A "Temperature Control Guide" is presented to you!

To control the temperature, a loop consisting of a thermistor, a thermoelectric cooler (TEC), and a TEC controller is required. The resistance of the thermistor changes proportionally with temperature (inversely or directly, depending on the thermistor type), and when configured as a voltage divider, it can be used to convert temperature into a voltage. The TEC controller compares this feedback voltage with a reference voltage representing the target temperature and then controls the current flowing through the TEC, thereby adjusting the amount of heat transferred by the TEC.

Figure 1. Temperature control system of laser module

The general block diagram of the above system is shown in Figure 1. The laser diode, TEC, and thermistor are located inside the laser module. The TEC controller ADN8833 or ADN8834 reads the feedback voltage from the thermistor and provides the drive voltage to the TEC. The thermal loop is monitored and controlled using a microcontroller. Note that the thermal loop can also be built in analog circuits. The ADN8834 has two built-in zero-drift chopper amplifiers that can be used as PID compensators.

This article will explain the components of laser diode thermal control systems in telecommunication systems and introduce the key specifications of the main components. The purpose of this article is to explain the various design considerations from a system perspective and provide designers with a global guide to build a high-performance system with good temperature control accuracy, low loss, and small size.

The thermoelectric cooler consists of two surface ceramic plates with P-type and N-type semiconductor arrays placed alternately between them, as shown in Figure 2.

Figure 2. TEC module with heat sink

When current flows through these conductors, heat is absorbed at one end and released at the other end; when the current is reversed, the heat transfer also reverses. This process is called the Peltier effect. The carriers in the N-type semiconductor are electrons, so its carriers and heat flow from the anode to the cathode. The opposite N-type semiconductor has hole carriers, and heat also flows in the opposite direction.

Take a pair of PN semiconductors and connect them with a metal plate as shown in Figure 3; when current flows, heat will be transferred in one direction.

Figure 3. Peltier effect: heat flow through a PN semiconductor pair

Changing the polarity of the DC voltage changes the direction of heat transfer, with the amount of heat transferred being proportional to the voltage amplitude. Thermoelectric cooling is widely used for thermal conditioning in telecommunication systems due to its simplicity and robustness.

When selecting a TEC module, many factors in the system need to be considered, such as ambient temperature, object target temperature, heat load, supply voltage, and physical characteristics of the module. The heat load must be carefully evaluated to ensure that the selected TEC module has sufficient capacity to pump heat out of the system to maintain the target temperature.

TEC module manufacturers usually provide two performance curves in their data sheets. One curve shows the heat transfer capacity at different temperature differences (ΔT) within the supply voltage range, and the other curve shows the cooling/heating current required for different combinations of supply voltage and ΔT. Designers can estimate the power capacity of the module to determine whether it can meet the needs of a specific application.

In order to use TEC to compensate temperature, the TEC controller should be able to generate a reversible differential voltage based on the feedback error and provide appropriate voltage and current limits. Figure 4 is a simplified system block diagram of ADN8834. The main functional blocks include temperature detection circuit, error amplifier and compensator, TEC voltage/current detection and limit circuit, and differential voltage driver.

Figure 4. Functional block diagram of the single-chip TEC controller ADN8834.

The TEC controller outputs a differential voltage so that the current through the TEC can either remove heat from the object connected to the TEC or smoothly change to the opposite polarity to heat the object. The voltage driver can be linear mode, switch mode, or hybrid bridge. Linear mode drivers are simpler and smaller, but have poor efficiency. Switch mode drivers have good efficiency—up to over 90%—but require additional filtering inductors and capacitors at the output . The ADN8833 and ADN8834 use a hybrid configuration with a linear driver and a switch mode driver, which reduces the number of bulky filtering components by half while maintaining high efficiency performance.

The voltage driver design is critical to the controller because it accounts for most of the power dissipation and board space. An optimized driver stage helps minimize power losses, circuit size, heat sink requirements, and cost.

Figure 5 shows the impedance of a negative temperature coefficient (NTC) thermistor over temperature. Because it is temperature dependent, it can be connected as a voltage divider to convert temperature to voltage.

Figure 5. NTC impedance vs. temperature curve

Typical connections are shown in Figure 6. As RTH varies with temperature, VFB also varies.

Figure 6. NTC thermistor connected as a voltage divider to convert temperature to voltage

By adding an Rx in series with the thermistor, the temperature-voltage transfer function can be linearized relative to VREF, as shown in Figure 7. It must be tightly coupled to the laser inside the module housing to isolate it from external temperature fluctuations so that it can accurately detect temperature.

