Designers of embedded systems–from boards and modules used in the latest telecom and networking equipment to a variety of cutting-edge designs–face a growing thermal-management challenge. Newer semiconductor devices are dissipating higher amounts of power, so that they run ever hotter and give off more heat within the system.
If the space available for cooling these components were also increasing, designers might make do with the usual approaches to cooling. However, the space allocated in new designs for thermal management isn't growing. At best, it's at a stand still.
In many cutting-edge embedded applications, this thermal-management dilemma is forcing designers to replace aluminum heat sinks (one innovative example is shown Figure 1 ) with much smaller, but heavier, copper heat sinks. Designers are finding that even large-sized aluminum heat sinks are often inadequate in these applications because these heat sinks have a limited ability to spread the heat, which restricts the overall heat sink's performance.
Because copper has nearly double the thermal conductivity than aluminum, copper heat sinks are much more effective at heat spreading. Copper also has 40% higher heat capacity, meaning devices with transient heat loads or spikes will have better temperature regulation when mated with a copper heat sink.
Nevertheless, copper heat sinks have two notable drawbacks–they're significantly heavier and more expensive than aluminum heat sinks. Fortunately, there is a third option that provides the heat spreading of copper heat sinks, yet is less expensive and significantly lighter. Hybrid heat sinks employ a hybrid copper-and-aluminum construction that combines the advantages of copper and aluminum, providing a superior cooling solution for a wide array of embedded applications.
The heat loads associated with state-of-the-art semiconductors in embedded applications demand heat sinks with substantial cooling capability. Typically, that means the heat sinks need to have large amounts of surface area. One way to achieve greater surface area would be to build heat sinks with taller fins. But that's not an option in many cases since popular standards such as PCI Express, Compact PCI, and ATCA impose height constraints on board-level components. So heat sinks need to be low profile.
To achieve large surface areas in low-profile environments, the footprint must be increased rather than the overall height. This leads to very large footprints for the heat sinks.
These types of large footprint, low-profile heat sinks are in fact becoming popular in embedded applications. They're being used to cool single integrated circuit (IC) devices as well as multiple ICs. Using one heat sink to cool multiple devices is a common solution for maximizing heat-sink surface area in embedded applications.
However, a problem arises when a heat sink that is substantially larger than the IC package it resides on. If the heat source is dissipating heat faster than the heat sink spreads the heat, portions of the heat sink that are far away from the device do not dissipate much heat. In other words, if the base is a poor heat spreader, much of the surface area of the heat sink will be wasted.
In cases where there is a large difference between the footprint of the heat sink and the footprint of the device being cooled, aluminum does not spread the heat quickly enough along the base, but copper does. To understand why this is so, consider the thermal properties of actual metals that are used to build heat sinks. CDA 110 and AL 1100 are copper and aluminum alloys, respectively, that are used to fabricate forged pin fin heat sinks.
The thermal conductivity of CDA 110 is 2,712 btu/in x ft2 x hr x °F while the thermal conductivity of AL 1100 is 1,510 btu/in x ft2 x hr x °F. Therefore, the copper used in the heat sink base is 80% more conductive of heat than its aluminum counterpart. As previously mentioned, this improvement in thermal performance comes at the cost of added weight. Specifically, CDA110 is 3.1 times heavier then AL 1100.
But weight aside, the higher thermal conductivity of copper accounts for its greater effectiveness in heat spreading and explains why aluminum heat sinks are often being replaced with copper in designs where the heat sink is much larger than the device being cooled. Copper heat sinks are also often used for multidevice cooling in which quick spreading of the heat from one side of the heat sink to the other is required.
Heat spreading versus thermal resistance
When discussing a heat sink's performance, a distinction must be made between the heat sink's overall cooling capability, sometimes referred to as cooling power, and its heat spreading capability. Cooling power is typically quantified in terms of the heat sink's thermal resistance , which is a measure of the temperature rise above ambient per dissipated watt on the top of the device. Thermal resistance is specified in degrees C per Watt (°C/W). The lower the value of thermal resistance, the higher is the cooling power of the heat sink.
Whereas, heat spreading capability is primarily a function of a metal's thermal conductivity, thermal resistance is a function of multiple factors including thermal conductivity, but also the heat sink's surface area, airflow, fin geometries and other issues.
For embedded system designers, it's not necessary to understand all of the complexities underlying a heat sink's design. It's sufficient for most designers to understand conceptually how a heat sink's cooling power or thermal resistance differs from its heat spreading capability and to know some basic rules of thumb regarding copper heat sinks.
