CFD simulations lower risk, pick up the pace of chassis design
StoryApril 26, 2010
Drew Castle
CP Technologies
In the design of custom computer chassis for mission-critical components, Computational Fluid Dynamics (CFD) software simulations allow accurate determination of the performance of cooling components, clarifying how specific configurations perform, thus cutting design costs and risks and increasing the speed at which concepts become products.
In computer chassis design, many basic needs can be tailored to motherboard, processor speed, communication, storage requirements, and size limitations by simply putting the correct off-the-shelf components into an appropriately shaped box. However, one integral design consideration that simply cannot yet be ignored or handled as a quick matter of routine (no matter how advanced the rest of the components are) is the cooling of the chassis. A simple temperature elevation in a poorly designed chassis can cripple even the fastest and most robust systems. And this becomes increasingly critical in the design of custom systems, where “off-the-shelf” is a four-letter word.
The new mantra in computer performance is: “More stuff – More power – Small box.” When these systems are required to function in ambient temperatures exceeding 100 °F and people are counting on them for necessary information, the stakes get even higher for effectively designed systems. Ruggedized chassis designed for use in the current battlefield environment are subjected to extreme conditions every day. There is no tolerance for failure and design, and production deadlines must be kept.
As a result, CFD simulation software has become an increasingly used tool as both programming and processor capabilities have become more advanced and refined. The software allows a set of conditions to be applied to an existing three-dimensional model. Then the software calculates resulting flow characteristics, thermal profile, and overall chassis cooling. Complicated Navier-Stokes equations describing the flow field of fluid – once shrugged off by engineers in favor of a best-guess or tried-and-true design dogma approach to engineering – are now embedded in software readily available to any engineer. The advantages: nearly instant access to answers to the cooling equation, plus the ability to monitor the effects of design changes throughout the design process rather than trying to fix underlying problems after a product has been designed, built, and evaluated. The result: design time reduction.
How CFD simulations work
CFD simulations operate by a method of Finite Element Analysis (FEA). In FEA, a control volume of solid, liquid, or gas is broken down into cells defined by a mesh grid. In the case of CFD, calculations of heat transfer through various media and fluid dynamics of the flowing air are done on a cell-wise basis, with each cell in the mesh grid influencing the other cells surrounding it. With detailed components, it is possible to optimize the mesh in a manner that provides smaller cells with more refinement around these components rather than in large spaces with bulk conditions.
Setting up a simulation with the aforementioned points in mind becomes a bit of an art. Developers determine how specific to make the mesh in certain regions of a model while negating this specificity in other regions while still extracting accurate results. In full simulations on complex chassis, it is not uncommon to create meshes comprised of more than half a million cells.
As with any tool, the quality of CFD results that can be achieved by a craftsman versus those achieved by a novice are a function of experience. Thankfully, CFD software developers are increasing the usability of their products to constantly accelerate that learning curve and help novice users quickly arrive at accurate results. An example illustrates the benefits and innerworkings of CFD.
Example: CFD simulation
The operations and advantages of CFD software can be seen through a simple exercise of varying the installation of a fan in a computer chassis. To simplify the images, the temperature field through the chassis has been hidden on the representative graphics herein, but it is important to know that CFD can determine the heat transfer as well as the flow field.
Many designers, when it comes to solving internal temperature issues, simply place enormous fans directly in front of important components, believing that increased air velocity is sufficient for cooling. While the increased velocity does have a cooling effect, it quickly wears off if the heated air is allowed to remain in the chassis. In Figure 1, a simplified chassis with one large 120 mm diameter internal fan, capable of pushing 120 cubic feet of air per minute (CFM), has been modeled in two chassis configurations to show airflow through the chassis. The streamlines in the two pictures in Figure 1 show the actual path of air throughout a three-dimensional chassis rendering.
Figure 1: An unsealed chassis bulkhead (left) compared with a sealed bulkhead (right) shows a net airflow improvement of 184%, a simple calculation using CFD software.
(Click graphic to zoom by 1.7x)
|
|
The left side of Figure 1 shows a typical design with an open bulkhead positioned mid chassis with the fan mounted to the bulkhead. As depicted, often a large area is left open to allow for easier routing of cables past the bulkhead. This is a common design element in low-end chassis built for consumers with basic needs operating in simple environmental conditions. In this simplified instance, there are minimal accessories in the chassis so the airflow is clearly visible. The only actual cooling airflow is represented by the streamlines shown entering the front of the chassis from the left side of the image, and exiting via the opposite wall. As this particular simulation was performed as an analysis of internal flow conditions, streamline behaviors after exiting the chassis were not calculated, and therefore do not appear in the diagram.
While the internal fan increases the air velocity in the chassis, the ability to transfer heat from the solid surfaces to the moving air is dramatically reduced as the temperature of the recirculating air increases. This is a key consideration in determining the resultant thermal profile of the chassis. While the empty chassis allowed the fan to push 112 CFM, the actual net airflow through the chassis was a mere 37 CFM. Once the chassis was sealed, though, as shown on the right side of Figure 1, the airflow rose to a much more respectable and effective 105 CFM through the chassis. The placement and design concessions made for cooling fan installation should not be an afterthought, but amongst the most important initial design concerns. While some internal fans are necessary to direct much-needed air to sensitive components buried within a labyrinth of wiring and circuitry, their implementation must be well thought out and tested before modifying designs to accommodate them.
The simple analysis offered by CFD software shows us not only the actual path air takes as it moves through the chassis, but also calculates for us the exact volume of air entering and exiting the chassis through each vent over time. While it is possible to predict streamline behavior based on a “common sense” design approach, accurately measuring the volumetric airflow through a chassis by experimental methods is a much more complicated and time-intensive process than utilizing CFD software. If we were to choose to pursue that method of design, it wouldn’t be until a fully assembled chassis were drafted, built, and bench-tested that we would discover the disparity between fan capability and the actual net flow rate. At that point the options would be a redesign of the entire product, or making last-minute concessions to improve the existing faults. Having this information available early in the design process allows design modification to achieve optimum performance.
CFD simulations: The result
In designing custom chassis for extreme environments, CFD simulations are often useful in saving time and helping verify configuration changes throughout the design process. Figure 2 shows a fully configured 4U chassis designed by Chassis Plans and similar to one used in Iraq and Afghanistan. Its redundant power supplies, multiple SBCs, CPU coolers, and inlet fans modeled for CFD analysis serve as an example of a chassis where the interaction between design geometry and fan performance is difficult to predict with so many different elements.
Figure 2: A more complex chassis with multiple flow variables is still a simple matter for CFD software simulations and analysis.
|
|
Though moderate simplification was done on this model to improve calculation times of the software, the calculation results still maintain a high degree of accuracy. The large inlet fans push air through the chassis, across sensitive components on the SBC, while smaller fans in the power supplies direct air over the heat-generating components of the power supplies. Meanwhile, CFD allows accurate visualization of flow trajectories, giving a clear picture of how this configuration performs in extreme conditions before engineers turn even one screw.
Drew Castle is a mechanical and thermal design engineer at Chassis Plans. His career spans a decade of varied positions in the high tech industry. He is currently pursuing a graduate degree in mechanical engineering with a thermal sciences specialization. He can be contacted at [email protected].
Chassis Plans 858-571-4330 www.chassisplans.com
