Coolit software identified two components insufficiently cooled in the original design. OptimizeIt repositioned existing vents and added another to bring temperatures within spec. |
The completion or case map of an optimization study produced several acceptable solutions. But the best one, highlighted in yellow, also had the lowest temperature for critical component C13A. |
The dialog box shows how users set temperature constraints. The maximum temperature is to be less than or equal to 280°K. Average surface temperature must be greater than or equal to 300°K. |
The Objective Function dialog box sets goals for the optimization, such as minimum or maximum temperatures, average surface temperatures, and minimum and maximum pressures. If users specify more than one objective, OptimizeIt forms a composite objective function that is a linear combination of the two with their relative importance determined by their weight value. |
Trying to squeeze a few extra watts of capability into a design can quickly turn into a frustrating task. For example, optimizing for thermal characteristics can involve setting up and solving dozens of variations, and then manually comparing results to find the best. The process is also time consuming. "Thermal engineers might spend 40 hours or more setting up test cases and analyzing results for new products," says Eldad Levy, engineering manager with CAS Ltd. in Israel (cas.co.il). "And they have no assurance they picked the right values to test." Levy says value-picking strategies are often hitormiss because design sensitivities are not always intuitive.
While thermal-optimization software is not new, most FEA programs do not compute airflow and work only on single components, such as a heat sink. But a new program, OptimizeIt, adjusts component sizes, material properties, and placement within enclosures to compute airflow as well as conduction and radiation. "This type of optimization promises to make thermal analysis faster, more effective and efficient, and it lets engineers examine many more design possibilities. If there is a combination of parameters that can meet a design goal, the software will find it," says Levy.
It works like this: Users identify design parameters that can vary, their ranges, and the objective of the optimization. For example, a vent could be 3 to 6-in. wide and located anywhere on the front of an enclosure. The objective might be to minimize the temperature of a particular component or maximize heat removal. Using an optimization algorithm, the software sets up and runs multiple cases and identifies the best design.
While any industry can benefit from thermal optimization, telecommunications, with its Network Equipment Building System (NEBS) requirements, can profit more than most. NEBS requires that equipment operate in a 55°C environment while maintaining electronic component case temperatures below the manufacturer-defined maximums, typically 80 to 95°C.
A telecomm chassis is a good test case for the software. The chassis is a 1U configuration, 1.75-in. high and 9.5-in. wide, dissipating about 35 W. The goal is to ensure several temperature-sensitive components stay within their manufacturer's limits, while at the same time optimizing the number, sizes, and locations of vents, and fan characteristics and location. It also uses standard, commercially available pin-fin heat sinks. "If optimizing the parameters proved insufficient to maintain the required component temperatures, we could have added expensive custom heat sinks, but preferred not to," says Levy.
The initial design had one fan and four vents. It also had several electronic components on both sides of the board with temperatures at or exceeding manufacturers' limits. The hottest device, a ball-rid array (BGA) was 6°C over its 90°C limit.
"We specified the maximum component temperatures in the optimizer as design constraints and objective functions. The number, size, and position of vents and fans became variable parameters," he adds.
The software, running on a 2.2-GHz Pentium 4 computer with 1-Gbyte RAM, ran 73 cases, each with about 1.3 million grid cells. Run time for the initial case was 3.5 hr, with later cases varying between 15 and 90 min. On completion, the software reported that the design goals could be reached with a single 40-mm exhaust fan placed roughly in the middle of the back wall. "It also resized and repositioned the vents, and added an additional vent to the front panel," says Levy. In the optimum design, the BGA temperature dropped from 96 to 88°C, while temperatures on print-side components fell from 95 to 89°C.
"Performing the optimization manually and solving the same number of cases would have taken on average 30 minutes per case for just setting them up and comparing results. It would have added almost a week to the project, and resulted in higher cost to the client," he says.
MAKE CONTACT:
Daat Research Corp.,
(603) 448-1302, daat.com