Original Title：How to Design Low-Power Always-On Wearables: Part 1 – Optimize the Microcontroller

　　As battery-powered wearable devices become increasingly popular, wearable manufacturers are adding greater functionality to stay competitive. This is especially true for fitness watches that are in constant use by the consumer. They are always powered on and users constantly seek new features and enhanced performance.

　　However, adding greater functionality often requires switching to a more powerful microcontroller to control and monitor the watch’s functions. This has the drawback of reducing battery life by requiring more frequent battery recharges, which compromises the user experience.

　　This article discusses the unique needs of microcontrollers for always-on wearables. It explains how to configure a microcontroller for an always-on wearable including the low power modes and autonomous peripherals. It then looks at a 16-bit microcontroller from Texas Instruments and a 32-bit microcontroller from Maxim Integrated and shows how their key features can be used to benefit a wearable design.

　　Unique needs of microcontrollers in always-on wearables

　　For wearables, a long battery life between recharges can be the most important feature to the end user. While online reviews might praise the accuracy and features of a wearable product, the time between recharges can be the difference between the disappointment of a one-star review and the satisfaction of a five-star review.

　　Poor battery life has more significance besides the inconvenience of frequent recharging. Lithium rechargeable batteries lose total capacity with frequent recharging, making maintaining battery health over time more difficult. Batteries for wearables will be discussed in Part 2 of this series.

　　Also, while the connector used for recharging is often robust, it does have a limited insertion/removal rate so every recharge results in wear.

　　Wearable electronics have different power needs compared to other consumer devices because the wearable is always on, requiring the microcontroller to be constantly powered. Usually there is also a Bluetooth Low Energy (BLE) connection that must always be ready and available to communicate with a companion mobile device. Note that wireless connectivity for wearables is discussed in Part 3 of this series.

　　However, while the wearable can sync its data with the mobile device when a connection is available, it must also allow for standalone operation without a mobile connection for either hours or days, depending upon the intended use.

　　The main purpose of a wearable such as a smart watch, besides telling time, is to constantly monitor and log inputs from external sensors connected to serial ports like I2C and SPI. These can include specialized accelerometers that can count steps for a pedometer, a GPS radio for location tracking and navigation functions, and a heart rate monitor. While most of these sensors can be individually turned on and off by the user, a good engineer should design the system for the worst-case scenario where all the sensors are on.

　　The data collected from these sensors must be constantly logged. Often in many Internet of Things (IoT) or consumer mobile devices, logged sensor data is stored in non-volatile memory such as flash or EEPROM. However, a write operation to flash or EEPROM draws a great deal of current that can quickly deplete the small battery in a wearable. A better solution is to store the sensor data in SRAM.

　　Writing to SRAM consumes much less current than writing to non-volatile memory. Since the microcontroller is always powered, the SRAM sensor data is constantly maintained and is safe unless the wearable is powered off, or the user fails to recharge the battery allowing it to become depleted. Stored sensor data is wirelessly transferred and stored in a mobile device, so even when powered down, sensor data is not lost.

　　An important feature for minimizing power drain on the microcontroller is autonomous peripherals. Exactly how autonomous varies by the microcontroller product family. Another common feature for conserving power is to disable power to an unused peripheral, independent of the rest of the microcontroller, by setting or clearing a bit in a power register.

　　Microcontroller low-power modes for a wearable

　　After the unique needs of a microcontroller in an always-on wearable are understood, it’s important to determine what the low-power modes should do, including which are useful and which are not.

　　Of course, the lowest power mode for a wearable is when it is turned off. Most wearables are turned on and off by holding down a software controlled momentary pushbutton for a specified time, preventing accidental power sequencing. This is superior to a mechanical switch, which is not only less cost-effective, but can be accidentally triggered. However, the engineer should assume that the user will rarely power their device down, so the wearable should be designed with two seemingly conflicting assumptions: that the device will never be turned off, and that it will also occasionally be turned off.

　　Usually a power management chip controls the charging of the battery and sequences the power on and off to the microcontroller and sensors. Power management is also discussed in Part 2 of this series. When a power management chip turns off the wearable, main power to the microcontroller is gated off except for separate power to the real time clock (RTC). This requires a microcontroller that can function with external power to the CPU, RAM, and most peripherals disabled with only the RTC running.

　　It is necessary to have the microcontroller RTC running while the wearable is turned off to maintain the correct time, so the microcontroller should have a separate power pin for the RTC that is constantly powered. An RTC is clocked by a low frequency 32.768 kilohertz (kHz) oscillator that draws mere nanoamps. A smart watch that loses the time when powered off would not be a satisfactory user experience, making any low-power mode that disables the RTC inapplicable for a wearable.

　　The CPU can be disabled to conserve power, as well as any unused peripherals. RAM contents must always be maintained, also making any low-power mode that disables the entire RAM array inapplicable for a wearable.

　　Configuring the microcontroller

　　A good example of a microcontroller optimized for wearables is the Texas Instruments MSP430FR2676TPTR 16 megahertz (MHz) microcontroller with ferroelectric random access memory (FRAM) (Figure 1). This is a member of Texas Instrument’s MSP430FR2676 16-bit MSP430™ CapTIvate™ capacitive-touch sensing microcontrollers, which contain a low-power peripheral that can sense touch through thick glass. The glass screens used on wearables must be thick and durable to withstand the punishment of regular use, making CapTIvate technology applicable for a wearable with a touchscreen.

