Modern ultra-low-power microcontrollers have reached such a level of integration and processing efficiency that many applications no longer require traditional batteries. These applications include complex and often power-intensive wireless sensor networks that may involve sampling various sensors and communicating wirelessly.
By harvesting miniscule amounts of wasted energy from the environment, such systems are enabled with near infinite up-time without a battery as its primary power source. Not only does energy harvesting enhance current applications by eliminating their dependency on the battery, but it also enables entirely new applications that weren't feasible given the finite lifetime and size of batteries.
Ultra–low-power embedded processing
Similar to Moore's Law, which defines the trend of digital technology to double in transistor count every two years, an inverse trend occurs for power consumption. Roughly every 18 months, the power dissipation of digital systems is cut in half.
Advancements in power efficiency already had dramatic results for small, ultra-low-power microcontrollers (MCUs) specifically designed for battery-powered applications and have resulted in designs where battery life has exceeded 10 years. For typical ultra-low-power MCUs, it's common for standby current to be in the < 1 µA range and active current consumption in the ~200 µA/MIPS range. Since the clock rate of these MCUs is typically in the order of 25 MHz or less, the peak current consumption is always relatively small and can be powered with simple power supplies.
Power consumption of a given application is rarely characterized by a single MCU's current draw. Analog conversion circuitry, power regulation, and communication devices each play a part in the system and consume power even when they're not active. By integrating the functionality of each of the devices into a single device using a single low-power fabrication process, it's not only possible to significantly reduce the leakage current of the overall system, but by giving a single MCU control to disable peripherals that are not in use, power consumption can be reduced even further. A single, highly integrated device will typically consume less power than separate discrete solutions; a single device also simplifies the design and reduces the cost and area required for a given function.
Flexible power requirements for ultra-low-power microcontrollers not only reduce power consumption by allowing lower supply voltages than typical embedded processors with a fixed supply voltage, they also allow a wider variety of energy sources. For example, some ultra-low power MCUs support a wide input voltage range between 1.8 V to 3.6 V.1 By allowing a lower operating voltage, this type of MCU can reduce the overall power consumption of the system and use power at the much lower voltage levels provided by the energy harvester.
The problem with batteries
Traditional batteries, such as lithium-ion cells, have been the default source for power in portable electronics for decades; however, traditional batteries place hard restrictions on products' usability, lifetime, and cost of ownership. While processing power roughly doubles ever two years, battery technology advances at a much more sluggish pace. Historically, battery capacity has doubled every 10 years. In addition to the very slow growth in their energy capacity, traditional batteries have a limit to the total practical energy density they can provide. Present-day lithium-ion batteries, which are popular due to their high energy-to-weight, have an energy density of 150 to 200 Wh/kg [Watt-hour/kilogram]. Research has shown that it's possible to increase their energy density by tenfold within a few years; however, even if this is achieved, we must still consider practical safety concerns. Given improper use, batteries with extremely high energy densities can become dangerous, explosive devices.
For most battery-operated devices, the cost associated with owning and operating the device is rarely limited to the initial cost of manufacturing it. In the long term, replacing the battery can have a significant impact on the overall cost of ownership. This is especially critical in applications where battery replacement is impractical or has high labor costs associated with maintenance. Take for example water meters that must be buried underground. Accessing the water meter would require digging it up, which in colder climates might be one meter or deeper underground. Thanks to this unavoidable inaccessibility, the replacement cost of the battery could be in the $100 to $200 range per water meter.
Miniaturization of products has been an ongoing trend in most application spaces, but the driving force has come from consumer electronics and medical applications. For consumer products, the demand for smaller and sleeker devices has driven innovation for more highly integrated electronics given the finite amount of space that products are expect to take up.
While integration at the IC level has kept up with consumer demand, the power source isn't benefiting from miniaturization. The space allowed for batteries is shrinking, the lifetime that they're expected to operate is longer, and the amount of power they're expected to provide has also increased.
The requirements for batteries in modern electronics have far exceeded what can be delivered. Despite the challenges with traditional batteries, it's possible to maintain functionality with today's rechargeable batteries, or we may even forgo the battery altogether if we couple an ultra-low-power embedded processor with power supply that harvests energy from its environment.
A new class of applications?
