Thermoelectric Cooler Driver Design Guide

A thermoelectric cooler moves heat rather than generating cold, and the amount it moves depends on the current through it in a way that is close to linear. That makes the drive circuit a precision current source rather than a power switch, and it makes the ripple, the polarity and the thermal path all part of the same design problem.

How a Thermoelectric Cooler Works

A thermoelectric cooler, often called a Peltier module, consists of many semiconductor couples connected electrically in series and thermally in parallel between two ceramic plates. Current through the couples carries heat from one plate to the other, so one side becomes cold and the other hot, and reversing the current reverses the direction of heat flow.

The heat pumping rate is proportional to the current, while the resistive heating inside the module rises with the square of the current. The difference between the two is the net heat removed, which is why the efficiency falls as the current rises and why a module driven far beyond its rating can end up heating the object it is supposed to cool.

Because the module is a series string of semiconductor junctions, its electrical behaviour is not resistive. The voltage across it rises with current and also with the temperature difference it is maintaining, so a driver that regulates voltage will produce a current that changes as the thermal load changes. Regulating current instead makes the control loop predictable.

Why Current Control Matters

The cooling effect follows the current, so the control loop should regulate current rather than voltage. A voltage controlled drive has an operating point where the module current is set by the temperature difference and by the resistance, and that operating point moves as the load changes, which makes the loop gain vary over a wide range.

Current control also protects the module. A short circuit or a thermal overload that reduces the module resistance would draw far more current from a voltage source than from a current source, and the current limit is what keeps the module inside its rating. Most driver designs therefore sense the module current and close the loop on that signal.

The thermal design matters as much as the electrical one. The heat removed from the cold side plus the electrical power dissipated inside the module must both leave through the hot side heat sink. If the heat sink cannot reject that total, the hot side rises, the temperature difference falls and the whole assembly reaches a new equilibrium in which the object is warmer than expected.

Driving a Module from a Switching Supply

A linear driver is simple and produces almost no ripple, but it dissipates the difference between the supply and the module voltage as heat, which is unacceptable in a cooling application. A switching driver dissipates far less, at the cost of a ripple current that has to be controlled.

The switching topology is usually a buck converter operating in continuous current mode, with the module as the load. The inductor value sets the ripple current, and the sense resistor in series with the module provides the feedback for a current mode control loop. A synchronous output stage is preferable at low output voltages because the freewheel diode drop would otherwise be a significant fraction of the output.

The converter has to handle a load that is partly inductive and partly resistive, and whose impedance changes with temperature. The compensation of the control loop should be evaluated at both extremes of the load rather than at a nominal condition, because the loop gain varies with the module operating point. A loop that is stable cold can oscillate when the module is hot.

Current Ripple and Its Effects

current ripple in a thermoelectric cooler has three consequences. It produces a small oscillating component in the heat pumping, which averages out; it produces resistive losses in the module that increase with the square of the ripple; and it couples into the temperature sensor and its wiring, which is much closer to the noise than the module itself.

Ripple is usually specified as a fraction of the direct current, and a figure below ten percent is a reasonable target for a cooling application. The inductor value, the switching frequency and the output capacitance all influence it, and the inductor is normally the parameter used to set it because increasing the switching frequency increases the switching loss.

The module is also a filter of sorts, because its thermal mass averages the heat pumping over seconds. That is why a ripple that would be unacceptable in a measurement circuit is tolerable here, and why the design effort belongs in keeping the ripple out of the sensor path rather than in making it vanishingly small.

Direction Control with an H Bridge

Many applications need both heating and cooling, which means the current through the module must be reversible. An H bridge provides that, either by driving the module directly or by driving a pair of switching cells that reverse the polarity of a single ended converter output.

Reversal should always pass through a controlled ramp rather than an abrupt change, because the module is a large capacitive and inductive load and its mechanical structure is bonded with solder. Some drivers include a brake phase that allows the current to decay through the low side devices before the opposite pair is enabled, which avoids a current spike at the transition.

Where the direction changes frequently, the switching losses and the module stress accumulate. A design that heats and cools alternately within seconds is a different proposition from one that changes direction once a day, and the driver and the module rating should both be selected with the actual cycle in mind.

Thermal Cycling and Reliability

thermal cycling is the dominant wear mechanism in a thermoelectric assembly. Every time the module changes temperature, the solder joints between the couples and the ceramic plates experience a shear stress because the materials expand differently. Modules that cycle between hot and cold thousands of times develop cracks that increase the resistance and eventually open the circuit.

Reducing the temperature swing per cycle lengthens the life more effectively than any other change. Where the application allows it, holding the module at a moderate temperature between operations rather than letting it return to ambient removes a large part of the swing. A soft ramp at the start and end of each operation has the same effect and costs nothing but software.

The mounting hardware contributes as well. A module clamped with a defined and even pressure transfers heat well and keeps the joints under a uniform load, while one held by a single screw concentrates the stress. The thermal interface material and the flatness of both surfaces should be specified together with the clamping arrangement, as described in our guide to PCB thermal management.

Sensing and Control Loop Design

The temperature sensor should be attached to the object being cooled rather than to the module, because that is what the loop is trying to control. A sensor bonded to the cold plate measures the module rather than the payload, and the resulting loop will overshoot when the thermal load changes.

The loop itself is slow. A thermoelectric assembly has a thermal time constant of tens of seconds, so a proportional integral controller with a modest bandwidth is the right choice, and derivative action is usually counterproductive because it amplifies sensor noise. The current loop inside the temperature loop should be much faster, which keeps the two loops from interacting.

Protection belongs in the driver rather than in the loop. A firmware loop cannot react quickly enough to a module that has failed short, so the hardware should include a current limit, an over temperature shutdown and a check that the sensor is still connected. A disconnected sensor usually reads as an extreme value, and the loop would otherwise drive the module to its limit.

Layout and Thermal Constraints

The power stage should be laid out with the same care as any switching converter: a small switching loop, a short return path and a local input capacitor. The module current is continuous and can be several amperes, so the traces and the connector have to be sized for it, and the sense resistor should be a four terminal part where the accuracy of the current loop depends on it.

Copper is the first heat sink. The switching devices dissipate heat, and the board also receives heat conducted from the hot side of the assembly, so the thermal path from the driver to the enclosure should be considered as part of the design. The general rules for that kind of layout are described in converter layout and routing, and they apply here with an additional thermal load.

The sensor wiring is the most sensitive part of the assembly. A thermistor or a thermocouple bonded to the payload runs close to a switching node and to the module leads, and the ripple current induces noise in it. Route the sensor wiring as a twisted pair away from the power path, filter it at the amplifier input, and treat the noise problem with the methods described in EMI suppression design.

FAQ

Can I drive a Peltier module from a fixed voltage supply? It works, but the current then depends on the temperature difference and on the module resistance, which makes the cooling effect unpredictable. A current controlled drive is the better choice.

How much ripple is acceptable? Around ten percent of the direct current is a reasonable target. More ripple adds resistive loss in the module and couples into the temperature sensor, while less ripple costs inductance and switching loss.

Why does my cooler work well when new and poorly after a year? Thermal cycling degrades the internal solder joints and raises the module resistance. Reducing the temperature swing per cycle and reviewing the clamping pressure are the two most effective remedies.