2D, silicon-based designs with monolithic integration are common in today’s aerospace, defense, and government applications landscape. Some designers in this space are also developing 2.5D and 3D designs, but for specialized purposes, with only a few layers, and at relatively low production volumes. In the next decade or two, 3D designs will likely become the norm in this sector, growing more complex and consisting of many more layers. As we look toward this future, we can anticipate disaggregated designs consisting of bare dies from multiple processes and multiple material types, all connected by dense interconnects. Packaging, along with assembly and test, will be considered in light of how the system will communicate and integrate with the outside world.
From an electronic design automation (EDA) standpoint, a new approach will be required to ensure that higher volume 3DHI designs can function reliably and successfully in aerospace, defense, and government systems. Such designs raise many new questions around considerations such as cross-coupling effects between die tiers and how to connect them; developing trustworthy assemblies; connecting wafers of disparate process nodes, types and even sizes; cooling disparate materials; and so on. It may be prudent to integrate a die test and error correction capability into 3DHI components to proactively address any issues before a system ships and adapt to issues that develop later. Also important are redundancy and resiliency. Due to aging effects typical of any type of silicon device, some of the dies in the system may wear out over time. Developing a design that’s able to adapt should an individual region fail could be one way to protect against an overall system failure.
While the tool flows to address these questions and concerns are still maturing, they do present an upgrade from manual package development. In addition, there are opportunities for research in various areas: multi-die; multi-technology integration and assembly; tools for architecture, design, simulation, and test; security; and thermal and power management, to name a few. The good news is, these considerations involve physics and, ultimately, can be modeled, often with modern digital twin technology. Now is the time to continue to explore and develop tools that enable enough abstraction of the various layers so that design teams can make informed decisions.