How to achieve higher power thermoelectric cooling using Peltier devices

summary

This article provides the thermoelectric cooler (TEC) concepts that must be understood before designing higher power TECs, explains the key Peltier characteristics that limit the cooling capabilities of thermoelectric coolers, and shows how to design around these limitations. Some driver examples illustrate the conditions required to control higher power TECs. Also included are issues that may prevent existing designs from achieving their intended cooling capabilities.

Introduction

TECs use Peltier modules to cool an object or provide accurate temperature control of an object and are used in a variety of applications. They are ideal for laser diode coolers1,2, microprocessor cooling, polymerase chain reaction (PCR) systems, and medical applications such as tomography, cardiovascular imaging, magnetic resonance imaging (MRI), and radiation therapy. Many applications such as laser diode temperature control use small, low-power TECs in the 5 W to 15 W range. Their drivers may run from a 5 V rail and provide 1 A to 3 A.

But what if we need more power? How do we do it? What should we focus on and what options do we have? We looked at it from two perspectives. The first case is that we already have a working TEC, but it's not enough and we need to increase the power by 10% to 20%. The second case is to build a higher power TEC from scratch. How much cooling power can we get from the Peltier device? What should we use to drive it?

Before we get started, let's understand a few key Peltier concepts.

Maximum heat absorption

The maximum heat absorption (Qc) of a Peltier module will be listed in the data sheet, but it applies to the case where ΔT is zero. ΔT is the temperature difference between the hot and cold ends of the Peltier. When the hot and cold ends are the same temperature, Qc will be as stated in the data sheet. However, it will decrease linearly as ΔT increases until a certain point where Qc = 0. This point, also called the maximum ΔT, varies widely, but a typical value for a single-stage module might be around 70°C. See Figure 1 for a general example.

Figure 1. Relationship between heat absorption and Peltier temperature difference.

Let’s say we want to keep the hot side of the Peltier at room temperature, +22°C, and we want the cold side to be at –5°C. The maximum current of the Peltier is 9 A, so we plan to use a 7 A driver. In our example graph, a temperature difference of 27°C at 7 A would give us a capability of 41 W. However, all interfaces have thermal resistance, so as heat flows from the Peltier, through the heat sink, and into the room environment, a temperature gradient will occur. This means that the hot side of the Peltier cannot be at room temperature, 22°C. Let’s say the hot side is at 30°C. This gives us a 35°C hot/cold temperature difference. Referring to Figure 1, following the 7 A line to the 35°C ΔT point indicates that our heat removal capability will be around 30 W—even if we purchased a 100 W Peltier!

Self-heating

Another important Peltier concept is that the module will generate a lot of self-heating when it is operating. The self-heating may be twice the heat absorbed from the target. For example, when absorbing 25 W of heat from the target, the Peltier may generate an additional 50 W of heat. Therefore, the hot end heat sink must be able to dissipate 75 W of heat.

Retrofitting existing TEC systems

In the first case, we have an existing TEC and just need to increase the cooling capacity slightly, for which we may have some issues to consider. Several obvious problem areas include the hot side temperature of the TEC, the thermal resistance of the TEC component interface, the voltage ripple across the Peltier device, ΔT, and the insulation of the component.

It is recommended to check the temperature of the hot end first, see Figure 2. One key point to remember from Figure 1 is that the small delta between the cold and hot ends of the Peltier is critical. As the temperature difference increases, the ability of the Peltier to extract heat from the target decreases.

One way to quickly get an idea of ​​the hot-end temperature is to check the heat sink temperature when the TEC is near maximum power. Thermistors work well simply by using a thermocouple or sending the measurement to a microprocessor. See “Thermistor-Based Temperature Sensing Systems—Part 1: Design Challenges and Circuit Configurations” and “Thermistor-Based Temperature Sensing Systems—Part 2: System Optimization and Evaluation” for two excellent thermistor articles. 3,4 If the hot-end heat sink temperature is significantly above room temperature, a larger heat sink and/or fan may be needed.

