Definition
LPWAN (Low Power Wide Area Network) is designed for large-scale, low-speed, low-power scenarios, with single-hop transmission distances up to several kilometers and terminal batteries lasting for years. Typical technologies include LoRa (license-free band) and NB-IoT (carrier-licensed band).
Technology Comparison
LoRa/LoRaWAN: license-free ISM band, private or public network, requires self-built gateways or cloud access, no monthly fee, suitable for campuses, agriculture, metering, etc.; NB-IoT: carrier-licensed band, wide coverage, strong deep coverage, requires SIM card and data plan, suitable for dispersed, wide-area scenarios with high coverage requirements. Additionally, there are Sigfox (in some areas), LTE-M (eMTC), etc. Selection requires comprehensive consideration of coverage, cost, power consumption, data rate, and downlink requirements.
Typical Applications
Smart meters (AMR), environmental monitoring, agricultural IoT, smart cities (manhole covers, streetlights, parking), industrial sensors, asset tracking. LPWAN is suitable for low-frequency, small-data, latency-insensitive applications, complementing high-speed 4G/5G scenarios.
Link Budget and Duty-Cycle Limits
Sensitivity is paid for with airtime
The core of low-power wide-area access is to push receiver sensitivity down while staying inside the transmit power the rules allow. Spreading stretches one symbol across a longer time on air, and correlation can lift a weak signal out of the noise. Each step up in spreading factor improves sensitivity and also makes one packet occupy the channel for noticeably longer. Narrowband cellular instead places the carrier in a very narrow resource block, and the base station buys coverage enhancement with repetitions. The device does not have to build its own gateway, but every attach, acknowledgement, and retransmission uses the cell's control plane. Both approaches write distance as time: extra margin in the link budget corresponds to a longer transmit window and lower network capacity. It is not bandwidth obtained for free.
When judging a link, put receive level, spreading or repetition level, antenna height, feeder loss, and the co-channel noise floor on the same budget. A metal cover, a basement, and a closed cabinet consume a large share of the margin. An antenna placed against metal distorts the pattern, and the stated range no longer holds. On the device, separate peak current from average current. Instantaneous transmit current depends on the battery's internal resistance and pulse capability; average current is set by the reporting period and the time on air. Recording only the farthest distance that once worked does not say how many more devices the network can still accept.
Built for reports, not for conversation
These networks fit points whose payload stays in the tens to a few hundred bytes, whose period runs from minutes to hours, and that can wait for the next cycle if one packet is lost: meter readings, discrete states, and environmental values that change slowly. If the service needs a round trip, continuous adjustment, or a parameter change at any moment, work out the downlink window first. On a private spread-spectrum network, a sleeping device gives the gateway no path to push data in. On licensed cellular, whether the device can be found while saving power still depends on whether the network accepts the PSM and eDRX timers. Firmware fragments, remote threshold changes, and frequent time synchronization pull average current from a yearly budget back to a monthly one.
Before rollout, check the local rules on band, power, and duty cycle. License-free bands often limit how much of the time a single channel may transmit, and retries written too aggressively break that limit outright. On a carrier network, confirm the module bands, what the SIM can attach to, and whether the cell has the matching coverage level enabled. If several private networks already share the same channel on one site, a gateway added later will not avoid them by itself. Only channel planning, spreading factor, and antenna isolation reduce collisions.
Capacity collapses when spreading is maxed out
A common field mistake is to copy a line-of-sight distance measured in the lab into the design, then pin every device at the highest spreading factor or the highest repetition count. Once a few distant nodes fill the air, nearby nodes start losing packets at random. It looks like an unstable signal. The cause is capacity. Another mistake is application-layer retransmission with no ceiling. A failing node keeps sending and pulls its neighbors' success rate down with it. The acknowledgement itself also occupies airtime. Turning acknowledgements on makes real capacity a step smaller than uplink-only reporting.
Gateway backhaul is a third limit that is easy to miss. Power saved on the device does not reduce what the gateway needs for supply, backhaul, and surge protection. When the backhaul is a single congested uplink, packets the radio has already received expire in the queue. Writing down battery life, gateway backhaul, band rules, and whether the service can accept delay — each on its own — comes closer to a condition you can run than comparing only the maximum range of one technology.