Chapter 16
Choice of IoT technology
Abstract
This chapter discusses the options of communication technologies for IoT. In a first part, the options of cellular IoT versus non-cellular IoT solutions are assessed.
In a second part, choices of cellular IoT technologies are compared. First, cellular IoT solutions for massive IoT are analyzed, comprising LTE-M and NB-IoT. The characteristics of those technologies are compared with regard to deployment, achievable data rates, latencies, spectral efficiency and battery efficiency. Second, cellular IoT solutions for critical IoT are analyzed, comprising Long-Term Evolution URLLC and NR URLLC.
In a third part, the considerations are described on how to assess the choices of cellular IoT technologies from the perspective of a mobile network operator and an IoT service provider.
Keywords
5G; Cellular internet of things (CIoT); Internet of things (IoT); Long-term evolution (LTE); Long-term evolution for machine-type communications (LTE-M); Machine-type communications (MTC); Narrowband IoT (NB-IoT); Massive machine-type communications (mMTC); Critical machine-type communications (cMTC); Ultra-reliable and low latency communications (URLLC); Massive IoT; Critical IoT; Broadband IoT; Industrial IoT
16.1. Cellular IoT versus non-cellular IoT
Chapters 3 to 13 present 3GPP cellular IoT solutions. Chapters 14 and 15 provide an overview of interesting unlicensed wireless connectivity solutions for IoT. In this section we discuss how cellular IoT solutions differ from unlicensed connectivity solutions and what benefits they can provide. For a further discussion on IoT connectivity options, see also [1].
One of the differentiators of cellular IoT connectivity is that it decouples connectivity provisioning from the IoT service realization. Cellular IoT is built on the high-level paradigm, that an independent operator provides suitable IoT connectivity essentially everywhere where an IoT service shall be realized. This means that when a new IoT service is established, no dedicated effort needs to be put into installing, managing and operating an IoT connectivity solution. Instead the connectivity is realized via the network of an operator. This is different from unlicensed IoT connectivity solutions. In this case, an installation of a connectivity infrastructure is needed to provide connectivity at the location where the IoT service is to be realized. This comprises installing of base stations or access points, establishing backhaul connectivity, providing authentication, authorization and accounting infrastructure, maintaining and updating the connectivity network with security updates, etc. Furthermore, the connectivity needs to be monitored and managed throughout the lifetime of the IoT service. There is the potential that the total cost of ownership for providing and managing a connectivity infrastructure for a wide range of IoT services is lower than the cost of ownership of separate connectivity solutions per IoT service. This is in particular the case when the IoT service and the participating devices are spread over larger areas and are not confined to limited deployments. From the unlicensed technologies, Sigfox provides an operator model for Sigfox-based end-to-end connectivity, where operators build up a dedicated Sigfox infrastructure and connectivity can be purchased by end users.
For critical IoT solutions, providing critical Machine-Type Communications (cMTC) services, the connectivity is often tighter integrated into the critical system. Critical systems can be local (e.g. a factory), and often a dedicated deployment of network components is required to guarantee performance in terms of capacity, latency, reliability and availability. Even for critical IoT solutions, the cellular IoT deployment and operation may be decoupled from the operation of the service realization of the end-to-end system (e.g. the automation of the production system). However, due to the tight relationship of the cellular critical IoT solution with the service performance and service assurance, a close coordination between the communication service provider and the end user are needed. This can be done in the form of a stringent service level agreement combined with service performance auditing. But it can also be realized in that the communication system is provided by the end user directly with own installation and operation of the cellular IoT system.
One major benefit of cellular IoT solutions is that they provide a reliable long-term and future proof solution. Cellular IoT is based on global standards with very large industry support by a large number of vendors, network and service providers. The technology outlook is independent from the outlook of few individual market players; this is in contrast to proprietary technologies which come with high risk concerning their long-term support.
Cellular IoT solutions are embedded into cellular communication networks, which are, and will be, an essential infrastructure for a society. Deployment plans are made over decades and systems are built to be highly reliable according to standards with high availability. Cellular IoT systems are built for a global market and allow roaming over multiple operator networks. Cellular IoT networks have full support for mobility of devices, which can also be handled over larger areas due to the wide area coverage and high availability. The rollout of cellular IoT capabilities, as well as future updates, takes mainly place as software updates to the installed network infrastructure.
One extremely important benefit of cellular IoT connectivity is that it provides reliable and predictable service performance also for future operation. For wireless critical IoT services, it is only cellular IoT that can guarantee reliable low latency services at scale. Cellular IoT uses dedicated spectrum. Radio resources are managed, interference is coordinated, and full quality of service is supported. Long-term guarantees are challenging to provide for any solution based on unlicensed spectrum. Both mobile broadband services, as well as IoT services are predicted to continue to grow. In particular for IoT devices, an extremely strong growth is predicted leading to hundreds of billions of communicating devices within a decade. Many of those mobile broadband and IoT services will be provided in unlicensed spectrum, which means that a significant increase in utilization of unlicensed spectrum can be expected. This will in particular provide challenges to long-range unlicensed technologies as described in Section 14.1.2 and Chapter 15; but also for critical IoT services where low latency and high reliability are required.
Cellular IoT also follows the continuous evolution of cellular network technologies, where new capabilities and features are continuously added to the networks. This evolution is designed for backwards compatible operation, so that devices that cannot be upgraded to new functionality can continue to operate long-term according to the original capabilities, while new services and devices can simultaneously benefit from newer features.
