IoT Sensor Connectivity
There are several technology options for connecting widely distributed IoT sensors. But system developers face challenges with the range and power of many of today’s wireless networking options. In this article, BehrTech’s Wolfgang Thieme does a deep dive on Low Power Wide Area Networks (LPWAN), explaining how the technology solves many of those design dilemmas.
As the global IoT connections exponentially grow to 22 billion in 2025 (IoT Analytics), Low Power Wide Area Networks (LPWAN)  are expected to be a prominent facilitator. The central value of IoT is the unprecedented visibility into physical assets, processes and people to enable informed decision making. Often times, this visibility comes from granular, battery-powered IoT sensors distributed over large, structurally dense campuses like factories, mine sites, oil fields or commercial buildings.
Legacy wireless technologies can’t keep up with the range, power and cost requirements in smart sensor networks. Traditional cellular connectivity (for example 3G, LTE…) and wireless local area networks (such as Wi-Fi) are too expensive and power hungry for transmitting small amounts of data from a large number of sensor devices. Other solutions such as Bluetooth, Zigbee, Z-Wave have highly constrained physical range. And even though many of them employ a mesh topology to extend their coverage, the multi-hop relaying nature is power-consuming while entailing complex network planning and management. As such, mesh networks are suitable for medium-range applications at best.
LPWANs are unique in that they overcome these pitfalls to deliver an efficient, affordable and easy-to-deploy solution for massive-scale IoT networks. They are ideal for low-bandwidth applications with small payloads, such as environmental monitoring, asset and facility management, worker safety and smart metering.
KEY QUALITIES OF LPWAN
LPWANs employ a star topology in which a base station collects data from numerous remote, distributed end nodes. With the exception of cellular LPWAN (such as NB-IoT), the connection between end nodes and the base station is non-TCP/IP to avoid hefty packet headers (Figure 1). After receiving and demodulating messages, the base station then relays them to the backend server through a standard TCP/IP backhaul link (for example Ethernet, cellular…). For public LPWAN services, data must be routed through the network operators’ server before reaching the end user’s applications, while in privately managed LPWANs, data can be directly transferred to the user’s preferred backend for complete data privacy and control.
The appeal of LPWANs exist within their two signature features that used to come as a trade-off in traditional technologies: long range and low power consumption (Figure 2). While Wi-Fi and Bluetooth can only communicate over tens or a hundred meters at best, LPWANs are able to transmit signals up to 15km in rural areas and up to 5km in urban, structurally dense areas. This wireless family also provides deep penetration capability to connect devices at hard-to-reach indoor and underground locations. On top of that, it comes with a simple, small-footprint transceiver designed to minimize cost and power consumption on the end node side. The idea is to leave all the heavy-lifting to the base station and keep the data frame as short as possible.
Long Range: Range is often measured in terms of receiver sensitivity—the lowest signal power for a message to be detected and demodulated. In LPWANs, receiver sensitivity can reach -130dBm, as compared to a moderate -70dBm sensitivity in Bluetooth. This high receiver sensitivity is typically achieved by reducing the signal bandwidth and thus experienced noise levels (such as (Ultra-)Narrow Band) or adding processing gain (like Spread Spectrum)—both come at the cost of lower data rates.
Besides these special modulation techniques, the use of sub-GHz frequency bands in most LPWAN solutions, instead of the popular 2.4GHz band, further improves range and penetration capability. As the wavelength is inversely proportional to free space path loss, the long radio waves in sub-GHz systems can travel over kilometers in open areas. Compared to 2.4GHz signals, they can also better penetrate through walls, trees and other structures along the propagation path, while bending farther around solid obstacles.
Low Power: LPWAN systems adopt multiple approaches to optimize power efficiency, securing many years of battery life on end nodes. First, outside the transmission time, the transceivers are put into deep “sleep” mode whereby very minimal power is consumed. In bi-directional communications, a listening schedule is defined so that the device is “awake” only at predefined times or shortly after an uplink is sent to receive the downlink message.
Second, though not all, many LPWAN technologies employ a lightweight asynchronous protocol at the Medium Access Control layer to minimize data overhead. Pure ALOHA—a very simple random-access protocol—is a common choice. In pure ALOHA, a node accesses the channel and transmits a message whenever data is available. There is no time-slotted coordination or carrier sensing, and even acknowledgment of received messages is often bypassed to further reduce the power footprint.
