Military and aerospace systems are generally designed on the basis of modular sub-systems. For example, line replaceable units (LRUs) simplify service and operational support. Interconnecting LRUs rely on standards such as the MIL-STD-1553 bus interface. Implementing these functions in hybrids, modules, ASIC macros or on standard format circuit boards has become the preferred approach and they are, in effect, application specific standard products (ASSPs).
This highlights two important factors. First, there is nothing to be gained by re-inventing the wheel and designers are more usefully employed focusing on the core intellectual property of the system. Secondly, the military and aerospace industry is a small user of semiconductors by today’s standards, and developing solutions at the module or board level is a more realistic proposition than a monolithic IC-level ASSP.
Traditionally, the performance requirements of power supply modules has also aligned well with hybrid module technology where the use of hermetic metal can packaging supports the power density and thermal management needs of high-temperature, high-reliability military applications. With increasing power requirements from large FPGAs and microprocessors, the drive for more efficient power architectures and point-of-load (POL) regulation has given rise to new module solutions.
Applications such as radar have also long-relied on hybrids and modules for RF and microwave solutions. Only in recent years have there been monolithic IC products that begin to address some of these needs.
Product obsolescence is an acute problem for the military. A 30- to 50-year program life is commonplace so military and aerospace equipment suppliers are continually looking for ways to mitigate risk. Hybrids and modules have been one approach to try and isolate the defense industry from the rapid pace of change in the semiconductor industry. Memory modules are a specific area of interest because of the particularly short lifecycle of DRAM and SRAM technology. The concept of a standard form factor and pin-out can be maintained, while the memory dice inside the module can be updated. This is a lot easier to write about than it is to do in practice, in part, due to continual advancements in access times, architectures and supply voltages. On another level, the use of standard format embedded processor cards provide an alternative higher-level approach where space permits. However, the notion of standard form factors is central to many obsolescence management strategies and has certainly been a major influencing factor in the long life of hybrid and module solutions.
Hybrids and modules is also benefit because a full-custom module can be used to conceal valuable intellectual property related to the hardware design, making it more difficult to reverse engineer. Simply looking at part numbers on packages will not be enough to decode the design hardware. Furthermore, some semiconductor dice are also not readily available to everyone. For example, Linear Technology controls the supply of dice very tightly and will not sell product without disclosure of the end customer and application.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.