{"id":"82c48989-8f36-4e2c-914d-0260b31f702e","arxiv_id":"2412.04563","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Field tests in rural Nepal show that a LoRa link at 62.5 kHz bandwidth with spreading factor 8 gives the best balance of signal strength, noise margin, and packet delivery over a 5 km line-of-sight path.","lead":"This paper measures LoRa radio performance across bandwidth, spreading factor, and coding rate on a 5 km clear-sight link in hilly Nepal. It recommends 62.5 kHz bandwidth, spreading factor 8, and coding rate 4/8 as the best settings for low-data-rate rural IoT links.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed optimum is underdetermined by the reported data: at SF=8 the 62.5, 125, and 250 kHz settings differ by <0.1 dB in RSSI and SNR, with no error bars or significance tests.","rationale":"The strongest claim requires two things to be true: (i) the measured configuration genuinely outperforms the alternatives on the tested link, and (ii) that link represents the terrain class. The reader's weakest assumption identifies (ii), but I find (i) to be the more immediate and load-bearing weakness. The reported tables show the candidate bandwidths at SF=8 differing by hundredths of a decibel, a scale far below typical sample-to-sample variation in outdoor LoRa measurements. Without uncertainty quantification, the phrase 'optimum configuration' cannot be supported even for the exact link tested. The paper deserves credit for performing a real outdoor deployment, publishing raw data and code, and reporting a concrete configuration; those are genuine strengths. However, the analytical step from tables to the single recommended triple goes beyond what the data can resolve. The reader's CONDITIONAL verdict remains appropriate: the paper should either add proper uncertainty analysis and significance testing or explicitly downgrade the headline from 'optimum' to 'one of several statistically equivalent settings'. The additional coding-rate labeling issue (5/8-7/8 are not valid LoRa coding rates) reinforces the need for revision but is secondary to the statistical underdetermination of the main recommendation.","tokens_in":14767,"tokens_out":5238,"duration_ms":53120,"concrete_test":"Using the raw per-packet measurements publicly deposited in the Zenodo repository [29], compute paired or unpaired two-sided 95% confidence intervals for the mean SNR and RSSI differences between SF=8 at BW=62.5 kHz versus BW=125 kHz and BW=250 kHz, averaging over all available repetitions. If the confidence interval for the 62.5-vs-125 kHz difference includes zero, or is wider than roughly 0.5 dB, then the claimed optimal bandwidth is not statistically distinguishable from its nearest competitor and the conclusion should be recast as 'statistically equivalent within measurement error' rather than 'optimal'.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central recommendation of BW=62.5 kHz, SF=8, CR=4/8 is not established by the reported measurements. At SF=8, Table 2 lists RSSI values of -91.8, -91.83, and -91 dBm for bandwidths 62.5, 125, and 250 kHz, and Table 3 lists SNR values of 10.2, 10.18, and 10.04 dB. The paper's own text says all three bandwidths can be chosen, and the decisive 62.5-vs-125 differences are 0.02 dB in SNR and 0.03 dB in RSSI. These differences are far below what 5-7 samples of a field link can resolve; no standard deviations, confidence intervals, or significance tests are provided anywhere. The packet-loss table also cannot discriminate among these candidates, because SF=8 shows 0% loss at every bandwidth. Thus, even granting that this single line-of-sight link is representative of hilly Nepalese terrain, the data do not uniquely select 62.5 kHz over neighboring bandwidths. The claim of an optimal configuration therefore rests on measurement noise, not on a demonstrated performance separation. In addition, Table 6 lists coding rates 5/8, 6/8, and 7/8, which are not valid LoRa SX1278 coding rates (the valid set is 4/5 through 4/8), so the coding-rate portion of the recommendation is also not supported as reported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports an outdoor experimental evaluation of LoRa (SX1278, 433 MHz) over a roughly 5 km clear-line-of-sight link in the hilly Kavrepalanchowk district of Nepal. The authors vary bandwidth, spreading factor, and coding rate, and record RSSI, SNR, and packet loss, along with a derived path-loss comparison. They also demonstrate a