Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Jingxiang Huang, Samer Lahoud

TL;DR

The paper analyzes LoRa's physical-layer robustness to narrowband interference (BPSK and GMSK) and demonstrates that treating interference as AWGN overestimates symbol errors. It derives a stationary-phase relationship between the interferer's complex envelope and LoRa's decision statistic, and provides a two-segment INR(SNR) model to predict tolerable interference across spreading factors SF7–SF12. The results reveal modulation-specific impairment: AWGN causes the strongest degradation, followed by BPSK and then GMSK, with clear implications for coexistence and carrier-sensing strategies in mixed-IoT environments. These insights enable more accurate interference modeling and motivate adaptive sensing thresholds keyed to interferer type to improve overall network throughput.

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.

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

TL;DR

The paper analyzes LoRa's physical-layer robustness to narrowband interference (BPSK and GMSK) and demonstrates that treating interference as AWGN overestimates symbol errors. It derives a stationary-phase relationship between the interferer's complex envelope and LoRa's decision statistic, and provides a two-segment INR(SNR) model to predict tolerable interference across spreading factors SF7–SF12. The results reveal modulation-specific impairment: AWGN causes the strongest degradation, followed by BPSK and then GMSK, with clear implications for coexistence and carrier-sensing strategies in mixed-IoT environments. These insights enable more accurate interference modeling and motivate adaptive sensing thresholds keyed to interferer type to improve overall network throughput.

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.

Paper Structure

This paper contains 10 sections, 24 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Related Works
  4. Contributions
  5. LoRa Demodulation Process
  6. Simulation Setup and Result
  7. Symbol Error Rate of LoRa with AWGN
  8. Symbol Error Rate of LoRa with AWGN and Interference
  9. INR Threshold for Non-zero SER
  10. Conclusion

Figures (7)

  • Figure 1: Frequency-domain example of $Y_s[k]$,$Y_n[k]$,$Y_i[k]$ and $Y[k]$.
  • Figure 2: Illustration of Assumptions.
  • Figure 3: Symbol Error Rate versus RSSI.
  • Figure 4: Curves of Symbol Error Rate versus INR under different SNR for SF7.
  • Figure 5: Curves of Symbol Error Rate versus INR under different SNR for SF12.
  • ...and 2 more figures