Minimizing Power Consumption in Energy Micro-harvesting Wireless Sensors

Powered by ambient energy, "zero-power" wireless sensors find application in nearly every market segment. However, engineers must exercise extra care in managing the scant power budget available with energy micro-harvesting. By combining ultra-low power MCUs and RF ICs with high efficiency power conversion devices, engineers can create wireless sensors capable of operating without battery replacement for the effective lifetime of the system's components. To design these highly efficient systems, engineers can take advantage of ultra-low-power devices and specialized ICs from manufacturers including Cymbet, Linear Technology, Maxim Integrated Products, Microchip Technology, Silicon Labs, and Texas Instruments, among others.

Along with an appropriate transducer for light, thermal, vibration, or RF energy, a zero-power wireless sensor system design typically comprises a power conversion and management unit, microcontroller, RF radio, and application sensor (Figure 1). In operation, the system converts a trickle of a few microwatts of ambient energy into sufficient power to allow the system to periodically wake up, collect application sensor data, perform required sensor signal processing and data formatting, and finally transmit the results.

Figure 1: A typical zero-power wireless sensor combines energy transducer, energy processing capability, MCU, RF radio, and application sensor (Courtesy of Cymbet).

A typical wireless sensor application will need a few hundred milliseconds or less to complete a sensor data collection and transmission event. For most applications monitoring real-world processes, sensor events are likely to occur every few minutes rather than every few seconds. Consequently, wireless sensors exhibit an activity profile characterized by prolonged quiescent states interrupted periodically by bursts of activity (Figure 2, also see the TechZone article "Ultra-Low-Power MCUs Enable Energy Harvesting Designs").

Figure 2: The activity profile for a typical wireless sensor exhibits long quiescent periods interrupted by bursts of activity with wake up transitions of diverse duration depending on oscillator type and device requirements (Courtesy of Texas Instruments).

The low duty-cycle operational characteristics of wireless sensors present unique challenges to engineers in creating efficient power, processing, and communications stages of a wireless sensor design. The power stage must be capable of harvesting ambient energy sources and generating sufficient power to supply downstream circuitry. In a wireless sensor, communications requirements lead to peak demands well beyond that available instantaneously from ambient sources. Consequently, the power stage must be able to efficiently trickle-charge storage devices such as thin-film batteries like the Cymbet EnerChip, or supercapacitors like the Taiyo Yuden LR Series or Eaton PowerStor series. During peak loads, the power management system must be able to switch to stored energy to power the activity burst, as well as activity associated with the return to quiescent state.

To design a suitable power stage, engineers can turn to specialized, highly integrated energy-harvesting devices such as the Cymbet CBC915 or Linear LTC3588, Maxim MAX17710, each of which offers power conversion functionality designed specifically for energy micro-harvesting applications (Figure 3, also see the TechZone article "Power Management ICs for Micro-harvesting Designs").

Figure 3: Specialized micro-harvesting ICs such as the Maxim MAX17710 can provide a ready solution for ambient-sourced power supplies in wireless sensors (Courtesy of Maxim Integrated Products).

These devices form the core of an ambient-sourced power supply, providing converted energy to the application stage in a wireless sensor system. With the wide availability of MCUs with integrated peripherals including analog-digital converters (ADC), the application circuitry can comprise simply the MCU and RF devices with a minimum of additional discrete components. For engineers, the challenge then becomes one of meeting a very tight power budget by minimizing wasted power and maximizing efficiency in processing and communication operations. To address these challenges, manufacturers equip MCUs and RF devices with multiple power-saving modes that enable the engineer to finely balance device functionality against power consumption. For a zero-power wireless sensor design, the key performance criteria focus on features and capabilities for maximizing power efficiency in standby, wake-up, and active modes.

Standby modes

With the low duty cycle operation typically found in wireless sensor applications, standby mode will tend to be the dominant operational state in these systems. Power consumption integrated across even a prolonged quiescent state will likely not rise to the instantaneous levels found during a single activity burst. Nevertheless, the power efficiency in standby mode will play an important role in determining a system's overall efficiency in using scarce ambient energy.

