Wireless technology is endemic in modern society; broadly speaking we rely on cellular technology to stay connected when we’re travelling, and probably WiFi when we’re not. Both offer bandwidths that are high enough to be considered broadband, giving us untethered and virtually unrestricted access to all of our essential content, whether for work or relaxation.
In the IoT, cellular and broadband do, of course, play a role but for many of the ‘things’ in the IoT, adding broadband would be an unnecessary expense in terms of power and cost. End-devices will come in an almost unlimited number of shapes and sizes but they will probably have many commonalities, first among them being cost, closely followed by size and weight. But all of those demands could easily be overshadowed by a fundamental need to be low power.
Balancing ‘Size, Weight, Power and Cost’ (or SWaP-C, as it is often referred to) is well understood in the military and aerospace industries and just as applicable in the IoT. Many end-devices in the IoT will be expected to operate for many years with little or no maintenance, which includes changing the batteries. As they will most likely use some form of wireless connectivity, the choice of wireless technology will be a major part of the SWaP-C calculations.
One of the simplest ways to reduce RF power is to limit either the range or the bandwidth. To put it another way, range and bandwidth will be determined by the available RF power. Accepting that the application will have a small power budget, selecting a technology that provides the required range and bandwidth presents a limited number of potential solutions. Most of them overcome the issue of range by employing some form of network; in this way they are more like cellular technology than Wi-Fi.
Looking at it from a single node’s point of view, the power needed to get a message to the network’s server is effectively distributed across all of the nodes actively participating in the network. For a mesh network that functions in a self-healing way, one message could be picked up and passed on by multiple nodes, in some cases consuming power unnecessarily. This is characteristic of wireless networking technologies that occupy the PAN (Personal Area Network) and LAN (Local Area Network) space and adopt a mesh topology, such as Bluetooth 5 and ZigBee. It is less characteristic of Low Power Wide Area (LPWA) wireless technologies designed to transmit over kilometres, as opposed to just a few metres for PANs and LANs. This includes technologies developed for the licensed spectrum such as LTE-M and NB-IoT, as well as LPWAs operating in the unlicensed spectrum, like Sigfox and LoRaWAN.
Creating and operating a cellular network in the licensed spectrum is expensive, so nascent solutions like LTE-M and NB-IoT could be precluded on the basis of cost, at least in the early stages of deployment. Similarly, mesh networks that rely on many local nodes to guarantee connectivity and extend range (Bluetooth 5, ZigBee) could also put the total cost of ownership (TCO) up, particularly when large areas and/or long distances are involved. The solutions presented by the likes of Sigfox and LoRaWAN attempt to overcome these limitations by offering long range connectivity (10s of kilometres) for low bandwidth messages (10s of bytes) at very low power (years of service on a single battery), without the responsibility of becoming a network operator.
While they can’t be considered cellular networks in the conventional sense, LPWANs operating in the license-free spectrum do require base stations to provide coverage. With Sigfox, the responsibility for these staging posts falls to third parties who operate local Sigfox networks; users must engage with these operators in order to use the network(s). Its lightweight protocol supports payloads of up to 12 bytes in the uplink (from the end-device to the network), and minimises transmit power requirements by removing most of the conventional network overheads such as handshaking. The device simply transmits its message three times on different frequencies, in the hope that it will be received by at least one base station within range. Acknowledgements can be configured but it isn’t a requirement of the protocol. As a global standard, end-devices on opposite sides of the world could theoretically communicate (coverage permitting), with a data rate of between 100 and 600 bits per second (region-dependent), using Ultra Narrow Band (UNB) technology based on shift-keying modulation schemes.
As it is operating in the same space (both application and RF spectrum; sub-1GHz), LoRaWAN is comparable to Sigfox but offers a number of nice features. It is based on a spread spectrum technology and uses a technique called Adaptive Data Rate (ADP) to automatically adjust the spreading factor for each end-device, in order to get the best performance (range, payload and data rate, battery life, and network capacity). The ‘star of stars’ network topology means it doesn’t need repeaters and avoids the overhead associated with mesh technologies. In addition, communications are bidirectional and acknowledged, it also supports multicast messages (useful for Over The Air updates, for example). The network topology also includes gateways, which can be put into service by users to provide backhaul to the widest of all networks; the internet.
Increasingly, modules are used to provide pre-certified wireless solutions that can be designed-in with minimal RF experience. One of the first modules to provide both a complete LoRaWAN and Sigfox solution is the dual-mode CMWX1ZZABZ-078 from Murata, which is a ‘drop-in’ standalone module for adding LPWA connectivity to practically any application. It includes an ARM Cortex-M0+ microcontroller for hosting embedded applications and an RF transceiver developed by Semtech, the company behind LoRa’s spread spectrum technology.
If you are thinking about developing an IoT application that needs long range, low power wireless connectivity, but aren’t sure which one is best for you, our webinar with Murata takes an in-depth look at both the LoRaWAN and Sigfox protocols and how to design nodes using their dual-mode module. You can watch it on demand here. If you want some direct advice on your options then why not Ask an Expert to get the full picture.
With the number of connected nodes that make up the Internet of Things (IoT) growing on a daily basis, it is universally accepted that the way in which machine learning (ML) inferencing is executed has to change.
