Excessive heat can destroy electronic systems. Although designers know that mitigation strategies such as heat sinks, fans, diverter fins, and conduction cooling solve temperature challenges in chassis enclosures, too often designs are based upon “best practice” or even trial and error. Today, computational fluid dynamics simulation discovers optimum thermal solutions before troubles occur. Additionally, “what if” software aids designers in making trade-off decisions.
Wouldn’t it be great if you knew what your thermal problems were before you wasted time and money building prototypes that melted down due to excessive heat? It’s possible, practical, and it doesn’t take clairvoyance. It’s simply a matter of using the right weapon.
(visit our sponsor below)
That weapon is Computational Fluid Dynamics (CFD) software. CFD applies the conservation laws of physics to create a model that predicts temperature, pressure, and other variables at every location in the design: something other methods cannot do. The simulations are relatively quick and easy to build, and the predictions highly accurate; they’re typically within 10% or better.
Prior to CFD analyses, an engineer would attack thermal issues through a cycle of prototype-test-modify-and-prototype again. Often, thermal solutions proved counterintuitive, and a so-called “fix” only made the problem worse. And there was no guarantee that an engineer would pick the right combination of values to investigate.
With CFD, those problems disappear. An engineer can eliminate the time and cost of prototyping, the software can pick the direction in which to move the parameters, and the user is guaranteed to find the optimum design.
User interface A CFD model can be constructed using smart on-line component and materials libraries, although the purist may want to construct their own from scratch. The design also can be imported from CAD, although it is usually quicker to build a new model rather than deal with superfluous CAD details that can bog down calculations.
While the mathematics behind CFD is complex, it is hidden from the designer by a user-friendly interface and built-in decision-making capabilities. The result is easy-to-use software that can have a new user up and running within hours of installation, completely bypassing formal training.
The output is a color-coded 3-D simulation that makes it easy for an engineer to eyeball hot spots. For greater detail, the user can access a table listing temperatures at any location.
Streamlines (or ribbons) indicate airflow rate and direction. When animated, the user can watch air move through and around devices and see which components interfere with the flow. By overlaying the temperature and airflow graphics, the user can observe any interactions between the two.
Solving Thermal Problems Thermal simulation is suitable for any size and shape task. It can manage heat and airflow around tiny chips or handle room-sized heating, ventilating, and air conditioning problems. It can investigate both steady state and transient conditions, examine co-existing regions of laminar, turbulent, and transitional flow, and compute natural and forced convection, conduction, and surface-to-surface radiation. With a few mouse clicks, the designer can rapidly develop “what-if” models, calculate the results, and compare performances.
The following designs, developed using Coolit CFD thermal and flow analysis software from Daat Research Corp, illustrate how effective thermal simulation can be.
Case Study #1: C2 console from design to production in three months.
Miltope Corp. had three months to take a US Army armored vehicle command and control (C2) Heavy Mortar Commander’s Interface console from design to production. The unit had to operate under extreme environmental conditions including desert ambient temperatures that climbed to 125 ºF. The sealed and ruggedized computer, designated the VCU-1600, drew 26 W steady-state and a hefty 67 W when its two removable batteries were charging (Figure 1).
Because of the severe shock and vibration conditions, as well as extremely high reliability requirements, fan cooling was impossible because fans could not survive the environment. The best cooling alternative appeared to be direct conduction cooling of the processor module components through the chassis, but which components needed sinking?
Solution:
Initially, engineering believed conduction cooling of the processor and video chips would be sufficient. But the CFD simulations revealed that conduction cooling just these chips wouldn’t do the job; additional chips would also require sinking to keep their temperatures below the specified maximums when operating in a 125 ºF ambient.
Miltope investigated its options. It examined conduction sinking of various chips, the benefits of thermal pads with different thicknesses and conductivities, and the impact of changing ambient temperatures. By building virtual models instead of physical prototypes, the company obtained highly accurate estimates while slicing over 30% from the development cycle. When the pre-production hardware was built, the actual component temperatures were found to be within 5% of the thermal model predictions. Four weeks were shaved from a 12 week development cycle. For more details, refer to sidebar: PC/104-based airborne maintenance computer.
Case Study #2: Develop a thermal design for multiple battery configurations on a hybridized Humvee.
Like the tank command console in Case Study #1, the high-powered, compact systems in a hybrid powered HMMWV (high-mobility, multipurpose wheeled vehicle) had to survive the stringent environmental conditions found in US Army applications.
