The motivation comes from a very concrete problem. Volcano observatories, seismic networks, and geophysical stations in regions without cellular or wired infrastructure need radio links that can carry uninterrupted sensor data across ten to twenty kilometres. A triaxial seismometer sampling at 200 samples per second with 24 bits per axis generates a raw stream of 14.4 kilobits per second, but telemetry protocols and encapsulation overheads such as MiniSEED formatting standardise on a 115,200 bits per second serial interface at the sensor output. To prevent buffer accumulation, the wireless link must sustain goodput at or above that serial speed. LoRa, the proprietary chirp spread spectrum modulation that has become ubiquitous in smart agriculture and volcanic monitoring networks, simply cannot deliver: at the high spreading factors required for robust long-distance links, its effective air rate collapses to one to five kilobits per second, making continuous high-rate streaming fundamentally impossible.

The research team, affiliated with monitoring efforts around the active Misti and Sabancaya volcanoes in southern Peru, validated the throughput limitation empirically. In comparative benchmarks against commercial Ebyte E22 modules at 433 MHz and 915 MHz, both sub-gigahertz LoRa systems saturated at a maximum sustained goodput of just 29.38 kilobits per second, enforced by a physical packet inter-arrival floor of 65.35 milliseconds. That is a shortfall of nearly four times against the 115.2 kilobits per second target, meaning such links can carry a 100-samples-per-second seismic stream but cannot handle a 200-samples-per-second stream with any robust margin. Previous work on the 16-kilometre Misti–Arequipa path had already confirmed the geometric feasibility of the radio link while highlighting exactly this throughput wall.

OpenFLRC’s answer is to move to the 2.4 GHz industrial, scientific and medical band and exploit FLRC, or Fast Long-Range Communication, a modulation mode native to the SX1280 transceiver family. In its fastest configuration, FLRC achieves a 1.3 megabit per second air rate with a receiver sensitivity of minus 97 dBm, sufficient for multi-kilometre links. The team’s carrier board integrates an Ebyte E28-2G4M27SX radio front-end delivering up to 27 dBm, or 500 milliwatts, of output power through a Skyworks power amplifier, alongside an ESP32-S3 dual-core microcontroller running at 240 MHz with 16 MB of flash and 8 MB of PSRAM. Crucially, the design grants users bare-metal control over the radio, exposing internal hardware buffers, explicit transmit and receive timings, and direct SPI interrupts through a custom C++ library, a level of deterministic control that closed commercial LoRa modules abstract away entirely.

The engineering details reveal why the platform performs as it does. Driving a 500-milliwatt power amplifier continuously under high-rate streaming draws rapid current pulses peaking at 550 milliamperes with nanosecond-scale rise times, which would destabilise a typical breadboard or prototype supply. The carrier board embeds a five-capacitor decoupling ladder directly adjacent to the radio module’s power pins: two 150-microfarad low-ESR tantalum capacitors acting as a local bulk energy reservoir, in parallel with three high-frequency X7R ceramic capacitors. A continuous 1-ounce copper ground plane on the board’s underside, stitched to the module with thermal vias, acts as a passive heat spreader for the roughly 1.3 watts dissipated as heat during full-power transmission. The board also routes the radio’s transmit and receive enable lines through dedicated low-capacitance traces, allowing the firmware to switch the RF front-end deterministically within 2.5 microseconds and preventing destructive amplifier reflections during transmit-to-receive turnarounds.

The firmware architecture is equally deliberate. Under FreeRTOS, the dual-core processor divides labour asymmetrically: Core 0 ingests the incoming 115,200 bits per second serial stream into a lock-free circular ring buffer in RAM, while Core 1 runs a dedicated radio scheduler that fetches 117-byte chunks and pushes them to the SX1280’s FIFO over an 18 MHz SPI bus with direct memory access, shrinking each payload transfer to roughly 56 microseconds. Radio state transitions are governed strictly by a hardware interrupt that triggers a RAM-cached fast interrupt service routine, eliminating the scheduling jitter that plagues polling-based designs. The team’s timing budget analysis shows a minimum theoretical packet cycle of 840.9 microseconds, and the empirical benchmark achieved 884.3 kilobits per second of goodput at a one-millisecond packet interval, corresponding to a 73.1 percent channel efficiency relative to the raw 1.3 megabit per second air rate. Notably, 82 percent of Core 0’s processing capacity and more than 470 kilobytes of SRAM remain free for user data acquisition and edge signal processing.

