How Semiconductor Manufacturing Benefits from Smart Fabs
Semiconductor manufacturing is a complex industry and small improvements anywhere in the process can make a big difference to the bottom line. This article discusses the benefits of continually optimizing manufacturing for the semiconductor industry by introducing smart manufacturing solutions.
Increasingly complex market demands are driving semiconductor fabs—whether integrated device manufacturers (like Intel or Samsung) or the fabless/foundry model—to find ways to improve manufacturing flexibility, time to market, and profitability. The pace of new product introduction is increasing, as is the complexity of developing, manufacturing, and deploying products to market. In order to meet these demands, companies are connecting product design and test with manufacturing to enable optimizations of both the product and the process in a closed-loop feedback cycle.
Digitalization makes it possible to meet the business demands of reducing cycle time, lowering costs and, improving yield while boosting creativity and innovation. Digitalization is helping to solve the limitations of scaling and coordinating across organizations and value chains. Likewise, the boundaries between engineering domains and vertical specializations are blurring as companies leverage the power of big data and analytics horizontally across previously siloed domains becomes better understood.
But what does it mean for a fab to become fully digitized and “smart?” How is smart manufacturing different from what most semiconductor companies do today? Each fab will have its own needs, with different entry points to digital transformation driven by well-quantified business goals. Smart manufacturing is comprehensively digitalized, allowing greater and more reliable information flows not just within the fab, but between all the players from design to distribution. Smart manufacturing in semiconductors is more than just connected devices in the fab; it is about the optimization of processes from concept and design through manufacturing and service.
Aspects of a Smart Fab
Realizing the cycle time, yield, and cost-benefit of the smart fab means digitally connecting the entire manufacturing process lifecycle. This spans well beyond the most modern semiconductor manufacturing execution systems (MES) and their track and trace functions. While improvements to track and trace are largely in place, driven by customers making ICs for regulated devices (in automotive, medical, aerospace and defense, security, and biometrics), technologies that connect design to manufacturing have only more recently emerged as important resources for fabs to leverage both history and advanced predictive analytics to improve fab performance.
One way to conceptualize a smart fab is as a unified collaboration platform that includes digital twins, not just static models, of the product, the production, and the performance.
Figure 1. Digital twins can be of product, production, or performance. These digital twins feed each other for insights and continuous improvement. Siemens Xcelerator portfolio of products covers the design, realization, and optimization of digital twins.
This concept of a digital twin is heavily used in industries such as aerospace, defense, and automotive to describe the as-designed product. Building on the concept of the digital twin, manufacturers can tie product information, decisions, and history together in a structured, integrated way that captures product innovation and knowledge throughout the product lifecycle. This is referred to as the digital thread and it enables traceability across the innovation lifecycle. It also makes it easier to reuse the design during downstream processes, which not only saves time but also improves quality as everyone is working off a single source of truth. Initially, smart manufacturing would have focused more on the digital threads of production and performance, but the use of product data is now considered an important element for process improvement as well.
A smart semiconductor fab has scheduling integrated with MES to create an environment for constantly updated, resource-constrained views of what to do next at every workstation. Today's advanced MES can model complex process flows, but they need to also manage specifications, recipes, masks, and tools in re-usable workflows.
All the components of a smart fab need to be customizable to adjust to the specific demands of any given organization. Low-code and no-code applications development platforms are a critical component. These platforms open up the ability to drive business transformation through custom software applications.
Today, designers at any experience level can create applications to fit their needs. This fosters a network of internal citizen developers, deriving insight and speeding digitalization by connecting new and legacy systems, automating processes, and facilitating data analytics that provides actionable intelligence.
Closing the Loop on Design, Build, and Test
There are opportunities to use design data to improve yield and other margin factors as part of the smart manufacturing realization. Fabs already connect product design with manufacturing to enable product optimizations through design-for-manufacturing (DFM) techniques. Data from manufacturing test results also provide valuable insights needed to reduce costs and improve quality and yield.
The connections between the design and manufacturing are accelerated through new generations of end-to-end software automation. Design automation software is perhaps the least obvious component of a smart fab, but it is becoming increasingly more critical in order to create more efficient IC manufacturing, controlling costs, and creating room for continued improvements in electronics capabilities. In one example, an IC design goes through physical verification before masks are generated. The data from that process is analyzed using machine learning algorithms and is fed back to the foundry to optimize the design flow.
Another example of the connection between design and manufacturing, one that also helps build an open, modern ecosystem, comes from IC test analysis. Semiconductor technology and process control require reproducible atomic-level performance across thousands of different product designs. The ability to consistently produce high yield on all products is daunting. In the past, yield was governed by process capability, the use of test chips for yield learning, and process monitoring in the fab. Today, the data from failed IC tests are analyzed with advanced machine learning algorithms to find yield limiters that would have remained hidden in the past.
Figure 2. Part of smart manufacturing for semiconductors involves using design data to improve the process.
The notion of capturing yield limiters can also be extended upstream into the actual design for manufacturing (DFM) flow, especially during the lithography process, which is a critical step in manufacturing. By the time a serious lithography-related problem is identified at the fab, it is too late in the design process to make simple layout changes. To avoid or reduce design delays, designers can use lithography simulation to detect weak points in a layout and analyze the effect of lithography on the design’s electrical performance.
Engineers can then implement any required adjustments early in the design flow. To do this, there must be a mechanism to enable designers to incorporate lithography process knowledge into their existing design flow. There is, in fact, in the form of lithography friendly design (LFD) kits. LFD kits let the IC designer gauge the response of the layout to a given process, and then correct those configurations that are most likely to fail during manufacturing. Machine learning algorithms are leveraged here to predict yield limiters with high detection accuracy and coverage while reducing overall runtime.
Collecting, analyzing, and using this data is a part of creating a comprehensive digital twin that is at the center of smart manufacturing. It also requires a tight ecosystem of partnerships based on trust and verified data security.
Design houses need ways to understand their design-specific manufacturing/yield challenges and the foundry needs design data to effectively address the product/process interactions. Manufacturing test data combined with design data can provide a powerful tool to solve some of the most challenging atomic-level issues. Without this information sharing, the ecosystem will not truly evolve into the smart manufacturing vision needed to compete in the global semiconductor market.
Where Smart Semiconductor Manufacturing is Now
Semiconductor fabs are ahead of the curve on implementing smart manufacturing, but the capabilities are myriad and quickly evolving. Facing competitive pressure, semiconductor fabs need catalysts to enable and accelerate their digital transformations. Barriers remain, but the ecosystem is actively developing solutions for all the components of a smart fab, from a complete digital twin that leverages not just IoT in the factory, but valuable design data, to a flexible platform and a unified ecosystem that brings buyers, developers, designers, production houses, sub-contractors, suppliers and more together. Working with dedicated and experienced industry partners will best position a smart fab to realize profit and innovation that would otherwise be left on the table.
Fabs should look to a supplier with ready-to-use solutions for both product development and manufacturing to accelerate the realization of a true smart fab. For example, Siemens’ Xcelerator portfolio integrates software, services, and an application development platform to help manufacturers fully digitize their operations to realize the smart fab goal. Taking the next step in a fully smart fab will take some institutional will, investment, and partnerships with ecosystem partners that have a rich digital software portfolio for both product development and manufacturing.
Industry Articles are a form of content that allows industry partners to share useful news, messages, and technology with All About Circuits readers in a way editorial content is not well suited to. All Industry Articles are subject to strict editorial guidelines with the intention of offering readers useful news, technical expertise, or stories. The viewpoints and opinions expressed in Industry Articles are those of the partner and not necessarily those of All About Circuits or its writers.