Ultra-low voltage energy harvester uses thermoelectric generator to power battery-free wireless sensors

The proliferation of ultra-low power wireless sensor nodes for measurement and control, coupled with the use of new energy harvesting techniques, is enabling fully autonomous systems powered by local ambient energy rather than batteries. Using ambient or "free" energy to power wireless sensor nodes is attractive because it can supplement or even eliminate the need for batteries or wires. This is an obvious advantage when replacing or servicing batteries is inconvenient, expensive, or dangerous.

Furthermore, the complete elimination of power wires makes it easy to scale monitoring and control systems on a large scale. Energy harvesting wireless sensor systems simplify installation and maintenance in areas such as building automation, wireless/automated instrumentation and predictive maintenance, as well as many other industrial, military, automotive and consumer applications.

The benefits of energy harvesting are clear, however, effective energy harvesting systems require the use of intelligent power management schemes to convert the tiny amounts of free energy into a form that can be used by wireless sensor systems.

It all comes down to duty cycle
The average power consumption of many wireless sensor systems is very low, making them prime candidates for powering with energy harvesting techniques. Many sensor nodes monitor slowly changing physical quantities. Measurements can be made infrequently, and measurements do not need to be sent often, so the sensor nodes operate at a very low duty cycle. Accordingly, the average power requirement is also low.

For example, if a sensor system requires 3.3V/30mA (100mW) when awake, but is only active for 10ms per second, the average power required is only 1mW, assuming that the sensor system current drops to a few μA when not active between transmit bursts. If the wireless sensor only samples and transmits once a minute instead of every second, the average power drops to less than 20μW. This difference in power requirements is important because most forms of energy harvesting provide very small steady-state power (usually only a few mW, sometimes even just a few μW). The lower the average power required for the application, the more likely it is to be powered by harvested energy.

Energy Harvesting Sources
The most common energy sources available for harvesting are vibration (or motion), light, and heat. The transducers used for all of these energy sources have the following common characteristics:

• Their electrical output is unregulated and unsuitable for direct use to power electronic circuits
 • They may not be able to provide a continuous and uninterrupted power supply
 • They tend to produce only very low average output power (usually between 10μW and 10mW)

If such energy sources are to be used to power wireless sensors or other electronic circuits, judicious and careful power management must be done to account for the above characteristics.

Power Management: The Missing Link in Energy Harvesting Until Now
A typical wireless sensor system powered by harvested energy can be broken down into five basic building blocks, as shown in Figure 1. All of these building blocks, except the power management building block, have been available for some time. For example, microprocessors that run on just a few μW of power, as well as small, cost-effective radio frequency (RF) transmitters and transceivers that also consume very low power, are widely available. Low-power analog and digital sensors are also ubiquitous.

Figure 1: Typical wireless sensor block diagram

The missing link in implementing this energy harvesting system link has always been a power converter/power management building block that can operate from one or more common free energy sources. The ideal power management solution for energy harvesting should be small and easy to use, operate well when operating from the unusually high or low voltages generated by common energy harvesting sources, and ideally provide a good load match to the source impedance for optimal power transfer. The power manager itself must consume very little current when managing the accumulated energy and should produce a regulated output voltage using very few discrete components.

The LTC3108, available in a 3mm x 4mm x 0.75mm 12-pin DFN package or a 16-pin SSOP package, solves the energy harvesting problem for ultra-low input voltage applications. The device provides a compact, simple and highly integrated monolithic power management solution that operates with input voltages as low as 20mV. With this unique capability, the LTC3108 can power wireless sensors using a thermoelectric generator (TEG) and harvest energy from temperature differences (ΔT) as small as 1ºC. Using a small, off-the-shelf (6mm x 6mm) step-up transformer and a small number of low-cost capacitors, the device provides the regulated output voltage required to power today’s wireless sensor electronics.

