Making it through the heat
Single-board computer vendors are looking to solve the challenge of managing the heat from high-power processors, while keeping the costs of their products as low as possible, and building them rugged enough for even the harshest of operating environments.
By John McHale
Single-board computers from DY 4 Systems ride on the U.S. Army M1-A Battle Tank
Military single-board computer designers love the high performance that commercial processors provide them. Still, they encounter headaches when they try to manage the heat that advanced processors generate. And so it goes, as engineers must perform a delicate balancing act between their twin needs of designing-in high performance while safeguarding sensitive electronic components from extreme temperatures and contamination.
Fortunately for them, creative designs in packaging and cooling techniques are keeping pace not only with advances in electronic technology, but also with ever-more-severe operating environments. Nevertheless, designers must work with growing frequency with commercial-off-the-shelf (COTS) components as COTS continues its popularity in military applications. These conditions are forcing military designers to stay on their toes to keep up with exponential growth in COTS performance.
"Power distribution and heat dissipation are the two biggest issues in packaging single-board computers today," says Ray Alderman, executive director of the VME International Trade Association (VITA) in Scottsdale, Ariz. "When you put more chips on a board, watt distribution becomes a problem because the more power you use the more heat is distributed around the board."
Traditionally, VME board designers have solved this problem in one of two ways — with convection cooling, which blows heat away from boards with fan-driven forced air, or with conduction cooling, which draws heat away from sensitive components with heat sinks.
Today, promoters of CompactPCI, a rival board form factor to VME, are tackling some of the same cooling issues that their brethren in the VME community have addressed in the past. A convection-cooled specification of CompactPCI was just recently standardized, while a conduction-cooled specification is still in the standardization phase.
The conduction-cooled specification for CompactPCI, VITA 30.1, is being prepared for a task group ballot this summer, says Doug Patterson, director of marketing for Ixthos Inc., a designer of single-board digital signal processors in Leesburg, Va. Patterson, in fact, is the writer of the specification. "I boldly predict it will be an [ANSI VITA] standard by the end of this year," he adds.
Conventional cooling techniques like these have been part of military systems designs for years, yet using these techniques can be expensive and waste a lot of board real estate, experts say. Fortunately, there are emerging alternatives.
For example, a relatively inexpensive expensive technique for cooling COTS VME boards, spray cooling from Isothermal Systems Research in Clarkston, Wash., is starting to gain acceptance among military designers, albeit slowly.
Isothermal experts are removing heat from VME systems with a liquid chemical spray, rather than with moving air or metal heat sinks. Spray cooling removes heat from boards more than 500 times more efficiently than air cooling, Isothermal officials claim. Conventional air-cooled systems are limited to a power density of about 1-watt per cubic inch, while spray cooling can cool systems running at more power than 500 watts per cubic inch.
The technique benefits the military not only by reducing system size and weight, but also by eliminating the potential for excess vibration that cold-plate cooling brings to systems, Isothermal officials say.
Isothermal's spray-cool system pumps 3M Corp. Fluorinert fluid from a reservoir to atomizers, which spray the fluid directly onto printed circuit boards. The fluid evaporates as it touches hot components on the boards, re-condenses on the walls of the system enclosure, and runs back down into the reservoir. Then the process begins again.
Isothermal officials claim this approach is enabling electronic systems designers to increase power density, reduce vibration, and lighten weight in military and commercial VME systems.
"The technology is still in its infancy in terms of the military," says Donald Tilton, president of Isothermal. However Isothermal's spray cooling is already being used for the U.S. Army Crusader future self-propelled howitzer system, as well as the U.S. Marine Corp. Advanced Amphibious Assault Vehicle (AAAV), Tilton says.
"Our system showed a size and weight reduction on the Crusader, which was one of the Army's biggest concerns on the vehicle," Tilton says.
"We're also doing all the command, control, communications, computers, and intelligence (C4I) electronics on the command variant on the AAAV," Tilton adds. Marine Corps officials were so pleased with the initial test results that they are considering using Isothermal to do the personnel variant of the AAAV as well. Specifically, Marine Corps leaders are considering Isothermal to provide electronics cooling in the personnel AAAV's command, control, and turret system, Tilton adds.
Marine Corps program managers really had no viable alternative to Isothermal's cooling technique on the AAAV, says Lt. Col. Harry Oldland, AAAV acquisition program manager at the Marine Corps Advanced Technology Center in Woodbridge, Va.