The battery systems deliver over 3 KW to power both the weapons system and the hybridized vehicle. While there are multiple configurations, typically, a system consists of 200, two-inch diameter, nine-inch long batteries tucked into one or more compact trays that mount beneath the vehicle frame. At no time can the batteries exceed their 140 ºF upper temperature limit, even when operating in the desert.
Solution:
Keeping the trays cool is tricky; the batteries are tightly stacked so there is little room for air movement and forced air pressure losses are high. Using CFD analysis, the contractor analyzed the thermal and airflow conditions under multiple design scenarios.
The entire simulation process moved from one scenario to the next in rapid succession without the need to do any physical testing in order to determine the best option. Even though this application was the contractor’s first thermal design using CFD, they were able to get up to speed quickly without formal training because of the software’s intuitive user interface.
Case Study #3: Maximize air cooling in an avionics package while minimizing EMI leaks.
Avionics developer Honeywell often finds its thermal and electronics engineers at odds over chassis design; thermal engineering wants to maximize air exchange with the ambient, while electronics engineering seeks to minimize chassis openings in order to maximize EMI protection. Optimal EMI design also means dividing the chassis into compartments, a feature that further restricts airflow. This is a common problem in many chassis designs.
Solution:
When “intuitive” fin locations failed to provide adequate cooling for a forced-air cooled avionics package, engineering investigated using CFD analysis. The simulations demonstrated that, regardless of fin size or spacing, the selected locations would create thermal problems instead of fixing them.
Thermal engineering then combined Design of Experiments techniques with CFD analyses and quickly identified alternate but suitable locations and optimized the fin height and spacing for those sites. Spacing proved especially critical because shrinking the fin spacing beyond a certain dimension produced temperature increases of 10 ºC.
The CFD analysis proved that the chassis required more vents than electronics engineering originally had been willing to allow. When testing verified that the simulation results were accurate to within a few percent, the chassis design was altered. Fortunately, this requirement was identified early in the design cycle, so venting changes could be incorporated into the first prototype. If the requirement had been discovered later–after the prototype was built and tested—it would have delayed the design schedule for the entire project by months.
The analysis further allowed engineering to check performance in flight. Designs that perform adequately at sea level often demonstrate significant problems at altitudes. Engineering was able to simulate the affects of altitude over a temperature range of 20 ºC to 70 ºC without having to climb into an airplane or employ a test chamber.
(visit our sponsor below)
Finding the best solution If the designer has to optimize performance by squeezing as many watts into the lowest cost and smallest package possible, it can easily take a week or more of constructing models and analyzing the results. It’s a tedious process at best, and there is no assurance that the designer will pick the correct values to test. "Value-picking" is often hit-or-miss because design sensitivities are not always intuitive.
Fortunately, the optimization process can be automated using a module, such as OptimizeIt that is integrated into Coolit. The engineer defines the design parameters that can be varied, their ranges, and the objective of the design optimization. Parameters might include types and sizes of heat sinks, vent sizes and locations, number and size of fans, and so forth. Using a sophisticated optimization algorithm, the module automatically sets up the problem and runs multiple cases to find the optimum solution. The results are ranked in order of thermal performance. If there is a combination of the parameters that can make the design goal possible, automated optimization will find it.
CAS Ltd used optimization to design a telecom chassis for operation at 55 ºC ambient (Figure 2). The chassis is a 1U configuration (1.75 inches high) and 9.5 inches wide, dissipating approximately 35 W. The design goals were to ensure that several temperature sensitive components stayed within their manufacturer’s limits, while at the same time optimizing the number, sizes, and locations of vents and optimizing the fan type and location. Standard, commercially available, pin fin heat sinks were chosen.
To compare the impact of optimization software, an initial thermal design was created consisting of one fan and four vents. In this design, several key electronic components on both sides of the PCB had temperatures that were at or exceeded their manufacturers’ limits. The hottest device, a BGA on the component side of the board, was 5 ºC over its 91 ºC limit.
Using OptimizeIt, CAS specified the maximum allowable component temperatures as the design constraints and objective functions. The number, size, and position of the vents and fans were set as variable parameters. The software automatically set up and ran 73 cases and reported that the design goals could be accomplished with a single Delta 40 mm exhaust fan positioned roughly in the middle of the back wall. It also resized and repositioned the existing vents, and added an additional vent to the front panel. In the optimum design, the BGA temperature dropped from 96 ºC to 88 ºC, while the case temperatures of the circuit side components fell from 95 ºC to 89 ºC (Figure 3).