A disadvantage of cellular IoT is the cost of licensed spectrum resources. This is a cost which solutions for unlicensed bands do not need to bear. Another potential disadvantage of cellular IoT for an IoT service provider can be if the cellular IoT coverage is insufficient for a specific IoT use case. In this case extra connectivity and corresponding network buildout may be needed to cover the entire IoT service area. If the additional coverage buildout is needed in few confined areas, such a buildout may be simpler and more flexibly arranged with a dedicated deployment rather than when an operator needs to be involved. One property that unlicensed long-range radio technologies have benefitted from is their fast time-to-market. Any proprietary technology has a timing benefit over standardized solutions that require harmonization and agreements within an entire industry segment. In case of unlicensed LPWAN versus cellular IoT, a benefit in time-to-market of unlicensed LPWAN has existed in the past while the cellular IoT standards were being developed. Since the first cellular IoT standards were finalized in 3GPP Release 13 and products became widely available, the time-to-market benefit of unlicensed LPWAN has disappeared. Instead the benefit shifts toward cellular IoT deployments, which can reach wide coverage quickly and at low cost due to the reuse of the installed cellular communication network infrastructure.
16.2. Choice of cellular IoT technology
16.2.1. Cellular technologies for massive IoT
Use cases for massive IoT are characterized by low complexity and cost, energy efficient operation, and ubiquitous deployments of massive number of devices. The scale of the deployments requires wireless coverage under the most challenging conditions, and devices that can operate on non-rechargeable batteries for years. The anticipated traffic profile for many massive IoT applications is characterized by small and infrequent data transmission. A good example of a massive IoT deployment is the Great Britain Smart Metering Implementation Program where the British government has decided to equip every home and small business in the country with advanced electricity and gas meters. At the end of 2018 12.8
million smart meters had been rolled out [2].
To support a design of the cellular IoT technologies that can meet the demands of massive IoT, 3GPP have agreed the following set of performance objectives:
- - Support for a coverage of 164 dB maximum coupling loss (MCL). This corresponds to 20 dB improved coverage compared to the 3GPP non-cellular IoT technologies.
- - Support for small and infrequent data transmissions on non-rechargeable batteries for up to 10 years.
- - Support for a connection density as high as 1,000,000 connections per square kilometer.
- - Support for low to ultra-low device complexity.
Coverage is meaningful first when coupled with a quality of service requirement. 3GPP therefore in addition requires that a cellular IoT system should be able to:
The earlier chapters of the book in detail covers the 3GPP Cellular IoT technologies EC-GSM-IoT, LTE-M and Narrowband IoT (NB-IoT). It also presents the work done in MFA to prepare LTE-M and NB-IoT for operation in unlicensed spectrum bands. NB-IoT and LTE-M have, since their introduction in 3GPP Release 13 until March 2019, been commercially deployed in 89 networks across Europe, North and South America, Africa, Asia and Australia [3]. The number of commercial networks is constantly increasing, and many operators are choosing to deploy both technologies. The technologies are supported by the major infrastructure vendors and a significant amount of chipset, module and device providers. EC-GSM-IoT is still waiting for a commercial uptake, while the MFA technologies are still, in March 2019
at the writing of this book, undergoing RF specification work. The following sections therefore naturally focus on the capabilities and performance of LTE-M and NB-IoT.
16.2.1.1. Spectrum aspects
LTE-M and NB-IoT are both LTE technologies and do as such support a long list of frequency bands in the range 450
MHz to just below 3
GHz. Cat-M1 and Cat-M2 devices are capable of operating in E-UTRA FDD bands 1, 2, 3, 4, 5, 7, 8, 11, 12, 13, 14, 18, 19, 20, 21, 25,
26, 27, 28, 31, 66, 71, 72, 73, 74 and 85 in both half-duplex (HD) and full-duplex (FD) FDD mode and in time-division duplex (TDD) bands 39, 40 and 41. Cat-NB1 and Cat-NB2 support E-UTRA bands 1, 2, 3, 4, 5, 8, 11, 12, 13, 14, 17, 18, 19, 20, 21, 25, 26, 28, 31, 41, 65, 66, 70, 71, 72, 73, 74 and 85 in HD-FDD, and band 41 for TDD. The list of bands with their associated frequency ranges is found in 3GPP TS 36.101 [4]. As 3GPP specifies the support for new bands based on market interest also these lists have been growing for every new release of the specification. Both technologies can operate in the NR bands corresponding to the same lists of E-UTRA bands. This is thanks to the forward compatible coexistence functions specified for NR. These are in detail described in Chapters 5 and 7. NB-IoT supports thanks to its configurable mode of operation and small spectrum footprint also operation in GSM spectrum.
LTE-M is natively supported on the LTE system bandwidths of 1.4, 3, 5, 10, 15 and 20
MHz. Cat-M1 devices operate on a channel bandwidth of up to 1.4
MHz, while Cat-M2 can make use of up to 5
MHz for its transmissions. These RF bandwidths are determined based on the spectrum within which 99% of the signal energy falls. In terms of physical resource blocks (PRB) the 1.4
MHz narrowband corresponds to 6 PRBs, while 5
MHz corresponds to 25 PRBs. NB-IoT operates at a minimal RF system bandwidth of 200
kHz, which is equivalent to 1 PRB of 180
kHz. The system capacity can be scaled by increasing the number of carriers. Carrier aggregation (CA) is not supported by NB-IoT, so it is system capacity, and not link capacity that scales with the number of deployed carriers.