The LTC3108 uses a small step-up transformer and an internal MOSFET to form a resonant oscillator that can operate from very low input voltages. With a transformer step-up ratio of 1:100, the converter can start up with input voltages as low as 20mV. The transformer's secondary winding feeds a voltage to the charge pump and rectifier circuit, which is then used to power the IC (via the VAUX pin) and charge the output capacitor. The output of the 2.2V LDO is designed to enter regulation first to power a low-power microprocessor as quickly as possible. The main output capacitor is then charged to the voltage set by the VS1 and VS2 pins (2.35V, 3.3V, 4.1V or 5.0V) and used to power sensors, analog circuits, RF transceivers, or even charge supercapacitors or batteries. When low-duty-cycle load pulses occur as wireless sensors operate and transmit data, the VOUT storage capacitor provides the burst energy needed. A switching output (VOUT2) is also provided that can be easily controlled by the host to power circuits that do not have a shutdown or low power sleep mode. The device has a power good output that is used to alert the host when the main output voltage is close to its regulation value. Figure 2 shows the block diagram of the LTC3108. The LTC3108-1 version of the device is exactly the same as the LTC3108 except that it provides a different set of selectable output voltages (2.5V, 3.0V, 3.7V or 4.5V).

Figure 2: Block diagram of the LTC3108

Once VOUT is charged and in regulation, the harvested current is directed to the VSTORE pin to charge an optional large storage capacitor or rechargeable battery. If the energy harvesting source is intermittent, this storage element can be used to maintain regulation and power the system. Output voltage sequencing during power-up and power-down can be seen in Figure 3. A shunt regulator on the VAUX pin prevents VSTORE from being charged above 5.3V.

Figure 3: Voltage sequencing during power-up and power-down

Using a standard square TEG with a side length of 40mm, the LTC3108 can operate with a ΔT as low as 1°C, making it suitable for a wide range of energy harvesting applications. At higher ΔTs, the LTC3108 will be able to provide a higher average output current.

Basic Principles of Thermoelectric Generators
A thermoelectric generator (TEG) is a thermoelectric module that uses the Seebeck effect to convert a temperature difference across the device (and the resulting heat flowing through the device) into a voltage. The reverse process of this phenomenon (called the Peltier effect) is to generate a temperature difference by applying a voltage and is commonly used by thermoelectric coolers (TECs). The polarity of the output voltage depends on the polarity of the temperature difference across the TEG. If the hot and cold ends of the TEG are reversed, the output voltage will change polarity.

TEG consists of a pair or couple of N-type doped and P-type doped semiconductor chips connected electrically in series and sandwiched between two thermally conductive ceramic plates. The most commonly used semiconductor material is bismuth telluride (Bi2Te3). Figure 4 shows the mechanical construction of TEG.

Figure 4: TEG structure

Some manufacturers differentiate TEGs from TECs. When marketed as TEGs, this usually means that the solder used to assemble the thermocouples inside the module has a higher melting point, allowing it to operate at higher temperatures and temperature differentials, thus providing higher output power than a standard TEC (which is typically limited to a maximum temperature of 125ºC). Most low-power energy harvesting applications do not encounter high temperatures or temperature differentials.

TEGs come in a variety of sizes and electrical specifications. Most common modules are square, with each side ranging from 10mm to 50mm in length and a thickness of 2mm to 5mm.

How much voltage a TEG will produce for a given ΔT (proportional to the Seebeck coefficient) is controlled by many variables. Its output voltage is 10mV/K to 50mV/K temperature difference (depending on the number of thermocouples) and has a source resistance of 0.5Ω to 5Ω. In general, for a given ΔT, the more series thermocouples a TEG has, the higher its output voltage. However, increasing the number of thermocouples also increases the series resistance of the TEG, resulting in a larger voltage drop when loaded. Manufacturers can compensate for this by adjusting the size and design of individual semiconductor chips to maintain low resistance while still providing a higher output voltage.

Load Matching
In order to extract the maximum available power from any voltage source, the load resistance must match the internal resistance of the source. This is illustrated in Figure 5, where a voltage source with an open circuit voltage of 100mV and a source resistance of 1Ω or 3Ω is used to drive a load resistor. Figure 6 shows the power delivered to the load as a function of the load resistance. In each curve, it can be seen that the power delivered to the load is maximized when the load resistance matches the source resistance. However, it is also important to note that when the source resistance is lower than the load resistance, the power delivered may not be the maximum possible, but may be higher (1.9mW in this case) than when a higher source resistance drives a matched load (0.8mW in this case). This is why choosing the TEG with the lowest resistance provides the maximum output power.

Figure 5: Simplified schematic of a voltage source driving a resistive load

Figure 6: Output power of a power supply as a function of load resistance

The LTC3108 presents a minimum input resistance of approximately 2.5Ω to the input supply. (Note: this is the input resistance of the converter, not the IC itself.) This is in the middle of the range of most TEG source resistances, providing an excellent load match for near-optimal power transfer. The LTC3108 is designed so that as VIN decreases, the input resistance increases (as shown in Figure 7). This feature allows the LTC3108 to adapt well to TEGs with different source resistances.

