Design Feature: October 12, 1995

To choose the best heat sink for your application, you should be aware of the different heat-sink technologies, have a good idea of available airflow, and learn how to estimate the performance you need. If you can perform thermal modeling and testing early in the project, you can minimize experimental testing and smooth product development.

You have many choices, including heat-sink type and size, interface material, and attachment method. Heat-sink types include stamped, extruded, bonded, and convoluted-fin heat sinks and fan sinks. Interface materials include thermal compound, tape, and conformable polymers. Common attachment methods include clips, epoxy, and pressure-sensitive adhesives.

In addition to these materials and physical configurations, the most important selection criterion is thermal performance. The key elements in thermal performance are temperature requirements, power dissipation, and airflow. Airflow is particularly important to heat-sink selection because the air carries away the heat. Fortunately, you can handle insufficient airflow in a variety of ways (see box, "Quick fixes for insufficient airflow").

Before reviewing the heat-sink types, realize that many factors influence a heat sink's performance. These factors include upstream heating and flow blockage due to upstream components, contact area to the hot surface, and sufficient airflow through the fins to remove the heat efficiently. In natural convection, heat-sink performance is a strong function of orientation. Performance is usually best when you orient the fins vertically and ensure plenty of room around the heat sink for fresh air to replace the air that's rising as it heats.

Table 1 summarizes the distinguishing characteristics of each heat-sink type, assigns a relative cost to each, and identifies the pros and cons of each type. The table includes the most common types of heat sinks, but other more exotic types of heat sinks also exist. These types include plastic and aluminum die-cast heat sinks, coolant-filled bags, heat pipes, and cast-base heat sinks.

Stamped heat sinks

Stamped heat sinks are usually light-gauge stampings of aluminum (Fig 1). Because they are stamped from a sheet of metal, these heat sinks have low fin density. The mounting surface is usually not flat, requiring high mounting torque or a conformable interface material.

Extruded heat sinks are either bidirectional or omnidirectional (Fig 2). Bidirectional sinks let air flow either way along the extrusion direction. Omnidirectional sinks let air pass through the heat sink at any angle, usually because the manufacturer crosscuts the heat sink. An extruded heat sink has a 2-D cross section, and the heat sinks typically come in 8-ft lengths. The heat-sink vendor buys these lengths from an extruder and cuts them to length, adding holes and other features as needed. Omnidirectional heat sinks start out as extrusions, and crosscut saws make the fins in the other direction. The result is a pin-fin grid arrangement. The crosscut heat sink in 45° angle flow has about 25% higher temperature rise than its lined-up flow twin. One advantage of the omnidirectional type is that the pin-fin arrangement allows for more precise positioning of the heat sink using clips.

Another important characteristic of an extruded heat sink is its aspect ratio. A high aspect ratio leads to a better-performing heat sink. The aspect ratio of an extrusion is roughly the fin height divided by the fin spacing. Typical heat sinks have a 3:1 to 5:1 aspect ratio. Extrusions can go up to an 8:1 or 10:1 ratio. Generally, fin heights are limited to 3 in. or less; heat-sink widths range up to 18 in.

Bonded-fin heat sinks are newer than the stamped and extruded types. In a bonded-fin heat sink, an extruded aluminum base holds plate fins in place, usually with an epoxy adhesive (Fig 3). Typical fin aspect ratios are around 10 or more; thus, if you can move enough air through these heat sinks, they perform well. However, moving enough air is a big if; most air movers move less air when confronted with a large impedance. Also, the fin material sometimes has lower thermal conductivity than does extruded aluminum.

So, in many applications, the performance of a bonded-fin heat sink compares with that of a good extruded heat sink. The advantage of the bonded fin's construction method is that its heat-sink base can hold much better dimensional tolerances, and, if the heat sink is very wide, the bonded fin's cost is comparable with that of an extrusion. In addition, fins taller than 3 in. are available, and their height can vary along the length of the heat sink. This height variation can even extend to leaving gaps in the fins where fins are not needed. In this case, bonded-fin technology can offer potential savings over an extrusion because it isn't necessary to remove any fin material.

Convoluted fins are basically thin folded fins (Fig 4). Add a base plate and a shroud for mechanical stability, and you have lightweight heat sink. These heat sinks pack a lot of heat-transfer surface area into a small space, but at the price of air-pressure drop. You need to duct the airflow into the heat sink to get good performance from the sink. As with the bonded fins, the fan cannot move as much volume because of the pressure drop. A typical fan curve (Fig 5) shows that as pressure increases, airflow decreases. Another consideration is that the convoluted fins don't perform well in natural convection, which is a potential problem if the system fan fails.