Specifically embedded systems designers should understand how the use of copper affects thermal resistance to a varying extent. Given two identically structured heat sinks–one made of aluminum and the other made of copper–the copper heat sink will always offer lower thermal resistance.
However, the actual reduction in thermal resistance will be more pronounced in those cases where the heat sink is larger than the semiconductor device being cooled. The larger the difference between the footprint of the device and the heat sink, the greater the impact of heat spreading, and the greater the difference in thermal resistance for copper and aluminum heat sinks.
Hybrid heat sinks
For those applications where heat spreading is the issue, hybrid heat sinks are an attractive alternative to the all-copper models. Hybrid heat sinks are available in various formats and configurations. But the concept is always the same: the portion of the heat sink that comes in contact with the semiconductor is made of copper while other portions of the heat sink are made of aluminum.
Because the spreading of the heat takes place along the base of the heat sink, hybrid heat sinks enjoy the same spreading properties as all-copper heat sinks and provide similar cooling power. At the same time, hybrid heat sinks are substantially lighter and generally less expensive than all-copper models.
For embedded applications, the weight differential is crucial as embedded systems are almost always weight sensitive. In addition to the impact on the system's weight, the use of heavy heat sinks makes it more difficult to mount the heat sink and to meet vibration requirements. Put it in another way, a heavier copper heat sink can severely complicate the assembly of the heat sink into the application while the use of a hybrid heat sink will not have such an adverse effect on the weight of the board.
As an illustration to the weight difference between copper and hybrid heat sinks we can look at typical identically structured copper and hybrid heat sinks of the pin fin technology that are featured in this article.
The weight of an 8.0-inch x 8.0-inch x 1.0-inch copper heat sinks is 9.5 lbs, while the weight of an identically structured hybrid heat sink is 4.6 lbs. In this example, 3.9 lbs, or 51% of the weight is saved by the switch to a hybrid heat sink.
To demonstrate the performance of hybrid heat sinks, we conducted an experiment that compared the performance of identically structured copper, aluminum (shown in Figure 2 ) and hybrid (shown in Figure 3 ) heat sinks. The experiment resembled a common embedded systems scenario in terms of the size of the heat sink, the size of the device being cooled, and the heat loads dissipated.
The heat sinks used in this experiment were identically structured pin fin heat sinks, all featuring a 4-inch x 4-inch footprint, a 0.4-inch overall height, and 900 pins with 0.07-inch diameters. In the hybrid heat sink model, the base consisted of two sections that were reflowed together. The bottom portion of the base was 0.10 inches thick, while the top aluminum portion was 0.30-inches thick.
The experiment was performed three times, each time with a different heat-sink style. In each case, the heat sink was placed on top of a 0.50-inch x 0.50-inch heat source that dissipated 40 watts. Care was taken to mount the heat sink so that the heat source was placed exactly in the middle of the base. The heat sink was positioned in front of a fan that provided 300 linear feet per minute (LFM) of airflow. After heat sink's temperature was allowed to stabilize, we took the heat sink's temperature readings. We then calculated the temperature rise above ambient.
When an all-aluminum heat sink was tested, a temperature rise of 23.2°C was recorded. This value corresponds to a thermal resistance of 0.58°C/W. For the all-copper heat sink, the temperature rise was just 20.3°C and the corresponding thermal resistance was 0.51°C/W. For the hybrid heat sink, the temperature rise was 20.9°C and the thermal resistance was 0.52°C/W.
The results show that the copper and hybrid had almost identical thermal results and outperformed the aluminum heat sink in a substantial fashion, thus showing the importance of quick heat spreading along the base. At the same time, the hybrid heat sink weighed 24% less than the all-copper model.
To further illustrate the effect of heat spreading, we placed a second thermocouple at the corner of each heat sink, the farthest location from the heat source. During each test, the temperature at the corner of the heat sink was recorded and compared to the junction temperature. The results are illustrated in Table 1 .
The results showed that the copper and hybrid heat sinks had significantly better temperature uniformities than the aluminum heat sink. A uniform temperature across a heat sink is a sign that it is operating efficiently as it is an indication that there is good heat spreading from the heat source. As a result, the junction temperatures of the devices when placed under the copper and hybrid heat sinks were significantly better than when the device was placed under the all aluminum heat sink.
Beat the heat
As heat loads grow in embedded system designs, it will become increasingly difficult to achieve adequate cooling using standard aluminum heat sinks. In many new designs, hybrid heat sinks will provide an optimum cooling solution, balancing the demand for thermal performance with size, weight, and cost objectives.
Barry Dagan, P. E., is chief technology officer of Cool Innovations, Inc. Barry has more then 25 years of experience in thermal-management design and holds multiple patents.