Validation was conducted with unusual rigour for open hardware. Rather than relying on uncontrolled over-the-air tests, the team built a conducted radio-frequency bench using a calibrated chain of SMA attenuators totalling up to 145 decibels of path loss, characterised segment by segment with a NanoVNA vector network analyser at 433 MHz, 915 MHz, and 2.44 GHz. This approach isolates true receiver sensitivity and firmware behaviour from the multipath fading, interference, and antenna pointing errors that make field measurements irreproducible. In attenuation sweeps, the sub-gigahertz LoRa links sustained near-unity packet delivery up to 130 decibels of conducted path loss, while the 2.4 GHz FLRC link held a 96.2 percent delivery ratio at 125.8 decibels before failing completely, a roughly five-decibel penalty consistent with the higher noise bandwidth of FLRC compared to narrow-band chirp spread spectrum.

Spectral measurements with a TinySA Ultra analyser confirmed the platform’s regulatory compliance. The occupied bandwidth measured between 1.336 and 1.347 megahertz across three operating channels, closely matching the configured FLRC parameters, and second-harmonic suppression ranged from 38.5 to 41.8 decibels below the carrier, comfortably within FCC Part 15.247 and ETSI EN 300 328 emission limits. The power control loop proved highly linear across an 18-decibel dynamic range. Link-budget emulation for the 16.4-kilometre Misti–Arequipa volcanic corridor, based on the ITU-R free-space model and the empirical sensitivity thresholds, indicates the link can theoretically close at that distance, though the authors recommend upgrading to 19 dBi parabolic grid antennas or adding an intermediate repeater to maintain a proper fade margin for field reliability.

The trade-offs are candidly acknowledged. The 2.4 GHz band incurs a 7.9-decibel free-space path loss penalty compared to 915 MHz, making the system dependent on line-of-sight conditions, and continuous full-power transmission demands careful thermal management in any field enclosure. The v1.0 firmware implements dedicated point-to-point streaming only, with multi-node contention protocols and the SX1280’s built-in time-of-flight ranging engine reserved for future releases. Yet the regulatory picture favours the chosen band: sub-gigahertz spectrum is fragmented regionally and subject to duty-cycle limits as low as 0.1 percent in Europe that legally prohibit continuous telemetry, whereas the 2.4 GHz band is globally harmonised with no duty-cycle restrictions, and rain attenuation over a 16-kilometre path is negligible at these frequencies.

Perhaps most striking is the price and accessibility. Each complete transceiver node costs 32.70 US dollars, built from commodity modules and a two-layer printed circuit board, and the entire project, including schematics, gerber files, firmware, bill of materials, calibration data, and benchmark videos, is published under a Creative Commons Attribution 4.0 license and certified by the Open Source Hardware Association under UID PE000005. For research groups monitoring volcanoes, earthquakes, or any high-rate phenomenon in infrastructure-poor regions, OpenFLRC demonstrates that the era of choosing between long range and sufficient bandwidth may finally be ending, and that a thirty-dollar open board can do what expensive proprietary telemetry systems have long monopolised.