LTE-M's limited spectrum footprint makes it an attractive option for deployments in narrow, or fragmented frequency bands, but it can also efficiently and dynamically make use of a larger system bandwidth if available. NB-IoT's even more limited spectrum footprint makes it 3GPP's most flexible system in terms of deployment. Its ability to operate in the guard-band of an LTE carrier is e.g. very attractive as it allows an LTE operator to make better use of its highly valuable spectrum assets. A detailed explanation of the NB-IoT modes of operation, i.e. stand-alone, in-band and guard-band, is provided in Section 7.1.2.5.
LTE-M-U and NB-IoT-U technologies are interesting alternatives for vendors lacking access to licensed spectrum. The current designs are expected to support operation in US and Europe according to the FCC and ETSI regulations, respectively. But it needs to be stressed that the coverage, link quality, reliability and capacity provided by LTE-M-U and NB-IoT-U are not comparable to that provided by LTE-M and NB-IoT.
16.2.1.2. Features and capabilities
LTE-M and NB-IoT have many similarities. This is the result of years of parallel development in 3GPP. But there are important differences. Low complexity and simplicity, both on the device and system side, runs through every aspect of the NB-IoT design. NB-IoT aims at being the low-cost massive IoT carrier with capabilities in terms of reachability, device power efficiency and system capacity. NB-IoT Release 15 does therefore not support features like voice, connected mode mobility, connected mode device measurements and reporting, and closed loop power control which were not deemed relevant for the intended usage of NB-IoT. The Control Plane Cellular Internet of Things EPS optimization feature (see Section 7.3.1.7), which is mandatory for Cat-NB devices, does not support RRC Reconfiguration of the access stratum for a device in RRC connected mode, nor does it support the MAC layer data radio bearer scheduling and prioritization developed to
guarantee a targeted quality of service. This is a consequence of routing the data over the control plane on a signaling radio bearer. This functionality is available for Cat-NB devices capable of user plane data transfer, which is an optional capability for NB-IoT.
LTE-M, being an LTE system, is naturally supporting a richer set of features than NB-IoT. LTE-M supports more advanced transmission modes using up to four antenna ports. It supports wideband transmissions, voice, connected mode mobility and full duplex operation. The set of massive IoT use cases to be supported by an operator goes in many cases beyond those characterized by small and infrequent data transmission. This observation has been a driver for keeping LTE-M more capable and advanced compared to NB-IoT. With that said, remember that LTE-M supports two coverage enhancements (CE) modes A and B, with CE mode B providing the most extreme coverage is an optional feature. So, when it comes to supporting operation in extreme coverage NB-IoT can be claimed to be the more capable system.
16.2.1.3. Coverage
The 3GPP 5G objective requires a 5G system for massive IoT to support a coverage of 164
dB MCL [5]. In Section 6.2 and Section 8.9 we provide detailed evaluations of the coverage supported by LTE-M and NB-IoT. The same evaluation assumptions, which are aligned with those agreed by ITU-R for the IMT-2020 evaluation [6], are used in both cases. It is shown that LTE-M and NB-IoT meet the 164
dB requirement. For both technologies the PUSCH can be seen as the limiting channel, i.e. the channel which requires the longest transmission times to reach the 164
dB coverage target. For LTE-M the MPDCCH needs to be configured with 256 repetitions to achieve the 1% BLER target set for the control channel transmission. This is the maximum configurable repetition number for the MPDCCH. So, for LTE-M the MPDCCH coverage is also a limiting factor.
The transmission times used for all channels to reach 164
dB, except for the MPDCCH, are not using the maximum configurable transmission times. We can therefore extend coverage beyond 164
dB if we can accept to reduce the data rates, latencies and battery life presented in in the next sections.
The performance evaluations for EC-GSM-IoT described in Chapter 4 follows slightly different assumptions compared to those used for 5G. The result shows that EC-GSM-IoT can offer similar cell-edge coverage and performance as NB-IoT and LTE-M. When looking at NB-IoT-U and LTE-M-U, described in Chapter 15, we can conclude that NB-IoT-U offers a coverage 10–20
dB below 164
dB. For LTE-M-U which only supports CE mode A a similar conclusion is expected to hold.
16.2.1.4. Data rate
Tables 16.1 and 16.2 summarize the range of data rates achievable for NB-IoT and LTE-M. For simplicity, here we focus on the MAC-layer data rates achievable at the 164
dB MCL, and the MAC-layer data rates observed under error-free conditions. We also present the physical layer data rates. Remember that the MAC-layer data rates correspond to the efficient data rate a device may experience, while the physical layer data rate is defined by the throughput during the actual transmission time interval of the PDSCH or PUSCH. The physical layer data rate, when normalized by the signal bandwidth, should be seen as an indication of the maximum spectral efficiency that can be achieved.
The two tables present Cat-NB1 and Cat-M1 Release 13 performance, and Cat-NB2 and Cat-M2 Release 14 performance. Cat-M1 PUSCH data rate can be further improved compared to the numbers in the table by means of the larger uplink TBS introduced in Release 14, and Cat-M1/M2 PDSCH data rates can be improved by means of the Release 14 feature for HARQ bundling and 10 HARQ processes in downlink. The NB-IoT results are based on the guard-band mode of operation. Chapters 6 and 8 give a richer set of results including data rates for the just mentioned LTE-M features and all three NB-IoT modes of operation.
The result speaks for themselves: The cell-edge MAC-layer data rates for NB-IoT and LTE-M are similar and meet the 5G requirement of 160
bps [5]. LTE-M can offer significantly higher data rates thanks to the larger device bandwidths and the lower processing times. Here we focus on error-free conditions, but this observation holds for a large portion of a cell.