Figure 7: LTC3108 input resistance vs. VIN (using 1:100 turns ratio)

Since the converter has a fairly low input resistance, it will draw current from the supply regardless of the load. For example, Figure 8 shows that with a 100mV input, the converter draws about 37mA from the supply. This input current should not be mistaken for the 6μA quiescent current (drawn from VAUX) required by the IC itself to power its internal circuitry. Low quiescent current is most significant when starting from very low voltages or operating from a storage capacitor.

Figure 8: LTC3108 input current vs. VIN (using 1:100 turns ratio)

Most thermoelectric
module manufacturers do not provide data on output voltage or output power versus temperature difference, which is exactly what thermal energy harvester designers want to know. The two parameters that are always provided are VMAX and IMAX, which are the maximum operating voltage and maximum operating current (when driven in a heating/cooling application) for a particular module.

When selecting thermoelectric modules for power generation, a good rule of thumb is to select the module with the largest (VMAX • IMAX) product for a given size. This will generally provide the highest TEG output voltage and lowest source resistance. There is a caveat to this rule of thumb, which is that the heat sink must be sized according to the size of the TEG. Larger TEGs require larger heat sinks for best performance. It is important to note that when manufacturers provide resistance parameters, this refers to AC resistance because it cannot be measured in the traditional way using DC current (DC current induces a Seebeck voltage, giving an erroneous resistance reading). Figure 9 is a graph of the power output of the LTC3108 for 13 different TEGs (fixed ΔT = 5ºC) versus the (VMAX • IMAX) product for each module. As can be seen, the LTC3108 generally delivers higher output power when the VI product is higher.

Figure 9: LTC3108 output power vs. TEG with different V and I products

Figure 10 shows the output voltage and maximum output power capability of a 30mm square TEG over a ΔT range of 1°C to 20°C. Within this ΔT range, the output power ranges from a few hundred μW to tens of mW. It should be noted that this power curve is derived assuming an ideal load match and no conversion losses. Finally, the output power available after boosting to a higher voltage using the LTC3108 will be lower than the value shown in the figure due to power conversion losses. The LTC3108 data sheet provides several graphs of the output power available under a variety of different operating conditions.

Figure 10: Open circuit voltage and maximum power output of a typical TEG

The size of the TEG required for a given application depends on the minimum available ΔT, the maximum average power required by the load, and the thermal resistance of the heat sink used to keep one end of the TEG at ambient temperature. The maximum power output of the LTC3108 is between 15µW/K-cm2 and 30µW/K-cm2, depending on the transformer turns ratio and the specific TEG selected. Table 1 lists some recommended TEG device models.

Table 1: Recommended TEG devices

Thermal Considerations
When placing a TEG between two surfaces at different temperatures, the "open circuit" temperature difference before the TEG is added is higher than the temperature difference across the TEG once it is in place. This is because the TEG itself has a fairly low thermal resistance between its ceramic plates (typically 1ºC/W to 10ºC/W).

Consider the following example, a large machine operating in an environment with an ambient temperature of 25ºC and a surface temperature of 35ºC. When a TEG is connected to this machine, a heat sink must also be added to the cooler (ambient) side of the TEG, otherwise the entire TEG will heat up to nearly 35ºC, eliminating any temperature difference. It is important to remember that the electrical output power is generated by the heat flowing through the TEG.

In this case, the thermal resistances of the heat sink and TEG determine what portion of the total temperature difference (ΔT) exists across the TEG. A simple thermal model of this system is shown in Figure 11. Assuming the thermal resistance of the heat source (RS) is negligible, if the thermal resistance of the TEG (RTEG) is 2ºC/W and the thermal resistance of the heat sink is 8ºC/W, then the ΔT across the TEG is only 2ºC. With the temperature across the TEG being only a few ºC, its output voltage is low, and the LTC3108’s ability to operate from ultra-low input voltages becomes important.

Figure 11: Thermal resistance model of TEG and heat sink

Note that larger TEGs generally have lower thermal resistance than smaller TEGs due to their increased surface area. Therefore, in applications where a smaller heat sink is used on one side of the TEG, the ΔT across the larger TEG may be smaller than the smaller TEG, and thus may not necessarily provide more output power. In all cases, a heat sink with the lowest possible thermal resistance should be used to maximize electrical output by maximizing the temperature difference across the TEG.