Standby power consumption arises largely from two principal factors: leakage current of devices and the minimum power needed to support required functions of the system in sleep mode. Leakage is inevitable on device pins (Figure 4), but today's ultra-low-power devices boast minimum leakage currents, typically exhibiting per-pin values in the microamps or even nanoamps, and will continue to fall to new lows with each new generation of process technology.

Figure 4: A simplified model illustrates leakage current at a representative input pin (Courtesy of Microchip Technology).

Engineers will also need to account for leakage from discrete components in the power budget, either directly using specified leakage ratings or through calculation based on the capacitor's insulation resistance (IR) specification:

I = V x C / IR

where IR is specified in megaohms or megafarads.

Engineers can further reduce leakage current by powering down unneeded circuits in their own circuits or within integrated devices that support this kind of selective power state. For example, application sensor and RF stages could be separately powered up only at the beginning and end of burst activity for data acquisition and communications, respectively. As described below, advanced RF circuits allow engineers to programmatically disable selected portions of the RF signal chain to reduce leakage current and overall power consumption.

During sleep mode, a wireless sensor system must nevertheless maintain sufficient functionality to wake itself according to programmed criteria or in response to interrupts triggered by external events. For a typical application, this minimum level of function would likely include retaining MCU state and memory contents between activity burst, rather than wasting power to write state to nonvolatile memory at the end of a burst period and restoring state at the beginning of the next burst period.

MCUs also need to be able to detect supply voltage brownouts and take appropriate actions, including safely returning to standby or even resetting themselves. For example, the Microchip PIC12LF1840T48A integrated MCU includes a programmable brownout reset (BOR) feature that causes the MCU to reset in the event of a brownout (Figure 5). If left unaccounted for, brownouts could lead to corrupted state, as the supply dips below the minimum voltage needed to retain MCU state, register values, program status, and memory.

Figure 5: MCUs such as the Microchip PIC12LF1840T48A can be programmed to reset on supply brownouts. Here, the device asserts Reset when supply falls below the brown-out voltage threshold, VBOR, maintaining Reset until VDD rises above VBOR plus a hysteresis value (Courtesy of Microchip Technology).

Beyond these minimum functional requirements, the MCU would need to retain the ability to respond to interrupts from external events, such as a change in temperature, pressure, or sudden acceleration. Alternatively, engineers could design a wireless system to wake up on a regular basis and perform sensor measurements. For this timed polling approach, an MCU in standby mode must be able to maintain real-time clock (RTC) function and the ability to respond to RTC alarms. Integrated MCUs, such as the Silicon Labs Si1030x and Texas Instruments MSP430F513x, provide an on-chip RTC and offer low-power modes that maintain RTC function and alarm-wakeup capability.

Wake up

In an energy micro-harvesting design where every microwatt is vital, the transition from standby to active mode represents wasted power as circuits re-energize to useful operating modes. Power required during wakeup begins to deplete precious stored energy, critically needed for peak loads associated with the application activity burst. Consequently, MCUs and RF devices targeted for these systems should feature very fast startup times. Furthermore, the devices should ideally support the ability to energize required sub-circuits in a defined sequence to avoid exceeding the instantaneous power budget, possibly resulting in collapse of the entire wireless sensor system. This sequential startup-up capability is particularly important in "cold start" situations, where a wireless sensor has been newly deployed or an existing system has been cut off from its ambient source for extended periods of time.

For MCUs, the wake up time is a critical performance characteristic. MCUs such as the Silicon Labs Si1030x wake up from sleep to active mode in as little as 2 µs. The Texas Instruments MSP430F513x MCU wakes up from low power modes in 5 µs, and even from brownout reset in only 2 ms. The MSP430F513x also provides a slow wakeup feature that provides for a wakeup sequence in lower power modes.

RF devices such as the Silicon Labs Si4420 allow engineers to selectively power separate stages of the RF signal chain. By setting or resetting bits in the device's power management control register, engineers can activate or deactivate specific blocks of circuitry needed at different times.

