Assessing Reliability of Thermoelectric Coolers
Thermoelectric coolers can be used for applications that require heat removal ranging from milliwatts up to several thousand watts. However, there has always been a general axiom in thermoelectrics: the smaller the better. A thermoelectric cooler makes the most sense when used in applications where even the smallest vapor compressor system would provide much more cooling than necessary. In such cases, a thermoelectric cooler can provide a solution that is smaller, weighs less and may be more reliable than a comparatively small compressor system. They have traditionally been used in applications such as temperature-controlled test fixtures, cooling small electronic enclosures and gas dehumidifiers that are placed before sensors.
However, despite the smaller-is-better axiom, the trend in recent years has been toward larger thermoelectric systems. Power supplies have become less expensive, which has pushed down the cost of a complete thermoelectric system (cooler, power supply and temperature controller) so that higher-power systems now are more attractive. Systems with capacities in the 200 to 400 W range are becoming more common although they still are not nearly as commonplace as smaller systems with cooling capacities below 100 W.
Thermoelectric systems are highly reliable provided they are installed and used in an appropriate manner. The specific reliability of thermoelectric coolers tends to be difficult to define because failure rates are highly dependent upon the particular application. Thermoelectric modules that are at steady state (constant power, heat load, temperature, etc.) can have mean time before failures (MTBFs) in excess of 200,000 hr. However, applications involving thermal cycling show significantly worse MTBFs, especially when thermoelectric coolers are cycled up to a high temperature. With thermal cycling, a more appropriate measure of reliability is not time but rather number of cycles.
Several things can cause poor reliability in a cooling system, but good design techniques can help alleviate them.
NemesesExpansion and contraction are the nemeses of thermoelectric modules. All materials expand or contract as they are heated or cooled, with different materials expanding or contracting at different rates. (The rate of expansion is given by the material property called the coefficient of thermal expansion.) Consequently, as the cold side of a module gets colder, it shrinks, and as the hot side gets hotter, it expands. This flexes the thermoelectric elements and their solder junctions. Furthermore, because the module is constructed of several different materials, there is added stress simply because the materials themselves are expanding and contracting at different rates.
After repeated thermal cycling, the solder junctions within the module fatigue, and the electrical resistance increases. Cooling performance is reduced, and eventually the module becomes inoperable. The failure point is thus a function of operating temperature, the amount of temperature cycling, and how much degradation the particular system can tolerate before performance becomes unacceptable.
All thermoelectric modules, regardless of manufacturer, experience the same stresses of operation, but how they tolerate these stresses is a question of "build" quality. It is essential to select a product with strong solder junctions. The best way to design around stresses is to use a larger number of small modules rather than a smaller number of large modules. Reducing the size of the modules reduces the overall expansion and contraction. If the system must undergo thermal cycling, the temperature range of the cycle and the number of cycles should be minimized as much as possible to keep the reliability as high as possible. Also, modules can be customized to better handle thermal cycling if required.
Only when a thermoelectric module is combined with a heat sink does it actually become a cooling assembly capable of effectively removing heat.
The Nemeses ReturnExpansion and contraction also can kill a cooling assembly. Many people want to solder or epoxy modules to a heat sink. This usually is not a good idea. If a module is soldered or epoxied to a heat sink, there is a zero-tension point where there is no internal stress resulting from mismatches in the coefficient of thermal expansion. The zero-tension point occurs at the temperature where the solder or epoxy becomes rigid and solidifies, and this is generally at some temperature that is different from the operating temperature. So, if you were to solder a module to a heat sink, enough stress to damage the module can be induced just by letting the module and heat sink cool back down to room temperature.
If an epoxy is used to bond the module, the zero-tension point may be shifted to room temperature, but this still does not solve the problem. Once power is applied, the rigid bonding induces thermal stress as the typical copper or aluminum heat sink expands at a different rate from the ceramic substrate of the module. The bond line between the module and heat sink also is stressed. This can be particularly troublesome because the bond could potentially fail at local spots. The module could overheat at these local spots, which would exacerbate the problem. Each of these scenarios is very similar to a bimetallic spring used in a household thermostat, but in this case, the parts are not made to flex.
If the entire cooling assembly (heat sink, cold sink and modules) is bonded together, there can be even bigger trouble. Remember, the entire cold sink is contracting and the entire heat sink is expanding. Distorting the modules, which are only 0.1181" (3 mm) thick on average, is the only way to make up the difference between the two. Because the heat sink and cold plate are many centimeters in length, it does not take a large change in temperature to crack the solder junctions and brittle thermoelectric material.
For these reasons, most experts do not recommend soldering (or epoxying) the module to its heat sink or cold sink. If you do solder (or epoxy) the modules, it is suggested that you use only one very small module in the cooler and then thermal-cycle the complete assembly to make sure you get adequate lifetimes. Of course, if your cooler goes into a ballistic missile, and it only has to work once, there is a good chance you will meet your reliability goal.
