October 18, 2021
Editor’s Note: This content is contributed by Mario LaMarche, Sr Prin Product Marketing Manager at Mercury Systems
Yesterday, as I was sitting in traffic on my way home from work and looking at the long lines of cars surrounding me, I had an interesting thought. Every car I could see was loaded with semiconductor products—in engine sensors, infotainment systems, smartphones, laptops, and even watches. Compared to a decade or two ago, the proliferation of modern computing power has changed nearly every aspect of our lives.
As I was slowly driving along, I passed by some of the industry leaders—giant tech campuses with teams of hundreds of engineers all working to invent the next generation of processing technology. And with such widespread adoption and high-volume sales, the business case for investing in the R&D is obvious. Just then, my thoughts were interrupted by the sound of a low-flying aircraft. Another clear example of an application requiring the latest in semiconductor processing technology. However, with so many fewer airplanes than smartphones, I had to wonder how the development costs are amortized over the lower volume.
The same challenges hold true for not just commercial airplanes, but for a wide range of high-performance, low-volume platforms. This list includes everything from commercial satellites to fighter aircraft to weather radar. These systems must process large volumes of sensor data in challenging environments directly adjacent to the sensor—typically referred to as edge processing. However, unlike commercial applications where the design costs can be spread over a quantity of millions, the sales volume for these applications is low.
This brings us to the challenge and point of this blog post. How can mission-critical edge processing systems access the same level of Silicon Valley technology that we’re all familiar with? At Mercury Systems, we address this challenge by combining close collaborations with semiconductor manufacturers with our experience operating in these highly regulated markets.
A perfect example is found in our relationship with Xilinx, a leading semiconductor company and inventor of the FPGA. Mercury and Xilinx have a long history of working closely together to deliver some of the most advanced, low-latency edge processing technology to some of the harshest environments imaginable.
The Versal™ adaptive computing acceleration platform (ACAP) is a processing device from Xilinx that incorporates the low-latency and efficiency of an FPGA, the programable processing of a CPU, and the AI processing of a GPU in a single product. This high-performance processing solution is ideal for a wide range of applications, which makes it worth the investment required for its development.
Thinking again about my traffic-inspired observation, automotive edge processing, with low-latency processing and AI-capable vector processing, makes the Versal ACAP exactly the type of device required for a vision processing system in a self-driving car. Additionally, one can easily imagine a multitude of other use cases that benefit from this technology—automated manufacturing, healthcare robotics, industrial sensor fusion, etc.
And then there’s aerospace and defense, another example of a market that needs high-performance edge computing. Whether processing adversarial radar data in the nose of an aircraft, using cognitive electronic warfare (EW) to mitigate radar threats in real-time, or performing electro-optical infrared (EOIR) pre-processing in an unmanned aerial vehicles (UAV), an increasing number of applications are dependent on AI processing at the sensor edge.
Taking the power of Xilinx’s Versal ACAP and delivering it to the tactical edge is no easy matter. Compact systems present thermal management challenges. Long program life cycles require high levels of interoperability. Mission-critical applications demand the highest levels of reliability. Additionally, selling processing components to government end-users requires careful attention to complex requirements around trusted manufacturing, data security, and supply chain traceability.
Overcoming these challenges requires a company with experience operating in this market. An engineering team needs to know how to predict the power dissipation and use advance cooling techniques capable of operating in an aircraft. To maximize interoperability, a company must understand how to build modules aligned with standards such as OpenVPX and SOSA. They must also understand the reliability ramifications of each design choice. Lastly, the entire manufacturing organization must be certified, for example through AS9100, as having the quality for mission-critical programs.
At Mercury Systems, we take these responsibilities seriously. We leverage our decades of experience collaborating with leading commercial semiconductor companies to take their latest processing solutions and utilize them in rugged edge computing environments. Through this approach, we’ve found a way to take advantage of the high-volume commercial and industrial processing markets and leverage those technologies for use in more specialized, lower-volume applications.
To learn about the latest technologies from Xilinx and Mercury, check out the webinar Bringing AI to the tactical edge with Xilinx® Versal™ ACAP.