Closer, more coordinated interaction between IC design and manufacturing teams is the key to improving product yield. Although many IC manufacturers are trying to institute true design for manufacturability (DFM), more effective collaboration with their design counterparts is necessary to overcome numerous issues that impede success.
Historically, three major issues have prevented a successful strategy for DFM implementation. First, because design and manufacturing engineers do not typically communicate or share the same working language, there is no cross-domain knowledge of their respective processes and problems. Second, DFM solutions in both design and manufacturing have been point tool focused, offering only a one-directional data flow targeting a single domain. They are not architected to pass and share information outside their unique domain. Finally and most importantly, the infrastructure to support DFM has been missing.
It’s clear a closer collaboration between manufacturing and design engineers is important and will offer long-term benefits to both groups. Fortunately, several DFM applications offer a number of practical solutions to improve yield and include an infrastructure that can support the seamless exchange of data between design and manufacturing.
Currently, the lack of communications between design and manufacturing is a typical scenario. This is a situation rife with miscommunication, misunderstanding, and missed opportunities.
In the manufacturing environment, both wafer throughput and yield are most important. In fact, this is how fabs are fundamentally measured. Maintaining high yields in manufacturing requires in-depth physical process knowledge and device knowledge. This produces a community that is driven by a unique “manufacturing and process” vocabulary and develops a unique culture that values hard-driving manufacturing at all costs. Conversely, the EDA community does not come from this background, has its own unique vocabulary and design culture, and uses statistical models to measure the success of a device being designed.
The trend today is to move away from point tools to more fully-integrated solutions to support the entire infrastructure. Although there are many point tools within the overall design-to-manufacturing area––including automated test equipment (ATE), there are areas where solutions for joint collaboration can and should be realized.
Synergy must exist between various vendors that provide solutions outlining where a device may be defective and sources contributing to the defects, either random or systematic. An example is the need to map both logical nets generated by an EDA diagnostic engine to physical nets used and navigated to within manufacturing. Once correct mapping is performed, physical failure analysis and validation can be performed. From there, device debug can be automated, the set of failed nets validated and X/Y coordinates returned to the fault diagnostic engine to optimize fault models. Various other applications can be realized once logical-to-physical mapping is performed.
Recent moves toward an open systems architecture help address some of the needs of the semiconductor manufacturer, yield management software provider, EDA vendor and the ATE manufacturer. Open systems architecture is defined as a system’s ability to provide an environment for tools to interface and pass/share diagnostic data within an advanced yield management system to enhance and ramp yield. It ensures modularity and interoperability across dissimilar operating systems, EDA solutions, and ATE systems. The ability to incorporate new analysis tools at an application layer transparent to the user is also essential.
An additional benefit of an open systems architecture is the minimization of the time required to develop additional analysis applications. If possible, an open systems architecture should be based on industry standards such as Open Access 2.2, STIL, and GDSII, because standards set guidelines for interconnectivity in a modular system. The relationship between the architecture and the standards should be interdependent, not a rigid bond. Today the need for interconnectivity also applies to the internal implementation of software application packages.
By implementing an open architecture, design and manufacturing teams are well on their way to improving product yield.