In principle, energy harvesting has been around for thousands of years. The first waterwheels have been dated back to as far as the fourth century B.C. The waterwheel effectively harvested the energy from flowing water and transferred it to mechanical energy. Similarly, present-day wind farms or solar arrays all use the same principle of operation and usually provide power back to the main grid. These large-scale applications can be referred to as macro energy harvesting.
On the other hand, micro energy harvesting, which we will be focusing on, is the principle that enables small, autonomous devices to capture energy from the environment and store it. While micro and macro energy harvesting may be similar in principle, their scope and applications are radically different.
The portion of the system that harvests energy consists of two main parts:
• The component that converts the ambient energy from the environment.
• A means of storing the energy for later use by the application.
Although the rest of the system can be defined in an infinite number of ways and is dependant on the task at hand, energy-harvesting systems typically contain similar components given that they are ideal for sensor network applications. An ultra-low-power MCU will be the heart of the system and is responsible for the majority of the processing, sensing, and communication. The MCU will interface to any number of sensors to collect data from its environment and will usually transfer or receive data via a wireless transceiver. Since energy harvesting systems are completely untethered, they each act as autonomous systems. A typical block diagram is found in Figure 1 .
The sources of energy to harvest are similarly numerous, and more esoteric systems continue to be introduced. However, the most common sources for ambient energy are light, thermal, radio frequency (RF), and vibration. Each has unique advantages and disadvantages, and the specific harvesting technology is dependent on the application and the use case. Clearly, a device outfitted with a solar panel wouldn't benefit if it sits in a dark cave all day. The key to an energy-harvesting system is to take energy that is readily and predictably available and collect what would otherwise be wasted power.
The output power from an energy-harvesting circuit varies by several orders of magnitude based on the technology being used, its efficiency, size, and the environment it's in. For example, solar panels, which currently provide the highest output power of the commonly available harvesting technologies, would generate about 1,000W in full sunlight if it were 100% efficient and one square meter in size. In reality, however, the best solar panels available in 2009 are only ~41% efficient, and since a one-square-meter solar panel would be impractical for small, autonomous devices, a typical solar panel only a few centimeters in size will generate a few 100 milliwatts. Typical output power for real-world energy-harvesting technology is listed in Table 1 .
Given the relatively small amount of power that is directly output from micro-energy harvesters, few applications could run directly from the energy generated. A typical ultra-low-power MCU and wireless transceiver consume a combined total of ~50 to 100mW while transmitting, which means that a very large harvester would be required to run the application, and the energy source would have to be present at all times when the device would be running.
Additionally, the peak output power would be too low to run power intensive tasks. It would rarely be practical to have a gigantic harvester to power small applications, and it's also unrealistic to assume that the energy source would be constantly available. Since the energy, in most cases, will trickle in at low voltages over a long period of time, the energy is typically stored first and made available for the application when needed. While the system collects energy, the MCU and any sensors and communication systems are put in a low-power standby mode minimizing their drain on the available power.
The element used to store the power would act as an energy buffer for the rest of the application. The size and technical properties of the buffer is dependent on the application. If the application requires long periods of time to elapse between when it accesses an available energy source, a very large buffer is required; however, if the application is constantly around the energy source and rarely needs to be active (low duty-cycle applications), a very small buffer would be sufficient.
In order to accommodate the widest possible uses cases, the ideal energy buffer would have the following properties:
• Negligible leakage (self discharge).
• Unlimited capacity.
• Negligible volume.
• No need for energy conversion.
• Efficient energy acceptance and delivery.
Unfortunately, the ideal storage element doesn't exist, but several options are available including rechargeable batteries (such as alkaline, nickel-cadmium, and lithium-ion), super capacitors, or thin-film batteries. While rechargeable batteries in various chemistries and super capacitors are well-established technologies that continue to improve, thin-film batteries have only recently begun to proliferate in the market and serve as a good alternative to super capacitors. Key parameters of each type of storage technology are listed in Table 2 .
A time and place for energy harvesting
While we'd all like to eliminate our dependency on wires and are ready to replace our last set of batteries forever, unfortunately, not all applications are good candidates be to fully autonomous energy-harvesting systems. In addition to hard requirements that must be met such as having an energy source available to harvest from, there are practical concerns such as set-up cost.
For example, if a given application only needs to be functional for two years and it's easily achievable with a set of batteries, and an energy-harvesting power source is more costly than the batteries and consumes more space, it probably doesn't make sense to switch the power source.
