The Art of Designing Remote IoT Devices-Technologies and Strategies for a Long Battery Life

Abstract

Long-range wireless connectivity technologies for sensors and actuators open the door for a variety of new Internet of Things (IoT) applications. These technologies can be deployed to establish new monitoring capabilities and enhance efficiency of services in a rich diversity of domains. Low energy consumption is essential to enable battery-powered IoT nodes with a long autonomy. This paper explains the challenges posed by combining low-power and long-range connectivity. An energy breakdown demonstrates the dominance of transmit and sleep energy. The principles for achieving both low-power and wide-area are outlined, and the landscape of available networking technologies that are suited to connect remote IoT nodes is sketched. The typical anatomy of such a node is presented, and the subsystems are zoomed into. The art of designing remote IoT devices requires an application-oriented approach, where a meticulous design and smart operation are essential to grant a long battery life. In particular we demonstrate the importance of strategies such as "think before you talk" and "race to sleep". As maintenance of IoT nodes is often cumbersome due to being deployed at hard to reach places, extending the battery life of these devices is critical. Moreover, the environmental impact of batteries further demonstrates the need for a longer battery life in order to reduce the number of batteries used.

Keywords: embedded design; energy management; energy-saving strategies; internet of things; low-power design; low-power wide-area networks; sensors.

Figure 1

The ‘IoTree’ node [1] is designed and deployed for remote monitoring of a tree’s health.

Figure 2

The typical architecture connects remote IoT nodes (a) via long-range connectivity to one of more gateways or base stations (b). The interconnected gateways relay the information to the server or cloud resources (c), where data can be visualized and interpreted.

Figure 3

Power consumption and energy consumption share of each stage of a Long-Range Wide-Area Network (LoRaWAN) node. The energy corresponds to reading out a sensor and send ing a 16 bytes payload packet (SF12) every hour [18].

Figure 4

Constellation diagrams of BPSK and 64-QAM modulated symbols at different SNRs including the percentage of successfully demodulated symbols. It demonstrates that low-order modulation schemes are vital in order to work in low-SNR environments or when the transmit power levels are low.

Figure 5

Measured waterfall spectrum of LoRaWAN, Sigfox and NB-IoT. The vertical axis represents the time and the horizontal axis the frequency domain. Due to the difference in time and frequency allocation, the figures have not the same time and frequency scale. The observed spectral leakage is due to the Fast Fourrier Transfomation (FFT) operation.

Figure 6

Typical MAC scheme for remote IoT devices. The communication is mostly device-induced with a limited payload and minimized protocol overhead. While downlink communication is supported, it is mostly restricted to a couple of messages per day.

Figure 7

Medium Access Control mechanism for LoRaWAN and Sigfox. (a) LoRaWAN opens two receive windows after a transmit message. The first window utilizes the same data rate as the transmit message. The second window employs more robust settings to increase the chance of reception. (b) Sigfox transmits three duplicates at different frequencies and different time instances. Limited downlink is possible and is received at the initial transmit frequency plus an offset.

Figure 8

Energy saving strategies employed in NB-IoT, i.e., Extended Discontinuous Reception Mode and Power Saving Mode, respectively. (a) Listen more infrequently to downlink message with eDRX. (b) Hibernate between data transmissions with PSM.

Figure 9

Generalized architecture of an IoT node. The complete device is typically battery-powered, optionally assisted by some kind of energy harvesting technology, e.g., a solar panel. Sensors periodically sample the environment for information, which is transmitted wirelessly. A microcontroller has all the amenities to manage the operation and behavior of the node. Actuators, e.g., a LED, allow the node to influence its environment or signal the user.

Figure 10

Modems combine RF front-end, transceiver and MAC layer implementation on a controller in one easy to use module. These examples show the internal components of two popular IoT modules. While the building blocks of these modules are independent of their manufacturer, these modules are chosen because of their high availability and adoption in IoT nodes.

Figure 11

The internal Power Management Unit (PMU) of a microcontroller supervises other Microcontroller Unit (MCU) building blocks. Depending on the current controller state, different voltage regulators are connected or disconnected. The firmware on the CPU can, in turn, enable or disable external subsystem, such as, sensors or the modem.

Figure 12

The potential power density in function of different energy harvesting approaches is shown, with the actual produce heavily depending on the stimulus. A distinction is made between three energy harvesting sources: mechanical, thermal and radiant [77,80,81].

Figure 13

Typical power profile of an IoT node. This periodically recurring profile generally consists of four states, i.e., sleep, wake-up, processing and connectivity. A node typically spends most of its time in the sleep mode. The area under the curve yields the energy consumed in each state.

Figure 14

Proposed energy saving strategies. The applicable strategies—or combination of strategies—strongly depends on the application. (a) Think before talk. Check the validity of the sensor measurement and optionally accumulate data prior to transmission. Some sensors do not require the MCU to be active yielding a lower power consumption as depicted by the border. (b) Race to sleep. Utilize dedicated hardware features to reduce time spend in high-power states. (c) Sleep as much as possible. Avoid powering unnecessary hardware and fall-back to a sleep state as often as possible.

Figure 15

Power graph depicting the workings of a ’forced’ gas measurement with a BME680 sensor [47]. A sensor remains in sleep and has no active operations until the controller commands, or forces, a measurement. A heated bed is required for an accurate gas value measurement. Heating the sensor element can take up to 92 $\mathrm{s}$, consuming large amounts of energy.

Figure 16

Illustration of possible power savings by using a fixed voltage, provided by an LDO voltage regulator [84].