16.2.1.5. Latency
As already mentioned, many massive IoT use cases are characterized by small data transmission. For these the importance of the data rates presented in the previous section is overshadowed by the latency required to set up a connection and transmit a data packet of limited size.
Table 16.3 shows the latency for small data transmission achievable for LTE-M and NB-IoT at the 164
dB MCL based on the two procedures Early Data Transmission (EDT) and RRC Resume. For EDT data is MAC multiplexed with the RRC connection resume request
message in Message 3. For RRC Resume the data is multiplexed with the RRC connection resume complete message in Message 5. Figure 6.4 illustrates this MAC layer multiplexing of data and RRC messages. Here we focus on procedures where data is transmitted over the user plane. Message 3 and 5 data transmission is also supported over the control plane. Those methods are expected to provide similar latencies as shown here for the user plane.
Table 16.3
| Technology | Method | Latency [s] |
|---|---|---|
| LTE-M | EDT | 5.0 s |
| RRC Resume | 7.7 s | |
| NB-IoT | EDT | 5.8 s |
| RRC Resume | 9.0 s |
LTE-M is seen to perform slightly better than NB-IoT. The reason is that the MPDCCH manages to achieve 164
dB MCL for a transmission time of 256
ms. For the NPDCCH, the same transmission time results in an MCL just below 164
dB. The results are therefore based on an NPDCCH transmission over 512
ms, which supports an MCL of 166
dB. The performance difference presented in Table 16.3 should therefore in practice be negligible.
Both technologies meet the 5G requirement of a 10 s latency with margin [5]. Chapters 6 and 8 present the detailed assumptions behind these results. They also add more results for different radio conditions to give a more complete picture of the massive IoT latency for small data transmission.
16.2.1.6. Battery life
5G massive IoT devices are required to support operation on a non-rechargeable battery power of 5
Wh for over 10
years [5]. Table 16.4 presents the estimated battery lives of Cat-M and Cat-NB devices when operating at the 164
dB MCL under the assumption that a 200-byte uplink report is sent to the network once per day. The RRC Resume procedure was used in these evaluations. The EDT procedure could in principle have been used if it had not been for a limitation of the maximum TBS to 1000 bits for the Message 3 data transmission. The 5G requirement of 10 years battery lifetime is met by both technologies.
The evaluation assumption behind these results are again found in Chapters 6 and 8. Noteworthy is the assumed device transmit power level of 500
mW, which has been deemed realistic by 3GPP [7]. Lab and field measurements performed on commercial NB-IoT and
LTE-M devices suggest that this may be a somewhat optimistic assumption. A higher transmit power level would naturally reduce the supported battery life, unless compensated for by a redimensioning of the battery power supply.
16.2.1.7. Connection density
5G requires that a massive IoT system can serve up to 1,000,000 connections per square kilometer for a traffic model where each device is triggered to send a 32-byte uplink report every second hour [8]. In practice this means that the system should support 1,000,000 connections across 2
h, or on average 139 connection establishments per second.
Fig. 16.1 presents the number of connection establishments LTE-M and NB-IoT can support per cell, second and PRB versus the 99th percentile latency the systems offer [9]. According to the 5G requirement the supported connection density is defined at the load where 99% of the connections are served with a latency of 10
s or less. The results are depicted for four different urban macro scenarios defined by Ref. [10]:
For LTE-M the minimum system bandwidth is a narrowband defined by 6 consecutive PRBs. To estimate the LTE-M capacity per narrowband the results in Fig. 16.1 should be scaled by a factor 6.
The knowledge that 1,000,000 connections are equivalent to 139 connection establishments per second allows us to use the simulated results in Fig. 16.1 to calculate the resources needed to support the data traffic generated by the 1,000,000 connections. Table 16.5 shows the results, i.e. the resources needed to support the connection density that 5G is required to support. For LTE-M two PRBs are added to the narrowbands. These are carrying the PUCCH transmissions which are sent outside of the narrowbands close to the cell edges of the LTE system bandwidth.
The general conclusion from the results presented in Fig. 16.1 and Table 16.5 is that both LTE-M and NB-IoT meet the 5G requirement for bandwidths in the range of 1 PRB to 20 PRBs, depending on the scenario. Remember that 1 PRB corresponds to a bandwidth of 180
kHz. The denser deployment making use of a 500
m inter-site distance can carry a significantly higher load that the 1732-m case. The difference in capacity between the two configurations is in parity with the difference in cell area between the two cases. NB-IoT is shown to provide slightly higher capacity, which is explained by the efficient use of subcarrier NPUSCH transmissions in challenging radio conditions. LTE-M is shown to provide better latency, which is due to the ability to use a higher bandwidth in good radio conditions. The LTE-M capacity can be further improved by using the PUSCH sub-PRB feature introduced in Release 15 for users in bad radio conditions.
16.2.1.8. Device complexity
The massive IoT technologies have introduced similar features to reduce device complexity, and thereby enabling low-cost IoT devices. The following design objectives have been pursued for low device complexity:
The frequency bandwidth used by the device for transmitting and receiving has been limited to avoid the high costs of wideband front ends. For LTE-M the RF bandwidth that
needs to be supported by a device is 1.4
MHz, which is significantly less than the maximum LTE channel bandwidth of 20
MHz. For NB-IoT the RF bandwidth that needs to be supported by the device is 200
kHz.
