Definition
LoRa is Semtech's physical layer modulation technology, using Chirp Spread Spectrum (CSS) spread spectrum, operating in 470/868/915MHz Sub-GHz ISM license-free bands. Single-hop transmission distance can reach several kilometers, and terminal batteries can last for several years. LoRaWAN is a network protocol built on LoRa, supporting Class A/B/C terminals.
Technical Features
LoRa uses Chirp Spread Spectrum; the spreading factor (SF) can be adjusted from SF7 to SF12. The larger the SF, the longer the transmission distance and the lower the data rate. Typical rates are 0.3–50 kbps, with receive sensitivity up to -137 dBm (SF12). It operates in unlicensed ISM bands: in China commonly 470–510 MHz, in Europe 868 MHz, in North America 915 MHz. Terminal transmit power is usually 14–20 dBm, and gateways can cover several kilometers to more than ten kilometers (depending on the environment). Compared with NB-IoT, LoRa is a private network that requires self-built gateways, but has no operator fees, making it suitable for self-organized network scenarios like campuses, buildings, and agriculture.
Use Cases
Smart metering (water, electricity, gas, heat) reading, AMR; environmental monitoring (temperature and humidity, air quality); agricultural IoT (soil, meteorology); smart city (manhole covers, street lights, parking); industrial sensor aggregation. A private LoRaWAN network can be deployed in factories and campuses, aggregating dispersed sensor data to the gateway, then uploading to the cloud platform via 4G/Ethernet. LoRa point-to-point or private protocols can also be used to enable direct communication between devices.
Comparison with NB-IoT
NB-IoT is a carrier network with wide coverage and no need for self-building, suitable for dispersed, wide-area deployment; LoRa is a private network, suitable for concentrated areas, self-controlled, and no monthly fee scenarios. NB-IoT requires a SIM card and operator data plan; LoRa requires a one-time investment in gateways and terminals. Both are LPWAN technologies, supporting low power consumption and long distances.
Spreading factor and airtime
Chirp spread spectrum represents a symbol with a chirp whose frequency changes with time. A higher spreading factor makes each symbol longer. Receive sensitivity is usually better, the channel is occupied longer, and the same channel can hold fewer devices. Bandwidth, coding rate, and transmit power together decide the link budget and how long the channel is busy. Turning the spreading factor all the way up does not buy distance in a straight line. It lets one device occupy the channel for a long time. The band has to be a sub-gigahertz range and a power the region allows, and the antenna has to be built for that frequency. The physical layer itself does not define a device address, a join, or encryption. Those belong to the protocol above it. A private point-to-point frame has to define its own check, sequence, and retransmission.
The conditions are a gateway antenna above the typical obstruction, a known feeder loss, and a device antenna that is not wrapped in a metal housing. Loss through a floor indoors is far larger than line of sight in the open. A laboratory line-of-sight figure cannot be turned into a number of floors. Duty cycle, or a polite wait, is a regulation or a coexistence requirement on some bands. A device has to leave the channel empty after it transmits, and a gateway downlink occupies the channel as well. Turn adaptive data rate on only after the link is stable. During installation, hold one set of parameters. Otherwise the device changes rate as it moves, and the site notes cannot be compared.
Band duty cycle and where the gateway is mounted
An antenna connector for the wrong frequency, or a feeder that is too long with the loss left out of the budget, leaves the receive sensitivity unchanged on paper and the real link a step worse. Another private network on the same frequency, or a wrong bandwidth, makes the gateway decode broken frames, which looks like random loss rather than a whole area offline. Turning both the spreading factor and the transmit power all the way up lets a nearby device drown a far one, and communication near the gateway gets worse. A device stuck inside a metal meter box warps the whole antenna pattern. A higher factor only partly makes up for it.
A gateway with its back against a metal wall, or installed in a basement, pulls coverage into a fan along the opening. The circle drawn on a map is not there. A confirmed frame makes the device open a receive window after it transmits. If the application confirms every packet, the downlink becomes the bottleneck and the device uses more energy. Retransmission with no limit on the count lets a device that cannot be heard keep occupying the channel. A gateway only receives the physical frame. An unsynchronized clock is not corrected automatically at the physical layer. Application time still has to be written into the payload by the device. After a change of region, change the band and the power limit. Keeping another region's channel table may mean nothing is transmitted, or that it is transmitted where it is not allowed.
Nearby power and channel occupancy
A nearby device with both spreading factor and power turned all the way up presses down a far device, and packet loss appears near the gateway. Hold adaptive data rate fixed while walking the site. Otherwise the parameters change every few steps and antenna positions cannot be compared. Confirmed frames make the downlink a shared resource. Confirming every packet exhausts the gateway downlink first, and the far devices fail first. Retransmission needs a limit on the count. A device that cannot be heard must not hold the channel indefinitely.
Match the antenna to the band, and put feeder loss into the link budget rather than looking only at the chip sensitivity. An antenna inside a metal meter box has a warped pattern. Bring the antenna out of the box before raising the factor. Do not mount a gateway against a metal wall or in a basement. Coverage is not a circle on a map. After a change of region, change the channel and the power limit together. Keeping another region's channel table can transmit on the wrong frequency. The physical layer does not handle application time. The timestamp is still written into the payload by the device. The moment the gateway receives the frame is only the arrival time.
Airtime grows with both the spreading factor and the payload length. With a very short payload and the factor at its highest, that one link still occupies most of the channel time, and other devices can only wait. The receive window on the gateway side is shared. Downlink replies have to count duty cycle. Confirmations cannot grow in a straight line with the number of devices. The benefit of antenna height is eaten by a metal bracket that is moved into the way and by a feeder that is too long. Write both into the link budget when the gateway is put up. Co-channel interference does not always make the gateway hear nothing. It raises the share of frames that fail the check. Count failed frames, not only successful ones. A device that changes rate while it is moving makes comparison difficult. Hold the parameters during a walk test, and turn adaptation on after the site is chosen. Changing region is not only rewriting a frequency number. The power cap and the allowed duty cycle change with it.