As military programs continue to push the limits of computing system requirements, rugged computer systems must continue to evolve to endure the toughest conditions. To meet these stringent requirements, rugged subsystems suppliers must produce innovative thermal management technologies that can withstand situations that may threaten a computer’s durability. Because of ever-increasing processor speeds, advances in thermal management will continue to rank as one of the most important trends in rugged computing design.
With heat issues often credited as the largest contributor to system failures, designing rugged systems to meet these thermal challenges is critical. The key to successful thermal design is getting heat into contact with ambient air for convection to the external environment as quickly as possible. By implementing a variety of thermal management techniques, heat can be dissipated fast enough to prevent thermal runaway — an increase in temperature (on a component or system) that changes the conditions in a way that cause exponential temperature increases, ultimately leading to destruction.
Relying on convection cooling inside a system has its limits as heat is dissipated rather slowly (high thermal resistance) to the surface of the chassis where it can be dissipated into the atmosphere. However, advancements in conduction cooling have had a tremendous impact on rugged system design. For example, custom-designed clamshell heat sinks can be fitted to encapsulate each printed circuit card assembly, using wedge locks or direct mechanical interfaces as the contact point with the card stack to dissipate heat more quickly (lower thermal resistance) to the chassis. The clamshells, which essentially work as heat spreaders, and the thermally conductive gap pad significantly reduce thermal issues. Heat spreaders are also being designed to accommodate a number of thermal options, such as top-mounted heat sinks, fan heat sinks and heat pipes to effectively cool microprocessors. Innovative heat pipe/heat spreader combinations are proving especially effective in the thermal management of stand-alone rugged boxes (Fig. 1).
When complete convection cooling is not possible ñ or when designing the cooling properties of system enclosures ñ thermal designs must consider the thermal boundary layer. The thermal boundary layer is a layer of warm stagnant air that builds up between the fins of heat sinks blocking air movement. This build up of air effectively reduces the useful dissipative surface area of a heat sink. The turbulence of the air affects the thickness of the boundary layer and therefore the rate of heat transfer. For example, the more air molecules that contact the surface of the heat sink, the greater the dissipative effect. The boundary layers of rugged devices need to be carefully evaluated, as it is not uncommon for embedded systems to be installed in a stagnant air environment. As such, boundary layers can greatly influence the effectiveness of heat sinks and other convection cooling methods. Since air is the final “fluid” to dissipate the heat into, it makes sense that the contact with this medium be the most efficient possible.
Rugged system designs often must specify extremely rugged components that have warranties guaranteeing performance up to +85°C. To ensure optimal performance for these components, derating can be employed. A technique where electronic devices are operated at less than their rated maximum power dissipation, derating takes into account the case/body temperature, the ambient temperature and the type of cooling mechanism used. Derating increases the margin of safety between part design limits and applied stresses, thereby providing extra protection for the part. By applying derating for an electrical or electronic component, its degradation rate is reduced and the reliability and life expectancy are improved.
Although thermal management mainly surrounds the task of moving heat to the outer chassis, when designing for rugged conditions, managing cold temperatures must also be considered. Cold temperatures can cause a variety of problems with electronic devices, including problems with voltage sags and clock frequencies, creating timing difficulties. Descending to cold temperatures also causes rapid contractions, which can cause problems with powering on a device. To help mitigate these problems, screening each electrical component to ensure it is rated to operate from -40°C to +85°C is imperative. Parvus also takes the additional step to screen each CPU board before it is installed.
Aircraft Thermal Management
For over a decade, we have supplied the U.S. military with rugged cockpit computer subsystems for an electronic warfare aircraft program. In its latest follow-on project, we developed a multi-mission computer to support both electronic warfare and laser targeting.
To provide enhanced thermal ruggedization for the computer, three thermal “envelopes” had to be managed. This included the stagnant air inside the main subsystem chassis, the air between the main chassis, and the Air Transport Rack (ATR) aluminum chassis and the ambient air. To ensure heat could rapidly transition these “envelopes”, Parvus engineers carefully managed the selection of rugged components, implemented conduction cooling where possible, plus provided components with the most heat direct conduction links to ambient air. For example, the main CPU has a direct heat pipe link to a heat sink exposed to ambient air. The main thermal management goal in this deployment was to keep the thermal resistance of each heat source to ambient air as low as possible and practical.
While energy-efficient single-core processors can still serve many applications, new technology developments are demanding higher processing power with low power consumption. Ultra-low power Intel Atom processors cannot satisfy performance requirements in all cases. Consequently, multi-core processing technology is seeing a significant boost in deployment within stand-alone rugged boxes. For demanding applications, mobile Core 2 Duo or Core i7 processors, as examples, provide attractive solutions. While presenting a challenge to manage the 10-55 W of Thermal Design Power (TDP) of these processors, they offer a level of performance that pushes the boundaries between traditional rugged computing and new tactical applications.