　　Figure 1: The Texas Instruments MSP430FR2676TPTR ultra-low-power 16-bit FRAM microcontroller has a wide variety of peripherals and can control a simple wearable with a minimum of external parts. (Image source: Texas Instruments)

　　The MSP430FR2676TPTR has 64 kilobytes (Kbytes) of Texas Instruments’ FRAM program memory to achieve higher read/write performance figures at lower power compared to flash microcontrollers. It has 8 Kbytes of SRAM and a full set of peripherals including I2C, SPI, and a UART for connecting to sensors. A 32 x 32 hardware multiplier speeds up multiplication, lowering power consumption.

　　The RTC on the MSP430FR2676TPTR can be configured to wake up the microcontroller in intervals from microseconds to hours. This is useful for waking up the CPU to perform tasks such as periodically processing sensor data and wirelessly sending it to a mobile device.

　　The oscillator and clock system of the MSP430FR2676TPTR is designed to lower system cost and provide low power consumption. The microcontroller supports four internally generated clock sources and two external high accuracy clock sources. These oscillators and clocks can be enabled and disabled under firmware control depending on the low-power mode selected and firmware configuration. For running peripherals, the MSP430FR2676TPTR has two clocks: a high speed subsystem master clock (SMCLK) that can run as fast as the system clock frequency, and a low speed 40 kHz auxiliary clock (ACLK).

　　Besides active mode, where the CPU and everything else is enabled, the MSP430FR2676TPTR supports configurable and complex low-power modes. Any on-chip peripheral that is active in a particular MSP430 low power mode can be powered off by firmware. This allows for custom low-power configurations. For an MSP430FR2676TPTR wearable, the following low-power modes (LPMx) are applicable:

　　LPM0 allows everything to run except the CPU. This is useful when autonomous peripherals need to be active and running at full speed without CPU intervention.

　　LPM3 disables the CPU, high speed oscillator, and SMCLK. All enabled peripherals can run off the power-saving 40 kHz ACLK. This is useful while the wearable is idle with no button presses. Serial peripherals like the I2C and SPI can run autonomously to collect sensor data while the direct memory access (DMA) transfers the data to RAM. The RTC can wake up the device to perform any needed tasks.

　　LPM4 turns off everything except the RTC. SRAM is powered down. This is useful for when the wearable is turned off by the user.

　　The MSP430FR2676TPTR can operate off 1.8 to 3.6 volts, making it suitable for use with 3.6 volt lithium batteries. With the RTC running and minimal peripherals, the microcontroller can draw less than 5 microamps (µA). With the main oscillator running, the MSP430FR2676TPTR draws 135 µA/MHz (typical).

　　For a higher performance wearable, Maxim Integrated has the MAX32660GWE 32-bit microcontroller (Figure 2). It is based on the Arm® Cortex®-M4 core with floating point unit (FPU). The MAX32660 has 256 Kbytes of flash and 96 Kbytes of SRAM. The SRAM is divided into four sectors. Any sector can be enabled for read/write, put in light sleep to disable read/write while retaining contents to conserve power, or completely disabled to remove power from that sector.

　　Figure 2: The Maxim Integrated MAX32660 is specifically designed for always-on wearable electronics. To conserve power, it minimizes the number of peripherals to only those needed to interface to external sensors in a wearable application. (Image source: Maxim Integrated)

　　The MAX32660 can run at up to 96 MHz, and with all peripherals operational draws only 85 µA/MHz. To minimize current draw and reduce package size it has a minimal set of peripherals used for wearables including two SPIs, two I2Cs, and two UARTs.

　　It supports two internal oscillators: a 96 MHz high speed internal oscillator that can be disabled by firmware, and a low power 8 KHz ring oscillator that is always on regardless of the low power mode. A 32.768 kHz oscillator uses an external crystal and is used for the RTC. Any of these three oscillators can be used to clock the CPU and peripherals.

　　Any peripherals can be powered off in firmware. In addition, firmware can also disable the clock to that peripheral, saving precious nanoamps.

　　Per wearable requirements, the RTC is always on in every low-power mode unless deliberately disabled by firmware when in Active mode. The RTC and clock are in a separate section, designated the “Always-On Domain”. This domain is isolated from the rest of the microcontroller, ensuring that in the event of a firmware malfunction or tampering the RTC is unaffected.

　　Besides Active mode, the MAX32660 supports three low-power modes specifically customized for wearable electronics:

　　In Sleep mode, the CPU is off while any enabled peripheral can run autonomously. This can be useful when the wearable is idle and sensor data is logged and stored with the DMA. Any active peripheral can wake the CPU into Active mode.

　　In Deep-Sleep mode, all the internal clocks to the CPU and peripherals are gated off except for the 32.768 kHz clock to the RTC. Firmware can configure the 96 MHz internal clock to automatically turn off when entering Deep-Sleep. All RAM contents are retained including data SRAM and all peripheral registers. This is useful for a wearable that requires a soft-off mode, where the wearable is powered off to conserve power, but needs to restart where it left off when powered back on.

　　Backup mode is the lowest power mode. Clocks and power to the CPU and all peripherals are gated off except for the RTC. By default all power to SRAM is disabled. This is useful for when the wearable is powered off by the user with no SRAM retention to save power. However, this mode can optionally keep any of four SRAM sectors in light sleep to retain the memory contents. This is useful for a wearable that needs to maintain a minimal state with a small additional current draw.

　　The MAX32660 requires between 1.71 and 3.63 volts, allowing it to run off 3.6 volt lithium batteries. The microcontroller also has a self-contained power management unit, reducing pin count by eliminating the external component. It also has battery gauge support that monitors the external battery and provides an accurate reading of the battery state of charge, which can be displayed on the wearable’s user interface.

　　Conclusion

　　Always-on wearable electronics provide unique challenges to engineers. Even when a wearable appears to be powered off, it still draws some level of power. However, as shown, designers can add functionality and features to their designs and use the configurable low-power modes to maintain and extend battery life.