IoT and Low-Power Wide-Area Networks

Introduction

The Internet of Things (IoT) connects physical objects—sensors, meters, machines, tags and actuators—to digital services. Unlike conventional mobile broadband, many IoT applications transmit only a few bytes, operate from batteries for years, tolerate moderate latency and require coverage across buildings, farms or cities. No single wireless technology optimizes range, data rate, energy, cost, mobility and reliability simultaneously. Connectivity must therefore be designed from application requirements rather than selected solely by advertised range.

An IoT system is an end-to-end control and information system. Measurements travel from devices toward applications, while configuration changes, acknowledgements and actuation commands travel in the reverse direction. Designers must evaluate both paths because downlink capacity is often more constrained than uplink capacity in low-power networks.

Learning Objectives

After studying this chapter, students should be able to:

1. Explain IoT architectures, sensor networks and RFID operation.
2. Compare LoRaWAN, NB-IoT and LTE-M for different applications.
3. Estimate LoRa airtime and construct a wireless link budget.
4. Estimate battery life and analyze capacity, reliability and scale.
5. Select MQTT, CoAP and related lightweight protocols appropriately.
6. Design secure and manageable wireless IoT connectivity.
7. Evaluate an IoT design using technical, regulatory and operational criteria.

1. End-to-End IoT Architecture

A complete IoT solution commonly contains five functional layers:

• Device layer: sensors, actuators, embedded processors and radio transceivers.
• Access network: RFID readers, wireless sensor networks, LoRaWAN gateways or cellular base stations.
• Transport and platform layer: message brokers, device registries, databases and digital twins.
• Application layer: dashboards, alarms, analytics and enterprise applications.
• Management and security plane: identity, provisioning, monitoring, updates and access control across all layers.

Figure: Bidirectional IoT information and control flow, including the reverse path to actuators.

Telemetry and control patterns

Periodic telemetry is suitable for slowly changing quantities such as daily water consumption. Event-driven reporting reduces traffic by transmitting only when a threshold or state change occurs. Closed-loop control additionally requires bounded latency and dependable downlink delivery. A technology appropriate for hourly meter readings may therefore be unsuitable for emergency shutdown control.

Edge computing processes data near the device or gateway. It can reduce latency and backhaul traffic, preserve privacy and maintain limited operation during cloud outages. Cloud platforms remain useful for fleet-wide analysis, long-term storage and model training.

2. Sensors, Actuators and Wireless Sensor Networks

A sensor node typically contains sensing elements, signal conditioning, an analog-to-digital converter, a microcontroller, memory, a radio and a power source. Measurement quality depends on calibration, resolution, accuracy, drift, sampling and placement—not merely connectivity.

Wireless sensor networks may use three broad topologies:

• Star: each node communicates with a central coordinator. It is simple and energy-efficient but depends on coordinator coverage.
• Tree: traffic follows hierarchical parent–child paths. It extends coverage but failures near the root affect many nodes.
• Mesh: nodes may relay packets along alternative routes. It improves path diversity but consumes energy and introduces routing overhead.

Low-power mesh protocols are useful for local-area sensing, whereas LPWAN technologies favor long single-hop links to infrastructure. Multi-hop communication is not automatically more energy-efficient: receiving and forwarding consume energy, and routers may need to remain awake.

Sampling and data reduction

If a sensor samples at rate f_s samples/s, uses b bits/sample and sends all samples without compression, the raw source rate is:

Here R_raw is in bit/s. Protocol headers, retransmissions and security tags increase the radio traffic above this value. Local aggregation, threshold detection and feature extraction can greatly reduce transmissions, but excessive filtering may remove diagnostically valuable information.

3. RFID and Identification Systems

Radio-frequency identification (RFID) identifies objects using tags and readers. An RFID tag stores an identifier and may also store sensor or lifecycle information.

RFID categories

Low-frequency and high-frequency systems commonly operate through near-field inductive coupling. UHF RFID generally uses far-field propagation and backscatter. Actual range depends on frequency, antenna orientation, polarization, materials, reader power and regional regulations. Metals detune antennas, while water-rich materials absorb RF energy; specialized tags and placement are often necessary.

RFID is primarily an identification and short-data technology rather than a general-purpose wide-area network. Readers bridge tag observations into enterprise or IoT platforms. Privacy controls are important because persistent identifiers can enable unauthorized tracking.

4. LPWAN Design Space

Low-power wide-area networks exchange low or moderate data volumes over large areas while minimizing device energy. Their long range is achieved through robust modulation, narrow bandwidth, repetition, coding or processing gain. The corresponding trade-offs include lower throughput, longer airtime and limited downlink capacity.

