When electrical engineers use the phrase “power management,” most of us tend to think about the various DC power rails via assorted converters, regulators, and other power-handling and power-transforming functions. However, there is much more to power management than just those functions. All power sources give off heat due to inefficiency, and all components must dissipate some heat.
Thus, power management is also concerned with thermal management, in particular how the dissipation of power-related functions affects the thermal design and heat buildup. Further, even if the component and system continue to function within specification, temperature increases induce changes in performance as component parameters drift. This may lead to eventual system malfunction if not outright failure. Heat also shortens component life and decreases mean time to failure, which makes it a long-term reliability consideration as well.
There are two thermal-management perspectives that designers must review:
- The “micro” view, where an individual component is in danger of overheating due to excess self-dissipation, but the rest of the system (and its enclosure) are within acceptable bounds.
- The “macro” situations, where the entire system is too hot due to aggregate heat accumulation from many sources.
One design challenge is to determine how much of the thermal management problem is micro versus macro, and to what extent the two are related. Obviously, a hot component—even one that exceeds its allowed limits—will contribute to heating of the system, but that does not necessarily mean that the overall system will be too hot. It does, however, mean that the excessive heat of that one component must be managed and reduced.
One question to always keep in mind when discussing thermal management and using phrases such as “getting the heat out” or “removing the heat,” is this: To where is that heat being moved? A cynic might say the designer’s challenge is to find somewhere to dump that heat and thereby make his or her problem into someone else’s problem.
While that view is indeed somewhat cynical, there is an element of truth to it. The challenge is to get the heat to a cooler location where it will not have adverse effects on the system. This could be an adjacent part of the system and enclosure, or it could be outside the enclosure altogether (possible only if it is cooler externally than internally). Also keep in mind one of the laws of thermodynamics: Heat will only travel from a higher-temperature location to a lower-temperature one, unless some sort of active pumping mechanism is used.
Thermal Management Solutions
Thermal management is governed by basic principles of physics. Heat transfer—here, in the cooling mode—occurs in a combination of three ways: Radiation, conduction, and convection (Figure 1):
Figure 1: There are three mechanisms by which heat is transferred, and all are often involved in a given situation to differing extents. (Source: Kmecfiunit / CC BY-SA 4.0)
In simplest terms:
- Radiation refers to heat that is carried away by electromagnetic radiation (mostly infrared) and can occur in a vacuum. In most applications, it is not a major cooling factor; an exception is in the vacuum of space, where it is the only way to draw heat away from the spacecraft.
- Conduction is the flow of heat through a solid or liquid, without actual movement of the heat-conveying material (although liquids do flow, of course).
- Convection is the flow of heat that is carried along by a fluid medium, such as air or water.
For most electronic systems, achieving the desired cooling is a combination of conduction away from the immediate source of the heat, and then convection to carry that heat elsewhere. The design challenge is to combine various pieces of thermal-management hardware—that is, hardware in the original, non-electronic sense of the word—to effectively implement the conduction and convection needed.
There are three elements which are most commonly used: Heat sinks, heat pipes, and fans. Heat sinks and heat pipes are types of passive, self-powered cooling, which also includes naturally-induced conduction and convection air flow methods. By contrast, fans are a type of active, forced-air cooling.
Start With Heat Sinks
Heat sinks are aluminum or copper structures that draw heat from the source via conduction and expose it to air flow (and in some cases, to water or other liquid flow) for convection. They come in thousands of sizes and shapes, ranging from small, stamped metal wings which attach to a single transistor, (Figure 2), to large extrusions with many fins (or fingers) which intercept the convection airflow and transfer the heat to that flow, (Figure 3).
Figure 2: The simple sheet-metal Aavid Thermalloy 574502B00000G heat sink is designed to slide onto a TO-220 package transistor and features 21.2C/W thermal resistance; it measures about 10×22×19mm. (Source: Aavid Thermalloy)
Figure 3: These larger, extruded multi-finned heat sinks (M-C308, M-C091, M-C092) from Cincom are designed for larger ICs as a well as modules. The smallest is about 60×60 ×20mm high, while the largest one is 60×110×25mm high. (Source: Cincom Electronics)
One of the virtues of the heat sink is that it has no moving parts, no operating cost, and no failure mode. Once the appropriate-sized heat sink is attached to the source, the convection occurs naturally as the warmed air rises, thus starting and continuing an airflow. Therefore, it is essential when employing a heat sink to allow for unimpeded air flow from source inlet to outlet. Also, the inlet must be below the heat sink and the outlet above; otherwise, the heated air will stagnate above the heat source and make the situation worse.