Subject of Research: Open-source 2.4 GHz FLRC wireless transceiver hardware for high-throughput, low-latency continuous scientific telemetry

Article Title: Open-source SX1280 2.4 GHz FLRC transceiver: Overcoming sub-GHz LoRa bottlenecks for high-rate, low-latency continuous telemetry

Article References: Choquepuma, K. J. C., Rodriguez, K. E. A., Arenas Mamani, F. A., & Hilario-Tacuri, A. (2026). Open-source SX1280 2.4 GHz FLRC transceiver: Overcoming sub-GHz LoRa bottlenecks for high-rate, low-latency continuous telemetry. HardwareX, Article e00848. https://doi.org/10.1016/j.ohx.2026.e00848

Image Credits: AI Generated

DOI: 10.1016/j.ohx.2026.e00848

Keywords: OpenFLRC, SX1280, FLRC modulation, LoRa bottleneck, wireless telemetry, volcano monitoring, seismic networks, ESP32-S3, open-source hardware, 2.4 GHz ISM band, packet delivery ratio, HardwareX

The first era, spanning 1990 to 1999, was a period of modest output and constrained ambition. Annual publication counts remained low, reflecting the limitations of sparse seismic networks, limited computational resources, and predominantly analog or early digital instrumentation. Monitoring during this decade was largely bound by data availability and processing capacity, so research clustered around waveform interpretation, source characterization, and station-based physical analysis within the classical seismological framework. It was, in essence, an instrument-centered discipline in which a handful of carefully placed stations and expert analysts defined the frontier of what could be observed about the Earth’s shaking.

The second era, from 2000 to 2018, brought sustained growth driven by an observational expansion that fundamentally changed what seismologists could see. Broadband seismic networks proliferated, dramatically improving the fidelity and coverage of ground-motion recordings. At the same time, space-based geodesy matured: continuous Global Navigation Satellite System (GNSS) observations enabled near-real-time tracking of crustal deformation, while Interferometric Synthetic Aperture Radar (InSAR) offered systematic mapping of how the ground shifted across vast regions after earthquakes. These techniques pulled geodesy and seismology into closer collaboration, broadening the research agenda from single-event analysis toward regional-scale monitoring, deformation tracking, and multi-parameter interpretation of seismic processes.

The third and most recent era, beginning in 2019, is defined by an unmistakable acceleration. Publication output rose sharply, and keyword and co-citation patterns became dominated by three emerging fronts: deep-learning-based earthquake detection and phase picking, Distributed Acoustic Sensing (DAS), and the joint use of GNSS and InSAR data. Benchmark studies such as PhaseNet and Earthquake Transformer, which apply neural networks to the automated identification of seismic phases in continuous waveform streams, have become among the most heavily co-cited works in the field. DAS, meanwhile, converts ordinary fiber-optic cables into dense seismic arrays capable of sampling ground motion every few meters over tens of kilometers, on land and across the seafloor. The authors are careful, however, to frame this surge as evidence of rising research attention rather than proof of mature, field-wide operational adoption, since the phase covers only six years and recent publications have had little time to accumulate citations.

Beyond technology, the analysis exposes a deeply uneven global landscape. Although researchers from 105 countries and regions contributed to the corpus, the United States and China together account for roughly half of all publications. The American lead reflects decades of investment in seismic and geodetic infrastructure, early adoption of open-data policies, and sustained methodological innovation. China’s rapid ascent, by contrast, corresponds to large-scale modernization of national monitoring networks, integrated GNSS-InSAR observation systems, and heavy investment in AI-enabled monitoring and earthquake early warning. A second tier of productivity is occupied by Italy, France, and Germany, and several European countries, notably France, Germany, and England, achieve higher citation rates per publication, suggesting an emphasis on methodologically integrative and internationally collaborative work.

The collaboration network analysis places the United States at the center of the field’s international architecture, with high betweenness centrality indicating a bridging role that connects otherwise separate regional research communities. At the institutional level, 2,372 organizations contributed to the corpus, led by national and transnational bodies that operate large observational systems, including the United States Department of Energy, the French National Centre for Scientific Research (CNRS), the China Earthquake Administration, and the Russian Academy of Sciences. Organizations such as INGV in Italy, GFZ in Germany, and the Helmholtz Association exemplify the fusion of operational monitoring mandates with strong research missions, occupying the interface between fundamental seismology and applied hazard mitigation. Notably, institutional productivity and average citation impact do not follow the same rankings, hinting at a structural distinction between data-producing operational agencies and conceptually influential research institutions.