Finally, the one-hop star topology introduces great power benefits. While certain mesh solutions (for example, Zigbee or WirelessHART) have been previously implemented for battery-operated, industrial sensor networks, they consume more power than an LPWAN solution by orders of magnitude. This is because, in a multi-hop mesh topology, a device must spend extra energy on listening for and relaying messages from other devices. On the other hand, a star network allows devices to “turn off” and stay most of the time in sleep mode (Figure 3).
All that said, power efficiency can drastically vary among LPWAN technologies. This is because transmission time or on-air radio time of each message is very different across systems, and transmission is technically the most energy-consuming activity. Short on-air time means that the transceiver can turn off faster to further reduce power consumption.
CURRENT LPWAN LANDSCAPE
The LPWAN landscape can be confusing at first sight, given the plethora of available solutions on the market. Nevertheless, if we take a look at the underlying technology, LPWAN solutions can be broadly grouped into four major types: cellular LPWAN, Ultra-Narrowband (UNB), Spread Spectrum and Telegram Splitting. Among these four, cellular LPWAN is the only category that operates in the licensed spectrum, while the latter three mostly leverage the license-free Industrial, Scientific and Medical (ISM) frequency bands.
While introducing low cost and quick deployment benefits, the use of the license-free spectrum raises considerable Quality-of-Service (QoS) and scalability challenges. In most solutions, there exists a persistent trade-off between QoS and power efficiency. As mentioned earlier, the lightweight asynchronous protocol at the MAC layer is widely used in LPWAN for its power advantage. Nevertheless, when multiple radio systems co-exist and share the same spectrum resource, uncoordinated transmissions in asynchronous networks significantly increase the risks of packet collisions and data loss.
Mitigation mechanisms like Listen-before-Talk, handshaking and acknowledgment to ensure QoS inevitably come with heavy overheads or frequent signaling, which means more power consumption. As wireless IoT deployments and radio traffic exponentially grow, warranting network reliability and scalability while optimizing battery life will be a major undertaking in many LPWAN technologies.
Standardization is another important consideration, given its critical role in enabling a robust and vibrant IoT ecosystem. A standardized technology provides a rigorous and transparent technical framework to fuel both vertical and horizontal interoperability. So far, there have been only two camps of LPWAN technologies that succeeded in standardization efforts and are endorsed by formal standard organizations. One is cellular LPWAN that implements 3GPP standards, and the other is the Telegram Splitting  technology based on the newly released ETSI standard on Low Throughput Networks—TS 103 357 .
Some industrial alliances have also been established around certain proprietary LPWAN technologies to promote standard development. However, these efforts do not ratify the viability of the technology and might not cover the whole network stack. It’s common that only the MAC layer is made open, while the physical layer remains entirely proprietary, like in the case of the LoRa Alliance. Having part of technical specifications publicly available on a royalty-basis doesn’t necessarily make the technology a truly open standard. Also, these industrial activities do not incorporate a stringent technology evaluation and quality testing process, as in an SDOs’ (standards developing organization’s) formal procedure.
FOUR MAJOR LPWAN GROUPS
After a quick glimpse into the existing LPWAN landscape, we’ll now dive into each type of LPWAN technologies and review their major technical features (Figure 4).
1. Cellular LPWAN (Licensed Spectrum): LTE-M and NB-IoT are the two major variants of cellular LPWAN. Both employ a narrowband approach, wherein the received signal bandwidth and data rates are reduced to improve range and building penetration ability. Compared to LTE, their transmission power and technical design complexity are also drastically reduced to achieve low-cost, low-power qualities. NB-IoT, however, uses a much smaller system bandwidth (200kHz) than LTE-M (1.4MHz) and is thus a better choice for underground and indoor applications.
Thanks to their operations in the licensed spectrum, cellular LPWAN solutions introduce great Quality-of-Service advantages. That’s because there is no co-channel interference from external systems. They additionally employ time and frequency synchronization alongside handshaking for very high transmission reliability and network scalability. That being said, these mechanisms come at the cost of power efficiency due to the required data overhead . Besides consuming extra energy, handshaking makes the battery life of a node unpredictable, since it’s difficult to decide how many times the process needs to be repeated for each transmission.
