English is currently available.
LoRa & LoRaWAN DISCUSSION
Two RFM95W modules (868 MHz) on Arduino Pro Mini boards talk reliably across the house, but outdoors I lose packets beyond roughly 300 m even with a clear view. Each module has a piece of wire soldered on as an antenna, the settings are SF7 and 125 kHz, and the library is set to 17 dBm.
Everybody quotes kilometres for LoRa. What should I check first: the antenna, the radio settings, or the power supply?
Start with the antenna, because that is where most short-range reports end. A quarter-wave wire is c / (4 × f): at 868 MHz the wavelength is 345.6 mm, so the wire should be about 86 mm, in practice a little shorter, around 82 mm. At 915 MHz it is about 82 mm before trimming and at 433 MHz about 173 mm. A wire cut for the wrong band, or a 433 MHz coil antenna on an 868 MHz module, costs many dB.
A quarter-wave monopole is only half of an antenna; it needs a ground plane or counterpoise. On a tiny board with little ground area, add a second 86 mm wire from ground pointing the opposite way to form a dipole, and keep both wires straight and away from the battery and other metal.
Check the output pin selection in software. The SX1276 has two transmitter outputs, RFO and PA_BOOST, and the RFM95W connects only PA_BOOST to its antenna pin. If the library is configured for RFO, almost nothing is radiated whatever power the register shows. With PA_BOOST the normal range is 2 to 17 dBm; 20 dBm needs the high-power setting and carries a duty cycle restriction in the datasheet.
Remember that the legal limit on the EU868 default channels is 14 dBm ERP, so 17 dBm into a good antenna is already more than you may radiate there.
Then the supply. At 17 dBm the SX1276 draws about 90 mA during transmit and at 20 dBm about 120 mA, in bursts. The 3.3 V regulator on small Arduino-style boards is often rated for 150 mA or less, and the 3.3 V pin of some 5 V boards supplies far less. If the rail sags during the burst, the output power drops or the module resets. Put 10 µF plus 100 nF right at the module's supply pins and watch the rail with a scope during a transmission.
The module is a 3.3 V part and is not 5 V tolerant, so with a 5 V microcontroller use level shifting on the SPI and control lines.
Do the link budget to see how far off you are. Free-space path loss in dB is 20×log10(d_km) + 20×log10(f_MHz) + 32.44. At 868 MHz that is 91.2 dB at 1 km and 80.8 dB at 300 m. With 14 dBm transmitted and a sensitivity of about -123 dBm at SF7, the budget is 137 dB, so at 300 m in the clear you should have more than 50 dB to spare. Losing packets there means tens of dB are disappearing in the antennas, the output path or the supply.
Log the RSSI and SNR the receiver reports at 10 m, 50 m and 100 m. In free space the level falls about 6 dB per doubling of distance; a level already near -100 dBm at 50 m confirms a hardware problem.
Once the hardware is right, the remaining limits are geometry and settings. Radio waves need the first Fresnel zone mostly clear, not just a visual line. Its radius at mid-path is sqrt(λ × d / 4): for 2 km at 868 MHz that is sqrt(0.3456 × 2000 / 4), about 13 m. Two nodes at waist height over flat ground have the ground well inside that zone, which is why raising the antennas a few metres often does more than any setting.
After that, step the spreading factor up. Each step adds roughly 2.5 dB of sensitivity at the cost of doubling the airtime, and SF12 gives about 12.5 dB more than SF7.
Back to LoRa & LoRaWAN community