Specifications evolve, and operator implementations differ. A design must confirm local bands, roaming, coverage, certification and service availability rather than relying only on nominal standard capabilities.

5. LoRa and LoRaWAN

LoRa is a chirp spread-spectrum physical-layer modulation. LoRaWAN defines medium access, device classes, security and network architecture above LoRa. Conflating the two obscures important design decisions.

A LoRaWAN end device broadcasts uplinks that may be received by multiple gateways. Gateways forward packets to a network server, which removes duplicates, performs network functions and routes application data. This “star-of-stars” architecture does not require end devices to relay other devices' traffic.

Spreading factor and bandwidth

For common LoRa parameterization, the symbol duration is:

where SF is the spreading factor and BW is bandwidth in hertz. Larger spreading factors increase symbol duration and receiver sensitivity but also increase airtime and collision exposure.

LoRa airtime calculation

For the commonly used explicit-header LoRa packet formulation, the preamble duration is:

The payload symbol count is:

and total packet airtime is:

The variables are:

• PL: physical payload length in bytes.
• CRC: 1 when the payload CRC is enabled, otherwise 0.
• IH: 0 for explicit header mode and 1 for implicit header mode.
• DE: 1 when low-data-rate optimization is enabled, otherwise 0.
• CR: coding-rate parameter 1, 2, 3 or 4, corresponding respectively to 4/5, 4/6, 4/7 or 4/8.
• N_pre: programmed preamble symbols, commonly 8.

This method applies to the LoRa packet configuration represented by these parameters. Regional LoRaWAN settings, explicit/implicit mode, optional headers and radio implementation must be checked. It is not a universal formula for every chirp-spread-spectrum system.

Worked example: LoRa airtime

A packet uses SF=10, BW=125000 Hz, PL=20 bytes, explicit header, payload CRC, coding rate 4/5 and an eight-symbol preamble. Low-data-rate optimization is disabled.

The ceiling term is:

Thus:

The packet occupies the channel for approximately 371 ms. Application data size alone therefore understates network load.

Device classes

• Class A: every uplink is followed by two receive windows. It has the lowest energy requirement and is mandatory for LoRaWAN end devices.
• Class B: scheduled additional receive windows improve downlink predictability but require synchronization.
• Class C: the receiver is open almost continuously except during transmission, reducing downlink latency at high energy cost.

Confirmed messages provide acknowledgements but should be used selectively. Excessive acknowledgements consume gateway downlink airtime and can reduce network scalability.

6. Cellular LPWAN: NB-IoT and LTE-M

NB-IoT and LTE-M integrate IoT devices into cellular infrastructure, providing managed spectrum, subscriber authentication and operator coverage.

NB-IoT

NB-IoT emphasizes coverage, small messages and stationary or low-mobility devices. Coverage enhancement may use repetitions, increasing latency and energy consumption. Power-saving mode can make a device unreachable for long periods, while extended discontinuous reception allows periodic paging at an energy cost.

LTE-M

LTE-M supports greater throughput, lower latency and stronger mobility capabilities than NB-IoT. It is suitable for mobile tracking, voice-capable designs in supported networks and larger firmware transfers. Its energy consumption may exceed that of an optimized NB-IoT or LoRaWAN device for very small, rare messages, but the actual result depends strongly on signal conditions and session behavior.

Cellular selection must account for module certification, SIM or eSIM management, subscription cost, operator sunset policy and roaming. Poor coverage can dominate energy consumption because the device performs repeated access attempts or transmissions.

7. Lightweight Application Protocols

Radio efficiency does not guarantee application efficiency. Application protocols determine message semantics, delivery behavior and overhead.

MQTT

MQTT uses a publish–subscribe model through a broker, normally over TCP. Publishers send messages to topics; subscribers receive matching topics. Quality-of-service levels are:

• QoS 0: at most once; no application-level acknowledgement.
• QoS 1: at least once; duplicates are possible.
• QoS 2: exactly once at the MQTT protocol level; greatest overhead.

MQTT is effective for persistent gateway or cellular connections. TCP setup and keep-alive traffic may be inefficient for highly constrained devices that wake briefly.

CoAP

The Constrained Application Protocol (CoAP) provides REST-like methods over UDP. Confirmable messages can be retransmitted until acknowledged; non-confirmable messages avoid this overhead. Observe supports asynchronous resource updates. CoAP is compact and suitable for constrained networks, but security and congestion control must be designed carefully.