Choosing the Optimal Transformer Turns Ratio
For applications that can accommodate a higher temperature differential (i.e., higher input voltage), a transformer with a lower turns ratio (e.g., 1:50 or 1:20) can be used to provide higher output current capability. As a rule of thumb, if the minimum input voltage is at least 50mV when loaded, a 1:50 turns ratio is recommended. If the minimum input voltage is at least 150mV, a 1:20 turns ratio is recommended. All of the turns ratios discussed are commercially available in Coilcraft parts (see the LTC3108 data sheet for more information, including specific part numbers). The curves in Figure 12 show the output power capability of the LTC3108 over a range of temperature differentials for two different transformer step-up ratios and two different TEG sizes.

Figure 12: LTC3108 output power vs. ∆T for two TEG sizes and two transformer turns ratios (VOUT = 5V)

Pulse Load Applications
A typical wireless sensor application powered by a TEG is shown in Figure 13. In this example, at least a 2ºC temperature differential is available across the TEG, so a 1:50 transformer step-up ratio is chosen to achieve the highest output power over a 2ºC to 10ºC ΔT range. When using the TEG shown (a 40mm square device with a 1.25Ω resistance), the circuit is able to start up and charge the VOUT capacitor with a temperature differential as low as 2ºC. Note that a large decoupling capacitor is connected across the input of the converter. Providing good decoupling between the input voltage and the TEG minimizes input ripple, maximizes output power capability, and starts up at the lowest possible ΔT.

Figure 13: Wireless sensor application powered by a TEG

In the example shown in Figure 13, the 2.2V LDO output powers the microprocessor, while VOUT is set to 3.3V using the VS1 and VS2 pins to power the RF transmitter. The switched VOUT (VOUT2) is controlled by the microprocessor to power the 3.3V sensor only when needed. The PGOOD output signals the microprocessor when VOUT reaches 93% of its regulated value. To maintain operation when the input voltage is not present, a 0.1F storage capacitor is charged from the VSTORE pin in the background. This capacitor can charge all the way up to the 5.25V clamp voltage of the VAUX shunt regulator. If the input voltage supply is lost, energy is automatically provided by the storage capacitor to power the IC and keep VLDO and VOUT regulated.

In this example, the COUT storage capacitor is sized to support a total load pulse of 15mA for 10ms duration, allowing a 0.33V drop in VOUT during the load pulse, according to the following formula. Note that IPULSE includes VLDO and VOUT2 as well as the load on VOUT, but the available charging current is not included as it can be very small compared to the load.

COUT(μF) = IPULSE (mA) • tPULSE (ms) / dVOUT

Considering these requirements, COUT must be at least 454μF, so a 470μF capacitor was selected.

Using the TEG shown, operating at a ΔT of 5°C, the average charge current that the LTC3108 can provide at 3.3V is approximately 560μA. Using this data, we can calculate how long it takes to first charge the VOUT storage capacitor and how often the circuit can pulse it. Assuming that the load on VLDO and VOUT during the charge phase is very small (relative to 560μA), the initial charge time for VOUT is:

tCHARGE = 470μF • 3.3V / 560μA = 2.77s

Assuming that the load current between transmit pulses is very small, a simple way to estimate the maximum allowable transmit rate is to divide the average output power available from the LTC3108 (in this case 3.3V • 560μA = 1.85mW) by the power required during the pulse (in this case 3.3V • 15mA = 49.5mW). The maximum duty cycle that the harvester can support is 1.85mW / 49.5mW = 0.037 or 3.7%. The maximum pulse transmit rate is therefore 0.01 / 0.037 = 0.27s or about 3.7Hz.

Note that if the average load current (as determined by the transmit rate) is the maximum current the harvester can support, there will be no harvested energy left to charge the storage capacitor (if storage capability is required). Therefore, in this example, the transmit rate is set to 2Hz, leaving almost half of the available energy to charge the storage capacitor. In this case, the storage time provided by the VSTORE capacitor is calculated using the following formula:

tSTORE = 0.1F • (5.25V - 3.3V) / (6μA + 15mA • 0.01 / 0.5) = 637s

The above calculations include the 6μA quiescent current required by the LTC3108 and assume minimal loading between transmit pulses. In this case, once the storage capacitor reaches full charge, it can support the load for 637s at a 2Hz transmit rate, or a total of 1274 transmit pulses.