Despite the heat sink’s ease of use, it does have some negative aspects. First, heat sinks that are sized to transfer large amounts of heat can get bulky, costly, and heavy. Also, they must be properly positioned, and thus can affect or constrain physical board layout. Their fins can get clogged with dirt and dust from the airflow, too, which greatly reduces effectiveness. They must be properly attached to the heat source so that the heat flows unimpeded from that source to the heat sink.
With so many heat sinks to choose from in terms of size, configuration, and other factors, making the choice can be overwhelming at first. Note that there are many general-purpose heat sinks, as well as those that are designed and sized for a particular IC, such as a specific processor or field-programmable gate array (FPGA) model.
There are also heat sink embodiments that are not discrete components. Some ICs use their pins or leads to conduct the heat away from their die and body to their PC board traces, which then function as heat sinks. Other ICs actually have a copper slug under their package; when it is soldered to the PC board, the slug acts as a pathway for removing heat from the die. This is a low-cost and effective way to conduct heat away, but it assumes that the rest of the PC board is cooler and that no other nearby components are also using the board for cooling. In effect, each device is trying to dump its excess waste heat into the neighbor’s yard, which is a zero-sum game.
Add Heat Pipes
Another important element in the thermal-management kit is the heat pipe (Figure 4). This passive component is as close to “something for almost nothing” as engineers get, as it moves heat from point A to point B without need for any sort of active forcing mechanism. In brief, a heat pipe is a sealed metal tube that contains a wick and a working fluid. The role of the heat pipe is to draw heat away from a source and convey it to a cooler location, but not to act as a heat dissipating sink itself. The heat pipe is used when there is not enough space at the source to place a heat sink or there is insufficient air flow. It acts as a highly efficient conduit to get the heat from the source to a place where it can be better managed.
Figure 4: This diminutive heat pipe from Wakefield-Vette (model 120231) measures just 6mm × 1.5mm diameter and is designed to convey heat loads up to 25W. (Source: Wakefield-Vette)
How does the heat pipe work? It’s simple and clever: It implements a phase change, which is a basic principle of thermal physics. The heat source turns the working fluid to vapor within the sealed tube, and heat-laden vapor carries that heat as it travels to the cool end of the heat pipe. At the cool end, the vapor condenses back into liquid, releasing its heat, and the fluid travels back to the warmer end. This gas-liquid phase-change sequence operates continuously and is powered only by the heat differential between the warmer and cooler ends.
Heat pipes come in many diameters and lengths, with most diameters between about one-quarter inch and one-half inch, with lengths of several inches to about a foot. As with a water pipe, the larger-diameter pipes have the capacity to convey more heat. When joined with a heat sink or other cooling apparatus at the cool end, it can solve the problem of getting heat away from a local hot spot that has impeded airflow.
Then Add a Fan
Finally, there are fans (Figure 5), which mark the first step away from the passive, self-powered cooling of heat sinks and heat pipes to active, forced-air cooling. Fans can be both problem solvers and a headache, and designers often have mixed emotions when using them.
Figure 5: A tiny fan, such as the 30mm diameter x 6.5mm deep ASB0305HP-00CP4 from Delta Electronics, operates from a single +5V pulse width modulator (PWM) signal and can deliver about 0.144m3/min (5ft.3/min) of airflow. It is driven by a PWM signal and includes a tachometer feedback signal. (Source: Delta Electronics)
Obviously, a fan adds cost, requires space, and increases the acoustic noise of a system. Also, as an electromechanical device, they are prone to failure, consume some energy, and affect overall system efficiency. However, they are often the only way to get sufficient airflow in many situations, especially when the airflow path is not straight, vertical, and unimpeded. Many applications use thermally controlled fans that run only if needed to reduce their rpm, thus cutting power use, and with blades that minimize noise at the optimum operating speed.
The key parameter that defines fan capacity is the airflow in linear or cubic feet of air per minute. Physical size is also an issue; obviously, a large fan running at lower rpm can produce the same airflow as a small fan at higher rpm, so there is a size/speed tradeoff. Some designs use internal baffles to direct the fan airflow past hot areas and heat sinks for optimum performance.
Models, Simulation Pull It All Together
The decision whether to use passive cooling alone or go to an active forced-air system is often a difficult one. A passive-only system is larger but more efficient and reliable, while a fan allows operation in circumstances that might otherwise be impossible using passive cooling alone.