two-node TDMA scheme with ThingSpeak cloud integration. Based on the measurements, the paper recommends bandwidth 62.5 kHz, spreading factor 8, and coding rate 4/8 as the optimal configuration for long-range low-data-rate rural connectivity.","tokens_in":15064,"tokens_out":3536,"duration_ms":38379,"significance":"If the recommended configuration were robustly established, the paper would provide a concrete, low-cost deployment guideline for LoRa links in a specific type of hilly terrain, together with a reproducible public dataset (Zenodo/GitHub) and a working IoT integration demo. The transparency of sharing raw data and code is a genuine strength, as is the real field deployment rather than a simulation. However, the central recommendation is currently underdetermined by the reported statistics, and the coding-rate experiment appears to use values outside the valid LoRa parameter set. With appropriate revisions that either add statistical support or substantially narrow the claims, the paper could serve as a useful empirical data point for rural LoRa deployment.","major_comments":[{"comment":"The selection of 62.5 kHz over 125 kHz and 250 kHz at SF=8 is not supported by the reported data. For SF=8, the RSSI values are -91.8, -91.83, and -91.0 dBm for 62.5, 125, and 250 kHz, respectively, and the SNR values are 10.2, 10.18, and 10.04 dB. These differences are far smaller than what 5-7 averaged measurements without error bars or significance tests can resolve, and packet loss is 0% for all three configurations. The paper's own text in Section 4.5 acknowledges that all three bandwidths are viable, so the unique recommendation of 62.5 kHz needs either a statistical justification, an explicit decision criterion based on application constraints, or a revised conclusion that presents a small set of equivalent choices instead of a single optimum.","section":"4.5, Tables 2-4"},{"comment":"The coding rates listed as 4/8, 5/8, 6/8, and 7/8 are not valid LoRa coding rates for the SX1278. The SX1276/77/78/79 datasheet defines the LoRa coding rate denominator as 4 to 8 with the numerator fixed at 4, yielding 4/5, 4/6, 4/7, and 4/8. As a result, the coding-rate experiment in Section 4.6 and the resulting recommendation of CR=4/8 are based on configurations that the hardware does not actually implement as LoRa coding rates. This portion of the experiment and the related conclusion must be redone with valid rates or removed from the recommendation.","section":"Table 1 and Section 4.6"},{"comment":"The experiment covers a single clear-line-of-sight link between two fixed hilltop nodes under one weather window (23.5°C, 86% humidity). The abstract and conclusion extend the result to 'hilly and mountainous terrain' generally. Since the dominant challenges in such terrain are typically shadowing and non-line-of-sight propagation, a single LOS link cannot support a terrain-class recommendation. The paper should either restrict its claim to the tested LOS conditions or supplement the study with non-LOS or variable-weather measurements; otherwise the generalized framing in the abstract and conclusion is not empirically justified.","section":"3.5 and 4.4"},{"comment":"The packet-loss metric cannot discriminate among the recommended candidates because every configuration with SF≥8 reports 0% packet loss, and even SF=7 shows 0% loss at 250 and 500 kHz. This means the 'optimal' choice rests entirely on RSSI and SNR differences that, as noted above, are within measurement noise. The paper should explicitly state this limitation and avoid presenting packet loss as a differentiating factor in the optimal-configuration decision.","section":"4.3"}],"minor_comments":[{"comment":"The packet loss table is incorrectly numbered as 'Table 1', which duplicates the parameter table in Section 3.3; it should be Table 4.","section":"4.3"},{"comment":"The interpolation method is called 'PHCIP' in the text but should be 'PCHIP' (Piecewise Cubic Hermite Interpolating Polynomial); correct this typo.","section":"4.1"},{"comment":"Section 4.2 states that the best SNR is obtained at SF=8 and BW=10.4 kHz, while Section 4.5 recommends BW=62.5 kHz; reconcile these two statements or explain why the SNR-optimal bandwidth was not chosen.","section":"4.2"},{"comment":"The definition of ESP = RSSI + SNR - 10 log10(1 + 10^{0.1 SNR}) is nonstandard and reduces approximately to RSSI for large SNR; the paper should justify this formula and clarify how it relates to the SX1278's built-in RSSI