We were recently tasked with designing a passively cooled solution that met the increased thermal demands of a multi-core processor, upon receipt of a contract to design a specialized version of the DuraCOR 810-Duo mission computer for a U.S. Navy special operations river craft. This specific military customer required a shipboard mission computer as part of its command, control, and communications system designed to operate in an ambient operating temperature of up to +71 °C per the requirements of MIL-STD-810G (Fig. 2).
Unlike our legacy-generation mission computer based on a Pentium M processor, this new mission computer integrated an Intel Core 2 Duo CPU — which consumes about 60 percent more power and generates significantly more heat. The challenge for engineers was to design a rugged mission computer with integrated application-specific payload cards capable of cooling a multi-core processor in an environment that can only support cooling by natural convection and possibly operate in a stagnant air environment.
To manage the thermal increase in the new mission computer, engineers initially designed the internal subsystem with an integrated heat pipe: a thermal management device that transfers heat by the evaporation and condensation of an internal fluid. Heat pipes don’t rely on any moving parts to dissipate heat from the system, plus they have been used successfully in the past for other single-core CPU system designs. Heat pipes are also well known for how quickly they can wick heat away from system hot spots. The heat pipe is a common and well-known device, which contains a copper outer layer with a hollow center that utilizes a wicking mechanism. The center is filled with a liquid capable of vaporizing, such as water, alcohol, ammonia, or methanol. When heat hits the pipe, it converts the liquid to a vapor. The vapor escapes down the pipe and comes in contact with the cool side of the heat sink. The vapor then cools and is condensed back to a liquid where it is absorbed by the wick. This transfers the liquid back down the heat pipe to start the exchange over again (Fig. 3).
Engineers soon learned from preliminary qualification tests that the heat pipe was actually “saturating” in high temperatures. The computer server was experiencing difficulty when performing system stress testing beyond +65 °C ambient air temperature before the CPU’s die would hit Intel’s rated thresholds. The internal temperature of the heat pipe was exceeding its design limits and the core temperature of the computer system enclosure was approaching +90°C. The heat pipe’s increased temperature would not allow the vapor to cool enough to convert back to a liquid — referred to as “saturation”. This heat pipe saturation effectively eliminated continued vapor-to-liquid conversion, halting convection to the exterior of the enclosure. The only effective method of heat transfer was the conduction of the copper pipe from the heat-producing processor to the exterior heat sink interface. This condition reduced the possibilities of the system operating past +65 °C ambient air for extended periods of time.
Through initial testing, the first suggested engineering answer to this challenge was to introduce a conduction-cooling mechanism through a structural heat-spreader plate. Rather than channeling the heat from the single board computer through a copper pipe and converting the liquid to a vapor, engineers designed an aluminum heat spreader to transfer heat to the exterior chassis heat sink. When the aluminum is exposed to a heat source, the heat is conducted through the metal. Unlike the heat-pipe approach where vapors move very quickly through the pipe, the aluminum takes a longer time to conduct the heat. However, Parvus engineers quickly found that this alone was not the overall solution as the thermal flow path to the exterior was too slow. The existing design utilizing a heat pipe was superior in removing the heat quickly and delivering it to the exterior.
Engineers solved the thermal management challenge by combining the benefits of heat pipes with an effective conduction method and delivering it to the exterior surfaces ñ which also needed to be sufficient in area for the final dissipation into the air. Designing this combination limits the risk of saturation, allowing the system to operate in higher temperatures as required by contract: to +71˚C. By using both heat pipes and aluminum heat spreaders, when the heat reaches the exterior heat-spreader plate, it interfaces with a thermally conductive material, which then transfers the heat to the chassis heat sink. With a high-finned protruding design, these extruded heat sinks increase the surface area of an enclosure housing, transferring potentially damaging heat away from a system and preventing performance degradation. By designing a direct conduction method from the internal heat-producing components to the natural convecting exterior of the enclosure, the heat can continue to move to the ambient air.
Though the DuraCOR 810-Duo described was designed for a specific application, other military customers are interested in deploying similar solutions in ground, naval, and airborne vehicles. Accordingly, Parvus intends to apply its lessons learned and similar thermal management analysis to address varying program requirements. Because at 40,000 feet (12,192 meters), roughly one-fifth of the atmospheric pressure (2.7 PSI) is available for convective cooling compared to sea level (14.7 PSI), the COTS dual-core subsystem will be fitted with an optimized heat pipe / heat-spreader plate as the default configuration to maximize application robustness (Fig. 4).
Engineers always need to be conscious of which issues are preventing a computing system from reaching its ability to meet a customer’s needs. Gaining a better understanding of which combination of thermal products and techniques help transfer heat while maintaining cost, weight, and system integrity will prove to be one of the most important elements in rugged computing design. As customers continue to push the limits of computing system requirements, rugged computer subsystems will continue to evolve to endure the toughest conditions as long as good engineering practices are implemented and maintained in the process.