The command variant of the AAAV runs on the Sun Solaris operating system, and needs two VME boards for the vehicle's C4I functions. No existing boards could do the job, Oldland says, and he estimates that customized boards, if they could be developed at all, would cost $25,000 to $30,000 each.
But by using Isothermal's liquid cooling, AAAV systems designers can use commercial-grade boards, which Oldland estimates cost $5,000 to $8,000 apiece. The Isothermal cooling technique also opens the door for technology refresh. The VME boards now contain 166 MHz militarized Pentium microprocessors, but he looks for migration to faster and potentially hotter-running 500 MHz devices.
Oldland explains that the use of the spray cooling technique, by reducing the mass of the board, also improves shock and vibration characteristics.
Other candidates include the U.S. Navy's EA-6B carrier-based aircraft, the U.S. Air Force's F-16 fighter, and the U.S. Army's M1A1 main battle tank, Oldland notes.
"I like it a lot," Ixthos's Patterson says. Although, some issues still need to be addressed such as single points of failure with the fluid pumps, he adds.
The technique may also have a hard time in avionics applications due to the constantly changing attitude of the typical combat aircraft — one minute the box is upside down and sideways the next, Patterson, explains. However, engineers are working to solve these problems. Patterson says it is only a matter of time before spray cooling in military applications really starts to grow.
Heat conduction is a primary focus at Vista Controls in Santa Clarita, Calif., says Gorky Chin, vice president of advanced technology at Vista. Chin and his engineers are currently "experimenting with different cocooning techniques," he adds.
Vista engineers are trying to wrap a card in an environmental cocoon to protect from extreme conditions in temperature, shock, and vibration, Chin explains. Some of their research involves looking at different metal matrixes as opposed to pure metal, which does not conduct as well, he adds.
Vista's engineers are primarily concerned with getting the most benefit at a given cost, reliability, and ease of use, Chin says. Although spray cooling is strong in many categories, Chin points out, it does have issues concerning ease of use and maintainability, he explains. It is somewhat difficult to open the lid to maintain the equipment, Chin points out.
Spray-cooled boxes are easier to maintain than conduction-cooled boxes, Tilton counters. There are difficulties in the way cards are wedged in conduction-cooled boards, he adds. When replacing a card with a spray-cooled device you do not lose much fluid and can add more if needed, Tilton explains.
Spray cooling also controls contamination more efficiently than does conduction cooling, Tilton continues. When contaminants such as water, salt spray, dirt, or dust get in a conduction-cooled box they stay there after the box is sealed, he says. However, with spay-cooled systems filters trap the contaminants and never get near the board, Tilton explains.
VITA's Alderman says he likes the way Isothermal's closed-loop system keeps contaminants out of the enclosure, but questions where the heat goes once the liquid captures it and moves it away from the board. The heat has to get out of the box somehow, he says.
Isothermal's answer to Alderman's challenge amounts a big heat sink that surrounds the box itself, Tilton says. The heat dissipates in the fins on the walls around the box or moves into a heat exchanger attached to the box, Tilton explains.
Officials at Themis Computer in Fremont, Calif., have been active in spray cooling for about the last four years and see strong possibilities for the technology, says William Kehret, president of Themis. However, there is not yet a large demand for it, but because Themis invested early they will be ready when there is demand, he adds.
Engineers from Radstone Technology recently produced an enclosure that uses a liquid-cooling technique in the walls of the enclosure, but not around the board, says Peter Cavill, managing director of Radstone Technology's Embedded Computing business in Towcester, England. Radstone experts tested it with water, but Cavill says he did not know what his customer used.
Radstone has no current plans to use spray cooling, but is keeping a close eye on the technology to see what develops, he says.
Cooling techniques can maintain temperature and keep out contaminants. Still, systems designers have other options as they strive to keep boards safe in extreme temperatures amid strong shock and vibration.
"We design boards to be rugged from the ground up," Radstone's Cavill says. Just how rugged, he says, depends on how Radstone engineers build it. Radstone offers boards that offer several different levels of ruggedness and temperature resistance ranging from straight commercial to full mil-temp. Radstone engineers also, as do most other board vendors, conformally coat their devices to keep moisture out and to prevent fungus from building up, Cavill says.
A common problem that engineers from Radstone and other board vendors face is the thermal coefficient of expansion. This phenomenon happens when the board and its components expand at different rates as they heat up. This can damage the board or its components, and in extreme cases can literally cause chips to pop off boards like popcorn.
It also can cause stress in the solder joints and reduce compliance in ball grid arrays, Cavill says. Radstone engineers solve the problem with a fault-resistant solder, which retains compliance more than a traditional ball grid array, he explains.
