37 Wide-Area Access: LPWAN and Cellular
37.1 Start With Reach Before Speed
LPWAN and cellular choices begin with reach, ownership, power, and service model before raw throughput. A tiny sensor that reports once an hour and a mobile gateway that streams diagnostics do not need the same wide-area network.
Follow the workload first: payload size, reporting interval, latency tolerance, battery target, coverage evidence, and who operates the network. The right technology is the one whose constraints match that record.
Phoebe’s Why
A radio’s power spreads outward over a growing sphere, so a fixed-size receive antenna catches a shrinking share of it as distance grows – the inverse-square law behind free-space path loss and the Friis equation. Real ground-level paths lose faster than free space, because reflections, foliage, and terrain add loss that grows with distance too, captured empirically by a path-loss exponent \(n\) that exceeds the free-space value of 2. The gap between what the link budget can tolerate and what the path actually costs is the fade margin – the cushion this chapter’s own “best, typical, and worst install locations” advice exists to protect. And because a lower data rate improves receiver sensitivity only by keeping the transmitter on the air longer, that margin is bought with battery energy, exactly as this chapter’s own airtime-versus-power warning says.
The Derivation
Free-space path loss from inverse-square spreading and antenna aperture:
\[\mathrm{FSPL_{dB}}(d) = 20\log_{10}\!\left(\frac{4\pi d}{\lambda}\right)\]
Real paths replace the free-space exponent with a measured path-loss exponent \(n\) beyond a reference distance \(d_0\):
\[PL_{dB}(d) = \mathrm{FSPL_{dB}}(d_0) + 10\,n\,\log_{10}\!\left(\frac{d}{d_0}\right)\]
Link budget and the fade margin remaining after every loss:
\[B = \mathrm{EIRP} + G_{rx} - S_{rx}, \qquad M = B - PL_{dB}(d)\]
Worked Numbers: This Chapter’s Own “Four Readings A Day For Years” Sensor
The chapter names no specific site or band for its soil-moisture example, so the worked numbers use catalog-typical figures: 915 MHz (\(\lambda=0.328\) m), a 14 dBm end node, 3 dBi gateway antenna, and a catalog-typical SF12 sensitivity of \(-137\) dBm.
- \(\mathrm{FSPL}\) at \(d_0=1\) m: \(20\log_{10}(4\pi(1)/0.328)=31.7\) dB. Link budget: \(B=14+3-(-137)=154\) dB – a number this large is the real reason LoRa-class links are called “long range,” not a marketing claim.
- At a catalog-typical suburban/vegetated exponent \(n=2.7\): path loss to 3,000 m is \(31.7+10(2.7)\log_{10}(3000)=125.6\) dB. After a catalog-typical 15 dB obstruction allowance, margin \(=154-125.6-15=13.4\) dB – comfortable.
- Push the same link to 8,000 m: path loss \(=31.7+10(2.7)\log_{10}(8000)=137.1\) dB, margin \(=154-137.1-15=1.95\) dB – on paper it still closes, but with almost nothing left for weather, foliage growth, or a moved antenna.
- Battery tie: a lower spreading factor raises sensitivity but keeps the radio on air longer. Using a catalog-typical 125 kHz-channel symbol time \(T_{sym}=2^{SF}/\mathrm{BW}\) and a 120 mA transmit current, a short packet costs about \(0.00096\) mAh at SF7 versus \(0.0306\) mAh at SF12 – roughly \(32\times\) more charge per report for the deeper-coverage setting, which is exactly the “longer airtime… consumes battery energy” trade-off this chapter’s own text names without a number attached.
Overview: Wide-Area IoT Trades Bandwidth For Reach
LPWAN and cellular IoT links are chosen when sensors must reach beyond a building, campus, or Wi-Fi cell while using small batteries. They are not just "longer Wi-Fi." They use different assumptions: small payloads, long sleep periods, constrained downlink, operator or gateway coverage, and careful link-budget planning.
The first design question is not which radio is newest. It is what the application really needs: reporting interval, payload size, battery target, mobility, ownership model, coverage proof, regulatory limits, and acceptable latency.
Unlicensed LPWAN and cellular IoT solve the wide-area problem in different ways. A private LoRaWAN deployment may let the owner place gateways where the sensors actually are, but it also makes gateway siting, network-server operation, shared-spectrum behavior, and regional airtime limits part of the design. NB-IoT and LTE-M move much of that infrastructure into a mobile operator network, but they add SIM or eSIM provisioning, subscription terms, roaming behavior, module certification, and operator lifecycle as real dependencies. Neither family is automatically better. Sparse meters, mobile trackers, remote alarms, firmware-update paths, and gateway backhauls each stress a different part of the trade space.
