NASHUA, N.H. - Rugged computers for the warfighter are hitting the market today with broad new capabilities in artificial intelligence (AI), innovative ruggedization and thermal management, distributed architectures for tight packaging for deployment at the edge, and a pursuit of new and emerging open-systems industry standards in efforts to "future-proof" the latest designs.
The innovations don't stop there. Not only are enabling technologies for AI like fast general-purpose graphics processing units (GPGPUs), field-programmable gate arrays (FPGAs), and multi-core central processing units (CPUs) taking the market by storm, but systems designers also are investigating future uses of additive manufacturing and 3D printing to tackle tomorrow's thermal-management and packaging issues.
Artificial intelligence
"As the increased number of sensors and increase data, we are seeing a requirement for a lot more centralized processing and capability to process that data," says Jason Wade, president of ZMicro Inc., in San Diego. "Customers are looking to leverage AI. From a hardware side, SWaP [size, weight, and power consumption] always will be important, but is no longer the sole focus and driver."
While AI may be overtaking SWaP as the central issue in aerospace and defense rugged computing, the reasons revolve around an ever-increasing number of sensors and sensor processing, and a driving need to support military decision-making in real time.
"AI is a common trend, and we see that constantly," Wade continues. "Collection of data is critical, and the ability for high-end GPGPUs to process data is important. ZMicro has seen evolution of GPGPU as an integral component -- sometimes even more important even than the CPU."
Among the most influential purveyors of enabling technologies for AI are Nvidia Corp. in Santa Clara, Calif., for GPGPUs; Intel Corp. of Santa Clara, Calif., for multicore CPUs; and Advanced Micro Devices (AMD) Inc. in Santa Clara, Calif., and Xilinx Inc. of San Jose, Calif., for FPGAs.
The military's demand for sensors and sensor-processing technologies is driving interest in companies that offer AI-related products. "All these sensors connected, and more sensor data transported from sensors to computers, are seeing speeds go from 1 gigabit per second to 10 gigabits per second," says Jim Shaw, chief technical fellow at Crystal Group Inc. in Hiawatha, Iowa "This is a trend that is exploding with sensor connectivity, and situational awareness, and the need to process a tremendous amount of data. The military loves real-time situational awareness."
That's good news for companies like Nvidia and Intel. "It's interesting to watch the Nvidia and Intel market space in GPGPUs and accelerators," Shaw says. "Intel is doing a pretty good job of coming up with an alternative to Nvidia, and we will see more of those products coming up. Nvidia has done a good job of carving out that market space, particularly with the Cuda coding. That drives a tremendous amount of volume in GPUs."
There's no end in sight for the military's appetite for AI. "We are seeing more and more the tremendous focus on AI and autonomy across the board, which is driven by the new capabilities we are seeing, such as the Nvidia GPGPUs, says Aneesh Kothari, president of rugged computer specialist Systel in Sugar Land, Texas. "This density and compute technology available on the market can be deployed in a harsh environment outside the data center."
AI is considered one of the chief enablers of the military's vision for real-time situational awareness. "We are swimming in sensors and drowning in data, so using AI to do the analysis piece is really helpful, and if it can be done in real-time you can take action faster," points out Jason Dechiaro, systems architect at the Curtiss-Wright Corp. Defense Solutions Division in Ashburn, Va.
"The military wants to transmit the results of the analysis rather than the raw data itself," Dechiaro says. "For this AI piece, Nvidia is in that game. At Curtiss-Wright we take advantage of Nvidia GPGPUs for AI processing."
Cooling and thermal management
AI and the powerful processing that enable it does have consequences, however. Powerful processors generate substantial waste heat, which systems designers must find ways to remove without compromising performance.
"We are seeing a lot of the Intel and AMD products power budgets are just exploding," says Crystal Group's Shaw. "It is becoming more and difficult to put two large-core-count processors in a rugged server rack because of power dissipation. The challenge continues to be the output from these GPUs in waste heat. It is the age-old argument -- always trying to dissipate more power. Things are getting smaller, and we plan for more capability into a smaller package."
Related: Lilium selects Honeywell thermal management tech for its eVTOL
Those designing rugged computers with simple convection cooling must make tough decisions on what they can and cannot install in a small space. "It's not really feasible to put two large CPUs into a single-server application; is just too much power draw, and thermal limitations are extraordinarily difficult to overcome in an air-cooled environment," Shaw says. "In a 1U server application, you struggle to see one CPU in that space because of the thermal load.