Compared to the unlicensed counterparts, cellular LPWAN provides relatively higher peak data rates (greater than 1Mb/s for LTE-M and 250Kb/s for NB-IoT), which further increases power budget requirements. Available as managed connectivity services from telecom providers, their coverage at remote locations might not be guaranteed, and network longevity is at stake due to the unforeseeable technology sunsetting. If your IoT end nodes are mobile, NB-IoT won’t be in your best interest as it’s mostly designed for stationary devices.
Given their pros and cons, cellular LPWAN options are best suited for higher data rate IoT use cases and in smart city scenarios where telecom infrastructure is mature. On the other hand, they aren’t optimal for applications where ultra-low power is at a high priority. The same goes for industrial deployments which often take place at remote locations and require the supported communications network to sustain over several decades.
2. Ultra-Narrowband – UNB (License-free Spectrum): To minimize the subjected noise level and optimize receiver sensitivity, Ultra-Narrowband solutions contract the signal bandwidth to as small as 100Hz. Besides extensive range and excellent penetration, UNB approach allows for high spectral efficiency as each signal occupies very minimal channel bandwidth. High spectral efficiency means that more messages can fit into an assigned frequency band without overlapping with each other, thereby improving overall system capacity and scalability. Sigfox and Telensa are representatives of UNB-based LPWAN technologies.
UNB signals, however, introduce very low data rates which translate into long on-air radio time. For example, systems like Sigfox feature a 100Hz signal bandwidth and a data rate of 100bps (EU mode), which means a 12 byte transmission could last for as long as 2 seconds. This presents several challenges. First, long on-air time inevitably comes with more power usage as the transceiver needs to be active for a longer period of time. Second, under EU duty cycle regulations (1%) imposed by ETSI, a device operating in the 868MHz band can “speak” for only 36 seconds per hour. As such, the longer each transmission takes place, the fewer total messages are allowed to be sent. In the US, FCC regulations limit the frequency occupation time of each message to 0.4 seconds, requiring a different network design with a higher data rate and shorter overall network range.
Another issue with long on-air time is impaired Quality-of-Service (Figure 5). Coupled with asynchronous communications, longer time in the air interface exposes a message to a higher chance of data collision, especially in a crowded license-free spectrum with heavy radio traffic from multiple co-existing systems. Certain solutions apply redundancy in which the same data is sent several times in an attempt to improve message reception. However, this measure proves to be counter-productive, as it increases total on-air time and energy usage per unique payload, while further limiting effective data amounts that can be sent per hour.
Another drawback of UNB networks is its sensitivity to multipath fading caused by Doppler effects in mobile end devices or those situated close to fast-moving objects (near a highway, for instance) . To avoid packet errors due to Doppler shifts, UNB nodes should be stationary or moving only at minimal speeds.
3. Spread Spectrum (License-free Spectrum): As a common LPWAN modulation technique, Spread Spectrum overcomes the very slow data rate and Doppler fading issues experienced by UNB solutions to a certain extent. In Spread Spectrum, a narrowband signal continuously changes frequency, resulting in a frequency ramp that occupies a much wider channel bandwidth. More bandwidth use essentially comes with a higher experienced noise level. As such, processing gain is added to improve the signal-to-noise ratio (SNR) and overall system range. Spreading Factors (SF) signify the level of processing gain with higher SF enabling longer range at a lower data rate.
Compared to UNB signals, Spread Spectrum signals are more robust against interception and eavesdropping attempts. Chirp Spread Spectrum (CSS) implemented in LoRa technology is a representative variant of this modulation scheme. A recent study shows that CSS systems can effectively support mobile nodes at a speed of up to 40km/h .
On the other hand, the major limitation of Spread Spectrum solutions is their inefficient use of the spectrum resource, since more bandwidth is required to transmit only a small data amount. This induces bad co-existing behavior and serious scalability problems. In the limited sub-GHz radio spectrum, high wideband data traffic combined with uncoordinated transmissions in pure ALOHA can cause message overlays and eventually packet errors. This challenge further intensifies in long-range applications using a high spreading factor, due to the low data rate and thus, longer on air-time of messages .