The practical distinction is therefore not only range. Ask whether the device needs confirmed delivery, frequent downlink, mobility between cells or gateways, large bursts, indoor penetration, private network ownership, or years of unattended battery life. A link that is excellent for four tiny readings per day can be wrong for streaming diagnostics or interactive control, even if both use the phrase "wide area."
Short-Range Mesh
IEEE 802.15.4 families such as Zigbee, Thread, and WirelessHART fit local low-power networks where mesh or scheduled operation matters more than kilometre-scale reach.
Unlicensed LPWAN
LoRaWAN and similar systems can support private or community infrastructure, but must live within shared-spectrum rules and duty-cycle or airtime limits.
Cellular IoT
NB-IoT and LTE-M use licensed operator infrastructure, which can simplify coverage in served areas while adding subscription, module, roaming, and lifecycle dependencies.
Selection doctrine: choose the radio family only after the workload is bounded. A daily meter read, a mobile asset tracker, a firmware-updated field gateway, and a low-latency factory controller are different access problems.
Practitioner: Build A Coverage And Cadence Record
A practical LPWAN or cellular decision should leave a record that a field team can test: where the devices are, how often they transmit, how much data moves in each direction, who owns the network, what coverage evidence exists, and what happens when the device cannot reach the network.
Build the record from representative sites, not from the easiest demonstration point. Include the best, typical, and worst install locations; the normal message; the largest exceptional message; the expected acknowledgement pattern; and the fallback when a report is missed. For battery devices, keep one measured current trace that includes wake, attach or join, transmit, receive-window behavior, retry, and return to sleep. For managed networks, keep the provisioning path, owner, renewal date, and support route beside the radio evidence.
Do not approve a wide-area access choice from a lab demo alone. Field signal, retry behavior, payload cadence, and provisioning friction often dominate the real deployment.
Under The Hood: Radio Physics Meets Service Model
Wide-area IoT is a stack of constraints. The radio layer controls link budget, airtime, interference, and energy per message. The service layer controls network ownership, SIM or device provisioning, roaming, downlink timing, firmware-update path, and long-term continuity.
The radio side starts with a link budget, not with an advertised range number. Transmit power, antenna gain, receiver sensitivity, path loss, building loss, terrain, foliage, interference, and fade margin decide whether a packet can be heard at the gateway or base station. Lower data rates can improve sensitivity, but they keep the transmitter on air for longer. That longer airtime can consume battery, reduce channel capacity, increase collision exposure, and run into regional duty-cycle or fair-use limits. A design that looks efficient for one sensor can fail when thousands of devices wake near the same reporting window.
The service side is just as technical. LoRaWAN classes define different receive-window and downlink expectations; cellular IoT devices may use features such as Power Saving Mode or extended Discontinuous Reception to trade reachability for battery life. Those settings affect command latency, firmware-update feasibility, and troubleshooting. Provisioning also becomes part of the system: keys, SIM profiles, roaming permissions, operator coverage, gateway backhaul, and network-server ownership decide whether the radio path remains usable after rollout. A wide-area design is complete only when radio evidence and service evidence agree, and both are tied to real install locations, payload schedules, and support owners.
Airtime Pressure
Low data rate increases time on air. That can help sensitivity but consumes duty-cycle budget, gateway capacity, and battery energy.
Downlink Boundaries
Battery devices often sleep. Confirm when a device can receive commands, acknowledgements, or updates before relying on downlink behavior.
Lifecycle Risk
Operator sunsets, subscription changes, regional band differences, SIM logistics, and provider continuity can matter as much as radio performance.
37.2 Summary
LPWAN and cellular IoT access choices are workload and lifecycle decisions. Short-range mesh, private LPWAN, public LPWAN, NB-IoT, LTE-M, and 5G service categories each optimize different combinations of range, rate, latency, power, ownership, mobility, and coverage. A defensible selection records the payload cadence, link margin, power profile, network ownership, provisioning path, and field evidence.
37.3 Key Takeaway
Choose wide-area IoT access from evidence, not slogans: prove the coverage, airtime, battery, downlink, provisioning, and lifecycle assumptions at representative deployment sites.
37.4 See Also
Wireless Access: Wi-Fi
Compare high-throughput local wireless access with low-power wide-area alternatives.
Network Physical Classification
Place LPWAN and cellular choices beside other physical access media.
Link Budget and Coverage Planning
Turn range claims into received-power, margin, and field-validation records.
Free-Space Path Loss
Build the open-air baseline before adding terrain, building, foliage, and fade losses.