Semiconductor industry is booming, which is driven by the need for chips, memory and overall AI acceleration. At the same time, scientists have made great strides in progressing semiconductor innovation as well, which drive novel architectures and better computing power. Yet, what remained a relatively unexplored area in semiconductors has been defect inspection.
Despite having in-built inspections for specific stages of production, most factories were not able to effectively capture the sampling rate 100% for excursion, minor drift and behavioral change. By the time a defect is identified, the production chain has suffered a huge yield loss. For instance, in the case of Sigenic’s recent engagement with a Taiwan case, it has been observed that there are a total of 6000 wafers due to robot cylinder fitting leak caused position drift, which resulted in US 17 million yield loss.
To resolve this industry-wide problem, Sigenic provides enhanced real-time machine condition monitoring solutions to factories. The in-house developed proprietary software has the capability to run high resolution multi-dimensional in-depth analysis of real-time data with its AI/ML driver, thereby enabling engineers to optimize their machine condition monitoring requirements. More precisely, it effectively achieves: Prevents excursions by prediction, Minimizes wafer scraps and improves production reliability, Real-time monitoring of machine behavior down to sub-millisecond level, Unleashes the full potential of existing factory host analytical performance and Cost-savings by avoiding expensive server upgrades
Before the invention of Sigenic, most factories either dealt with the problem reactively by setting an acceptable loss, or relied on conventional approaches toward machine performance monitoring, which often centered around increasing the data sampling rate, typically ranging from 1 Hz to 100 Hz. However, such an approach is costly and can result in higher network traffic, leading to latency issues, and may not effectively address some persistent problems.
Sigenic’s solution, on the other hand, could handle vast amounts of real-time data and provide highly accurate analyzed data to the host without requiring the host to increase the sampling rate. This combination reduces capital expenditure, provides more meaningful data and enhances real-time machine quality control.
Diving further to the architecture, Sigenic's software bridges the gap between machines and control systems as an intermediary layer, processing sensor data directly at the edge for unparalleled analysis. This enables efficient handling of massive data and sends only essential insights to the host, leading to improved machine performance and solving previously unmanageable problems.
Overall, Sigenic's solutions have been successfully adopted by numerous chip manufacturers and has demonstrated its potential to enhance client’s FDC control systems, including E3, Bistel, Camline Space, etc. The use case has now expanded to the automotive and transportation/logistic and warehousing industry. This superior capability provides users with a significant number of additional options to improve overall machine performance by preventing excursions as well as predict failure that has the potential to lead to production losses. Some have recognized the Sigenic system as their Best-Known Method (BKM) for machine condition monitoring.