Protocol selection

Delivery guarantees at one layer do not prove end-to-end actuation. An MQTT acknowledgement, for example, may indicate broker receipt rather than successful physical action. Critical systems need application-level command identifiers, authorization, execution confirmation and safe timeout behavior.

8. Energy Engineering

For a device that passes through operating states, average current is:

Here I_i is current in a state and t_i is its duration. If usable battery capacity is C_u in mAh, an idealized lifetime estimate is:

The estimate assumes that the state pattern repeats and that current and capacity use consistent units. Practical capacity must be derated for self-discharge, temperature, pulse-current limitations, voltage cut-off, aging and regulator losses.

Worked example: battery-life estimate

A node sleeps at 0.010 mA for 3595 s and transmits at 40 mA for 5 s each hour. Other states are neglected.

A 2400 mAh battery is derated to 80%, giving C_u=1920 mAh.

This is a planning estimate, not a guarantee. Receive windows, retries, sensing, startup and battery self-discharge would reduce the result.

9. Link Budget and Coverage

A link budget accounts for gains and losses between transmitter and receiver:

All quantities are expressed in decibels or dBm consistently. P_t is transmit power in dBm; G_t and G_r are antenna gains in dBi; L_path is propagation loss in dB; and L_misc includes feeder, body, polarization, penetration and implementation losses.

The link margin relative to receiver sensitivity S_r is:

Because sensitivity is normally negative in dBm, this subtraction must be performed carefully. A positive margin indicates that the predicted received power exceeds sensitivity, but adequate design margin is still required for fading, interference and model uncertainty.

Worked example: link margin

Suppose P_t=14 dBm, G_t=2 dBi, G_r=2 dBi, L_path=137 dB and miscellaneous losses are 3 dB.

If sensitivity is −137 dBm:

The predicted link margin is 15 dB. The design must also comply with regional limits on transmit power, effective radiated power, channel access and duty cycle. Antenna gains cannot be added without checking those limits.

Free-space path loss is useful only as a reference for unobstructed propagation. Buildings, terrain, foliage, antenna height and interference require appropriate models and field measurements. Packet-delivery maps are often more informative than received signal strength alone.

10. Capacity, Scale and Reliability

Offered airtime load

For N devices, each transmitting m packets during an observation interval, with airtime T_p per packet, aggregate offered airtime is:

Dividing by the observation interval gives a channel-time fraction. This is not automatically the achieved utilization because overlapping transmissions, multiple channels, capture effects, spreading factors, retransmissions and regulatory restrictions alter capacity.

For 500 devices sending 24 packets/day with the 0.370688 s airtime calculated earlier:

Relative to 86,400 s/day, this equals approximately 5.15% of one channel's time. It must not be interpreted as a 94.85% probability of success; random-access collisions and traffic concentration require separate analysis or simulation.

Planning growth correctly

A phrase such as “10% additional load” is ambiguous unless its basis is stated. If baseline load is 0.50 and growth is 10% relative, then:

If the requirement is an absolute increase of 10 percentage points, then:

Capacity plans must explicitly distinguish relative percentages from percentage-point changes.

Reliability

If one attempt succeeds with probability p and attempts are statistically independent with unchanged success probability, the probability of at least one success in n attempts is:

The independence assumption may be invalid during persistent interference, deep fading or network outage. Repetition also increases energy use and congestion. Reliability should therefore be improved first through coverage, antenna design, scheduling, interference management and robust application logic—not unlimited retries.

Useful operational indicators include packet-delivery ratio, command completion rate, latency percentiles, duplicate rate, energy per delivered message and outage duration. Averages alone may conceal unacceptable tail behavior.

11. Security, Governance and Regulation

IoT devices often remain deployed longer than their cryptographic or software assumptions. Security must cover the complete lifecycle:

1. Assign each device a unique identity; avoid fleet-wide default credentials.
2. Protect keys in suitable secure storage and support controlled rotation.
3. Authenticate devices and servers and protect message integrity and confidentiality.
4. Authorize commands according to least privilege.
5. Use signed firmware and secure boot where risk justifies them.
6. Provide rollback-safe, recoverable update mechanisms.
7. Maintain asset inventories, vulnerability processes and retirement procedures.

Encryption does not hide all metadata. Timing, location, identifiers and communication patterns can reveal behavior. Data minimization, retention limits, purpose restriction and controlled access are therefore governance requirements as well as technical practices.

Radio designers must comply with national spectrum rules, certified power limits, permitted bands, duty cycles or channel-access procedures. Cellular products additionally require operator and device certification. Safety-critical actuation should fail to a defined safe state when connectivity, cloud services or authentication fails.

