It lets designers relatively quickly evaluate concepts and correct thermal problems long before the proto type phase. Our company provides enclosures, backplanes, and subracks, as well as accessories such as power supplies, front panels, end monitors, and extender boards. We use Flotherm to calculate air velocities and flow rates over slots and cards and for detailed heat-transfer analysis of embedded-chassis designs.
When developing a chassis to be used with high-capacity, high-power-dissipating 6U VME or Compact PCI-based electronics, the first step was to model the enclosure and circuit card. Flotherm's extensive materials library came in handy while modeling the enclosure because it contains a wide range of common alloys.
We primarily design around plug-in cards specified by customers. So we frequently exploit the software's capability to generate compact models or library assemblies. After a detailed analysis of the card in question, I used volume-resistance elements, computationally efficient representations of component flow and thermal characteristics. These elements are basically detailed behavioral models that react to the environment in which they are placed. I then save everything as a simplified model. This substantially reduces computational time on subsequent enclosure-level analyses containing the same board.
Most chassis we design meet IEEE 1101.1 and 1101.2 standards that specify a 0.8-in. spacing between cards. The software makes it easy to use volume utilization, the percentage of space between the parallel cards consumed by the components, and facilitates easy correlation to the pressure-drop characteristics generated by the card at a specific flow rate.
Power in this case was assumed to be 50 W for cards in the main cage and 20 W for cards in the rear. We have also designed chassis with higher power dissipations using the same Flotherm features.
This chassis application used six tubular fans to generate target flows determined by mass flow heat-transfer equations. The fans were placed in the model using SmartParts supplied by the fan manufacturer and posted to Flomeric's online library at www.SmartParts3D.com . Many companies that make fans, heat sinks, components, and thermal-interface materials have contributed to the library. Version 7.1 lets users launch a Web browser inside the software and simply click on a model to add it to their thermal simulation.
Once fans were put in the chassis, the software predicted individual flow rates and power dissipation for each slot. The software's monitor points efficiently measure junction temperatures of critical components, when located at critical points of interest.
Chassis flow rates in the example averaged 10.4 CFM. At this rate, the cards dissipate a maximum of 84 W when operating at mean sea level, and 58 W when operating at 10,000 ft, with an average temperature rise across the cards of 15˚C.
Flotherm also predicts slot velocity-profiles and temperature distributions. In our case, the profile proved to be fairly uniform across the cards, but it was evident that better cooling distribution was possible by changing the shape of the enclosure. Analysis assumed unused cards have blockers in stalled that ensure cooling air goes to the slots.
Simulation showed the chassis exceeded initial design targets. In the example, as in most applications, the full cooling capacity of the system will not be required. We typically add motor-speed controls to reduce noise, fan speed, and power consumption.
Performing a detailed heat-transfer analysis of an embedded chassis can take a lot of time. So I generally begin analyzing new designs by creating simple models that can be run relatively quickly. I use these models to find design factors with major influence on thermal-performance variables such as the temperature of critical components and air flow velocity over them.
Next I specify a reasonable range for variables. Flotherm's Command Center selects values of these variables over a series of runs to explore designs using the least possible computational resources. When I leave work at night, I activate the software's Command Center, and by the next morning, it has completed all runs and determined optimal values for each variable.
The latest version goes one step further. It adds a Response Surface Optimization feature that lets engineers see interactions of design parameters on design goals and identify best values with more accuracy. Users can see response surfaces through either a 2D or 3D chart window, making it easy to understand the sensitivity of cost to changes in particular parameters. Optimization provides better design insight and intuition, lets the user rapidly determine which parameters are crucial, provides instant assessment of the effects of manufacturing variations, and so on.
The software comes from Flomerics Inc. , 4 Mount Royal Ave., Suite 450, Marlborough, MA 01752, (508) 357-2012, flomerics.com.
— Michael Palis
Michael Palis is Senior Simulation Engineer at Hybricon Inc., Ayer, Mass., hybricon.com
Edited by Leslie Gordon