What is LR-FHSS, and how does it let LoRaWAN handle far more devices?

LR-FHSS (Long Range Frequency Hopping Spread Spectrum) is an uplink-only physical layer for LoRaWAN that chops each packet into short fragments hopping across many narrow frequencies, so a single gateway or satellite can hear vastly more sensors at the same range as LoRa.

The short answer: LR-FHSS trades per-device speed for network capacity

LR-FHSS, short for Long Range Frequency Hopping Spread Spectrum, is an alternative physical layer for LoRaWAN, announced by Semtech, designed for very large sensor networks and for sensors that talk directly to satellites. It keeps roughly the same radio link budget as classic LoRa, so the range is similar, but it sends each packet as a series of short pieces that hop across many narrow frequencies. Because two devices rarely land on the same frequency at the same moment, far more of them can share one receiver before collisions take over.

It is used only for the uplink, from sensor to gateway. Downlink traffic still uses ordinary LoRa modulation. Devices still access the channel the simple LoRaWAN way: they pick an enabled channel at random and transmit whenever they have data, with no coordination.

How an LR-FHSS packet is built: headers, fragments and coding

Every LR-FHSS packet starts with several copies of a small header. In the European modes there are three copies for the most robust data rate and two for the faster one. Each header is 5 bytes including a checksum, protected by a rate-1/2 convolutional code, and lasts about 0.233 seconds. The header tells the gateway the hopping sequence, payload length, data rate and coding rate. The gateway only needs to catch one header copy to know where to look for the rest.

The payload is then split into fragments of about 50 milliseconds, each sent once on a different frequency. The payload bits are convolutionally encoded, at a coding rate of 1/3 for the robust modes and 2/3 for the faster ones, and decoded with a Viterbi decoder. That redundancy is what makes hopping safe: the original analysis notes that a packet can usually still be rebuilt even if one header copy and about a third of the payload bits are lost to collisions or interference.

Channels, subcarriers and the hopping pattern

LR-FHSS works inside a wider band called an Occupied Channel Width. In Europe that is either 137 kHz or 336 kHz; in the US 902 to 928 MHz band there are eight channels about 1.523 MHz wide. Each channel is divided into 488 Hz subcarriers, and successive fragments must jump at least a minimum distance apart, grouping the usable subcarriers into grids. The hop order comes from a hash of a random per-device value and the fragment number, so different devices follow different paths. A gateway does not need to know any device's sequence in advance, because the header announces it.

European data rates are about 162 bits per second for the robust modes and about 325 bits per second for the faster ones, with maximum MAC payloads of 58 and 123 bytes respectively. On the hardware side, newer Semtech radios such as the SX1261, SX1262 and SX1268 support it fully, and gateways built on the SX1302 chip can decode it after a firmware upgrade.

How much more capacity, and what it costs

In simulations under ideal channel conditions and a 1 percent duty cycle, the original performance analysis found LR-FHSS network goodput about one order of magnitude above fast LoRa and two orders above the slowest, longest-range LoRa setting. It estimated total European LR-FHSS capacity at roughly 36 times (robust mode) and 15 times (faster mode) that of a slow LoRa network spread across all its channels and data rates. That is the reason it suits low-Earth-orbit satellite IoT, where one receiver in the sky sees thousands of devices at once.

The price is paid at each device. A single node can send fewer packets per hour than with fast LoRa, and packets spend longer on the air. A 2025 urban field campaign in Halifax, Canada, on the US915 band measured 1,713 milliseconds on air for the robust LR-FHSS mode against 330 ms for LoRa SF10, which means more energy per message.

What real-world tests show, and what is still hard

That Halifax campaign walked about 85 km of city streets with a single rooftop gateway. The robust LR-FHSS mode delivered up to 20 percent more packets than LoRa SF10 in dense urban areas, while maximum distance was nearly the same, 3.97 km versus 3.88 km. The authors credit frequency hopping plus stronger coding for getting through cluttered urban paths rather than reaching further. The faster LR-FHSS mode did not consistently beat LoRa.

Open problems remain. Detection at the receiver is more complex than for LoRa, small hopping grids can still cause fragment collisions, and researchers are still working on smarter hopping sequences, coexistence with existing LoRa traffic, and how adaptive data rate should work in mixed networks. At the time of the field study, the gateway firmware was not public and commercial LR-FHSS products were scarce.