The optimization software found the best solution, and it also found it faster. Solving the same number of cases (73) without the aid of optimizing software would have taken an average of 30 minutes per case to set up and compare results. In total, it would have added almost a week to the project (assuming no human-driven manual calculation or boundary condition errors were made).
“Big picture” thermal approach As the aforementioned case studies have shown, CFD analysis creates a virtual world where the engineer can detect and analyze thermal problems, and develop a solution without building and testing prototypes. The software frequently pays for itself with savings in time and materials on the very first project. And an optimum thermal solution is guaranteed on each design—something that could not be done before.
PC/104-based airborne maintenance computer
While increased functionality is driving up the wattage required to power designs, compact sizes are making it difficult to dissipate what heat is generated. Furthermore, compact packages rarely leave adequate room for mounting temperature sensors and airflow detectors, so it may be impossible to determine what is happening inside a system design.
PC/104 systems may have connectors on all four sides of the board that effectively block airflow, leaving no way to get heat out. Other times, the cards are stacked creating a sandwich that retains heat unless some way is found to force air through the stack.
This was the situation in an airborne computer designed by VT Miltope. The Portable Maintenance Access Terminal 2 (PMAT2) unit provides on-board and ground maintenance and diagnostics capabilities for business jet avionics systems. The PMAT2’s aluminum chassis houses an embedded processor board with a stack of three PC/104 cards which provide aircraft communications system interfaces. Also inside the unit are a power supply, hard disk drive, battery pack, DVD-ROM, and touch screen LCD panel.
The first PMAT2 design concept proposed a sealed chassis, with processor board chips cooled by conduction to chassis walls, while PC/104 board chips were left to “fend for themselves”. It was physically impossible to add heat spreader plates to remove heat from the PC/104 chips because of the close spacing of the card stack and bulky card connectors. A CFD analysis (using Coolit) showed that, at the maximum external operating temperature of 55 ºC, many chips exceeded—or were too close to—their maximum temperature specifications. Engineers knew it was time to revisit the cooling approach. The temperatures for key components are shown in Table 1.
Chip Description
Temp WITHOUT FAN, degC
Temp WITH FAN, degC
Temp Improvement With Fan, degC
Processor Board
U4 Processor Die
91.1
80.5
10.6
U7 Northbridge
86.8
75.1
11.7
U10 Southbridge
87.7
75.5
12.2
U9 Ethernet Controller
87.2
75.0
12.2
U11 Ethernet Controller
87.4
75.3
12.1
ARINC 429 PC/104Card
FPGA
96.4
79.8
16.6
Synch RAM
92.0
74.8
17.2
CPLD
93.7
76.3
17.4
Microprocessor
97.5
80.3
17.2
ASCB PC/104 Card #1
U3 PC104 Logic
101.6
84.8
16.8
U2 Flash Mem
100.0
80.9
19.1
U1 AMD AM186TWEM-40V0
95.0
77.0
18.0
U5 IDT 70261 S35PF 79810P
92.4
74.8
17.6
ASCB PC/104 Card #2
U3 PC104 Logic
96.2
81.5
14.7
U2 Flash Mem
96.2
79.4
16.8
U1 AMD AM186TWEM-40V0
91.4
74.6
16.8
U5 IDT 70261 S35PF 79810P
89.1
72.9
16.2
Ambient External Air Temperature
55.0
55.0
na
Table 1
A small fan was added to the chassis model, and the revised CFD analysis showed all processor board and PC/104 chip temperatures were now within their rated limits, even without optimizing internal airflow via internal baffling. Table 1 shows processor board chip temperatures fell by 10 to 16 °C, and PC/104 board temperatures dropped by 13 to 19 °C (refer to Figure). In addition to reporting lower chip temperatures, the CFD model graphically showed how cooling air snaked through the chassis and between PC/104 boards, providing visualization of hot spots, and enabling further optimization of internal airflow routing.
Ben Kuster is a Senior Mechanical Engineer with nearly 15 years experience in mechanical engineering design and thermal management. For the last six years, he has applied his expertise to ensuring the extreme ruggedness and reliability of military and commercial aviation computers designed by VT Miltope, a company of Vision Technologies Systems.
Peggy Chalmers is Marketing Manager for Daat Research Corp. and author of numerous technical articles. Previously she held marketing and product management positions at Digital Equipment Corp. She holds an M.S. in Mechanical Engineering from Drexel University.
)