Other cooling approaches, such as liquid cooling, can offer efficient solutions, but the drawback often is higher costs, system complexity, potential compromised reliability, and increased size and weight to accommodate the pumps and plumbing necessary for liquid cooling.
Still, there are instances where liquid cooling or something equally exotic might be among the only solutions. "We are seeing liquid cooling as one technology that is gaining traction," Crystal Group's Shaw says. "We find that it is much more likely that liquid-cooled application will be successful if the integrator plans for that; it is not something you add later. Dealing with liquid cooling and where that heat is going has to be at the system level or you will not be very successful."
Planning for cooling has to be at the top of the list for designers of rugged aerospace and defense rugged computing. Our business, ruggedized computing, is really about cooling," says James Tierney, vice president of aerospace and defense at Atrenne Computing Solutions, A Celestica company in Brockton, Mass. "We do classic conduction-cooled computing; that's where the technology is today."
Choosing the right enabling technologies for each separate operating environment is key, Tierney says. Our role in this industry is to bring reality to the requirements," Tierney continues. "We bring reality into what you can do; you can't ask the enclosure manufacturer to put the card where it can't operate."
Atrenne engineers also are trying to design off-the-shelf high-performance rugged computers suitable for a range of military applications. "Our customers are trying to leverage their hardware in a number of platforms," Tierney points out. "They are looking for an environment they would like to meet to fit in multiple platforms. It used to be very specific requirements, but they are trying to future-proof technology. Customers are pushing beyond, to accommodate multiple platforms. It comes down to cooling, and meeting those standards for a much broader set of applications."
Designing rugged computing today "has become this thermal dynamics physics problem," says Systel's Kothari. "Each application has its own technology, and we make sure we take the time for rigorous testing to make sure it will work in those extreme environments. I have this much space, this much heat, and must dissipate that heat with the air flow I have. We go through the rigor and going through the testing and the analysis. We need to get the heat out of the box at the end of the day. More and more performance is more and more Wattage, but in the field it needs to be smaller and smaller."
In rugged computing, "the hardest problem to solve is thermal management," says Curtiss-Wright's Dechiaro. "As processors get faster they tend to generate more heat. What's available on the platform drives what we can do. Liquid flow-through on the chassis can cool everything on the platform -- if it's available."
Liquid cooling "takes a tremendous amount of work to get that on a platform," points out Dominic Perez, chief technology officer at Curtiss Wright. Using liquid, thermally efficient though it may be, is available only in selected environments.
"Generally the platform has an approach," says David Jedynak, vice president of strategic planning at Curtiss-Wright. "A pod has an inlet and outlet, which is essentially forced air, and maybe some fans to help channel the air. Other platforms may run liquid, like a fuel, through everything, when that is available. It is very much on the platform design."
Ruggedized designs
One of the central challenges of ruggedized military computer systems is capitalizing on commercially available technology and using specialized manufacturing to create systems able to withstand the environmental rigors of military applications -- shock, vibration, temperature extremes, and exposure to dust and other contaminants.
"Ruggedization is what we do; it's our business," says ZMicro's Wade. "there are a couple different approaches we are taking. One is we transitioned away from the process of conformal coating to nanocoating technologies. Instead of using traditional conformal coating of sensitive components, we use nanocoating."
For this technique, ZMicro relies on nanocoating technology from Nanoflow X LLC in Carrollton, Texas. Nanocoating uses coating for electronic components with a thickness of a few tens to a few hundreds of nanometers to improve boost protection from corrosion, water, ice, friction, and bacteria. Nanocoating is self-cleaning, and resists the effects of heat and radiation.
"You dip your electronics in, and it prevents intrusion of liquids and dust into the system, but does not inhibit the flow of electrons," Wade says. "It has taken the labor-intensive and relatively expensive process of conformal coating, and replaced it with a more seamless process of dipping electronics and providing the same level of ruggedization for mil-spec environments. Nanocoating streamlines the whole process."
Rugged computer design at ZMicro also involves distributed computing -- or placing different parts of a computer system in different parts of a military platform and connecting them by optical fiber. "We are seeing a push to more distributed computing, with servers in the back end, and distributed thin client to interface back to the enterprise-class server," Wade says.
