What does a water meter really need to say? One number, once a day. A soil probe in a vineyard? A moisture reading every few hours. Neither device needs a video call, and neither can be recharged every week. That gap — tiny messages, long range, battery life measured in years — is exactly what low-power wide-area networks were built to fill. Three names come up again and again when you start planning: LoRaWAN, NB-IoT and Sigfox. On a spec sheet they look like siblings. In the field they behave very differently, and picking the wrong one is expensive to undo.

What Low-Power Wide-Area Actually Means

LPWAN is an umbrella term for radio technologies designed around small payloads and long battery life rather than speed. Compare the alternatives you probably already know. Wi-Fi is fast and short-range. Zigbee and Bluetooth mesh are cheap and short-range. Standard cellular LTE is fast, wide-reaching and hungry. LPWAN pushes three levers at once: distance up, energy consumption down, throughput down.

The lever you cannot escape is throughput. An LPWAN device sends bytes, not megabytes. Once you accept that, the rest of the design follows: you compress your payload, you send on a schedule, and you sleep the radio most of the day. A well-built node can run for years on a couple of AA cells, which is the whole point when the sensor is inside a concrete pit or on a hillside.

The Three Big Options Side by Side

Before the detail, here is the rough shape of each one. Treat these as general characteristics, not guarantees — real range depends on terrain, antenna placement and building materials.

  • LoRaWAN — built on the LoRa radio modulation in unlicensed bands. Effective data rates sit in the low tens of kilobits per second, range is typically a few kilometres in a city and much further in open country. You or a public community network run the gateways.
  • NB-IoT — a licensed cellular standard. Higher throughput than the others, strong penetration into basements and underground spaces, but you pay a mobile operator per device and the modem draws more power.
  • Sigfox — an ultra-narrowband network with a fixed 12-byte uplink payload. Extremely frugal on power and simple in the radio, but you cannot send more than the format allows.
  • LTE-M — a close cellular relative of NB-IoT, worth knowing about if you need higher bandwidth or voice-adjacent features.

One clarification worth holding onto: LoRa is the physical radio layer, LoRaWAN is the networking protocol on top of it. People mix the two up constantly, and it matters when you are reading chip datasheets.

LoRaWAN: Own Your Network

LoRaWAN's appeal is control. You can buy a gateway, put it on a roof, register your devices with a network server such as The Things Network or a self-hosted option, and start receiving data without a per-device subscription. Gateways are relatively cheap and community coverage exists in many cities, so a small pilot can be running this week.

The trade-off is that somebody has to keep it running. Gateways need power, internet backhaul and occasional attention. Unlicensed bands also come with duty-cycle limits in most regions, which caps how often a device may transmit — fine for readings every fifteen minutes, painful for anything chatty. Interference from other radios is real, and you should plan your antenna and placement rather than hoping.

Typical fits: agricultural monitoring, campus and building sensors, asset tracking where you control the site, and any project where avoiding a monthly connectivity bill matters more than guaranteed uptime.

NB-IoT: Licensed Spectrum, Predictable Coverage

NB-IoT rides on existing cellular towers, so if your devices live where phones get signal, coverage is usually someone else's problem. That is a genuine advantage for fixed assets: gas and water meters, parking sensors, waste bins, agricultural pumps on a leased site. Penetration is a strong point — the signal reaches into basements and metal enclosures that would defeat a LoRa link.

Higher throughput also means you can do things LoRaWAN makes awkward, such as occasional firmware updates or richer alarm payloads. The costs are power and paperwork. NB-IoT modems consume more energy, so battery life tends to be shorter unless you are drafting a mains-powered design. You also need a SIM or eSIM arrangement per device, and roaming across networks is rarely as smooth as it sounds. Confirm coverage and commercial terms with the operator for your specific site before you commit a product design to it.

Sigfox: Small Messages, Wide Coverage

Sigfox takes minimalism to its logical end. Uplinks are capped at 12 bytes, daily message counts are limited, and there is no assumption of an acknowledgement. In exchange, the radio is about as simple and low-power as wide-area radio gets, and the network is operated for you.

That makes it a good match for one-way telemetry: a bin that reports fill level, a button that reports a press, a tank that reports a level. It is a poor match for anything needing two-way control, large payloads or rapid retransmission. Network ownership and coverage have shifted over the years — Sigfox's operations changed hands in 2022 — so check current availability and support in your country rather than assuming a map from an old blog post still holds.

How to Pick the Right One

Work through these in order. Each answer narrows the field quickly, and the exercise takes an afternoon rather than a redesign.

  1. Count your bytes and your messages. Write down the actual payload and how often it must arrive. Twelve bytes twice a day is Sigfox territory. Two hundred bytes every ten minutes is not.
  2. Decide who operates the network. If nobody on your team wants to run gateways, choose an operator-managed option. If you want to own the infrastructure and skip subscriptions, LoRaWAN wins.
  3. Test coverage at the exact install spot. Postcode-level maps lie. Put a node in the actual pit, basement or field corner and log signal quality for a week.
  4. Do the power budget on paper. Sleep current, transmit current, duty cycle and battery chemistry determine lifetime far more than the radio standard does.
  5. Plan for the long term. Devices in the field for eight years need a story for firmware bugs and for networks that change terms. Ask what happens if you need to update ten thousand nodes.

Pilot a Dozen Before You Order a Thousand

Buy five to ten nodes of one or two candidate technologies. Log received signal strength, missed messages, and battery drain for a month, including a cold snap and a wet week if you can. The first pilot almost always changes the design — an antenna moves, a transmit interval stretches, a gateway gets added. That is far cheaper to discover with ten devices than with a production run.

Then keep a fallback in mind. Many sensible deployments are hybrids: LoRaWAN for site-owned sensors, NB-IoT for the handful of devices in awkward locations outside your own coverage. Radio choice is an engineering decision, not a loyalty programme.

Photo: stanias / Pixabay