12. Applied Case: Smart Irrigation Network

Background

This is an illustrative case. A cooperative manages 500 soil-moisture nodes and 40 irrigation valves across 12 km². Sensor reports are required hourly, while valve commands must normally arrive within five minutes. Most sensors are solar-exposed, but valves have larger batteries and safety interlocks.

Problem

Public cellular coverage is inconsistent at ground level, and recurring subscriptions for every moisture sensor are undesirable. The cooperative also needs authenticated remote commands and operation during Internet outages.

Analysis

Hourly sensor traffic is small, making private LoRaWAN attractive. Gateway surveys identify two elevated sites with overlapping coverage. Class A is appropriate for moisture nodes because immediate downlink is unnecessary. Valve commands, however, cannot reliably wait for an arbitrary future uplink, so the valves need scheduled receive opportunities or a different connectivity arrangement.

Capacity calculations include payload, LoRaWAN overhead, airtime, retransmissions and downlinks—not only sensor bytes. Confirming every sensor report would create avoidable downlink load. Link budgets are validated by seasonal field tests because wet foliage and low antenna placement alter propagation.

Approach and solution

The design uses two backhauled LoRaWAN gateways and one local edge controller. Moisture nodes send unconfirmed reports with sequence numbers. Missing reports are detected by the application rather than acknowledged individually. Valves use externally powered or larger-battery Class C devices where feasible; each valve also has local pressure and maximum-runtime interlocks.

Commands contain a unique identifier, expiry time and authenticated payload. Valves report acceptance and final physical state separately. During Internet failure, the edge controller applies conservative preapproved schedules. MQTT carries data between the edge controller and cloud platform, while the constrained radio payload uses a compact binary schema.

Outcome

The architecture reduces subscriptions and supports local resilience. Gateway diversity improves reception, while limited confirmation traffic preserves downlink capacity. The five-minute command objective is validated through latency-percentile testing rather than inferred from signal strength.

Learning points

• Separate telemetry requirements from actuation requirements.
• Design and test the reverse control path explicitly.
• Use airtime and link budgets before deployment, then validate them in the field.
• Treat acknowledgements as scarce network resources.
• Combine communications security with local physical safety controls.

13. Contemporary Developments and Best Practices

Edge artificial intelligence can classify vibration, detect anomalies or compress measurements before transmission. It is beneficial when models are small, energy savings exceed computation cost and false alarms are controlled. Model versioning, drift monitoring and explainability become part of device management.

Satellite IoT is extending coverage to remote infrastructure, although device power, antenna visibility, latency and service cost remain constraints. Integrated terrestrial and non-terrestrial connectivity may improve resilience but complicates roaming and application behavior.

Recommended engineering practice is to begin with measurable requirements: payload, reporting interval, latency percentile, mobility, lifetime, environment, coverage probability, security level and total cost. Prototype with production antennas and enclosures; calculate both uplink and downlink capacity; retain margin for growth; instrument deployed devices; and plan secure updates before mass deployment.

Chapter Summary

IoT connectivity is an end-to-end, bidirectional system linking sensors and actuators to applications. Sensor networks, RFID, LoRaWAN, NB-IoT and LTE-M serve different identification, coverage, mobility and energy requirements. Sound designs quantify airtime, average current, battery lifetime, link margin and offered traffic rather than relying on nominal range claims. LoRaWAN provides private low-rate wide-area connectivity, whereas cellular LPWANs provide managed spectrum and operator infrastructure. MQTT and CoAP offer complementary application models. Scalability depends on downlink constraints, random access, retries and traffic concentration. Security, privacy, spectrum compliance, maintainability and safe local behavior are essential throughout the system lifecycle.

Key Terms

Actuator: A device that changes a physical process in response to a command.

Backscatter: Communication produced by varying the reflection of an incident radio signal.

CoAP: A compact REST-oriented application protocol commonly carried over UDP.

Device class: A LoRaWAN operating mode defining receive-window behavior and energy use.

Duty cycle: The fraction of time for which a transmitter or process is active.

Edge computing: Processing performed near data sources rather than only in a remote cloud.

Link budget: An accounting of transmitter power, gains and propagation losses.

Link margin: The amount by which predicted received power exceeds receiver sensitivity.

LoRa: A chirp spread-spectrum physical-layer modulation.

LoRaWAN: A network and medium-access protocol using LoRa radios.

LTE-M: A cellular IoT technology supporting mobility and moderate data rates.

MQTT: A broker-based publish–subscribe messaging protocol.

NB-IoT: A narrowband cellular technology optimized for low-rate IoT services.

