Why does an always-listening LoRa receiver drain batteries so fast?

An always-listening LoRa receiver drains batteries because a standard LoRa radio runs its whole receive chain, from amplifier to FFT, just to check whether a packet has arrived, so new designs split cheap detection from full demodulation.

The short answer: listening costs almost as much as receiving

An always-listening LoRa receiver drains batteries because a commercial LoRa radio cannot tell that a packet has arrived without running its full receive chain. The amplifier, the oscillator, the analog-to-digital converter and the digital dechirp and FFT stages all stay powered, so idle listening costs nearly as much as decoding real data.

A 2026 SenSys paper on a receiver called Longan puts numbers on it: a common SX1278 radio draws up to 42.4 mW in always-on mode, which would empty two AA cells of 2000 mAh in about 6.3 days. That is why battery LoRa devices almost never listen continuously, and why several research groups are now trying to break that limit.

How LoRaWAN device classes trade latency for battery life

LoRaWAN handles the problem by rationing listening. A Class A device sends when it chooses and then opens two short receive windows, RX1 and RX2; outside those windows the network cannot reach it. These devices are usually battery powered.

Class B adds scheduled receive slots, called ping slots, timed by beacons the gateways broadcast, so the network can reach a device at known moments. Class C leaves the second receive window open almost all the time except while transmitting. It gives the fastest response, which suits pumps and valves, street lighting and alarms, but Class C devices are normally mains powered because the radio rarely sleeps.

Where the power goes inside a LoRa receiver

Commercial LoRa chips use a superheterodyne design. A low-noise amplifier and a PLL-based local oscillator sit at the front, an ADC digitizes the intermediate frequency, and the baseband dechirps the signal and runs an FFT to find each symbol. The Longan authors point out that this same pipeline is shared between packet detection and demodulation, so the radio only knows a preamble is present after the FFT has already run.

Datasheet figures for commercial LoRa receivers collected by a 2026 University of Louvain paper range from about 9.9 to 35.6 mW. Module makers quote receive currents of several milliamps, compared with about a microamp in sleep, which is why the time a radio spends listening often matters more to battery life than how long it spends transmitting.

Wake-up receivers: detect cheaply, demodulate on demand

The common research answer is to split the job. A tiny, low-power circuit watches the channel for a LoRa preamble, and the full radio and microcontroller wake only when one is found. Longan does this with a tunnel-diode amplifier and oscillator, an analog dechirp stage and an off-the-shelf low-frequency wake-up chip that handles detection at about 5.75 microwatts. The whole detector listens continuously at about 1.16 mW, which the authors report as roughly 36 times lower than an SX1278 in receive mode.

The Louvain team took a chip-design route: a 65 nm LoRa-compatible receiver whose amplifier, mixer and filter settings can be dialed between sensitivity and power. In main mode it reaches -138 dBm at 4.57 mW, it can save up to about 62.6 percent power by giving up 8.3 dB of sensitivity, and a dedicated wake-up mode listens at about 1.5 mW with -126 dBm sensitivity and 32 ms latency. Another 2026 paper, Hark, targets the same Class C problem with the stated aim of keeping the standard LoRa physical layer and working with existing commercial transmitters.

What is still hard

Every design pays somewhere. Cheaper detection usually means less sensitivity, and sensitivity is what gives LoRa its range: earlier sub-milliwatt receivers reached only hundreds of meters. Longan measured a 700 m line-of-sight link at low transmit power and only estimated, with a link calculator, about 1.6 to 2 km at 20 dBm, with more field tests still planned. The Louvain chip found that its oscillator still dominates wake-up power, and some of its noise results come from simulation rather than measurement.

Until such receivers ship in commercial chips, the practical rule stands: battery LoRa nodes should listen as little as possible, use Class A or Class B where latency allows, and keep Class C for devices with a steady power supply.