The proper design technique is to use a thermally conductive paste between the module, the heat sink, and the cold sink, and clamp them all together using stainless steel screws. This allows the components to "float" within the assembly. Some sort of physical retainer also may need to be used to keep the modules from migrating away from their original positions as the parts expand and contract over the lifetime of the cooling assembly.
Thermoelectric modules exhibit relatively high mechanical strength in compression mode but comparatively low tensile and shear strength. Consequently, a thermoelectric module should not be used to support weight that would subject it to tension or shear stress. This is another good reason to clamp the module between two plates as opposed to using solder or epoxy to secure it. Furthermore, multiple modules in an assembly should share common heights to within 0.025 mm. Otherwise, uneven clamping forces could crack a module or create poor thermal contact between the module and its heat sink.
Water, Electricity Don't MixMoisture should not be allowed to enter the inside of a thermoelectric module or it will cause a reduction in cooling performance and corrosion of the electrically conductive materials. Moisture will try and make its way inside a module anytime the module is cooling to a temperature below the dewpoint. Epoxy potting compounds (sealants) placed around the perimeter of the module generally eliminate any moisture problems within the module. Silicone rubber, or RTV, is a poor sealant, though, and can actually make the situation worse. Silicone rubber is permeable to water vapor but impermeable to liquid water. So, the water vapor gets inside the module, condenses, and then is trapped inside.
That said, having a sealed module would do you no good if the wires on the outside of the module get wet, corrode and fall off. The entire area surrounding the module also should be sealed. A good gasketing material or other sealing system that blocks the water vapor must be used if the system will operate below dewpoint. The potting around the modules serves as the secondary barrier, providing extra insurance against corrosion.
Exposure to high temperatures should be minimized as much as possible to extend reliability. As temperatures rise, there is an ever-increasing tendency for solder and copper to diffuse into the thermoelectric material and destroy its cooling properties. Going well past the module's rated temperature can even cause the internal solder to melt. Standard modules are rated for a maximum of 176oF (80oC). High-temperature modules are rated for 302oF (150oC). There are also 392oF (200oC) modules. However, these temperature limits are somewhat arbitrary. All modules, regardless of manufacturer, will be affected by operation at high temperatures.
Temperature control methods also have an impact on thermoelectric module reliability. Linear or pulse-width-modulated (frequency of at least 400 Hz) control should always be chosen to ensure higher reliability. The thermostatic or on/off type controller should be avoided because they cause thermal cycling, even if it is just at the module level and not at the object being cooled.
Always test, then test again. There are numerous application parameters and conditions that affect reliability: module size and quality, cooler assembly, mounting methods, ambient temperature and humidity, temperature control systems and techniques, and temperature cycling. These factors can combine to produce failure rates ranging from extremely low to very high. The failure point also is a function of how much degradation the particular system can tolerate before performance is unacceptable. Unfortunately, there is no magic equation that takes into account all of the parameters and predicts the reliability of a design with absolute certainty. It is specific to each application. Therefore, it is necessary to test the reliability of the specific system in applications where reliability is critical. PCE
Sidebar:There are practical limits to the individual size of a module or cooling assembly. Micro-modules, for example, are more expensive to produce because they are less suitable for automated processing. For larger modules, coefficients of thermal expansion and costs tend to limit thermoelectric modules to within a certain physical footprint.
Checking into Size, Power
For cooling assemblies, the minimum size might be limited by the minimum requirements needed to provide sufficient heat sinking. The maximum size is limited by the requirements of the mounting plates. If the plates get too large, then it becomes too difficult to maintain sufficient surface flatness. Usually, when more cooling capacity is required than what the typically largest size cooler can provide, multiple coolers are used rather than one giant device. In general, the largest individual cooler has a footprint of approximately 10 x 7" (254 x 177 mm).
Powering a thermoelectric cooler also has its special considerations. Ideally, thermoelectric coolers should operate on purely direct current for the best performance. However, a ripple factor of 10 percent will result only in 1 percent degradation in the temperature across the module. Most power supplies have better filtering than that, so ripple is not likely to be a concern.
Care should be taken not to overpower the cooler, which at best case actually reduces the cooling performance, and at worst case could lead to exceeding the temperature ratings and damaging to the cooler.
Treat the module's maximum rated voltage and current as if they were for informational purposes only. These parameters are only valid when the module is attached to a perfect heat sink operating at one specific temperature. In real-world operation, when maximum efficiency is desired, the applied power is typically one-third to two-thirds of its rated maximums. When maximum cooling is desired, the applied power is typically two-thirds to three-quarters of its rated maximums.
If a temperature controller is used, it should be of the linear type or the pulse-width-modulated (frequency of at least 400 Hz) type. These controllers bring the cooler down to the desired temperature and steadily hold it there, minimizing any detrimental effects of temperature cycling. Do not use thermostatic or on/off controllers.