and SNR registers.","section":"Section 3.4.4, Eq. (5)"},{"comment":"The spreading factor column formats 7, 8, 9, ... as 27, 28, 29, ... in a way that could be read as exponents; use superscript notation (2^7, 2^8, ...) to avoid confusion.","section":"Section 3.3, Table 1"},{"comment":"The abstract refers to a 'LoRa mesh network', but the experiment is a point-to-point link with two nodes sharing a channel via TDMA; either implement an actual mesh or use terminology such as 'multi-node LoRa network'.","section":"Abstract and Section 3.6"},{"comment":"The text says 'Table 4 illustrates the effect of the coding rate', but the referenced table is Table 6; update the cross-reference.","section":"4.6"}],"recommendation":"major_revision","confidential_remarks":"The paper's main value is as an openly documented field measurement from a region where such data are scarce. However, the headline recommendation is not statistically supported, and the coding-rate part reveals a likely misunderstanding of the LoRa physical layer. The authors should be encouraged to revise the claims to match the evidence (e.g., present a shortlist of equivalent configurations and clearly scope the result to the tested LOS link), and to either correct the coding-rate experiment with valid rates or drop it. The journal fit is acceptable for an applied experimental study if the claims are appropriately narrowed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look if you work on LoRa deployments in hilly terrain. The real content is a new public field dataset: RSSI, SNR, packet loss for 36 SF/BW combinations on a 5 km line-of-sight link in Kavrepalanchowk, Nepal, including bandwidths (10.4, 20.8 kHz) that previous field studies mostly skipped. The data and code are on Zenodo/GitHub, which makes this a replicable empirical data point rather than a simulation. The TDMA + ThingSpeak demo is straightforward but shows the system works end to end.\n\nThe soft spots are real but not fatal. The central recommendation—62.5 kHz, SF 8, CR 4/8—is not uniquely selected by the data. At SF 8 the RSSI values for 62.5, 125, and 250 kHz are -91.8, -91.83, and -91 dBm; SNR 10.2, 10.18, 10.04 dB. Those differences are below what 5-7 samples per configuration can resolve, and no error bars or significance tests are reported. The paper itself says all three bandwidths can be chosen, then picks 62.5 based on a marginal SNR edge. That's a judgment call, not a measured optimum.\n\nThe coding-rate experiment is the weakest part. Table 6 lists CR 4/8, 5/8, 6/8, 7/8, but valid SX1278 coding rates are 4/5 through 4/8. As reported, those settings aren't real LoRa configurations, so the CR=4/8 recommendation lacks a valid experimental basis. Also, the abstract says a 'mesh network' and claims the homemade antenna shows 'superior gain' compared to commercial options; neither is supported—the setup is a star/TDMA link, and the antenna gain is theoretical, never measured.\n\nThe single-link, single-weather-window limitation is acknowledged in the text, though the title and abstract overgeneralize to hilly/mountainous rural terrain. Still, the core RSSI/SNR/packet-loss measurements are honestly reported, the general SF/BW trends match prior work, and the paper is transparent about what it did and didn't measure.\n\nThis is a paper for practitioners deploying LoRa in similar terrain, not a conceptual breakthrough. It deserves peer review—the dataset is citable and the configuration recommendation is actionable if reframed as a case study rather than a general optimum. I'd send it out with a request for error bars, corrected CR values, and a title that doesn't overclaim.","headline":"Useful field dataset from hilly Nepal, but the headline 'optimum' isn't supported by the reported measurements.","tokens_in":15642,"tokens_out":2510,"would_cite":true,"duration_ms":19965,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Field measurements over a clear 5 km hilltop link in Kavrepalanchowk, Nepal, show that LoRa's best configuration for hilly rural low-data-rate links is 62.5 kHz bandwidth, spreading factor 8, and coding rate 4/8.","keywords":["LoRa","RSSI","SNR","packet loss","spreading factor","bandwidth","coding rate","rural IoT"],"falsifier":"Repeat the parameter sweep on a non-line-of-sight link between two slopes in the same district during monsoon fog, and compare spreading factor 10 or 12 at 125 kHz with spreading factor 8 at 62.5 kHz on packet loss and SNR; if the higher spreading factors