How does Doppler shift affect LoRa direct-to-satellite links?

The short answer: the satellite moves faster than LoRa expects

Doppler shift affects LoRa direct-to-satellite links by pushing the signal off its expected frequency and then dragging it across the band while a packet is still on the air. LoRa was designed for sensors talking to a fixed gateway on a hill or a rooftop. When the gateway is a satellite in low Earth orbit, travelling at about 27,000 km/h, the relative speed between sensor and receiver is enormous, and the carrier frequency moves with it.

A LoRa receiver can cope with a fixed frequency offset of up to about plus or minus 25 percent of the signal bandwidth. At 868 MHz with a satellite around 518 km up, the Doppler shift can exceed 20 kHz, and at 2 GHz it can pass 47 kHz. That is far more than a narrow channel can absorb, and studies of real passes show legacy LoRa losing packets whenever the satellite sits too low in the sky. Packets sent at the wrong moment simply never decode.

Doppler shift versus Doppler rate

Researchers separate two effects. Doppler shift is the offset itself. It is largest when the satellite is low on the horizon, at the start and end of a pass, and close to zero when the satellite is directly overhead.

Doppler rate is how fast that offset changes. It does the opposite: it peaks near the overhead point, where the satellite swings from approaching to receding. A changing frequency challenges the receiver's tracking loop, because the chirps at the end of a packet no longer sit where the chirps at the start did. Together the two effects shrink the useful part of every pass: the edges are lost to shift and the middle can be hurt by rate.

Why spreading factor, bandwidth and packet length matter

Each LoRa chirp lasts 2 to the power of the spreading factor divided by the bandwidth. Higher spreading factors give longer chirps and better range, which is why they are attractive for space links, but they also produce longer frames, and a longer frame gives the drifting frequency more time to wander. Higher spreading factors also pack symbols closer together in frequency, so a smaller error is enough to confuse them. Narrow bandwidths are hit hardest by the static shift, because 25 percent of a narrow channel is a small number of hertz.

LoRaWAN's low data rate optimization helps. It spaces symbols four times further apart, which makes the link more robust to Doppler rate at the cost of throughput. An analytical study from the University of Oulu and VTT found that even the most vulnerable setting, spreading factor 12 at 125 kHz and 433 MHz with a 59-byte payload and a satellite at 560 km, can still deliver more than 82 percent of packets, which shows the link is workable when parameters are chosen with Doppler in mind.

How engineers work around it

The most direct fix is pre-compensation: the ground device shifts its own transmit frequency by the opposite of the expected Doppler. The usual way to predict it uses a GNSS position plus the satellite's published orbit, but that costs energy, hardware and processing on a tiny sensor. One alternative has the satellite broadcast a short beacon made only of the LoRa preamble; the device measures the beacon's frequency error with the radio's built-in frequency error indicator and corrects its uplink. In simulations based on real satellite reception data, that approach removed Doppler-shift losses in the tested cases for about a 2.5 percent battery-life cost.

Other work fixes the problem at the receiver. Researchers have shown that the downchirps and extra pilot chirps inside a packet can be used to estimate shift and rate on the fly, needing only a software or firmware update. A third route avoids long chirps altogether: LR-FHSS breaks a packet into short narrowband hops, so the frequency barely drifts within any one hop.

Commercial systems have already flown LoRa to orbit, with operators adding their own protocols to make it work over long distances. Newer research keeps refining the picture, including reinforcement-learning methods that pick transmission settings to reduce Doppler-driven packet loss. The practical lesson for anyone building a satellite-linked sensor is the same: treat Doppler as a design input, choosing spreading factor, bandwidth, payload length and timing within the pass, not as noise to be tolerated.