Unfortunately, the quick check above doesn't tell us anything about the Peltier-to-heatsink interface. This interface can be difficult to access, so disassembly of the device is usually required. Thermal paste is often used at this interface, and we want to check it to determine if there are air pockets that could interfere with heat conduction. Since air is a poor conductor of heat (0.026 W/(mK)), the job of thermal paste is to eliminate air pockets. Don't use a very thick layer, though, because at 0.2 W/(mK) to 0.3 W/(mK), thermal paste is not a good conductor either, although metal types may be in the 4 W/(mK range. However, this paste still performs 10 times better than air. By comparison, aluminum is 200 W/(mK), PCB copper is about 380 W/(mK), PCB FR4 is about 0.3 W/(mK) to 0.8 W/(mK), water is 0.6 W/(mK), and glass is about 1.0 W/(mK).

Note that it is possible to reach a point where increasing the current through the Peltier will have the opposite effect than intended, and will make the cold junction warmer! This is because the Peltier may be close to its maximum ΔT, and increasing the current will make the hot junction warmer due to inadequate heat dissipation. When the hot junction warms, it pushes the cold junction upward.

Also, check how the voltage ripple on the TEC can reduce the efficiency of the Peltier. The ripple should not exceed 10%, but 5% or less is recommended. Reducing the effective series resistance (ESR) of the load capacitor is probably the safest change. However, no matter what changes are made, whether increasing the frequency, increasing the output capacitance, or increasing the inductance, care needs to be taken to prevent affecting the efficiency of the switch and its control stability.

New Design

The first thing one might think about for a new high power design is whether to use a Peltier module or a Peltier assembly. The module itself is the Peltier, bismuth telluride sandwiched between a ceramic substrate and the hot end (+ end), with two wires soldered to it. In this case, it is up to the customer to design the heat sink and thermal interface. An assembly, on the other hand, consists of a Peltier module with a heat sink already attached. A typical setup might include two heat sinks and two fans with wiring leading out to the connection ports. Heat sinks come in different flavors, such as air-cooled, water- or glycol-cooled, and direct-attached, and may also include some type of frame for attaching to a cabinet or other equipment. The customer simply connects a power source to the fans and can then focus on the driver design.

Either way, whether starting from a module or components, if you want to build a high power TEC, you need to make trade-offs and choices. For example, for roughly the same power, the current and voltage of various Peltier modules (TEC modules) can vary greatly. It may be more advantageous to use multiple modules in an application, or you can choose multi-stage modules to increase ΔT. To drive higher power modules, ADI offers the LT8722 and the new LT8204 full-bridge power chips. Next, let's take a closer look at these issues.

The first thing to realize when choosing a Peltier module is that their current/voltage tradeoffs can vary greatly. For example, in the case of modules available in the 95 W to 105 W range, the resistance can vary from 0.34 Ω to 4.4 Ω. Furthermore, the maximum specs for a 95 W module at 27°C are 19 A and 7.7 V, while another 105 W module has a maximum spec of 7.6 A and 21.2 V at 27°C. While they are not exactly the same power, the point is that there can be a tradeoff between current and voltage, which in turn determines your driver requirements.

Alternatively, you can use multiple modules, however, they must be connected electrically in series because their resistance changes with temperature. This makes current sharing between parallel devices a challenge. Of course, with series devices, the voltage drop increases and a higher voltage driver will be required. However, the Peltier devices in electrical series will still have the parallel function of heat dissipation. If higher voltages are not possible, but two modules are still needed, each module must be driven by its own driver. However, a single temperature feedback can be used for both modules.

Figure 2. Simplified diagram of air-to-air TEC assembly.

Figure 3. LT8722 application circuit.

Another option is to use multi-stage modules. These modules consist of one to five modules stacked together by the manufacturer. In other words, the heat dissipation will be in series, so the total ΔT increases and it can cool to a lower temperature. However, this is not a one-size-fits-all approach. Remember that the hot side of each module must dissipate the heat absorbed from the target as well as the self-heating. Therefore, the cold side of the next connected module must transfer the self-heating and the target heat from the first device, and the third module in the series must be able to dissipate the heat from the target as well as the self-heating from all three devices. This additional temperature capability comes at the cost of significantly increased heat dissipation. Multi-stage modules often look like a pyramid because the modules farthest from the target need to dissipate the most heat and must be larger.