Ultra-low power applications with battery backup
Some applications may not have pulsed loads but may need to operate continuously. Traditionally, such applications are powered by a small primary battery (such as a 3V lithium coin cell). If the power requirements are low enough, these applications can be powered continuously using thermal energy harvesting, or they can use thermal energy harvesting to greatly extend the battery life, thereby reducing maintenance costs.

Figure 14 shows an energy harvesting application driving a continuous load with a backup battery. In this example, all electronics are powered entirely from the 2.2V LDO output, with a total current consumption of less than 200μA, and the LTC3108 can continuously power the load as long as there is at least a 3ºC temperature difference across the TEG. Under these conditions, there is no load on the battery. When the available harvested energy is insufficient, the 3V lithium battery seamlessly “takes over” and powers the load.

Figure 14: Energy harvester with battery backup

Energy Storage Alternatives
For applications where a rechargeable battery is used as an alternative to the primary battery for backup or energy storage, the diode in Figure 14 can be removed and the lithium battery replaced with a rechargeable nickel battery or a lithium-ion battery (including the new rechargeable thin-film lithium battery). If a rechargeable nickel battery is used, its self-discharge current must be less than the average charge current that the LTC3108 can supply. If a lithium-ion battery is used, additional circuitry is required to protect it from overcharge and overdischarge. Another storage alternative is a supercapacitor with a rated voltage of 5.25V, such as the Cooper-Bussman PB-5ROH104-R. Supercapacitors have the advantage of having more charge/discharge cycles than rechargeable batteries, but the disadvantage is much lower energy density.

Thermal Harvesting Applications Require Auto-Polarity
Some applications, such as wireless HVAC sensors or geothermally powered sensors, present another unique challenge to energy harvesting power converters. Such applications require the energy harvesting power manager to be able to operate not only from very low input voltages, but also in either polarity, since the polarity of the ∆T across the TEG can change. This is a particularly tricky problem, and at voltages of tens or hundreds of mV, a diode bridge rectifier is not an option.

The LTC3109 is uniquely suited to overcome the challenge of harvesting energy from sources of either polarity. The LTC3109 operates with input voltages as low as ±30mV using a transformer with a 1:100 step-up ratio. The LTC3109 has the same functionality as the LTC3108, including an LDO, a digitally programmable output voltage, a power good output, a switching output and an energy storage output. The LTC3109 is available in a 4mm x 4mm 20-lead QFN package or a 20-lead SSOP package. Figure 15 shows a typical example of the LTC3109 in an auto-polarity application. The converter’s output current vs. VIN curve, shown in Figure 16, illustrates that the device works equally well with input voltages of either polarity.

Figure 15: Wireless sensor node powered by auto-polarity energy harvester

Figure 16: Output current vs. VIN curve of the converter in Figure 15

The LTC3109 can also be configured for unipolar operation, using a single transformer (similar to the LTC3108) to accommodate applications that require the lowest possible startup voltage and highest possible output current. The circuit shown in Figure 17 can start up with only 15mV, generated with the TEG shown at a temperature differential of less than 1°C. At a 10°C temperature differential, it can provide a regulated 5V (at 0.74mA), delivering 3.7mW of regulated steady-state output power. This is nearly twice the output power of the LTC3108 under the same conditions, as shown in Figure 18.

Figure 17: A unipolar converter using the LTC3108 can start up with only 15mV

Figure 18: Comparison of LTC3108 and LTC3109 output power

Note that in a unipolar configuration, the LTC3109 presents a load resistance of approximately 1Ω to the TEG, so it is important to select a TEG with very low source resistance for good load matching, otherwise there is no advantage to using the LTC3109 in a unipolar configuration. The TEG used in this example has a nominal source resistance of 1.0Ω for optimal power transfer.

Conclusion
The LTC3108 and LTC3109 can uniquely operate from input voltages as low as 20mV, or at very low voltages of either polarity, providing a simple and effective power management solution that enables thermal energy harvesting to power wireless sensors and other low power applications using common thermoelectric devices. Available in 12-pin DFN or 16-pin SSOP packages (LTC3108 and LTC3108-1) and 20-pin QFN or SSOP packages (LTC3109), these products offer unprecedented low voltage capability and high integration to minimize solution footprint. The LTC3108, LTC3108-1 and LTC3109 can seamlessly interface with existing low power building blocks to support autonomous wireless sensors and extend battery life in critical backup battery applications.