Added Silent Disco HackRF Tool using Liquid DSP

[daltoncox]

Silent Disco multi-channel receiver (HackRF + liquid-dsp) — simultaneous multi-channel monitoring & live audio

Challenges addressed:

1. Monitors the DEF CON Silent Disco system and plays the audio, with a channel selection feature supporting more than two Silent Disco channels.
2. Monitors multiple channels at the same time — 5+ channels played/recorded simultaneously.

What this adds

A standalone C application (build/sdisco) that captures the entire 908–927.5 MHz Silent Disco band with a single HackRF One and FM-demodulates up to 10 channels simultaneously in software on one stream, mixing them into a single live stereo audio output. No GNU Radio / flowgraph required — pure C, built with CMake.

• Full 45-channel map (5 transmitters A–E × 9 slots) from the TX1345 datasheet, plus the Ultra-900 (-u) map (CH1–CH10, A1–A3)
• Select any subset at launch: -c A1,B3,E5 (comma/space separated, max 10)
• Runtime channel switching: 1-9 mutes/unmutes each channel in selection order, 0 toggles all, +/- master volume
• Auto LO selection nudges the mix off the HackRF DC spur; capture rate auto-picks the smallest 8–20 Msps that covers the selection
• All selected channels are demodulated in parallel, at the same time, from one 20 Msps capture — the band spans 19.4 MHz, so the full 45-channel set fits in a single stream
• Up to 10 channels played at once (MAX_DEMOD), each with its own DSP chain (NCO mix → Kaiser FIR channel filter → FM demod → de-emphasis → audio LPF → AGC)
• Verified 5 simultaneous chains decode distinct audio cleanly: tests/sim_test.c reports RESULT: PASS for 1/2/3/5 channels carrying independent tones
• One DSP worker thread per channel keeps 10 chains at 20 Msps in real time with audio production held at 48 kHz

Demo-able without hardware

./build/sdisco -s -c A1,B3,E5 synthesizes FM transmitters at true RF frequencies carrying unique 400 Hz-spaced tones — verifies per-channel mixing, routing, and simultaneous multi-channel decode by ear or with tests/sim_test.c.

Signal chain (per channel)

NCO mix-down → Kaiser-windowed-sinc FIR channel-select/decimator (60 dB stopband, ±110 kHz) → FM discriminator (kf from --dev) → de-emphasis (0/50/75 µs) → resample 500 kHz → 48 kHz → 15 kHz audio LPF → peak-envelope AGC → per-chain gain → summed stereo output via PortAudio.

Files

CMakeLists.txt build setup (sdisco + sim_test)
src/main.c CLI, sim source, runtime keys, diagnostics
src/channel.c/h TX1345 A-E + Ultra-900 channel maps
src/hackrx.c/h HackRF int8 I/Q capture ring
src/dsp.c/h multi-channel FM demod DSP engine (parallel chains)
src/audio.c/h PortAudio playback (priming, underrun stats)
src/ring.c/h SPSC rings
tests/sim_test.c offline DSP validation (1/2/3/5+ chains)
tools/ dsp_from_file.c, iqdump_analyze.py (debug aids)
README.md usage, channel maps, architecture, bug log

Tested

• sim_test 5 → PASS: 1/2/3/5 simultaneous chains each decode their distinct tone with the correct slot routing
• Verified against live DEF CON RF captures — each selected channel's message is audible and routes to the correct slot; residual in-band static traced to 39+ co-active carriers (RF noise floor), not the DSP

[ZeroChaos-]

Come up to the CTF table and show us it works.