Table 16.5
The physical layer peak data rate has been limited to reduce processing and memory requirements for a device. For Cat-M1 the peak rate has been reduced to 1
Mbps; for Cat-NB1 the peak rate has been limited to below 300
kbps.
Both technologies are specified so that the devices are not required to use more than one antenna to fulfill the performance requirements.
The technologies have been specified to support half-duplex operation. This avoids the needs for a device to integrate one or more costly duplex filters. LTE-M and NB-IoT devices can be implemented with support for half-duplex frequency-division duplex or TDD operation, and LTE-M devices can optionally be implemented with support for full-duplex frequency-division duplex.
The technologies have defined User Equipment categories with lower power classes. This enables a device to use cheaper power amplifiers. It can become an option to implement the power amplifier on the modem chip, and thereby avoiding the costs of a separate component. Three device power classes of 14, 20 and 23
dBm output power are supported.
The features above enable reducing the device costs for the IoT devices. However, it must be noted, that the device cost is not exclusively depending on the communication standard. The cost of the device depends also on what peripherals are added to the device, such as power supply, CPU, or the real-time clock.
In the end, the costs of the device depend on the market success and the market volume of the devices. A large economy of scale will help to reduce the production costs. Due to this many chipset vendors have decided to develop a multi-mode implementation supporting both LTE-M and NB-IoT. For devices targeting the lowest cost, NB-IoT only implementation appears to be the main choice. When it comes to EC-GSM-IoT, commercial chipset and module support remains unavailable on the cellular IoT market.
16.2.2. Cellular technologies for critical IoT
Use cases for critical IoT, or critical Machine-Type Communications (cMTC), appear mainly in the field of the industrial internet of things (see also Chapter 17) and can generally be grouped into wide-area and local area use cases. 3GPP has characterised several use cases and derived their corresponding requirements in Refs. [11–14]. Wide-area use cases include remote driving of vehicles and automation in intelligent transport systems, automated trains and rail-bound mass transit, or automation of the medium- and high-voltage energy distribution. The upper end of maximum tolerable end-to-end latency latencies for reliable communication over larger distances is in the range of ∼100
ms and even ∼500
ms for some use cases of train communication, and ∼50–100
ms for some of the smart energy grid use cases. On the lower end, 5
ms end-to-end latency is needed for remote driving, 10
ms for critical rail communication over distances up to 2
km, and 5
ms for fast switching and isolation in the power grid over distances up to ∼30
km (Fig. 16.2).
Other cMTC use cases are focused on local areas. Most prominent are applications of manufacturing for smart factories: motion control, distributed control and mobile robot applications may require ultra-reliable communication within end-to-end latencies between 0.5 and 10
ms. In these cases, the range of communication is within a confined area, e.g. the factory, and is often indoor. For process control used in the process industries – like for example chemical industry, mining, oil and gas industries – the latency requirements are a bit more relaxed compared to manufacturing. Latency requirements for closed-loop process control can get as low as 10
ms and for process monitoring they are at 100's of ms or even up to seconds. For process control, the confined areas are larger, up to 10's of km2 including outdoor areas.
What is common in the cMTC use cases above, is that latency plays a critical role and is required to stay within maximum delay bounds. Exceeding a maximum latency in cMTC services can lead to severe consequences, for example a production system comes to a halt, or a short-circuit in the energy grid is not detected and isolated in time.
When using cellular IoT for cMTC services, a suitable spectrum deployment needs to be considered, see Table 16.6. Mobile networks are deployed over different spectrum bands, including low spectrum bands below 1
GHz, and in the higher range 1–6
GHz referred to as mid-band. For 5G NR, also high spectrum bands above 6
GHz are added, which are initially mainly available in the range 24–40
GHz. Mobile network operators have typically licensed a number of spectrum bands and make use of those multiple bands in their deployments. To use them efficiently, mobile networks allow the integration of multiple bands for the transmission with a user equipment, with two defined mechanisms [15]. Dual connectivity (DC) allows to set up and aggregate two communication paths that can use radio links of different carriers with separate scheduling entities per carrier on the network side. CA allows to pool the radio resource of multiple carriers for the transmission and scheduling
of a user; a joint scheduling entity assigns the resources of the carriers enabling a higher resource efficiency compared to DC. For DC feedback, signaling is needed for each carrier separately, splitting the devices transmission power among those control channels; the control channel can thereby limit the achievable coverage. For CA, a combined feedback is provided on a single control channel which provides better coverage compared to DC.
Table 16.6
In sub-urban and rural areas, a mobile network buildout is typically not so dense; low frequency bands up to 1
GHz provide good coverage. Those can be complemented with spectrum carriers in mid-band, but for those bands full coverage is more challenging due to propagation. In denser network deployments, mid-band spectrum bands are used, and in very dense deployments also high band millimeter wave spectrum will be used in future. These spectrum bands differ in their characteristics. In spectrum bands up to 2.6
GHz, FDD allocations dominate, even if some TDD allocations can also be found in the lower mid-band spectrum. In these bands the bandwidth of the carriers available to an operator are typically small, rarely exceeding 20
MHz and often even less. The higher mid-band range of 2.6–6
GHz spectrum is based on TDD, and here larger bandwidth allocations up to e.g. 100
MHz are available. Future spectrum for 5G in the high-band millimeter wave spectrum will also be based on TDD and provide carrier bandwidths that are significantly larger than 100
MHz.