A fan sink is a heat sink with a fan mounted on it (Fig 6). For ICs, the fan usually mounts on top. For the fan to perform at its rated capacity, leave some headroom for air supply to the fan (check the manufacturer's specifications.) A fan sink can help in situations in which you have insufficient local airflow. However, if you can put your fan in a place where there is fresh, cool air, do so. This air is better than potentially recirculated warm air. If you are buying a fan sink, get the fan with the smallest hub diameter. Air cannot go under the hub, so the hub heats up. This hot spot is usually in the center, right over the chip. The fan hub heats, reducing the fan's life expectancy. So, don't skimp on the fan cost: Buy the best. Also, make sure the system has an overtemperature sensor.

The main purpose of an interface material is to fill the tiny air gaps between two dry surfaces. The flatter and smoother the surfaces, the thinner and softer the interface material required to fill the gaps. Because perfectly smooth and flat surfaces don't exist, you almost always get better performance from an interface material than from a bare metal-to-metal interface.

Key characteristics of an interface material are thermal performance, compliance, adhesive properties, and reworkability. In some cases, electrical conductivity is also important. The materials include thermal compounds, solid-filled polymers, epoxies, and tapes. If your application is critical, weigh the costs of applying the material against the thermal performance you need. For example, if you must remove the heat sink from the component, choose a solid material. To get the best performance with the lightest weight, select a thermal compound. If the surface on which the heat sink must sit is uneven, a thicker conformable polymer is necessary to make up for height variations. For best performance, choose a polymer filled with a thermally conductive powder (alumina or metal, for example). Finally, if you use a clip, material adhesive properties are unimportant.

Whatever material you choose, make sure that you can install it without air bubbles, because the point of using an interface material is to eliminate air pockets. Some materials are available preinstalled from heat-sink vendors, saving you the cost of dispensing or cutting to size and then applying it in your shop.

The key features of any attachment method are ease of heat-sink installation and removal, the ability of the assembly to withstand shock and vibration tests, and thermal performance. Thermal performance depends on the quality of the interface, and the attachment method affects the interface if external pressure on the heat-sink assembly is absent.

The clip is by far the most popular heat-sink attachment method. The stamped clips fit over the heat sink between the fins and attach to holes in the board or to protruding features on the IC component or socket. Clips apply a preload to the interface material, which generally helps thermal performance by eliminating air pockets. Make sure that the clip holds well enough that a heat sink doesn't dislodge during shock and vibration. Clips are also removable and relatively simple to install and sometimes help to align the heat sink.

The one drawback of using a clip is that you must design in the attachment features. Also, once you complete this design, you probably don't want to change it for the life of your product. Sockets with multiple attachment features and clip standardization by manufacturers address that concern, however.

For smaller heat sinks, a pressure-sensitive adhesive or epoxy that attaches the sink to the component case works well. Some exceptions, however, include gold-plated, copper-tungsten heat spreaders and plastics that are cast in molds using release agents. The gold and the release agents inhibit the adhesive action.

If you use all surface-mount components with low-power dissipations, you may want to consider a stamped, surface-mount heat sink. You can use this type of heat sink with pick-and-place machines, and the solder reflows in the same cycle as the components.

Although all of these physical characteristics are important, the most important characteristic of a heat sink is its thermal performance. To determine this performance, first identify component-temperature requirements. Then, find out the power dissipation. Typically, you don't know a component's power dissipation until after you test it, so it's best to work with some safety factor. Another important factor contributing to thermal performance is airflow. Remember that taller fins, a larger base, and more air all mean better heat-dissipation ability (within limits).

Manufacturers specify many heat sinks and interface materials in terms of theta, which is the thermal resistance in degrees Celsius per watts. The thermal resistance is analogous to electrical resistance: The potential is temperature difference, and the "current" is the heat flow (power dissipation) in watts. Thus, you can relate theta, the temperature difference, and power dissipation with the thermal analog of Ohm's law.

