HK.SDR turns a cheap RTL-SDR 820T2 dongle into a continuously-scanning, multi-mode signal-analysis lab you drive entirely from your browser. It was built as an educational, receive-only project to find, log, classify, decode and transcribe the radio signals around you — across aviation, maritime, paging, sensors, broadcast and digital bands — without any cloud service. Everything runs locally on your machine.
The whole web backend is pure Python standard library (http.server + sqlite3, no pip
dependencies); only optional features (speech-to-text, modulation classification) use extra tools.
- Continuous categorized scanning — sweeps band plans with
rtl_power, logs every signal (frequency, power, bandwidth, category, band) to SQLite with a live, newest-on-top history. - Live audio — click any audio-capable channel to stream live WB-FM / AM / NFM, with an optional WebAudio noise filter; every listen is auto-archived as a WAV recording.
- Decoders (one shared dongle, mutually exclusive modes):
✈️ ADS-B aircraft tracking on a live map (dump1090)- 🚢 AIS ship tracking (rtl_ais)
- 📡 APRS stations (multimon-ng)
- 🛩️ ACARS & 🛰️ VDL2 aircraft data-link messages (acarsdec / dumpvdl2)
- 📟 Pager POCSAG/FLEX — content masked for privacy (type + capcode + length only)
- 📻 RDS FM station name / radiotext (redsea)
- 🌡️ ISM 433 MHz sensors — grouped live device table (rtl_433): TPMS, thermometers, remotes…
- Waterfall / spectrum — live canvas heatmap; click to listen, Shift-click to classify, scroll to zoom.
- Speech-to-text (STT) — offline, on-device transcription with whisper.cpp (multilingual,
auto TR/EN detection); optional Turkish-tuned
distil-whisper-large-v3-tr. Activity-gated; shows one sentence per audio frequency in the scan table and a live flowing transcript while listening. - Modulation classification — feature-based (bandwidth / envelope CV / spectral flatness / SNR) rule classifier (FM-wide, constant-envelope digital, AM/NFM, OOK, noise) + likely-service lookup.
- Analytics & search — most-active frequencies, category & hourly distribution, full-text search, CSV/JSON export.
- Frequency grouping — collapse repeated detections of the same frequency into one row with a count and last-seen.
- Scan profiles & scheduling, gain / PPM calibration, CSV bookmark import, alarms (browser notification + beep + toast on keyword/category/frequency match).
- PWA — installable on a phone; bilingual UI (English / Türkçe) with a one-click flag toggle; optional Basic-auth and LAN binding.
| Desktop — scan view | Mobile (PWA) |
|---|---|
 |
 |
| Item | Requirement |
|---|---|
| SDR | RTL-SDR dongle (RTL2832U + R820T2 / "820T2", USB id 0bda:2838). RTL-SDR Blog V4 recommended for stability (TCXO, better filtering). |
| Antenna | Band-appropriate antenna. ADS-B (1090 MHz) works with a short whip; VHF/UHF (airband, AIS, APRS, ACARS/VDL2, ISM 433) needs a proper ~½λ dipole or ~46 cm ¼λ. Reception — not code — is the main limitation on weak bands. |
| Power/USB | A powered USB hub is strongly recommended — the 820T2 can drop off the bus under sustained load (heat / marginal power). |
| Host | Any Linux PC / SBC (Raspberry Pi class and up). |
📷 Tip: add your own photo of your setup at
docs/images/hardware.jpgand reference it here.
| Component | Needed for |
|---|---|
| Linux (Debian 12 tested) + Python 3.11+ | core (web backend is stdlib-only) |
rtl-sdr tools (rtl_power, rtl_fm, rtl_sdr, rtl_test) |
scanning & live audio (required) |
numpy |
modulation classification |
rtl_433 |
ISM 433 MHz sensors |
multimon-ng |
Pager (POCSAG/FLEX) & APRS |
dump1090 |
ADS-B |
rtl_ais |
AIS |
acarsdec / dumpvdl2 |
ACARS / VDL2 |
redsea |
RDS |
whisper.cpp (+ model) · optional faster-whisper |
speech-to-text |
git clone https://github.com/haliskilic/hksdr.git
cd hksdr
# 1) Core: RTL-SDR drivers + common decoder tools (apt; needs sudo)
sudo bash setup/setup.sh
# 2) Optional decoders (build from source into ~/.local)
bash setup/build_dump1090.sh # ADS-B
bash setup/build_rtl_ais.sh # AIS
bash setup/build_acarsdec.sh # ACARS
bash setup/build_dumpvdl2.sh # VDL2
bash setup/build_redsea.sh # RDS
# 3) Optional speech-to-text (whisper.cpp small, multilingual)
bash setup/setup_whisper.sh
# Optional Turkish-tuned model (heavier, opt-in via USE_DISTIL_TR=1):
bash setup/setup_distil_tr.shThe web app itself has no Python dependencies — only the optional STT/classify features use
numpy / faster-whisper.
scripts/run_webapp.sh # → http://localhost:8000
# options:
PORT=9000 GAIN=45 scripts/run_webapp.sh
HOST=0.0.0.0 scripts/run_webapp.sh # serve on the LAN (use AUTH for security)
AUTH=user:pass HOST=0.0.0.0 scripts/run_webapp.sh
USE_DISTIL_TR=1 scripts/run_webapp.sh # enable the Turkish distil-whisper engineOpen the page and scanning starts automatically. Use the left menu (☰) to open a decoder, the map, the waterfall, recordings, analytics or settings. Click a frequency's ▶ Listen to hear it. Switch language any time with the 🇬🇧/🇹🇷 button (top-right).
A single Device manager serializes the one RTL-SDR across mutually-exclusive modes
(scan / listen / adsb / ais / aprs / ism / pager / acars / rds / vdl2 / wf). The scanner runs
in a background thread; decoders are subprocess wrappers that parse tool output into SQLite /
in-memory state. The frontend is a dependency-light vanilla-JS SPA that polls JSON endpoints.
See docs/ROADMAP.md for the full design, status and roadmap.
HK.SDR is receive-only and intended for education and lawful monitoring of open signals. You are responsible for complying with the radio/privacy laws of your country.
- Pager content is masked (only type/capcode/length stored) — privacy by design.
- Decoding encrypted / public-safety voice (e.g. TETRA/P25/DMR) is out of scope.
- STT is meant for legal, open analog and unencrypted-digital signals only.
HK.SDR is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0) — see
LICENSE. In short: you may use, study, modify and self-host it freely, but if you
run a modified version as a network service you must make your complete source available to its
users under the same license.
For use that is incompatible with the AGPL-3.0 (e.g. closed-source / commercial products),
a separate commercial license is available — see COMMERCIAL_LICENSE.md.
Copyright © 2026 Halis KILIÇ. All rights reserved.