win, the claimed optimum is a property of this one clear link rather than of hilly rural Nepal.","tokens_in":14576,"feed_emoji":"📡","tokens_out":10377,"duration_ms":100408,"temperature":0.7,"pith_summary":"This paper tries to establish which LoRa radio settings give the most dependable low-data-rate link in Nepal's hilly rural terrain. The authors built two low-cost sensor nodes and a gateway with SX1278 LoRa transceivers, measured RSSI, SNR, packet loss, and path loss over a clear 5 km hilltop link in Kavrepalanchowk, and swept bandwidth, spreading factor, and coding rate. They conclude that 62.5 kHz bandwidth, spreading factor 8, and coding rate 4/8 offer the best balance for long-range telemetry. If this holds, rural Internet-of-Things deployments in similar terrain have a concrete starting configuration and a demonstrated path to cloud-connected multi-node monitoring.","feed_headline":"For Nepal's hills, LoRa's sweet spot: 62.5 kHz, SF 8","feed_subtitle":"A 5 km hilltop field test balances signal strength, signal clarity, and zero packet loss.","key_machinery":"The measuring machinery is a parameter sweep over bandwidth (10.4 to 500 kHz), spreading factor (7 to 12), and coding rate (4/8 to 7/8), with RSSI and SNR read from the SX1278 chip's packet registers and packet loss counted by comparing sent and received packets. The path-loss comparison uses the effective signal power $\\mathrm{ESP} = \\mathrm{RSSI} + \\mathrm{SNR} - 10\\log_{10}(1 + 10^{0.1\\,\\mathrm{SNR}})$ and compares the resulting loss with the Friis free-space loss to isolate environmental effects. The selection rule is to keep packet loss at zero, prefer higher SNR, and preserve acceptable RSSI, which leads the paper to SF 8 with 62.5 kHz bandwidth and the lowest tested coding rate.","core_discovery":"On its own terms, the paper's central claim is that the LoRa physical-layer configuration matters measurably for link quality, and that for a specific hilly, mountainous rural setting the optimum is bandwidth 62.5 kHz, spreading factor 8, and coding rate 4/8. The measurements show RSSI falling sharply once the spreading factor rises above 8, SNR highest at low bandwidth with SF 8, zero packet loss for every tested spreading factor from 8 upward, and packet loss up to 54 percent at SF 7 on the narrowest bandwidths. Path-loss analysis shows the gap between actual loss and free-space loss grows with spreading factor, while wider bandwidths keep the loss closer to the ideal. The recommended configuration is presented as a trade-off: it sacrifices some data rate to keep signal clarity and reliability over long distances, and the authors note that data rate, latency, and bit-error-rate requirements may shift the choice.","pith_inferences":["The near-universal zero packet loss above spreading factor 7 suggests the test link had a large link margin; deployments at shorter range could trade reliability for higher data rate by dropping to spreading factor 7 with wider bandwidth.","Because temperature and humidity shift absolute signal strength, repeating the sweep in monsoon fog could change RSSI and SNR levels even if it preserves the ranking; this paper only sampled one weather window.","The gap between measured path loss and free-space loss could be converted into a local terrain correction factor for hilly Nepal by repeating the same sweep on several hilltop links with different lengths and orientations.","The demonstration used two ultrasonic sensor nodes, so the TDMA and cloud results should be read as a feasibility check; scaling to many nodes would stress synchronization and time-slot allocation beyond what was tested."],"forward_implications":["Rural IoT projects in similar hilly terrain can start from 62.5 kHz bandwidth, spreading factor 8, and coding rate 4/8 instead of running a full parameter sweep.","Because every spreading factor from 8 upward produced zero packet loss on the 5 km link, reliability does not force the maximum spreading factor; data-rate and power budgets can decide.","A coding rate of 4/8 gives measurably better SNR than the higher rates, so the extra redundancy is worth its throughput cost on noisy rural links.","A TDMA schedule with per-node sync words lets several low-cost LoRa nodes share one channel, and the gateway can push parsed data to a cloud platform for remote monitoring.","The measured path-loss gap grows with spreading factor, so raising the spreading factor to extend range can reduce signal margin on links that already reach