The Growing Need for [ Real-Time] AI and ML in Semiconductor Testing
 Pete Singer
23 hours ago
ERAN ROUSSEAU, VP of Enterprise Software, Emerson Test and Measurement Group
The rapid advancement in technology and the ever-growing demand for more sophisticated electronics have made semiconductors indispensable. As demand for higher performance and more rapid development cycles in semiconductor manufacturing increases, so does the pressure to improve processes. The complexity and intricate architecture of next generation Artificial Intelligence (AI) and High-Performance Computing (HPC) chips, along with the evolution of 3D packaging, has created orders of magnitude more interconnect density, exponentially increasing the number of potential failure points.
Traditional testing methods, while essential, have struggled to keep up with the pace of technological advancements. Manual processes and slow, centralized testing systems are struggling to cope with the sheer volume of data that modern semiconductor devices generate during the manufacturing process. To meet these growing challenges, including the need for traceability across the entire process not only offline but also in-use, semiconductor manufacturers must look to new, innovative testing methodologies—those powered by Artificial Intelligence (AI) and Machine Learning (ML), and the shift to edge-based decision-making. By leveraging historical data and predictive models, AI and ML systems can identify patterns that help to optimize and automate testing flows in real time, thereby making testing more efficient and effective. This shift can improve yields and reduce time-to-market, ensuring that products meet stringent quality standards.
Real-time decision-making at the edge
A fundamental aspect of this innovation is the shift from centralized testing systems to edge-based decision-making. Traditionally, data generated during testing is sent to centralized cloud servers for analysis, which adds latency and reduces the speed at which decisions can be made. Edge computing solves this problem by processing data directly at the test site, close to the equipment where the data is generated. This approach reduces the time spent transmitting large datasets across networks and enables real-time decision-making during testing.
Emerson Test and Measurement and Advantest have formed a strategic partnership to combine Emerson’s domain expertise in data analytics and algorithm development with Advantest Cloud Solutions’ Real-Time Data InfrastructureTM (RTDITM). Through this collaboration, the companies have created an AI-driven test ecosystem capable of real-time, edge-based decision-making that drives faster, more efficient, and more reliable production cycles. Combining the Emerson NI OptimalPlus Global Operations (GO) software with the ACS RTDI platform, we can push computational and analytical capabilities to the edge, allowing for faster and more accurate decisions that ultimately improve yields, decrease time-to-market, and reduce costs.
By processing data on-site, AI and ML systems can immediately adjust the test parameters. For example, if an AI model detects that a particular device is showing signs of failure during testing, it can automatically modify the testing process to address the issue, thereby preventing faulty products from advancing through the production line. This level of real-time decision-making reduces test time, increases throughput, and enhances product quality.
Leveraging historical data for smarter testing decisions
One key advantage of AI and ML in semiconductor testing lies in the ability to use historical data to guide real-time decision-making. By analyzing previous test results, AI systems can predict the outcomes of future tests with greater accuracy, enabling engineers to make more informed decisions during the testing process.
Emerson’s deep understanding of product analytics and algorithm development is complemented by the ability to leverage historical data to feed-forward insights into the testing process. This bi-directional infrastructure allows historical data from past operations to optimize current tests, making the process more efficient and reliable.
For example, suppose a test program encounters a particular issue seen in past tests. In that case, the AI system can proactively adjust the current test parameters to address that issue, ensuring that the test continues smoothly. By using historical data, manufacturers can improve test accuracy, reduce errors, and identify potential problems before they cause production delays.
Click here to read the full article in Semiconductor Digest magazine.
PS
Edge (and Endpoint) bit bonanza bigger than Cloud?
Hmmm....
ASML
Village |