Of course, there are cases where a passive system alone will be inadequate or impractical. An analogous situation is the management of engine heat in automobiles. The earliest cars with their small engines were passively cooled via fins on the tops of their cylinders as heat sinks. As engines became larger and the heat load increased, these fins became larger and unwieldy, so circulating fluid was added to draw heat away from the fins and bring it to a radiator as a heat sink, through which air would flow as the car moved. This, too, was a passive system. Eventually, though, as engines became even bigger, the passive approach was inadequate and the car would overheat unless it was moving. Therefore, a fan was added behind the radiator to force air through it, regardless of the car’s speed.
Modeling and simulation are critical to an effective thermal-management strategy to determine how much cooling is needed and how to achieve it. The good news is that this activity is far easier and more accurate than most other types of electronic modeling, such as for RF or electromagnetic fields with their parasitics and anomalies.
For micromodeling, the heat source and all thermal paths from it are characterized by their thermal resistance, which is determined by the material used, its mass, and size. This shows how heat will flow from the source and is the first step in assessing that a component that may be in thermal distress due to its own dissipation, such as a high-dissipation IC, MOSFET, and insulated-gate bipolar transistor (IGBT), and even a resistor. Vendors of these devices often supply thermal models that provide details of the thermal path from source to surface (Figure 6).
Figure 6: The mechanical model of the installed FET (left) is used to develop an equivalent thermal-resistance model (right), which is used for simulation of the device’s thermal situation. (Source: International Rectifier/Infineon)
Note that for some components, their various surfaces may be at different temperatures. For example, the underside of a die may naturally be hotter than the top-side cover of its package, so vendors may design the package to convey more heat to the top to make better use of a top-side heat sink.
Once the heat load represented by various components is known, the next step is macrolevel modeling, which can be both simple and complex. As a first-order approximation, the airflow that passes by the various heat sources can be sized to keep their temperatures below allowed limits. Basic calculations using air temperature, amount of unforced flow available, fan air flow, and other factors will give a rough idea of the situation.
The next step is a more-sophisticated modeling of the entire product and its package, using models of the various heat sources, their locations, the PC board, the enclosure surface, and other factors. This type of modeling is based on computational fluid dynamics (CFD) and can be quite accurate in showing the temperature at every location in the box (Figure 7).
Figure 7: Using computational fluid dynamics (CFD) analysis, the detailed thermal profile across a system or circuit board can be seen, as shown by this PC board with three major heat sources (red) and the flow of heat both left and right on the extended board. (Source: Texas Instruments)
By making “what if” adjustments, designers can see if more air is needed via larger air ports, determine if a different air-flow path is more effective, identify differences in using larger or different heat sinks, inquire about the use of heat pipes to move heat spots, and much more. These CFD modeling packages produce tabular data as well as false-color images of the thermal situation. Changes such as the effects of fan size, airflow, and placement are also easily modeled.
Finally, modelling should address two additional factors. First, there are issues of peak versus average dissipation. A component which dissipates 1W steady state has a different thermal impact than one that dissipates 10W but with a 10 percent intermittent duty cycle. The reason is that the associated thermal masses and heat flow will result in different thermal profiles even if the average heat dissipations the same. Most CFD applications can take this static versus dynamic reality into the analysis.
Second, component-level micro models must take into account imperfections in the physical attachment between surfaces, such as between the top on an IC package and its heat sink. If that meeting has minute gaps, there will be a higher thermal impedance in that path. For this reason, thin thermal pads are often used between these surfaces to enhance the path’s conductivity (Figure 8).
Figure 8: To minimize thermal impedance between an IC and its heat sink, often due to microscopic air voids, users can interpose a thermally-conductive but electrically-insulating pad, such as the AP PAD HC 5.0 thermal interface high-compliance silicon-based pad, with 5.0W/m-K resistance. (Source: Bergquist Company)
Conclusion
Thermal management is an essential aspect of power management that’s needed to keep components and systems within their temperature limits. Passive solutions begin with heat sinks and heat pipes, and may be enhanced with active cooling using fans. System modeling at the component level and the complete-product level allows designers to begin a first-order, approximate analysis of the cooling strategy. Further analysis using computational flow dynamics enables comprehensive insight into the complete thermal situation and the effect of changes in cooling tactics. All thermal management solutions involve tradeoffs in size, power, efficiency, weight, reliability, and, of course, cost, and they must be assessed with respect to project priorities and constraints
