Beat the Heat
Any good embedded system engineer knows that heat is the enemy of reliability. With that in mind, it’s no surprise that an increasing amount of engineering mindshare is focusing on cooling electronic systems. Here, George examines the math and science around cooling, and looks at several cooling technologies—from cold plates to heat pipes.
Every engineer worth his salt considers heat the reliability killer. And as new systems cram more functionality at higher speeds into ever smaller packages, it is no wonder that a whole engineering discipline has arisen specializing in electronics packaging and cooling. In this article we’ll review some of the essentials of this engineering expertise.
The reduction of reliability of typical integrated circuits (ICs) is illustrated in Figure 1. The red trace represents the reliability of a digital CMOS, and the blue trace of an analog IC plotted as the Mean Time To Failure (MTTF) hours with respect to the operating temperature. I also plotted the respective failure rates λ (Greek letter Lambda) per operating hours, where λ is the reciprocal 1/ MTTF. The MTTF is more intuitive for the majority of readers, but reliability engineers prefer to think in terms of the failure rate. The orange trace is for the digital, and the violet trace is for the analog ICs. Clearly, the loss of reliability is significant.
For the purpose of this analysis, we assume the product reliability is affected by its operating temperature only. We do not consider parameter drifts, overloading, insufficient derating or otherwise poor design, potentially leading to inferior performance or a failure. By treating reliability as a function of the operating temperature only, we can appreciate how it can be improved by cooling.
THREE WAYS TO COOL
Electronic components can be cooled by conduction, convection and radiation. Conduction is the most common method of heat transfer, occurring through a physical contact—where the heat is transferred through direct molecular collision. The amount of transferred heat depends on the temperature gradient, cross-section of the material, the length of the travel path and physical properties of the material. The transfer ceases when the two materials are at the same temperature. The rate of conduction is shown by the equation:
where Q is the heat transfer rate, k is the thermal conductivity, A is the heat transfer area, ΔT the temperature difference and d is the thickness of the barrier. A typical example of conduction is heat transfer from a semiconductor to a heatsink.
Convection occurs when a fluid substance (air, water and so forth) absorbs energy from a source and removes it. For example, the air heated by a heatsink rises and carries the heat energy away. Therefore, a typical heatsink employs conduction on one side and convection on its other side.
where Q is the heat transfer rate, hc the heat transfer coefficient, A is the heat transfer area, Ts temperature of the surface and Tf the temperature of the fluid—which can be the air or a liquid.
Thermal radiation emanates from all hot objects and is carried away through media—such as vacuum, air and so forth—as is defined by the Stefan-Boltzmann law:
where P is the net radiated power, A is the radiating area, Tr the radiator temperature and Tc the temperature of the surroundings, e is emissivity and σ (Greek letter Sigma) the Stefan-Boltzmann constant. Radiation requires Blackbody and is rarely efficient. The most popular cooling methods used by the industry are listed in Table 1.
Natural convection by heatsinks has been around for a long time. It is easy to implement, robust, highly reliable due to the absence of moving parts and inexpensive. All that said, it is limited in its use by its ability to dissipate no more than about 300 W of heat—and that is quickly making it inadequate for the high-power densities of present-day equipment.
When natural convection cannot do the job, Cold Plate (Base Plate) cooling can dissipate up to about 500 W. Cold plates are essentially heatsinks that are a part of the chassis, extending over the entire electronic assembly and exposed to the ambient temperature (Figure 2). Forcing air or a cooling liquid through the cold plate can increase its cooling capacity manifold.
By adding internal fans or using an external source of air, the cooling effectiveness of a system can be increased up to approximately 2,000 W. A desktop PC is a good example of such a cooling method. Remember, though, that to cool the entire assembly evenly without creating hot spots requires careful placement of fans and internal components. The unfortunate side effect of forced-air cooling is noise.
Forced-air cooling capacity can be doubled by adding to the enclosure liquid conduction cooling. There, enclosure walls are cooled by liquid circulating through tubing embedded in the chassis. The flow channels in the chassis walls are optimized for the available liquid pressure and the internal heat source distribution.
Air flow-through (AFT) and liquid flow-through (LFT) methods provide a sealed environment, which also serves to protect critical processing modules from contamination by dust, humidity and other factors. Each processing board is physically separated from the others and has its own inlet and outlet for the cooling medium. To maximize the cooling efficiency, cooling channels bring the coolant to the high-power components first, and the coolant is chilled in an external heat exchanger. The exchanger can be air-to-air or liquid-to-air. On some aircraft, for example, the fuel tank is used as a coolant for the heat exchanger.
A single specific component may sometimes need effective cooling in excess of what the existing cooling method can provide without addition of a costly, often complicated mechanical channeling method described above. That’s when thermoelectric cooling with a Peltier module may be the right solution. Peltier modules—also called thermoelectric coolers (TECs)—are like electronic heat pumps with no moving parts. TECs are available from a variety of vendors, including Digi-Key and Mouser Electronics. At some point the amount of heat they generate cannot be dumped easily without significantly increasing the ambient temperature. Meanwhile, TECs consume considerable amounts of energy, which may not be an acceptable solution for the given system.
Enter heat pipe technology—a two-phase cooling system popular in laptop computers (Figure 3). To begin, a tubular heat pipe is filled with a cooling fluid. Its thermal conductivity is several hundred times that of a copper rod of the same dimensions, but it weighs less. The heat pipe transfers the heat from a component, such as a microprocessor, to an external heat sink radiator. It consists of a vacuum-tight pipe, a wick and a working fluid. The evacuated pipe is filled with a small amount of the working fluid—just enough to saturate the wick—and sealed. In consequence, an internal equilibrium of vapor and liquid develops.
Heat from the cooled component warms up the evaporator end of the pipe. This upsets the equilibrium by generating more vapor, which is at a slightly higher pressure. The higher-pressure vapor travels to the condenser, which is at a lower temperature, where it condenses, giving up its latent heat. The condensed fluid then travels back to the evaporator through the capillary action of the wick made of sintered powder. It works even against gravity.
The heat pipe operation is passive—there are no moving parts. The pipe is energized by the heat extracted from the cooled device, requiring no additional source of energy. It is simple to design and manufacture with the MTTF estimated well in excess of 100,000 hours. There are various coolant fluids in use, water and ethanol (alcohol) being the common ones. Fluorocarbon FC-77, for example, is suitable for high performance operation from about -95°C to +200°C (-139°F to +392°F) ambient. The heat pipe, however, is not a heat pump like the TEC. Its condenser must always be at a lower temperature than the evaporator for it to work.
We’ve touched on the mainstream methods for keeping your electronic controllers cool. Driven by ever increasing cooling demands, new cooling methods are being developed every day for embedded systems. Large systems—such as communications centers—are a different story. They are often furnished with independent air or liquid cooling generation systems, and are often located in fully air conditioned buildings.
For detailed article references and additional resources go to:
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • MARCH 2019 #344 – Get a PDF of the issue