Small form factor designs balance performance, powerStory
March 12, 2013
Whether next to the sensor on a UAV payload, in a vetronics mission computer, or in a helicopter's avionics bay, computer architectures are shrinking, requiring smaller form factors with the same performance punch as larger designs. The embedded COTS community is offering various Small Form Factor (SFF) flavors - off-the-shelf or customized - while industry groups such as VITA are developing standards to provide some order to the growing SFF demand.
“Smaller, faster, cheaper” was NASA’s space exploration motto a few years back, as they looked to send more affordable unmanned missions to Mars and other celestial targets. This phrase equally applies to the U.S. military’s current approach to mission-critical electronics. The Department of Defense’s (DoD’s) program managers want to reduce Size, Weight, and Power (SWaP) in vetronics, avionics, and unmanned systems to take advantage of commercial processing performance and open architectures. In response, embedded Commercial Off-the-Shelf (COTS) suppliers are developing Small Form Factor (SFF) standards and products that leverage these processors while managing the power and thermal challenges that come with them.
Tight spaces in military platforms such as tanks drive much of this demand, as they have little room for ever-expanding electronics bays. Legacy systems already take up more than enough space, leaving little room for warfighters and their gear. To take advantage of high-performance computing architectures, the electronics systems will have to be reduced in size and power. Modern systems and SFFs can provide as much as 10 times the performance in a quarter of the space compared to systems from 5 to 10 years ago.
“In last year’s tactical vehicles, there is equipment bolted to any available surface, with more equipment than available real estate. It is so bad in some vehicles that a reservist cannot fit into the crew seat wearing body armor,” says Bill Ripley, Director of Business Development, Tactical Systems at Themis Computer in Fremont, CA. In some Army vehicles it is possible to easily achieve a reduction in components of around 50 percent by simply switching to modular, standards-based architectures, such as those outlined in the VITA SFF specs currently under development, he adds (see Sidebar 1).
Sidebar 1: Three VITA standards aim to meet military SFF challenges
(Click graphic to zoom)
Packing performance into SFFs
Defense customers want solutions that reduce the electronics footprint for mobile platforms such as Unmanned Aerial Vehicles (UAVs) and wearable computers. These systems require considerable processing power for compute-intensive applications such as data acquisition and analysis and graphical imaging.
“The biggest military market request is reliable equipment that solves size, cooling, and other design problems,” says Ray Alderman, Executive Director of the VITA Standards Organization (VSO). “The military would love to use small form factors everywhere, but it’s a challenge to get performance, enough memory, and enough processing power.”
“Military customers are requiring PCI Express channels for high-speed subsystems such as video processing,” explains Gary Schulz, Director of Marketing at VersaLogic in Eugene, OR. Mini PCIe cards also are being used for system I/O expansion and memory functions, he adds.
“We’re really talking about putting better processing at the edge, where you need more in-field intelligence, so you need localized processing capability,” says Jack London, Embedded Products Business Unit Product Manager at Kontron in Poway, CA. Improved graphics performance is needed for better mapping and better field information, he adds. The technology is at a point where much of this processing can be done locally rather than having it sent back upstream, London continues.
“The military market is looking for higher-performance products from a computing architecture and networking architecture standpoint,” says Michael Smith, Lead Engineer at Parvus Corporation in Salt Lake City. “They’re asking for multicore architectures, more RAM, high-capacity SSDs, and they want it all in the smallest footprint possible.” Equipment manufacturers are integrating GPUs into their CPU chipsets, an architecture that generates considerably more power consumption and, in turn, constrains the thermal limitations of systems using only passive cooling, he adds.
Small and cool
Higher performance in a smaller footprint is often referred to as the functionality-to-size ratio, and designers want this ratio to be as high as possible (Figure 1). However, the higher-performing processors generate an excessive amount of heat that is difficult to dissipate using only conduction and convection cooling methods. “Military embedded systems are facing the same kinds of problems that there are in cell phones – trying to get the heat out of individual modules,” Ripley says.