the destination."],"supporting_citations":[{"why":"Supplies the Friis free-space-loss equation and the LoRa parameter trade-off analysis the experiment builds on.","marker":"[7]"},{"why":"Establishes that tuning physical-layer parameters is central to LoRa reliability and energy efficiency.","marker":"[8, 9]"},{"why":"Provides the earlier indoor LPWAN RSSI, SNR, and packet-loss analysis this study extends to rural terrain.","marker":"[3]"},{"why":"Offers the field-evaluation framework of RSSI, SNR, packet ratio, and latency that the Nepal test adapts.","marker":"[12]"},{"why":"Guides the choice of the SX1278 module by comparing low-cost LoRa modules for IoT use.","marker":"[13]"},{"why":"Supports the claim that lower spreading factors and narrower bandwidths increase packet loss through collisions.","marker":"[22]"},{"why":"The chip datasheet defining the RSSI offset and SNR register readings used for the metrics.","marker":"[28]"},{"why":"Antenna-theory reference for the monopole gain and ground-plane behavior assumed in the link budget.","marker":"[30]"},{"why":"Reference for the theoretical quarter-wave monopole gain used in the path-loss calculation.","marker":"[31]"},{"why":"Documents temperature and humidity effects on radio signal strength, used to frame the single-weather test.","marker":"[35]"}],"fun_headline_variants":["LoRa's best for Nepal hills: 62.5 kHz, SF 8","Field test finds LoRa sweet spot in Nepal's hills","LoRa tuning for rural Nepal: zero packet loss at SF 8","Nepal LoRa test: low bandwidth, SF 8 wins"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a single clear line-of-sight link between two fixed hilltops, tested in one moderate-weather window, represents hilly and mountainous rural Nepal generally.","fun_headline_variants_meta":{"raw":{"variants":["LoRa's best for Nepal hills: 62.5 kHz, SF 8","Field test finds LoRa sweet spot in Nepal's hills","LoRa tuning for rural Nepal: zero packet loss at SF 8","Nepal LoRa test: low bandwidth, SF 8 wins"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000532,"raw_usage":{"total_tokens":2602,"prompt_tokens":1025,"completion_tokens":1577,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":641,"completion_tokens_details":{"reasoning_tokens":1498}},"tokens_in":641,"tokens_out":1577,"duration_ms":10850,"temperature":1.0,"reasoning_tokens":1498,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T21:23:05.287105+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the parameter sweep on a non-line-of-sight link between two slopes in the same district during monsoon fog, and compare spreading factor 10 or 12 at 125 kHz with spreading factor 8 at 62.5 kHz on packet loss and SNR; if the higher spreading factors win, the claimed optimum is a property of this one clear link rather than of hilly rural Nepal.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Friis free-space-loss equation and the LoRa parameter trade-off analysis the experiment builds on."},{"cited_title":"RegPktSnrValue","cited_arxiv_id":null,"evidence_quote":"Provides the earlier indoor LPWAN RSSI, SNR, and packet-loss analysis this study extends to rural terrain."},{"cited_title":"IoT Real -Time Soil Monitoring Based on LoRa for Palm Oil Plantation,","cited_arxiv_id":null,"evidence_quote":"Guides the choice of the SX1278 module by comparing low-cost LoRa modules for IoT use."},{"cited_title":"Beginners Approach to the Open -Source Programming Case Study Arduino with ESP32,","cited_arxiv_id":null,"evidence_quote":"Supports the claim that lower spreading factors and narrower bandwidths increase packet loss through collisions."},{"cited_title":"Long -range communications in unlicensed bands: The rising stars in the IoT and smart city scenarios,","cited_arxiv_id":null,"evidence_quote":"The chip datasheet defining the RSSI offset and SNR register readings used for the metrics."},{"cited_title":"Collision and packet loss analysis in a LoRaWAN network,","cited_arxiv_id":null,"evidence_quote":"Antenna-theory reference for the monopole gain and ground-plane behavior assumed in the link budget."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reference for the theoretical quarter-wave monopole gain used in the path-loss calculation."},{"cited_title":"arduino -LoRa: An Arduino library for sending and receiving data using LoRa radios,","cited_arxiv_id":null,"evidence_quote":"Documents temperature and humidity effects on radio signal strength, used to frame the single-weather test."}],"review_version":1}