Physical-Layer Analysis of LoRa Robustness in the Presence of Narrowband Interference

Jingxiang Huang, Samer Lahoud · 2025

Read Physical-Layer Analysis of LoRa Robustness in the Presence of Narrowband Interference by Jingxiang Huang, Samer Lahoud (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-01088.txt.

Introduction

Jingxiang Huang and Samer Lahoud analyze LoRa wireless links under interference in a 2025 paper. The Internet of Things is filling the sub-gigahertz unlicensed bands with protocols like LoRa, Sigfox, and LR-FHSS. They must share the same frequencies, so they interfere with each other.

The authors study LoRa's robustness at the physical layer when narrowband interference is present. Physical-layer analysis looks at the raw signal, before any protocol fixes are applied. Knowing how much interference LoRa can tolerate helps planners deploy sensors in crowded spectrum. Their results inform choices about channels, spreading settings, and expected range.

Catalog record

CREATORJingxiang Huang, Samer Lahoud
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE27 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-01088

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Physical-Layer Analysis of LoRa Robustness in the Presence of Narrowband Interference Jingxiang Huang, Samer Lahoud Abstract With the rapid development of Internet of Things (IoT) technologies, the sub-GHz unlicensed spectrum is increasingly being shared by protocols such as Long Range (LoRa), Sigfox, and Long-Range Frequency-Hopping Spread Spectrum (LR-FHSS). These protocols must coexist within the same frequency bands, leading to mutual interference. This paper investigates the physical-layer impact of two types of narrowband signals (BPSK and GMSK) on LoRa demodulation. We employ symbol-level Monte Carlo simulations to analyse how the interference-to-noise ratio (INR) affects the symbol error rate (SER) at a given signal-to-noise ratio (SNR) and noise floor, and then compare the results with those for additive white Gaussian noise (AWGN) of equal power. We demonstrate that modelling narrowband interference as additive white Gaussian noise (AWGN) systematically overestimates the SER of Chirp Spread Spectrum (CSS) demodulation. We also clarify the distinct impairment levels induced by AWGN and two types of narrowband interferers, and provide physical insight into the underlying mechanisms. Finally, we fit a two-segment function for the maximum INR that ensures correct demodulation across SNRs, with one segment for low SNR and the other for high SNR. I Introduction I-A Background Coexistence among Internet of Things (IoT) devices in unlicensed spectrum has become an important challenge, with more systems sharing the same frequency bands. In contrast to licensed spectrum, where regulators allocate dedicated channels, unlicensed bands rely on simple rules such as duty-cycle limits and carrier sensing mechanisms (also known as Listen Before Talk, LBT). As device density grows in urban, industrial, and agricultural settings, transmissions from different technologies can overlap in time or frequency, causing interference that degrades performance for all users. Low-Power Wide-Area Networks (LPWANs) are a family of radio technologies for battery-powered devices that need to send small amounts of data over kilometers of range. LPWANs operate in the sub-GHz unlicensed bands and employ frequency hopping or spread-spectrum modulation, extremely low duty cycles, and simple MAC protocols to minimize power consumption and cost. LoRa (Long Range) is a widely adopted LPWAN protocol that achieves long-range coverage by using chirp spread spectrum (CSS) at the physical layer [ 12 , 14 ] . Other LPWAN protocols, such as Sigfox and Long-Range Frequency-Hopping Spread Spectrum (LR-FHSS), adopt narrowband approaches in the sub-GHz unlicensed bands [ 13 , 2 ] . When their transmissions overlap a LoRa channel, the interference can corrupt the demodulation process and increase the symbol error rate (SER). Understanding how different types of narrowband interference affect LoRa demodulation is helpful for developing targeting LBT strategies and improve coexistence among diverse IoT technologies. I-B Related Works Intra-technology interference happens when multiple LoRa devices use the same spreading factor (SF) simultaneously. The work [ 5 ] presents a mathematical model and simulations of the interference between two LoRa symbols with the same SF when they are chirp- or phase-aligned, re-evaluating the coverage for the LoRa network. Furthermore, the work [ 1 ] expanded on the previous work, proving that assuming full chirp-level alignment represents a worst-case scenario. From another perspective, another research [ 3 ] questioned the orthogonality