Engineers at DY 4 Systems in Kanata, Ontario, prefer to use column grid arrays to solve the thermal coefficient of expansion problem, says Duncan Young, director of marketing at DY 4.
To improve thermal density DY 4 engineers are also looking at exotic materials on their boards to deal with high-density component placement, says Lorraine Byerley, marketing specialist at DY 4. One of them is Dycostrate from Dyconex in Switzerland, she adds. The improved thermal density on the entire board makes it more rugged and stable but also more expensive, she says.
DY 4 experts are also using a conduction adhesive instead of glue to fill thespaces within the boards, Byerley says. The adhesive stabilizes components and helps conduct heat away, she adds.
Shock and vibration
To improve stability in heavy shock and vibration conditions, board vendors use a variety of combinations of stiffener bars on their boards to separate them from their competitors.
Engineers at VMETRO in Houston have designed a PCI mezzanine (PMC) card that screws into the stiffener bar to increase its stability and add ruggedness, says Tom Bohman vice president of business development for Midas RX and MDR products at VMETRO.
The VGM5 and KGM5 multiprocessor boards from Synergy Microsystems in San Diego straddle a central stiffener bar to guarantee the least amount of board flex, explains Ron Marcus, marketing director at Synergy.
Engineers at Themis do not design their Sun-based products to full military specifications, yet are used by many military applications because of their inherent rugged design, Themis's Kehret says. Themis experts use heat sinks, combinations of stiffener bars, and test their boards to 0 to 40 Gs of shock, he adds.
Themis pushes their 14-way symmetrical multiprocessing devices through rigorous shock and vibration testing, such as Class A barge testing, Kehret says.
Macrolink designs new CompactPCI serial I/O card
ANAHEIM, Calif. — Engineers at Macrolink Inc. of Anaheim, Calif., recently released their MCCR Compact PCI serial communications controller, which supports individual line rates as fast as 12.5 megabytes per second with eight or 16 asynchronous serial ports for high density applications.
The 6U module has a maximum aggregate throughput of 30 megabytes full duplex. Independent baud rates for each port and a variety of I/O interfaces enable developers, integrators, and end-users to tailor the MCCR for their specific requirements, Macrolink officials say.
Macrolink's 4-port Serial Line Interface Conditioner (SLiC) modules enable each MCCR to handle diverse line level requirements. One MCCR can be configured with any mix of RS-232, RS-422, RS-485, MIL-STD-188C, and buffered TTL in 4-port groups. Options exist to route all serial line signals to either the front panel, to the rear via a standard 80 millimeter deep I/O transition panel, or a combination of both. Macrolink also supports multiple cable and connector options from this I/O transition panel to the external serial interfaces.
The standard MCCR is a single-slot 6U CompactPCI bus board with a PICMG 2.1-compliant Slave Interface. The sixteen port rear panel I/O version of the MCCR is priced at $795 in OEM quantities, and includes RS-232 SLiC modules. The MCCR uses high-speed quad UARTs that include an on-chip PCI interface and 128-byte FIFOs.
Deep FIFOs reduce interrupt overhead, and hence the load on the host computer, resulting in higher system performance and efficiency, Macrolink officials explain. FIFO depth can also be read directly from the host computer.
The MCCR also acts as a replacement for host-based serial ports, or for a low-density mezzanine board, and provides a low-cost-per-port alternative for existing legacy systems as well as emerging requirements, Macrolink officials say. The low parts count used on the board results in reduced mean time between failure and a lower board cost, Macrolink officials claim. Device driver support for VxWorks, Solaris, Windows NT, as well as other UNIX and real-time operating systems is available.
For more information on Macrolink contact Jack Davis by phone at 714-777-8800, by fax at 714-777-8807, by email at firstname.lastname@example.org, or on the World Wide Web at http: www.macrolink.com. — J.M.
DAWN VME helps package electronics for missile applications
FREMONT, Calif. — Engineers at Computer Science and Applications (CSA) Inc. (CSA) in Shalimar, Fla., needed a VME card cage that could handle the heavy stress of seeker technology on missiles for the U.S. Air Force, Army, and Navy. They found their answer with a card cage from Dawn VME.
CSA experts design and build the Infrared Projector Control Electronics, which contains a ruggedized chassis and many custom system boards. All of this is used to create a complex and powerful "hardware in the loop" simulation system, Dawn VME officials say.