Active mode

Power efficiency in sleep and wake-up modes is essential for ensuring that sufficient energy can be accumulated to supply peak loads generated during activity bursts. In active mode, minimizing active power consumption is critical for ensuring that peak demand (and return to quiescent state) will not exceed available power, typically drawn from the system's stored power source. One of the most straightforward methods for reducing active power consumption is using the lowest possible supply voltage. Supply voltage is a dominating factor in the following equation for dynamic power consumption in a CMOS logic gate:

Active mode power = C x V² x f

Where

• C is a function of the processing technology,
• V is supply voltage
• f is switching frequency of the gate.

MCU manufacturer datasheets typically quote dynamic current relative to 1 MHz. Recast in those terms, the active mode equation becomes:

Active mode power = V * I
where dynamic current I = C x V x f

Low power MCUs suitable for energy micro-harvesting applications feature dynamic current ratings typically below 300 uA/MHz (again, see the TechZone article “Ultra-Low-Power MCUs Enable Energy-harvesting Designs"). Devices in this class typically operate across a supply voltage range from 1.8 V to 3.6 V, an essential characteristic for low-voltage, micro-harvesting designs.

The use of highly integrated devices also helps eliminate wasted power. Mixed-signal system-on-chip (SoC) devices, such as the previously mentioned Microchip PIC12LF1840T48A, Silicon Labs Si1030x and Texas Instruments MSP430F513x, combine a full complement of capabilities required in a wireless sensor, integrating RF, MCU, ADC, GPIO, clocks, voltage regulators, and power management units on a single device, thereby eliminating delays and power inefficiencies associated with off-chip access.

These devices typically provide programmable RF output power, as well as flexible operating modes that allow the engineer to selectively disable individual peripherals. In the Silicon Labs Si103x integrated MCU, for example, engineers can set operating modes of the on-chip transceiver to disable portions of the RF signal chain, just as with the standalone Si4420 RF transceiver chip mentioned earlier. In the Si103x, engineers can set RF modes that disable portions of the signal chain including the power amp, receiver unit, PLL, and others, cutting dynamic current from 18.5 mA in full operation to 450 nA in a register-saving standby mode. Of course, the engineer needs to balance these power savings with the additional wake up time needed to re-energize these stages.

Even with the most highly integrated components, radio communications typically account for a disproportionately large share of power consumption in a wireless sensor system. Engineers can minimize power by optimizing RF power and communications protocols. Individual wireless sensor nodes can be operated at very low output power levels when receivers are physically nearby. For example, engineers can use RSSI output (Figure 6) provided by transceiver-equipped SoCs, such as the Silicon Labs Si1030x and Texas Instruments MSP430F513x, to estimate proximity to the wireless network receiver and adjust transmitter output power accordingly.

Figure 6: Engineers can use RSSI output available in SoCs such as the Silicon Labs Si1030x to optimize RF power output (Courtesy of Silicon Labs).

Engineers can also design wireless sensors to adapt to available energy resources. If the wireless sensor system is operating with minimal energy reserves, it can be designed to vary RF output power levels to scale proportionately with available energy reserves, transmitting at full output power only when sufficient reserve energy is available to ensure completion of transmission and return of the system to standby mode.

The use of low overhead communications protocols can also significantly alleviate communications-related power requirements. Wireless sensor data communications is typically well constrained in the type of required communication transactions. Without the need to support a long list of transaction types, engineers can shave packet message envelopes down to the bare minimum overhead required to complete the data transfer reliably.

Opportunities for maximizing efficiency in a wireless sensor system extend to the software architecture. Engineers can also optimize active periods by opting for different data handling approaches. For example, an application that uses sensor data to compute trends can likely tolerate a certain amount of delay in receiving historical data. In this case, the wireless sensor can wake up, collect instantaneous sensor data, and return immediately to sleep. Only after accumulating a number of data points would the system need to remain awake long enough to energize relatively power-hungry RF circuits and complete data transmission.

Summary

Powered by ambient energy sources, an efficient wireless sensor system can operate continuously, with a lifespan determined only by the reliability of its components. Achieving this ideal of a perpetually powered wireless design, however, requires optimizing power characteristics across standby, wakeup, and active modes. Available MCUs, RF ICs, and RF mixed-signal SoCs feature a rich set of power-control capabilities needed to achieve this level of power optimization. By combining these energy efficient ICs with specialized energy-harvesting devices, engineers can meet sophisticated application requirements with wireless sensor designs able to run indefinitely.