Adapting commercial off-the-shelf (COTS) computer technology to the rugged military environment can be a difficult challenge, says Systel's Kothari. "Customers want that COTS technology, but when you actually have to deploy that with limited power draw and cooling, we find these interesting problem sets."
Systel and other rugged computer designers face harsher and harsher shock and vibration requirements to survive and operate through military operations at the edge of the battlefield. "We use MIL-STD-901, and we are tested to that," Kothari says. "We have been around for a long time, and we have experience, with products deployed across all domains. We understand very deeply shock and vibration."
Many, if not most, of computer ruggedization challenges are customer-driven, says Atrenne's Tierney. We follow what the board and processors guys are doing, and we put them into a rugged environment in the chassis; the customer will put on us their own challenges."
Customer requirements often can confront rugged computer designers with their toughest challenges, echoes Chris Ciufo, chief technology officer at General Micro Systems in Rancho Cucamonga, Calif.
"We are seeing an increase in the willingness to use 'barely rugged' new technical products as a way of getting much-needed technology onto the battlefield," Ciufo says, "but this equipment still needs to be rugged, and still needs to pass mil-spec qualifications. It's a 'damn-the-torpedoes' moment, where they'll deploy whatever since it's cheap, and not worry about things breaking or having shorter lives."
Systems designers may overlook some crucial issues if they pursue using barely rugged technologies. "If one chooses to use a perfectly good product like Nvidia's Jetson Orin, what do you do about the connectors? Helicopters, jets, pods, and armored vehicles are harsh environments, and plastic, flimsy connectors won't cut it." Ciufo points out.
There is trend emerging called "cocooning," in which systems designs use commercial-grade components and connectors, and design ruggedized enclosures to protect vulnerable components from shock, vibration, and temperature extremes, Ciufo says.
Open-systems industry standards are going a long way to help systems designers tame the challenges of ruggedization, says Curtiss-Wright's Jedynak. "Ruggedization really comes down to physics and different techniques. What's more important that we've seen over the years is the VITA community creating good definitions of what the ruggedization levels are. These levels are well defined on up the harshest levels."
Additive manufacturing
One of the most notable trends in rugged computers involves additive manufacturing, also known as 3D printing. "Additive manufacturing will enable our customers to speed time to market, since I'll be able to do design jobs in weeks, and not months," says Atrenne's Tierney.
"For all these cooling and weight requirements, additive manufacturing enables you to put metal only where you need it and want it. You can put in pockets and cooling channels. You can stay within the realm of machining, but additive manufacturing helps you with tooling, and with weight.
Examples of potential advantages of additive manufacturing include custom very intricate cooling approaches for directed liquid and air flow. "We have done some very creative enclosure systems using additive manufacturing that are much lighter, and time to market is very quick, while maintaining all the ruggedized characteristics, Tierney says.
Related: The world’s first metal 3D printer for space is on its way to the ISS
Is additive manufacturing a viable option today for production-level rugged computing? Probably not, Tierney admits. "Today additive manufacturing in our business is almost negligible -- about one percent of our business. I would expect in next three to five years it could be 15 or 20 percent, and from there it just will overtake traditional machining."
3D printing in the future could enable rugged computer manufacturers to use even more intricate designs than they can today with traditional manufacturing, points out Curtiss-Wright's Jedynak. "Looking way out in the future, I could conceive that things like additive manufacturing might allow you to create some geometries in the materials that could allow you to solve a thermal or mechanical problem in a new way."
For some tasks, additive manufacturing is a valuable design approach in rugged computing even today, says Curtiss-Wright's Perez. "We are using additive manufacturing for the duct work inside our 400 chassis, and we have for a number of years."
The future, Perez predicts, will see even more. "Additive manufacturing will be a fairly big deal. We use CAD and digital engineering, and using additive manufacturing to prototype is what we use frequently today. Metal additive manufacturing is still somewhat exotic. No there yet for additive metal, but will be something that will be widely used 5 and 10 years from now."
It's an open question when additive manufacturing will have a big influence on rugged computing design. We are aware of additive manufacturing to improve efficiency in components and products," says ZMicro's Wade. "ZMicro does not use additive manufacturing yet because we haven't hit that sweet spot in cost. It's a trend that always seems like it's tomorrow, and tomorrow eventually will come when we can meet the price points."