of different spreading factors. It quantified imperfect orthogonality through simulation, which highlights that LoRa’s physical-layer orthogonality cannot prevent collisions in dense networks. LoRa also suffers from inter-technology interference. The impact of inter-technology interference has been measured under different scenarios. The work [ 4 ] investigated the impact of broadband jamming and single-tone jamming and found that LoRa’s chirp spread spectrum can mitigate the effect of tone jamming, whereas broadband jamming simply raises the noise floor. The papers [ 11 , 6 ] used network models to investigate and simulate Sigfox interference on LoRa, examining how transmit power, distance, and the number of nodes affect LoRa’s error rate and throughput. However, neither of them discusses how the interference signal would affect the demodulation process at the physical layer. LoRa also supports operation in the 2.4 GHz band, where it coexists with Wi-Fi, Bluetooth Low Energy (BLE) and Long Term Evolution (LTE), all of which act as sources of broadband interference for LoRa. The work in [ 9 , 10 , 8 ] studied the interference brought by Wi-Fi, BLE, and LTE at the physical layer, respectively. The authors demonstrate that LoRa signals exhibit robustness against broadband interference and further show that this resilience depends on factors such as the signal-to-interference ratio (SIR), the modulation schemes used by the interferers, and the degree of overlap between the LoRa and interference bandwidths. Given the various types of interference impacting LoRa network performance, detecting and avoiding interference at the physical layer is crucial for improving coexistence. Energy-based carrier sensing is a traditional approach. The work [ 7 ] highlights that the choice of sensing threshold is critical: if the threshold is too high, devices may overlook weak but valid signals; setting the threshold too low can make the device overly sensitive to random noise. Although energy-based carrier sensing is simple to implement, it cannot distinguish between noise and different types of interference, which limits its potential for more nuanced approaches. Many studies have examined how inter-technology interference degrades LoRa reception through measurements and simulations. However, physical-layer analyses that elucidate the mechanisms by which different interferer types uniquely affect LoRa demodulation are still lacking. Furthermore, in the context of LoRa networks, it is still unclear if white noise and interference can be treated equivalently in network-performance assessments or when designing carrier sensing schemes. I-C Contributions The main contributions of this paper are as follows: i) We demonstrate that modeling narrowband interference as AWGN results in excessively conservative performance predictions through Monte Carlo simulations. ii) By applying the principle of stationary phase, we derive an explicit relationship between the complex envelope of interference and its impact on the CSS decision statistic. iii) Comprehensive Monte Carlo experiments for spreading factors 7-12 are condensed into a two-segment analytical fit that yields the maximum tolerable interference-to-noise power ratio (INR) as a function of signal-to-noise ratio (SNR). II LoRa Demodulation Process In this section, we present the mathematical analysis of the LoRa demodulation process in the presence of AWGN and interference. Suppose the bandwidth of LoRa is denoted as B , the duration of a LoRa symbol is determined by its bandwidth and spreading factor (SF) according to the following equation T_{s}=\frac{2^{\mathrm{SF}}}{B}. For a SF \in\{7,\dots,12\} , each LoRa symbol encodes SF bits, yielding N=2^{\text{SF}} distinct symbols. The symbol indices range from 0 to N-1 , and the symbol at index 0 is called the upchirp , which is defined as \sigma_{0}[n]=\exp\!\Bigl(j\pi\frac{n(n-N)}{N}\Bigr). (1) According to the datasheet, when a Semtech’s LoRa chip SX1276/7/8/9 processes an incoming signal, the signal is down-converted to baseband and goes through an ADC for analog-to-digital conversion [ 12 ] . We assume the sampling rate equals the bandwidth of the LoRa signal. Denote the received signal \sigma[n] as an addition of the LoRa signal \sigma_{s}[n] , AWGN \sigma_{n}[n] and interference \sigma_{i}[n] for each sampling point n \sigma[n]=\sigma_{s}[n]+\sigma_{n}[n]+\sigma_{i}[n]\quad n=0,1,...,N-1. (2) The first step of demodulation is dechirping , which consists of multiplying the received signal by the complex conjugate of the upchirp \sigma_{0}^{*}[n] for each sampling point n y[n]=\sigma_{0}^{*}[n]\,\sigma[n]=\sigma_{0}^{*}[n](\sigma_{s}[n]+\sigma_{n}[n]+\sigma_{i}[n]). (3) An N -point Discrete Fourier Transform

Read the complete text