LTE can only be deployed in low-band and mid-band spectrum. NR covers all mentioned spectrum ranges, and new spectrum bands have been identified for 5G in low-band, mid-band and high-band. There are some regional variations among countries about the exact band availability, and also timelines of spectrum auctions vary. In 2019, NR networks are commercially deployed in low, mid and high bands. In addition, NR can be used in spectrum bands already allocated to mobile communications, which allows for a migration of spectrum carriers from 2G, 3G or LTE to NR. A very effective way for fast NR roll-out that reuses the installed network infrastructure and reaches quickly large NR coverage, is to use LTE-NR spectrum sharing [16]. The NR and LTE specifications allow efficient sharing of the radio resources of a carrier among the two radio technologies, so that the same carrier appears as an NR carrier to NR devices, and as an LTE carrier to LTE devices. Radio
resources are pooled and can be dynamically distributed between LTE and NR according to instantaneous needs. LTE-NR spectrum sharing is a means of technology migration that is significantly more flexible to traditional spectrum refarming, where the usage of a carrier is first terminated for one radio access technology before a new one is introduced. For NR, spectrum sharing with LTE allows a (slim) NR carrier to be introduced at a low band carrier to provide very quickly wide-area coverage and apply CA to combine low-band coverage with high NR capacity provided by mid-band or high-band NR deployments.
Some of the cMTC use cases are confined to local areas, such as a factory. For such use cases, typically a dedicated network deployment is needed, in order to provide sufficient service coverage and availability for the high QoS requirements. Such a dedicated deployment may be limited to the radio network, extending the outdoor macro network. But most common will be network deployments were larger parts of the cellular IoT network are deployed locally, and possibly even the entire network. A local termination of the user-plane path is motivated for several reasons. By terminating the gateway function locally, the end-to-end latency is significantly reduced compared to routing it via a macro network. This local break-out can be combined with edge computing where application functions can be executed at, or close to, the cellular IoT network gateway, e.g. the user plane function of the 5G core network. This has further benefits, that for example, business critical information of the operation of the industrial system is not leaving the premise; also, the integration of the cellular system with existing communication systems and IT infrastructure is simpler. A further discussion on local industrial deployments can be found in Refs. [17,18].
For cMTC services, wider bandwidth allocations are desireable for several reasons. The need for being able to transmit an arriving data packet in short time, implies that instantaneous access to radio resources are needed and cannot be deferred in time. But in several cMTC use cases also a high traffic demand exists, for example, in manufacturing where hundreds of industrial controllers and industrial robots may need to communicate in real-time with ultra-low latency. For those use cases, the upper mid-band spectrum or high-band spectrum are beneficial since they provide larger spectrum allocations, see Table 16.6. In low bands, and largely also in the lower mid bands, only smaller carrier bandwidths are typically available, which limits the capability to support cMTC services, in particular if higher cMTC capacity is required. Methods like carrier aggregation are needed to add additional spectrum bandwidth to the communication system.
As shown in Chapters 10 and 12, both LTE and NR are capable to achieve the IMT-2020 Ultra-Reliable and Low Latency Communications (URLLC) requirement of achieving a 32
B data transmission over the RAN within 1
ms with a reliability of 99.999%. Owing to its design, NR can achieve much lower latencies than LTE, in particular if higher subcarrier spacing (SCS) is used. The choice of SCS is coupled to the network deployment and carrier frequency. In high bands, phase noise limits the SCS to higher SCS. At the same time, higher SCS implies shorter cyclic prefix of the OFDM symbols making the communication more sensitive to time-dispersion of the radio signal. Therefore, macro deployments with larger cell sizes, as often used in low band spectrum, require lower SCS. Suitable NR SCS values are listed in Table 16.7.
Table 16.8 shows the latencies that are achievable with high reliability for LTE and NR in different bands; configurations achieving the lowest latencies are selected according to the
features described in Chapter 9 and 11, and the results are from the evaluations done in Chapter 10 and 12. For LTE, we assume subslot operation; the NR configuration assumes for FDD a 30
kHz SCS and 2-symbol mini-slot, and for TDD a 120
kHz SCS, 4-symbol mini-slot and alternating uplink-downlink TDD slot configuration.
Table 16.7
| Low-band (up to 1 GHz) | Lower Mid-band (1–2.6 GHz) | Higher Mid-band (2.6–6 GHz) | High-band (above 6 GHz) | |
|---|---|---|---|---|
| Suitable NR SCS | 15, 30 | 30, 60 | 30, 60 | 60, 120 |
Table 16.8
| Low/mid-band FDD | Mid/high-band TDD | |||
|---|---|---|---|---|
| Downlink | Uplink | Downlink | Uplink | |
| LTE | 0.86 ms | 0.86 ms | – | – |
| NR | 0.48 ms | 0.48 ms | 0.51 ms | 0.64 ms |
Besides the achievable lower latencies, NR has the benefit that it supports URLLC in both FDD and TDD bands, or combinations of such bands. NR also achieves a higher spectral efficiency and service coverage for URLLC services than LTE, owing to its more flexible configuration, as also presented in Chapter 10 and 12. An additional analysis of both LTE and NR performance for URLLC can also be found in Ref. [19].
Many cMTC services have advanced requirements going beyond only the transmission latency. cMTC for industrial systems often require an interworking with industrial Ethernet or IEEE 802.1 time-sensitive networking (TSN), see also Section 17.2. Other requirements are that the communication system must be able to provide high-precision timing information toward devices for being able to synchronize devices with 1
μs precision. LTE provides a basic time synchronization for devices in 3GPP Release-15. For NR and the 5G core network, advanced synchronization features are added in the standardization from 3GPP Release-16, which enables devices to time-synchronize to multiple external time domains, which is needed for industrial automation scenarios. Also, mechanisms for enabling redundant transmission paths for a device throughout the core network are being defined. No activities are currently ongoing for introducing similar features into LTE and the 4G evolved packet core.