One of the original design problems faced by CSA engineers was developing a ruggedized chassis that could be mounted on a Flight Motion Simulator Table (FMST) and survive the associated shock, vibration, and "G" forces.
Rick Glattke, a CSA engineer, decided to try using a standard card cage and J1/J2 backplane, Dawn VME officials say. Since it was a standard card cage, a random VME products vendor was chosen for the initial prototype. Glattke built his special ruggedized chassis, mounted a customized shock isolation system from the chassis to the card cage, and installed the boards into the card cage.
The entire system had to function under the extremely energetic movements that the simulator produced. The cards had to stay in their slots, they had to remain attached to the backplane, the backplane had to function under extreme stress, and the card cage had to remain attached to the shock isolators without any shearing, ripping, or any kind of distortion, Dawn VME officials explain. In order for Air Force, Army, and Navy officials to be satisfied, this system had to be installed and demonstrated on a FMST.
Unfortunately, the card cage that Glattke purchased from another vendor could not handle the necessary stress, Dawn VME officials claim. He had to find a manufacturer with a more reliable product.
For his first production run of this ruggedized chassis, Glattke purchased the card cages and the J1J2 backplane from Dawn VME Products. So far, out of all of the systems installed, there have been no failures due to shock, vibration, or "G" force loads, Dawn VME officials claim.
For more information on Dawn VME rugged card cages and backplanes contact the company by phone at 800-258-3296, or on the World Wide Web at http://www.dawnvme.com. — J.M.
Ruggedizer flies on French fighter
BURLINGTON, Mass. — Officials for the French military are using the Ruggedizer from Cetia to bring commercial-off-the-shelf (COTS) performance to their Rafaele fighter jet.
The Ruggedizer acts as an "equalizer," to lower operating temperatures of the hottest points on the board by re-routing the heat to cooler areas. All the components then remain well within their standard operating range. By minimizing the effects of physical and thermal constraints, the Ruggedizer enhances the mean time between failure for conduction and convection cooled boards.
"I think the Ruggedizer is the perfect solution" for military applications such as the Rafaele that want the performance of commercial-off-the-shelf boards but have strict environmental requirements, says Hubert Canuet, vice president of worldwide sales at Cetia. Cetia engineers have made the device easier to manufacture, Canuet says, and continue to look for ways to improve performance.
Cetia's Ruggedizer, an aluminum heat drain, is milled out to fit the whole topside topology of the board. It is attached to the board using several mounting holes and a thermal film, with no electrical conduction that provides maximum heat transfer from the components to the drain, while at the same time it allows easy removal and re-application when servicing the board.
The aluminum used to produce Cetia's Ruggedizer is machined from aluminum 6082 in accordance with the standard QQ-A-2501/II TEMP T6. The aluminum is then subjected to a surface treatment in accordance with the standard MIL-C-5541, Cetia officials explain.
CETIA boards using the Ruggedizer have been extensively validated for military use, and successfully deployed in the field, Cetia officials claim. All of Cetia's hardware products are also covered by a twelve-month return-to-factory warranty.
For more information on the Ruggedizer contact Cetia by phone at 781-229-7930, by fax at 781-229-6926, by mail at 8 New England Executive Park, Burlington, Mass. 01803, or on the World Wide Web at http://www.cetia.com. — J.M.
FPDP used in Virginia-class submarine sonar simulator
FAIRFAX, Va. — Engineers at Digital System Resources (DSR) are using Front Panel Data Port (FPDP) technology from Interactive Circuits and Systems (ICS) in Ottawa, Ontario, for simulating high frequency sonar arrays on the future U.S. Navy Virginia-class new attack submarine.
The FPDP is very fast, simple, and good at distributing data, says Mike Beasley, chief engineer for submarine systems at DSR.
ICS's FPDP simplifies intra-system communications with ribbon connections between circuit boards, ICS officials say. The FPDP is a 32-bit parallel data bus that runs at 160 megabytes per second. It connects circuit cards in the front where they are readily accessible, rather than in the back like a traditional backplane does. The VME International Trade Association in Scottsdale, Ariz., has adopted FPDP as a standard.
ICS engineers are currently working the next generation of the FPDP specification, dubbed FPDP II. The new spec will have a 400 megabyte-per-second sustained data transfer rate, double edge clocking, and will be backward compatible with the current FPDP standard.
DSR engineers will not be upgrading to the FPDP II, only because the current FPDP is just right for their application, Beasley says.
For more information on FPDP contact ICs by phone at 613-749-9241, by fax at 613-749-9461, or on the World Wide Web at http://www.fpdp.com. — J.M.