This engineering assessment describes current capabilities and the evidence still needed for regional RF surveys. It is not an instrument certification or a promise that every survey requirement is satisfied.
The application is a useful experimental survey and analysis tool. Its strongest foundation is retained, frequency-resolved RF measurement with explicit gaps, original receiver-position association, reproducible filters and manageable reports. These are more useful for surveying than packet counts alone. Existing test coverage and known limitations are summarized in validation scope. Field evidence remains limited. The application is not a calibrated instrument, exhaustive mesh inventory or field-qualified regional channel-selection system.
The narrative report makes existing measurements easier to interpret; it does not remove their measurement or validation limits. Implementation presence, passing synthetic checks and field acceptance are separate statuses.
| Area | Available now | Gap and practical consequence | Priority / acceptance evidence |
|---|---|---|---|
| Frequency occupancy | SDR fixed-threshold FFT activity per bin, exact selected-bin busy-time union, exposure and known gaps | Threshold is uncalibrated dBFS. Receiver/antenna/gain differences can change what counts as activity | Before regional comparisons: calibrated signal-level sweeps across frequency and gain, documented detection probability/false-alarm behavior and reference setup |
| Concentrator sampling | One/two RAK5146 boards collect complete RSSI histograms at assigned scan centers and receive configured LoRa profiles | Revisit gaps, nominal 234.3 kHz filter, uncalibrated RSSI; no continuous occupancy, blind preset search or measured packet width | Before regional comparisons: characterize scan timing/sampling bias, receiver offsets, sensitivity, motion and endurance; keep sampled statistics separate from SDR occupancy |
| Channel widths | Arbitrary interval selection, retained fine bins, experimental inferred LoRa BW/SF | Energy span is not configured modem BW or a measured 99% occupied/emission bandwidth. No field-validated census of all channel widths | Before width/identity claims: independent known-signal tests at different strengths, offsets, widths and overlapping signals; report edge/resolution uncertainty |
| Time and representativeness | Stored UTC anchor, elapsed time, exposure, known gaps and selectable time groups | No rolling busy-hour analysis, occupancy confidence interval, minimum representative observation plan or reliable packet access-delay estimate | Before planning claims: repeat sites over relevant times/days; choose an estimator that accounts for correlated samples and nonstationary traffic |
| Receiver metrology | Gains, signed Offset, sample rate, FFT/Hann ENBW, threshold, descriptions, clipping and center diagnostic | No traceable power/frequency calibration, antenna/cable response correction, validated passband, dynamic-range or uncertainty budget | Before quantitative power claims: compare to a calibrated source/instrument and measure linearity, adjacent-channel rejection and analog overload |
| Transmitter power/model | SDR dBFS, RAK nominal RSSI in dBm, and relative comparisons with stable setup | Cannot infer transmitter watts, RAK model, EIRP or excessive power from received strength. Near weak transmitters can look stronger than distant powerful ones | Outside current survey inference: controlled calibrated conducted/radiated test with known path/setup; assess authorization separately |
| GPS and maps | Original fixes, fixed/mobile display, missing-position exposure, spatial filters and approximate geographic CSV cells | Indoor drift, sparse visits, tile-end attribution and time/location confounding; no offline basemap, surveyed accuracy or multi-route fusion | Before regional coverage claims: repeated routes, stationary checks/reference positions, exposure-by-area review and multi-session comparison |
| Protocol attribution | Generic energy, experimental LoRa waveform observations, automatic/optional-profile Meshtastic key-scoped classifications | Automatic dispatch is implemented but bounded and not field-qualified; incomplete PHY success; MeshCore, recipient PKI and reliable LoRaWAN/other-device attribution are missing | Before mesh/non-mesh inventory claims: labeled independent captures across supported parameter ranges; measure precision/recall and error stages |
| Collision/interference analysis | Overlapping energy and CRC failures may be observed | Neither proves a packet collision or harmful interference; no validated per-protocol airtime subtraction | Later specialized work: synchronized independent receivers/known transmissions and controlled interference tests |
| Regional comparison | One source recording with frequency/time/area filters | No multi-session merge/comparison, receiver-calibration normalization or repeat-visit reconciliation | High next product priority: compare preserved source sessions with explicit setup and coverage compatibility |
| Reporting | Narrative HTML, frequency/time/geographic CSV, separate waveform/GPS data, source ID/schema/settings and privacy choices | No signed provenance chain/source digest in the report, interactive frequency-by-time heatmap, formal uncertainty or automated recommendation model | Before external decision packages: review sample reports with regional operators; define versioned shared methodology and evidence requirements |
| Reliability/platform support | HackRF/RTL-SDR and experimental one/two-board RAK adapters, native regression suite, private/exclusive files, bounded workers and bounded receiver evidence | Receiver interruptions and discovery queue loss have occurred; no qualified long mobile/overnight runs, Windows/Linux acceptance or broad RTL tuner-model qualification | Before broad rollout: sustained and interruption/recovery tests, disk/resource limits, separate receiver-backend qualification and native platform matrix |
These priorities deliberately keep recording/reporting ahead of automatic channel recommendations. A frequency that looked quiet on one receiver at one time is a candidate for further study, not a proven safe assignment.
ITU-R SM.2256-2 (June 2026), especially sections 3, 6, 7, 9 and 10, discusses detector/selectivity, threshold, timing, presentation, uncertainty and interpretation. The report distinguishes channel occupancy from frequency-band/resource occupancy and discusses geographic averaging. This reference informs the review categories; the specific gaps above describe application behavior and validation scope. No ITU method conformance is claimed.
HackRF's RX gain documentation describes multiple adjustable analog receive stages. That supports recording setup with every comparison; it does not supply a calibrated dBFS-to-dBm conversion for this application. Received power cannot by itself establish unknown transmitter power or identity.
The following occupancy/report details describe SDR recordings. RAK schema-7 reports instead summarize sampled RSSI histograms, per-board configured reception and health, with optional receiver positions. They do not manufacture busy-time or FFT coverage. See concentrator reporting semantics.
Generate analysis report... in Analyze captures the latest completed selection, then allows full-session reset and local destination/privacy review. After reception stops, an asynchronous worker writes a new private .html file using a consistent read-only snapshot. It shares the existing CSV aggregation code and introduces no dependency, network request, script or message-content export. The HTML is bounded to 16 MiB; aggregation retains existing limits and fails explicitly rather than silently discarding later source data.
The report includes complete selected frequency-bin rows, peak-preserving frequency visualization, selected-scope metrics, ten highest-occupancy time groups, optional twenty receiver cells ordered by their busiest bin, grouped waveform evidence, selected reception/classification counts and whole-session health counters clearly labeled as such. Limited lists disclose their scope and point to full CSV categories. Message contents are excluded unconditionally. Coordinates and free-form titles/notes/descriptions are excluded by default. Opt-in text is HTML-escaped; a restrictive Content Security Policy disallows active content and external resources. Printing/PDF uses the user's browser.
Report fixtures check alternating versus simultaneous bin activity (100% versus 50% any-bin occupancy, both 50% mean frequency-time occupancy), source immutability, selection/no-exposure behavior, location opt-in, HTML escaping, missing-content defaults, overwrite refusal and removal of incomplete output. Previously recorded targeted test and synthetic browser-render results support those bounded cases; they are not a new test run of this source snapshot.
The report does not supply calibrated RF accuracy, exhaustive protocol attribution, a complete geographic survey, or a reliable estimate of unknown transmitter power. Its summaries preserve measurement limits rather than resolving them. See validation scope, measurement definitions and survey methodology.