Figure 1: Themis Computer’s RES-mini SFF high-density servers aim for a high functionality-to-size ratio by managing the thermal challenges that come with Intel Xeon E5 1600 and 2600 devices.
(Click graphic to zoom by 1.9x)
Although advanced solutions for cooling embedded systems exist, the methods can be very expensive, exotic, and even inefficient for some airborne applications, Alderman says. Methods such as “liquid cooling cannot be used in avionics due to G forces and the uncertainty of the orientation of the aircraft,” he continues. “Heat pipes that use a wick also can’t be used due to G forces. Even nuclear submarines can pull Gs like an airplane, so liquid cooling can’t be used there either. The optimal method for cooling is a hybrid system that depends on the application, but generally it involves a chassis that is conduction cooled, with liquid running through the end plates. SFFs mostly use conduction cooling or air cooling (which isn’t viable for military applications due to the low MTBF of fans and their low reliability in rugged environments) that keep the cost low. More advanced cooling options raise costs.”
Therefore, manufacturers have turned to more conventional ways of removing heat from SFF systems, most of which are targeted at modification to the board itself. “It all starts at the Integrated Circuit (IC) and works its way out to the external environment,” says Doug Patterson, Vice President of the Military and Aerospace Business Sector at Aitech Defense Solutions in Chatsworth, CA. “Internal thermal impedances need to be minimized – from the active devices out to the real world – in the most cost-effective and least complicated ways.”
These cost-effective and uncomplicated cooling solutions have largely presented themselves through convection and conduction, the former of which is typically accomplished through removal of active components on the SFF board. “Our military customers will say that they are trying to cut down on power, and we’ll work with them very closely to contain Thermal Design Power (TDP) with depopulation of the module to reduce power consumption,” London says. “We use System-on-Chip (SoC) solutions that minimize the amount of ICs on the board.”
For suppliers that provide higher-end products, conduction-cooling strategies are often employed via modifications in board layout. “Since the thermally limiting component in many SFF systems is the processor, we have attached the die of the processor directly to the wall of the chassis [in the XPand6000],” says Jeff Porter, Senior Systems Architect at Extreme Engineering Solutions (X-ES) in Madison, WI (Figure 2). “Simply put, there is no better way to get the heat off the processor.”
Figure 2: The XPand6000 Series from Extreme Engineering Solutions (X-ES) attaches the die of the processor directly to the chassis wall to maximize natural conduction cooling.
(Click graphic to zoom by 1.9x)
Thermal management considerations can also differ based on application requirements – airborne platforms might have different thermal restrictions than a vetronics application. Complex thermal load definitions – created by processors and FPGAs that generate more or less heat depending upon utilization and clock rates – also vary greatly depending on where the system is deployed, says Dave Caserza, Embedded Computing Architect at Elma Electronic in Atlanta. “For example, at very high altitudes, the temperature is quite cold and conduction cooling to an airframe works very well. When that same vehicle is on the ground in the desert, the conductive cooling is not as effective and some of the application features may need to be turned off.”
Customizing COTS around requirements
Despite the work on SFF standards and open architectures, there will never be a one-size-fits-all approach for military electronics systems, as space constraints in aging helicopter cockpits, ground vehicles, small UAVs, and other platforms often require a somewhat custom approach. “Everything in the military market is customized to some degree, but based off of standards,” Alderman says.
Attempting to mold the industry to standardized SFF platforms is a move contrary to the needs of the defense space, Porter says. “We feel that trying to get everybody’s system requirements and capabilities to fit into the same-sized packages with the same connectors may be too difficult of a task to actually accomplish without limiting performance, capabilities, and innovation.”
“Customers generally have a wide and diverse mix of sometimes-conflicting requirements,” Patterson says. “Vendors periodically canvas the customer base to garner insight and direction for their next-gen products. Once we collect these requirements, we then develop products that can meet the majority of our customers’ needs and demands with more off-the-shelf products and less-costly customization. Let’s face it, money is tight right now. In just about every industry, customers want to maximize every dollar spent to position themselves with the latest technology at the highest Technology Readiness Level (TRL) to provide market leadership and value to their end customers.”
(Click graphic to zoom)