In conclusion, both LTE and NR fulfill the 5G requirements from ITU-R (IMT-2020) on URLLC and are able to transmit small packets reliably within 1
ms. NR is more flexible in its design than LTE and has higher performance. NR can not only support URLLC in FDD but also in TDD bands. Further, NR can achieve significantly lower latencies than
LTE, in particular in local deployments, and has also higher spectral efficiency. Finally, NR is more future-proof in its evolution, providing capabilities needed for many cMTC use cases, such as support for redundant transmission paths, time synchronization over the radio interface, and interworking with TSN networks.
16.3. Which cellular IoT technology to select
The choice of a cellular IoT solution is a decision that needs to be taken by different market players. On one hand, it is the mobile network operator that has to decide which cellular IoT technology to add to its existing network. On the other hand, it is the IoT device manufacturer and service provider that have to select for which IoT connectivity options they develop their IoT service. Finally, it is the end customer that may demand what technology to use. The latter is in particular the case for enterprise customers in need of critical IoT services. It can be expected that different options of solutions will coexist.
16.3.1. The mobile network operator's perspective
For a mobile network operator, the decision about which cellular IoT technology to deploy and operate has multiple facets. There are in particular two sides that need to be considered:
A typical mobile network operator has one or more cellular networks deployed. Increasingly, different radio technologies are provided via a single multi-radio technology network. For example, the same base station can be used for GSM, UMTS/HSPA, LTE or NR transmission. But there are also deployments where the networks for 2G, 3G, 4G and 5G are rather independent in their deployment and operation.
In addition, a mobile network operator has a spectrum license from typically a national regulator, which gives rights to operate a network in the assigned spectrum. Spectrum licenses are long-lasting, e.g. 20
years; this is motivated by providing network operators with an economic safety. A return on investment for an extremely high network installation cost of a new technology can be planned over a long time period. On expiry of a spectrum license, a spectrum licensing contest, like a spectrum auction, is initiated by the regulator for providing a new spectrum license. In general, any network build-out roadmap by an operator is a long-term decision and needs to consider at least the following elements:
- - How long are the existing spectrum licenses valid, what technologies are allowed to be operated in the spectrum, and when is a new spectrum re-allocation process planned by the regulator?
- - What is the status of the existing network buildout for different radio technologies, in particular GSM and LTE, and what is the number of mobile devices and projected growth for each specific technology?
- - What is the network buildout of competing operators and what is their market share?
- - What is the strategic intent of an operator concerning IoT?
- - What services are planned to be provided, and on what roles does the operator intend to take (e.g. as a connectivity provider or also as a service provider/enabler)?
It shall be noted that the above questions are raised from a perspective of the operation of an operator network in a specific country. However, several operators are active in multiple countries and even on multiple continents. Even if the decision is largely made per country, an operator may want to harmonize decisions over multiple regions in which it operates networks.
When looking at the cellular IoT technology options, the following characteristics can be identified which will influence an operator's decision.
As a baseline, we assume that there is a very large incentive by an operator to reuse existing mobile network infrastructure for deploying any of the cellular IoT technologies. EC-GSM-IoT can be deployed based on a GSM infrastructure and by using GSM spectrum. The GSM network resources and the GSM spectrum would be shared between GSM usage and EC-GSM-IoT usage. LTE-M and NB-IoT can be deployed based on LTE infrastructure and by using the LTE spectrum; LTE network and spectrum resources would be shared between LTE, LTE-M and NB-IoT usage. In most network configurations, it can be expected that the deployment of EC-GSM-IoT, LTE-M and NB-IoT can be realized as a software upgrade to the deployed GSM or LTE networks. This implies that the introduction of the cellular IoT into the market can be realized by operators rather quickly and at a low total cost of ownership. NR has been designed to allow efficient interworking with LTE (including LTE-M) and NB-IoT. This means that LTE-M and NB-IoT can be embedded into an NR carrier in a similar way as they are today integrated into an LTE carrier. In general, it is possible to share a carrier between NR and LTE, where the resources being used for LTE or NR transmission can be dynamically adapted [16]. The LTE-NR coexistence flexibility also enables to migrate an LTE carrier to NR, while continuing LTE-M or NB-IoT devices with long device lifetime to continue operation within the NR carrier after the migration. For IoT, it is expected that many services expect a long lifetime of e.g. a decade. This expectation should be addressed with a cellular IoT network. As a result, the decision of the cellular IoT technology is also coupled to the operator's long-term strategy for mobile networks focusing on telephony and mobile broadband services. If an operator intends to transition GSM deployments to e.g. LTE or NR in the coming future, an introduction of EC-GSM-IoT seems a questionable choice, as any long-term EC-GSM-IoT users would require maintaining the GSM infrastructure operational for a long time. A general trend that is seen globally, is that 2G and 3G spectrum allocations are stepwise migrated to LTE [20]. With the market introduction of NR now and the superior NR capabilities, a refarming of spectrum toward NR is expected for the future. In this step, the compatibility of LTE and NR which allows for e.g. LTE-NR spectrum sharing [16] will make the transition from LTE to NR very smooth and it can be flexibly adapted according to the gradual increase of NR capable devices. Within the light of this migration, it can be noted that
EC-GSM-IoT has as of today not managed to attract significant market interest, while significant deployments of NB-IoT and LTE-M has happened during the last two years [3].