Note a few important points when using the thermal Ohm's law. First, know the location of the reference temperatures: component case, heat-sink base, ambient directly upstream, or outside the system. Second, you must know the power dissipation. However, this power is not necessarily the total power of the package but that which normally comes through the heat sink. For good heat sinks and poor alternate thermal paths (for example, spreading into the board and then into the air), the heat sink does dissipate most of the heat. In this case, you can use the total power dissipation. In other cases, though, using the total power dissipation gives you only an estimate that depends on the setup of the experimental measurements. To make matters worse, theta sometimes depends on the dissipated power. This dependence is especially common at low airflows, in which radiation can significantly affect performance.

To find the temperature difference for a heat sink or interface, multiply theta by the dissipated power. This temperature difference for a heat sink is usually the maximum sink temperature minus the surrounding ambient temperature. In this case, theta would carry the subscript "s-a," denoting sink-to-ambient temperature difference. For interface materials, the temperature difference is from one side of the interface to the other—from the case to the sink (thus the subscript "c-s"), for example. The air-exit temperature is a lower limit on the component-case temperature, so it's a good feasibility check for your airflow numbers.

Airflow is often the biggest uncertainty and the most important influence on heat-sink thermal behavior. Heat-sink performance is a strong function of local air speed between the fins. The air would usually rather flow around than through the heat sink. If other components are nearby, they partially block that escape path and force more air through the heat sink. What gets into the fin area then runs into flow resistance and goes up and out of the heat sink. To complicate matters further, trying to channel the air too much causes high flow impedance. Also, your fan moves less total air. It's a good idea to check out the fan curve for your air mover and to find its operating point through experimental testing. This testing doesn't have to be involved or expensive; a foam-core mockup or a closely related mechanical structure puts you into the right range. If the experimental approach isn't feasible, stick with estimation.

To estimate the airflow, start with the fan curve for the fan you select. (Fig 5 is one example.) Select a fan that fits your geometry and stay with it until you find that it won't do the job. The fan curve in Fig 5 shows static pressure on the y-axis and airflow on the x-axis, with a general negative slope. For minimum noise, operate in the right third (high cubic ft/minute) of the fan's operating range.

Measuring how much air the fan moves involves knowing the system impedance curve: a roughly quadratic curve starting at (0,0) and going up to the right. The system operating point is where the fan curve and impedance curve intersect. To find the impedance curve, obtain a point by flow-friction calculations, thermal modeling, or experimental measurements. If you lack the facilities for these strategies, consult a knowledgeable expert.

To illustrate the importance of knowing the airflow, look at a typical heat-sink performance curve (Fig 7), which is actually two curves. The curve with positive slope is for natural convection. For this curve, the left y-axis (heat-sink temperature rise) and the lower x-axis (heat dissipated) apply. Natural-convection performance depends so strongly on the dissipated power that it doesn't make sense to calculate a theta. So, you can read the temperature rise directly from the graph.

For the curve with a negative slope, the upper x-axis (air velocity) and the right y-axis (thermal resistance) apply. The curve is steeper for lower airflows and flattens as you get into the upper airflow range. Most applications are in the low airflow range, from around 100 to 400 lfm. This is the same range for which the performance is most sensitive to airflow.

A word of caution is in order here. An airflow of 200 lfm in one case may give different performance than 200 lfm in another case. It depends on where you measure the air speed in the tests and on the existence of "bypass," any parallel airflow paths. Typically, the heat-sink manufacturer tests the heat sink in a wind tunnel. The air-speed measurements are on average far upstream of the test setup. The velocity in the fin cage can differ significantly from that measurement, especially if the fin spacing is dense and there is wide bypass space around the heat sink. If your bypass is bigger than the wind tunnel, which you may not know, the heat sink doesn't perform as well. If your bypass is smaller than the wind tunnel, the heat sink should perform better. So, don't be surprised if the performance in your application differs from that of the catalog value.

Another word of caution about catalogs: If the performance trends in the catalog don't make sense, question them. Misprints do happen, but most vendor application engineers know about them. Use the catalog information as a guide, but keep your common sense and double-check everything. The bottom line is that you still need to do some work to make sure the heat sink you choose works for you.

Thermal modeling can help remove much of the guesswork. Using airflow measurements from your airflow mockup and some modeling, you can quickly identify the trouble spots. Once you do, you can easily modify the model to test various sensible alternatives and identify the most promising one. The advantage of this strategy is that it significantly cuts lab time, because it requires fewer iterations of the prototype and test setup. In addition, the insight you get from the model helps you make better design decisions. If you cannot do the modeling yourself, get a competent consultant to do it for you. Remember, however, the model is only as good as the inputs you give it.