Perhaps the most intellectually interesting finding concerns continuity. Despite three decades of technological upheaval, the co-cited author network remains anchored by foundational contributors to seismic source theory, wave propagation, event location, and earthquake mechanics. Modern monitoring workflows, in other words, still rest on classical seismological principles; the new tools have expanded the scale and speed at which physically grounded analyses can be performed rather than displacing the underlying physics. At the same time, a growing cluster of machine-learning researchers has become prominent in the co-citation structure, indicating a rebalancing of methodological priorities toward automated detection, phase picking, denoising, and large-scale catalog construction. The authors also caution that the growth of AI-related literature reflects not only genuine algorithmic innovation but also the diffusion of established machine-learning methods across new regional datasets and case studies.

The co-cited reference analysis identifies three benchmark directions that now structure the field’s intellectual agenda. The first is automated detection and phase-picking, exemplified by deep-learning models capable of processing vast continuous waveform archives. The second is dense and distributed sensing, where DAS benchmark studies demonstrate that existing fiber-optic infrastructure, including submarine telecommunications cables, can be repurposed as large-scale seismic observation arrays. The third is integrated and shareable data resources: curated benchmark datasets such as STanford EArthquake Dataset (STEAD) have enabled systematic model development, cross-dataset comparison, and reproducible evaluation, while GNSS-InSAR integration increasingly links seismic signals with deformation observations. Together, these clusters suggest that earthquake monitoring is converging on automation, dense observational systems, and multi-sensor data infrastructure as interconnected pillars.

The study’s authors are candid about the limitations of bibliometric evidence. Relying exclusively on the Web of Science may underrepresent regional and non-English journals, and the deliberately broad search query captured some methodologically adjacent industrial and reservoir monitoring work; a conservative sensitivity analysis flagged 376 such records, or 13.17 percent of the corpus, and confirmed that the main growth trajectory and core themes remained stable in the trimmed subset. Citation bias, time-lag effects, and database-driven visibility all complicate interpretation, and the authors stress that co-occurrence and co-citation indicators measure visibility and intellectual influence, not operational performance. No bibliometric signal can establish whether AI genuinely improves detection across all networks, whether DAS outperforms conventional arrays in practice, or whether GNSS-InSAR fusion improves warning performance.

What the analysis does offer is a roadmap. The convergence of seismic, geodetic, and fiber-optic themes points toward testable opportunities in heterogeneous monitoring: combining waveform detections with GNSS displacement for rapid magnitude characterization, using InSAR to map coseismic deformation when cloud-free acquisitions permit, and deploying DAS to densify observations along terrestrial and submarine fiber routes. The authors call for rigorous validation of AI models on geographically and temporally external networks, transparent comparison against classical baselines, error stratification by magnitude and noise regime, and reproducible benchmarks. They also highlight a representation gap: bibliometric visibility concentrates in well-funded systems, while many high-risk developing regions remain underrepresented, underscoring the need for interoperable regional catalogs, shared archives, and benchmark datasets suited to sparse networks and complex terrain. As earthquake monitoring moves toward hybrid paradigms that fuse computational intelligence with physical understanding, this study provides the clearest quantitative picture yet of where the field has been, and where its attention is now heading.

Subject of Research: Bibliometric analysis of technological evolution and research trends in earthquake monitoring from 1990 to 2024

Article Title: Technological evolution and emerging research hotspots in earthquake monitoring from 1990 to 2024 revealed by bibliometric evidence

Article References: Liu, S., Wu, L., & Liao, C. (2026). Technological evolution and emerging research hotspots in earthquake monitoring from 1990 to 2024 revealed by bibliometric evidence. Discover Geoscience, 4(1), Article 387. https://doi.org/10.1007/s44288-026-00744-7

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00744-7

Keywords: earthquake monitoring, bibliometrics, artificial intelligence, Distributed Acoustic Sensing, GNSS, InSAR, deep learning, seismic networks, earthquake early warning, crustal deformation, Web of Science, research trends