The uses of different spreading factor and bandwidth combinations (such as orthogonality) and a higher number of base stations are common approaches to partly remedy this issue. However, tuning each base station to different frequency entails complex network management and requires radio system expertise.
4. Telegram Splitting (License-free): Telegram Splitting is the latest and so far, the only standardized LPWAN technology in the licensed-free spectrum. Introducing a new radio transmission approach for UNB signals, the technology aims to surpass the trade-off between Quality-of-Service and power efficiency commonly faced in previous LPWAN solutions. MYTHINGS by BehrTech is the only solution that implements Telegram Splitting and fully complies with the ETSI TS 103 357 standard.
Telegram Splitting systems feature a data rate of 512bit/s. At the physical layer, the technology divides a UNB telegram into multiple equal-sized sub-packets, each of which is randomly sent at a different time and carrier frequency. As each sub-packet has a much smaller size than the original telegram, its on-air time is drastically reduced to only 16ms. The accumulated on-air time of a 10-byte as an example is only 390ms. Short on-air time combined with the virtually random distribution of sub-packets over time and frequency significantly mitigate their risk of being hit by interferers (Figure 6). On top of that, even if up to 50% of sub-packets are affected, Forward Error Correction ensures that the full message can be retrieved at the base station.
As such, although asynchronous communication is used for ultra-low power benefits, Telegram Splitting delivers very high interference immunity and system capacity. Specifically, a single base station is able to handle more than one million messages a day as specified in the ETSI TS 103 357 standard . Also, in an Industrial IoT-equivalent scenario, Telegram Splitting has been proved to drastically outperform Chirp Spread Spectrum in LoRa in terms of message deliverability and network reliability .
In addition to Quality-of-Service, the characteristics of Telegram Splitting, at the same time, offer great power benefits. After the transmission of each sub-packet, there is a significantly longer transmission-free period in which the node goes into “sleep mode”. Short on-air time and longer off-air time minimize power consumption while giving the battery time to recover, which in turn significantly extends battery life.
Short time in the air interface of sub-packets combined with coherent demodulation additionally diminish Doppler fading effects. And, even if some sub-packets suffer from deep fades, FEC ensures that message detection and retrieval is minimally affected. With this, Telegram Splitting systems can connect end nodes moving at up 120km/h —a feature not available in previous LPWAN technologies. Figure 7 quickly recaps how the four major LPWAN technology types compare in different network criteria.
Providing a unique combination of long-range, low-power and low-cost advantages, LPWANs are poised to become the backbone of battery-operated IoT sensor networks across verticals. Nevertheless, not all LPWAN technologies are created equal, and there exists a persistent trade-off between Quality-of-Service and battery life among most solutions. At the same time, the lack of standardization and limited mobility support are other challenges not to be overlooked. Recognized for its versatile technical design, Telegram Splitting represents a new LPWAN generation to surpass these limitations and provide a robust, scalable and power-efficient architecture for massive-scale IoT deployments in the industrial and commercial marketplaces.
 European Telecommunications Standards Institute, “Short range devices; Low Throughput Networks (LTN); Protocols for radio interface A”, European Telecommunications Standards Institute, ETSI TS 103 357, 2018. https://www.etsi.org/deliver/etsi_ts/103300_103399/103357/01.01.01_60/ts_103357v010101p.pdf
 K. Mekki, E. Bajic, F. Chaxel, and F. Meyer. “A comparative study of LPWAN technologies for large-scale IoT deployment, in ICT Express, 2018.
 J. Bardyn, T. Melly, O. Seller and N. Sornin, “IoT: The era of LPWAN is starting now,” ESSCIRC Conference 2016: 42nd European Solid-State Circuits Conference, Lausanne, 2016.
 J. Petäjäjärvi, et al., “Performance of a low-power wide area network based on LoRa technology: Doppler robustness, scalability and coverage”, International Journal of Distributed Sensor Networks, 2017.
 A. Lavric, “LoRa High-Density Sensors for Internet of Things”, Journal of Sensors, 2019.
 T. Lauterbach, “MYTHINGS vs LoRa: A comparative study of Quality-of-Service under external interference”, 2019. https://behrtech.com/resources/lora-vs-mythings/
Behr Technologies | www.behrtech.com
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • MARCH 2020 #356 – Get a PDF of the issue