In addition to having an energy source available to harvest from, applications that could benefit from energy harvesting share some of the following characteristics: it is difficult to install or access for maintenance, cords for power or communication are too costly or cumbersome, environmental friendliness is required, or very high uptime is demanded. If one or more of these characteristics apply to a given application, it would benefit from energy harvesting.
Best practices for energy harvesting system design
In an ideal world, it would be possible simply take your existing power supply, replace it with an energy harvesting system, and the system would function. However, in reality, this is rarely the case.
This would only be possible if the energy harvester was guaranteed to meet the applications power requirements at all times as if it were using the original power supply. In practice, energy harvesting systems are finely tune for ultra-low power operation and must accommodate for wide fluctuations in input power including the possibility that the power source becomes unavailable for some period of time.
Traditional ultra-low power design practices should be the foundation for operation. Using a microcontroller and RF transceiver optimized for minimum power consumption such as a 16-bit MSP430 MCU and CC2500 2.4GHz transceiver is a must. The MCU's low-power standby modes should be properly utilized and the time spent in a higher power active mode should be minimized. This requires a full understanding of the applications activity profile. Based on the activity profile, an appropriate storage technology can be selected to ensure that enough power will be available when the application is active and running.
For a real-world system, in low-power standby mode, a typical combination of an MSP430F2274 MCU and CC2500 RF transceiver would consume 1.3µA. Although, this is a nearly insignificant load, the power source or storage element must be able to deliver this minimum power level at all times.
If the application were responsible for sampling a temperature sensor and wirelessly transmitting the information to a central access point, for a short period of time, the power source must be able to sustain this significantly higher peak load of ~25mA, which is several orders of magnitude higher than the standby current. A detailed activity profile of this system is visible in Figure 2 .
By integrating the area under the curve, we would arrive at a total energy consumption of ~36µA*s when active and ~1.3µA*s when in standby mode. If this application had a one-second duty cycle, the average current consumption would be ~37µA ((36µA*s + 1.3 µA*s) / (1 s)).
If the energy-harvesting source could support a constant 37-µA load at all times, no energy buffer would be required; however, if a solar panel were used and it wasn't guaranteed to be in direct light at all times, then the storage element would need to be large enough to support the maximum amount of estimated time that the system would be in the dark.
Alternatively, if the sensor data wasn't needed at all times, the system could be modified to only transmit when a readily available energy source was present; thus the energy buffer would only need to support the system's required standby power.
Enabling longer system life
Micro-energy harvesters from various sources such as light, motion, thermal, or RF will open the doors to engineers to decouple themselves from the physical burden of batteries and applications no longer have to be limited to their accessibility for maintenance. Low-cost, autonomous sensor networks will not only enrich our lives by providing valuable data about the status of our environment, but they will do so with no long term, reoccurring cost or impact on the environment. Energy harvesting will extend the usable life from existing products and will enable design options that weren't possible before.
Some applications may sound like science fiction today, but the technology exists to enable a new generation of applications. By harvesting the vibrational energy, intelligent sensors will be able to be implanted in roads, bridges, and buildings at the time of construction providing real-time feedback on the structural integrity guaranteeing our safety. By harvesting energy from the sun, farmers will be able to monitor the health of their crop using low-cost, disposable sensors producing greater crop yield with lower maintenance. By harvesting the heat from the skin, smart band-aids no larger than a quarter will be able to monitor a patient's vital signs and transmit the information wirelessly to a central medical base station without having to tether a patient to a machine. We just need to break the design engineer's mold of what a traditional power source needs to look like and embrace the benefits of perpetually powered energy harvesting systems.
1. A specific example of an ultra-low power MCUs that support an input voltage range between 1.8 V and 3.6 V (and the device used to research this article) is Texas Instruments' MSP430.
Editor's note: Catch Adrian Valenzuela's class “Implementation of Autonomous Energy Harvesting Wireless Sensor Application” at the Embedded Systems Conference in Boston on September 22, 2009.
Adrian Valenzuela received his BS and MS degrees in electrical engineering from Rice University. During his time at Rice, he was an active researcher in areas of wireless sensor networks and low-power embedded system design. As a product marketing engineer for Texas Instruments, Adrian oversees mass-market activities that include new product development strategy for TI's microcontrollers.