP to obtain a regression model and rank important parameters

Figure 4: Finding the optimal parameter combination in the model

In summary, unlike other traditional statistical analysis software, JMP can bring customers an excellent user experience and provide faster, more comprehensive and more practical results. The advantages of JMP are obvious. Here are some of the advantages that are often mentioned:

• Interactive dynamic analysis chart
• Link data charts to easily find correlations
• Easy-to-use statistical analysis platform
• A variety of predictive modeling tools to find key factors and optimize parameter combinations
• Quick and easy report scripting

JMP has a 97% market share in the semiconductor industry in Taiwan and has strong credibility. It has become a benchmark language for discussing issues in the industry. Many companies use JMP to output analysis results to avoid lengthy result verification time. For users who need statistical analysis at work, the easy-to-use JMP statistical tool can help users significantly improve their work efficiency by more than 30%.

If you want to learn about the complete semiconductor analysis mentioned above or listen to how Taiwan’s senior semiconductor analysis consultants view the performance improvement of big data analysis, you are welcome to sign up for the online seminar [ 2 major process case analysis: statistical thinking and analysis practice essential for semiconductor engineers ] hosted by JMP Taiwan on May 20, 2025 (Tuesday) at 2 pm. In this seminar, senior semiconductor analysis consultants will demonstrate how to use interactive analysis, modeling and visualization methods to help semiconductor engineering teams make more accurate process and yield decisions in a shorter time. Semiconductor industry-related process, testing, quality assurance, R&D engineers and mid-level managers are welcome to register. (This seminar will be taught in Traditional Chinese)

Missed this webinar? Below are some of JMP’s online resources for the semiconductor industry. Please click to read or contact our team directly:

• JMP in the semiconductor industry
• Learn more about how JMP is applied in the semiconductor industry
• How to quickly identify wafer defects? Data analysis for high-tech semiconductors
• Data-driven, improving semiconductor high-tech process yield and product quality
• The key to improving product yield: 3 tips for production yield calculation and JMP specification setting
• Learn how to create a wafer map using JMP in 7 minutes