Prabhu Sathyamurthy
Electronics Industry Business Unit Manager

Rajesh Nair
Icepak Product Manager
Fluent Inc.
Lebanon, N.H.

The best way to model electronic designs in thermal-management software uses minimal component detail. Too much detail produces simulation models large enough to bring workstations to their knees and answers come too late to be useful. Simple models, on the other hand, provide useful information fast, especially early in the development phase when designs are still flexible and there’s time to pursue promising alternatives.

The best approach is to start with a simple system-level model to determine a general layout. Then gradually increase the model’s complexity as the design stabilizes. Recent advancements in CFD-based software have made it easier and faster to develop relatively simple system-level designs.

Thermal management software also holds the potential to dramatically streamline the thermal management process. Simulations let designers investigate more alternative designs than they could with traditional methods such as hand calculations, spreadsheets, or physical models.

Software tuned for thermal analysis of electronic packaging differs from general-purpose CFD codes in its use of software “objects,” devices that model vents, openings, fans, resistances, or heat sources. When a model needs a particular component, users simply enter its characteristics such as size, thermal conductivity, or specific heat into a macro program rather than modeling it from scratch as is required with general-purpose CFD software. The resulting object becomes the thermal component.

Don’t get fancy
Modeling electronic equipment with every possible detail greatly increases the time required to solve the model. One company, for example, generated a 600,000-cell CFD model to resolve basic cooling issues in an enclosure. The large model took several weeks to solve on a midrange workstation with 128-Mbytes RAM. By the time the computer reached a solution, the design changed, making the results useless. A smaller model, with 30,000 to 50,000 cells, provided enough accuracy to resolve design questions on component placement, fan sizing, and vent location. The simpler model let designers modify and solve different component arrangements in less than 30 min.

System-level models can show temperature, pressure, and velocity at every point in the cabinet. What’s more, quickly evaluating multiple designs makes it possible to determine a design’s sensitivity to changes in design parameters.

Recent thermal-management software aimed at the special problems of cooling electronics dramatically reduces the time and cost of solving thermal-management issues. Several case histories show how the software can be applied to different products.

System models lead to quick solutions
The problem of dissipating heat from a typical desktop computer in a tower enclosure shows how a system-level model helps early in design phases of electronic packaging. The original design specified 300 W of heat from sources including a CPU, power supply, hard-disk drive, motherboard, PCI cards and DIMMs. The baseline design incorporated a large vent, fans for the CPU and power supply, but no forced cooling of the PCI cards. System-level analysis showed an intolerable maximum temperature of 500°C in the cards. This called for a new component arrangement.

At this stage it’s easy to change the vent configuration and relocate fans. A simple change surprisingly showed that just reducing the size of the vents and moving them to rear of the card area dropped the maximum temperature to 200°C, according to the software. Analysis revealed that smaller vents directed air over the cards. The larger vent generated no such airflow.

A second configuration retained the smaller vents, eliminated a CPU fan, and added a system exhaust fan to change the direction of flow in the area of the PCI cards. This reduced the maximum temperature to 140°C. Finally, engineers modified the configuration further by adding a baffle to channel flow. Analysis showed this approach increased resistance to the system exhaust fan, actually dropping the flow rate through the PCI cards and pushing the maximum temperature up to 150°C. Consequently, designers went back to the previous case.

At this point the system-level model has done its work. Modeling efforts can now turn to component-level designs to evaluate individual device temperatures and consider the effect of other cooling options, such as heat spreaders and heat sinks. The completed component-level model can be incorporated into the system-level model to determine the impact of any changes on the overall design. n

Blocks and plates represent cards, fans, drives, and power supplies in a system-level model of a workstation. Software such as Icepak from Fluent Inc., lets packaging engineers identify and fix the hot spots. Such software usually includes macro programs for generating the blocks that accurately represent real components.

An easier way to model heat sinks

A heat sink provides an example of a software object and how to produce its model with a macro. Heat-sink objects represent the heat-dissipating device with two parts: a solid conducting block for the base plate and a volumetric resistance for the finned region.

Users produce block models for a heat sink by specifying the block and fin region characteristics using simple menus included in thermal management software. Designers characterize a heat-sink base by simply entering its dimensions as 3D coordinates and then specifying the material, such as die-cast aluminum. Programs and macros often include libraries of heat-sink materials for easy access to thermal values such as conductivity.

The second part of the heat-sink macro requires a value for the flow resistance in the finned region. It can be found several ways. Handbooks, for example, hold a few flow-loss coefficient values for specific heat-sink configurations. When those are not available, engineers can either gather information (test data) from the vendors or characterize it themselves by measuring it in their laboratory or model it in detail using CFD. The estimate must account for differences between the actual heat sink and the heat-sink object with respect to the available heat transfer area and the characteristics of flow past the fins. In actual heat sinks, heat moves from the base and through the fins before convecting into air streams. The end result is a simple model for a complex component.

Heat pipes, plates, and exchangers keep portables cool

Laptop computers provide special challenges for packaging engineers because they carry many heat-producing devices in small enclosures. Cooling strategies for three different notebook systems show different ways of predicting and handling the heat.

The thermal design for a 20-W system starts with a heat-pipe assembly to conduct heat from the 8-W CPU and to an extruded heat sink in front of a fan. Thermal-management software showed the CPU at an acceptable temperature but the hard drive was too hot because no airflow reached its corner of the case. Engineers showed through thermal analysis that a plate extending over the drive redistributed heat from the drive and lowered temperatures to below critical levels.

A second computer system had the additional components of a graphic chip set, PCI chip set, floppy drive, PCMCIA controller, and audio driver. Thermal engineers proposed a fan and heat sink to cool the CPU. Localized airflow from the fan kept the CPU at acceptable temperatures, but the other side of the notebook was significantly hotter. Analysis showed that adding a combination heat pipe and plate assembly connected to the heat-sink fan would improve system cooling.

The third notebook represents a next-generation computer with a total system power of 30 W. The CPU alone generates half of that. Dissipating this much power calls for a separate air duct, extruded heat sink, and fan to pull in outside air, transfer heat to it, and immediately remove the air from the system through a vent. Heat pipes direct thermal energy to the heat sink and an aluminum plate. The resulting temperature distribution is remarkably uniform and well within the required range.

© 2010 Penton Media, Inc.