Figure 7. VFB vs. Temperature

The analog thermal feedback loop consists of two stages, formed by two amplifiers, as shown in Figure 8. The first amplifier accepts the thermal feedback voltage (VFB) and converts or regulates this input into a linear voltage output. This voltage represents the object temperature and is fed into the compensation amplifier, where it is compared with the temperature setpoint voltage to produce an error voltage proportional to the difference between the two. The second amplifier is typically used to build a PID compensator, which consists of a very low frequency pole, two different higher frequency zeros, and two high frequency poles, as shown in Figure 8.

Figure 8. Thermal feedback loop diagram using the two chopper amplifiers inside the ADN8834.

The PID compensator can be determined mathematically or empirically. To mathematically model the thermal loop, accurate thermal time constants of the TEC, laser diode, connectors, and heat sink are required, which are not easy to obtain. It is more common to tune the compensator empirically. By assuming a step function at the temperature set point and varying the target temperature, the designer can adjust the compensation network to minimize the settling time of the TEC temperature.

An aggressive compensator will react quickly to thermal disturbances but can easily become unstable, while a conservative compensator will settle more slowly but will tolerate thermal disturbances with less likelihood of overshoot. A balance must be struck between system stability and response time.

Sometimes, even if the PID compensator is designed properly, steady-state error still exists. The following are several factors that cause this error.

• TEC thermal power budget: When designing this system, the TEC and supply voltage are the first things to be selected. However, since the heat load is not easy to estimate, the selection may be incorrect. In some cases, if the maximum power is applied to the TEC and the target temperature is still not reached, it may mean that the thermal power budget is not enough to handle the heat load. Increasing the supply voltage or picking a TEC with a higher power rating can solve this problem.

  
  

• Reference consistency: References drift over temperature and time, which is usually not a problem for closed thermal loops. However, especially in digital control systems, the references of the TEC controller and the microcontroller may drift differently, causing errors that are not noticeable to the compensator. It is recommended that the same reference be used for both circuits, with the voltage with higher drive capability overwriting the other.

  
  

• Temperature Sensing: To minimize temperature errors, it is important to accurately sense the load temperature. Any error from the feedback will enter the system and the compensator will not correct for this error. Using a high-precision thermistor and an auto-zero amplifier can avoid errors. The placement of the thermistor is also important. Make sure it is mounted to the laser so that it can read the actual temperature we are trying to control.

Most of the power dissipation in the TEC controller is consumed by the driver stage. In the ADN8833/ADN8834, the power dissipation of the linear driver is directly derived from the input-to-output voltage drop and the load current. The losses of the switch-mode driver are more complex and can be roughly broken down into three parts: conduction loss, switching loss, and transition loss. The conduction loss is proportional to the RDSON of the FET and the DC resistance of the filter inductor. The conduction loss can be reduced by selecting low resistance components. The switching loss and transition loss are highly dependent on the switching frequency. The higher the frequency, the higher the loss, but the size of the passive components can be reduced. To achieve the optimal design, a careful trade-off between efficiency and space must be made.

The switch mode driver in ADN8833/ADN8834 switches at 2MHz. The fast PWM switching clock edge contains a wide frequency spectrum, which will generate voltage ripple at the TEC terminal and generate noise in the whole system. Adding proper decoupling and ripple suppression capacitors can reduce the noise and ripple.

For the buck topology commonly used in switch-mode power supplies, the ripple on the supply voltage rail is mainly caused by the discontinuous current chopped by the PWM FET. Using multiple SMT ceramic capacitors in parallel can reduce the ESR (equivalent series resistance) and locally decouple the supply voltage. At the switch-mode driver output node, the voltage ripple is caused by the current ripple of the filter inductor. To suppress this ripple, multiple SMT ceramic capacitors should be used in parallel from the driver output to ground. The ripple voltage is mainly determined by the product of the capacitor ESR and the inductor ripple current: ΔV_TEC = ESR × ΔI_LUsing multiple capacitors in parallel can effectively reduce the equivalent ESR.

Designing a TEC controller system for laser diodes in telecommunications systems is a complex task. In addition to the challenges of thermal accuracy, the package size is usually very small and the power dissipation tolerance is also very low. In general, a well-designed TEC controller should have the following advantages:

• Precise temperature regulation

• high efficiency

• Small board size

• Low noise

• Current and voltage monitoring and protection