While the reuse of existing network infrastructure and spectrum is an important aspect for an operator, a specific benefit of NB-IoT shall be pointed out in its spectrum flexibility. It is generally expected that existing operator spectrum deployments are extended to also include LTE-M traffic, so the IoT traffic will be on the same spectrum that is already deployed for telephony and mobile broadband services. For NB-IoT, the narrow system bandwidth of NB-IoT makes it suitable to be deployed also in spectrum that is not used for mobile broadband services today. Examples exist where operators have spectrum allocations that do not fit with exact carrier bandwidths provided by LTE. As a result, a remainder of the spectrum allocation remains unused. NB-IoT provides the flexibility to make use of even small portions of idle spectrum resources that an operator may have. Such portions of spectrum resources can even be created by an operator, e.g. by emptying individual GSM carriers from GSM operation and re-use them instead for NB-IoT usage.
For critical IoT services, NR has benefits over LTE, for both wide-area and local (often industrial) use cases. These benefits stem from the higher flexibility, and better performance and spectral efficiency of NR over LTE for URLLC. As a market segment for critical cellular IoT services is at its very beginning, a wide range of innovations and optimizations can be expected for cellular IoT as this segment develops. Already now, several novel features have been identified for future NR standard releases to better address e.g. industrial use cases, by providing time synchronization over the radio interface or interworking with IEEE 802.1 TSN. We foresee that in particular the NR evolution will address these features in future standard releases.
There is a segment of cellular IoT services which has higher demands in performance and capabilities than massive IoT services, but yet without a need for the URLLC required by critical IoT services. This segment has been referred to as broadband IoT (see Ref. [21] and Figure 1.5) and it combines requirements on high data rates with massive IoT features like extended coverage and battery saving. There are numerous examples of this category, like advanced wearables, connected vehicles and telematics, video monitoring systems, augmented and virtual reality systems, or connected drones as described in Chapter 13 and [22,23]. Both LTE and NR are well suited to address this segment. Here it should be noted that the LTE-M features for extended coverage and battery saving described in Chapters 5 and 6 can be implemented not only by the massive IoT device categories Cat-M1 and Cat-M2 but also by more high-performing ordinary LTE devices, and that LTE-M can coexist seamlessly with both LTE and NR, so LTE-M can be seen as a suitable candidate solution for use cases requiring extended coverage and battery saving in both the massive IoT and the broadband IoT segment.
The considerations for choosing a cellular IoT technology by a mobile network operator is based on what spectrum and what radio access technology the operator use or plan to use in future. The driving force in this regard is to reuse existing or planned mobile networks in order to achieve a low capital expenditure and operational expenditure for the deployment and operation of cellular IoT connectivity. Another major component in an operator assessment of cellular IoT is the IoT service strategy of the operator. Does the operator target specific IoT market segment? And if so, what are the service requirements in this segment and what connectivity requirements does it imply? In this case the operator decision is
largely based on how well a cellular IoT technology fulfills the service requirements, as discussed in Section 16.2.
16.3.2. The IoT service provider's perspective
An IoT service provider targets a set of particular IoT services with its offering. For example, a focus may be on smart city applications, or precision agriculture. The targeted IoT service implies a certain location where the service will be realized, i.e. where IoT devices will be located. For smart city services, this will be in urban areas, for precision agriculture this will be primarily in rural areas, and for industrial IoT solutions it will be at industrial sites. The IoT service characteristics determine what kind of traffic profile needs to be supported. For a smart city this may be regular monitoring of available parking spaces, or notifications when waste containers have reached a certain fill level. For precision agriculture, it can be the monitoring of humidity and fertilization on fields or in green houses, or the tracking of cattle. For industrial IoT solutions, it can be monitoring and control of the industrial precesses and operations. Other IoT service characteristics besides the traffic profile can be the maximum time that a device must operate on a battery.
Based on an analysis of the targeted IoT service, the connectivity requirement of the service becomes clear:
- - What data rates need to be supported by the communication?
- - Are critical IoT services targeted, and what are the required latency bounds and levels of reliability and availability?
- - Do devices need to run on battery for extended time periods?
- - What device density is expected?
- - Where are the devices located?
- - Are devices in particularly hard to reach locations (e.g. in enclosures underground)?
- - Are devices mobile over larger areas, possibly even across national borders?
Based on this review a service provider can determine:
- - Which cellular IoT technologies provide sufficient performance for the targeted service, see Section 16.2.
- - At what locations network coverage is needed.
It can be expected that coverage of multiple cellular IoT technologies is provided by one or more network operators at various locations. In an increasing number of mobile networks, both LTE-M and NB-IoT will be found. For URLLC capabilities, NR is starting to be deployed, but availability will vary between different locations. An IoT service provider will want to select a network operator that provides coverage and connectivity via a suitable cellular IoT technology at the targeted deployment area at a fair price.
Since IoT devices may be deployed and operated over long time spans, flexibility in re-selecting a network provider is desirable. Embedded Subscriber Identity Modules that enable to remotely re-provision devices and re-select network providers will play an increasing role for cellular IoT devices, see Ref. [24].
For critical IoT systems, several IoT use cases require dedicated deployments and installations. The IoT service provider may be a system integrator, potentially a mobile network operator, or the industrial enterprise end user. Even shared responsibilities are imaginable. A specific plan for the system solution is needed, that analyses in detail the requirements and desired capabilities. A specific solution is needed, that considers the local availability of spectrum and is built on standard cellular IoT components.