Packet-delivery ratio: The proportion of transmitted packets successfully received.

RFID: Radio-based identification using tags and readers.

Sensor node: An embedded device that measures a physical quantity and communicates data.

Spreading factor: A LoRa parameter affecting symbol duration, sensitivity and airtime.

Telemetry: Remotely communicated measurements or device-state information.

Review Questions

Short Answer Questions

1. Distinguish LoRa from LoRaWAN.
2. Why may a Class A device be unsuitable for low-latency actuation?
3. Define link margin and state its unit.
4. How does passive RFID obtain operating energy?
5. Why can confirmed messages reduce LPWAN scalability?

Long Answer Questions

1. Explain the layers and bidirectional flows of an end-to-end IoT architecture.
2. Compare LoRaWAN, NB-IoT and LTE-M in terms of spectrum, energy, mobility, latency and deployment ownership.
3. Discuss the factors that determine practical battery lifetime in a wireless sensor node.
4. Explain MQTT and CoAP, including their delivery mechanisms and suitable applications.
5. Develop a lifecycle security plan for a city-wide smart-metering system.

Analytical / Application Questions

1. A LoRa packet uses SF9, 125 kHz bandwidth, a 15-byte payload, explicit header, CRC, coding rate 4/5, eight preamble symbols and no low-data-rate optimization. Calculate its airtime using the method in this chapter.
2. Construct a link budget for a sensor with 14 dBm transmit power, 1 dBi transmit gain, 3 dBi receive gain, 128 dB path loss, 4 dB miscellaneous loss and −130 dBm receiver sensitivity. Interpret the result.
3. Design connectivity for 2,000 mobile cold-chain containers that report every 15 minutes and require remote firmware updates. Justify the radio technology, protocol, security controls and capacity-validation method.

Multiple Choice Questions with Answers

Question 1

Which statement correctly distinguishes LoRa from LoRaWAN?

A. LoRa is a physical-layer modulation, while LoRaWAN defines networking and medium-access functions.

B. LoRa is an application protocol, while LoRaWAN is an antenna standard.

C. LoRa requires licensed cellular spectrum, while LoRaWAN uses satellite spectrum.

D. LoRa is a routing algorithm, while LoRaWAN is a sensor type.

Correct Answer: A. LoRa is a physical-layer modulation, while LoRaWAN defines networking and medium-access functions.

Question 2

Which LoRaWAN device class normally has the lowest energy consumption?

A. Class B

B. Class C

C. Class A

D. All classes necessarily consume equal energy.

Correct Answer: C. Class A

Question 3

A link has received power of −122 dBm and receiver sensitivity of −137 dBm. What is its link margin?

A. −259 dB

B. 15 dB

C. −15 dB

D. 259 dB

Correct Answer: B. 15 dB

Question 4

A baseline channel load is 0.50. If the load increases by 10% relative to the baseline, what is the new load?

A. 0.51

B. 0.55

C. 0.60

D. 0.65

Correct Answer: B. 0.55

Question 5

Which protocol is based on a brokered publish–subscribe model?

A. RFID

B. CoAP

C. MQTT

D. LoRa

Correct Answer: C. MQTT

Question 6

In the LoRa airtime example, the symbol duration at SF10 and 125 kHz bandwidth is:

A. 0.8192 ms

B. 4.096 ms

C. 8.192 ms

D. 10.240 ms

Correct Answer: C. 8.192 ms

Question 7

Why should acknowledgements be used selectively in LoRaWAN?

A. Acknowledgements eliminate receiver sensitivity.

B. Gateway downlink airtime is constrained and can limit scalability.

C. Acknowledgements make encryption impossible.

D. Acknowledgements convert LoRaWAN into a mesh network.

Correct Answer: B. Gateway downlink airtime is constrained and can limit scalability.

Question 8

Which technology is generally the best fit among the listed options for mobile assets requiring cellular handover and moderate-rate firmware transfers?

A. Passive RFID

B. LTE-M

C. LoRaWAN Class A only

D. Near-field inductive coupling

Correct Answer: B. LTE-M

Question 9

A device has 1920 mAh of usable battery capacity and an average current of 0.06554 mA. Its idealized lifetime is closest to:

A. 1,920 hours

B. 12,600 hours

C. 29,300 hours

D. 65,540 hours

Correct Answer: C. 29,300 hours

Question 10

Under independent attempts with single-attempt success probability p, which expression gives the probability of at least one success in n attempts?

A. pⁿ

B. (1-p)ⁿ

C. 1-(1-p)ⁿ

D. 1-pⁿ

Correct Answer: C. 1-(1-p)ⁿ