BufferedReader.read(n) on non-interactive streams (Python 3.14) blocks
until the full n bytes accumulate, starving the decoder of real-time
PCM data. Use os.read() on the raw file descriptor instead, which
returns as soon as any data is available. Falls back to .read() for
file-like objects without fileno() (e.g. BytesIO in tests).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace select.select()+os.read() with a blocking reader thread feeding
a queue, matching pager's working pattern. The select() approach fails
to detect available data on Python 3.14's BufferedReader-wrapped pipes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
rtl_fm prints device info, tuning, and errors to stderr but the morse
route only logged these server-side. Now stderr lines are forwarded to
the morse queue as info events, displayed in a compact diagnostic log
below the scope canvas. After 10s with no audio data, the scope text
escalates to prompt the user to check the SDR log.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace blocking rtl_stdout.read() with select()+os.read() so the
decoder thread emits diagnostic heartbeat scope events when rtl_fm
produces no PCM data (common in direct sampling mode). Add waiting-state
rendering in the scope canvas and hide the generic placeholder/status
bar for morse mode.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add automatic gain control (AGC) before Goertzel processing to normalize
quiet audio from direct sampling mode where the -g gain flag has no effect.
Fix broken adaptive threshold bootstrap by adding a 50-block warm-up phase
that collects magnitude statistics before seeding noise floor and signal peak.
Lower threshold ratio from 50% to 30% for better weak-CW sensitivity.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The -D flag is only available in newer rtl_fm builds. Docker and distro
packages use the older -E direct / -E direct2 flags instead, which are
universally supported.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Enable direct sampling (-D 2) for RTL-SDR at HF frequencies below 24 MHz
so rtl_fm can actually receive CW signals. Add startup health check to
detect immediate rtl_fm failures. Push stopped status event from decoder
thread on EOF so the frontend auto-resets. Add frequency placeholder and
help text. Fix stop button silently swallowing errors.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Correct modulation parameters (1200 bps, 2100/1300 Hz tones), replace
invented format codes with the six ITU-defined specifiers {102, 112,
114, 116, 120, 123}, accept all valid EOS symbols (117, 122, 127),
add parser validation (format, MMSI, raw field, telecommand range),
and fix truthiness bugs that dropped zero-valued fields.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add synchronous startup verification after Popen() — sleep 0.5s and poll
the process before returning to the caller. If SatDump exits immediately
(missing device, bad args), raise RuntimeError with the actual error
message instead of returning status: 'started'. Keep a shorter (2s) async
backup check for slower failures.
Also fix --source_id handling: omit the flag entirely when no serial number
is found instead of passing "0" which SatDump may reject. Change start()
and start_from_file() to return (bool, str|None) tuples so error messages
propagate through to the HTTP response.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
TSCM RF scan now auto-detects HackRF via SDRFactory and uses
hackrf_sweep as an alternative to rtl_power. Also includes
improvements to listening post, rtlamr, weather satellite,
SubGHz, Meshtastic, SSTV, WeFax, and process monitor modules.
Fixes#154
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New signal mode for decoding Morse code (CW) transmissions via SDR.
Includes route blueprint, utility decoder, frontend UI, and tests.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace hardcoded "rtl_fm" references in wefax.py with the actual SDR
tool name so error messages correctly show "rx_fm" for non-RTL devices.
Use get_tool_path('rx_fm') in all SoapySDR command builders to match
the pattern already used for rx_sdr.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add pre-flight check for I/Q capture binary before spawning process
- Capture stderr from I/Q process for better error diagnostics
- Sync effective span value back to UI when backend adjusts it
- Use get_tool_path('rx_sdr') in Airspy, HackRF, LimeSDR, and SDRPlay
command builders to support custom install locations
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace hardcoded rtl_fm with SDRFactory abstraction layer so WeFax works
with any supported SDR hardware, matching the pattern used by APRS and
other modes. RTL-SDR direct sampling flag preserved for HF reception.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The default 1 MHz sample rate was too low for SatDump's meteor_m2-x_lrpt
pipeline, causing NOSYNC and 0.000dB SNR. Bumped to 2.4 MHz (SatDump
recommendation) and wired up the WEATHER_SAT_SAMPLE_RATE config value
so it actually gets passed to decoder.start() from both the auto-scheduler
and manual start route.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Timer threads now log on fire and catch all exceptions so scheduled
captures never die silently. Frontend connects SSE when the scheduler
is enabled (not only on manual Start) and polls /wefax/status every 10s
as a fallback so the UI stays in sync with auto-fired captures.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
stop() was returning before the decode thread could save any partial
image to disk, so the frontend loadImages() call found nothing new.
Join the decode thread (2s timeout) before returning — with select()-
based reads the thread exits within ~0.5s so this stays responsive.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replace blocking stdout.read() with select()-based non-blocking reads
so the decode thread responds to stop within 0.5s
- Make stop() non-blocking by releasing the lock before terminating the
process and removing the redundant wait()
- Move initial scanning SSE event from start() into the decode thread so
it fires after the frontend EventSource connects
- Update frontend stop() to give immediate UI feedback before the fetch
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
rtl_fm subprocess failures (missing tool, no SDR hardware) were silent —
add tool-path check and post-spawn health check in _start_pipeline(),
show errors prominently in the strip status bar (red text + red dot),
and include error detail in scheduler skip events.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fix silent failure when starting without station/frequency selected by
flashing amber on status text and dropdowns. Add auto-capture scheduler
that uses fixed UTC broadcast schedules from station data to
automatically start/stop WeFax decoding at broadcast times.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implement HF radiofax decoding with custom Python DSP pipeline
(rtl_fm USB → Goertzel/Hilbert demodulation), 33-station database
with broadcast schedules, audio waveform scope, live image preview,
and decoded image gallery. Amber/gold UI theme for HF distinction.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Two root causes for the waterfall/monitor lockup when scrolling past the
2.4 MHz RTL-SDR span:
1. safe_terminate() sent SIGKILL but never called wait(), leaving a
zombie process that kept the USB device handle open. The subsequent
capture restart failed the USB probe and the monitor could not use
the shared IQ path, falling back to a process-based monitor that
stole the SDR from the waterfall.
2. When the frontend created a new WebSocket after a failure, the old
handler's finally block called _set_shared_capture_state(running=False)
which could race with the new handler's running=True, making the
shared monitor path unavailable. Added a generation counter so only
the owning handler can clear the shared state.
The slant correction was severely under-correcting because bwd=50 caused
the sync deviation measurements to saturate after only ~25 lines (for a
2-sample/line SDR clock drift). Lines 25-256 all reported deviation=-50,
pulling the linear regression slope toward zero.
Increase bwd and fwd to 800 samples each — sufficient to track cumulative
drift from up to ~±200 ppm SDR clock offset across the full 256-line image.
Also use a full-sync-length (432-sample) Goertzel window instead of 1/3
length, giving ~111 Hz frequency resolution to cleanly separate the 1200 Hz
sync tone from 1500 Hz pixel data. Search is stepped at 5 samples (~0.1 ms)
for efficiency, keeping the goertzel_batch batch size at ~320 windows/line.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Previous attempts to correct slant by altering R-channel placement and
buffer consumption caused cascading failures: a false positive in B pixel
data would misplace R, then the wrong consumed value misaligned the next
line's G, and the error compounded across all 256 lines.
New approach (safe by design):
- Sync search is measurement-only: never touches pos or consumed, so
a noisy or wrong measurement cannot corrupt the current or future lines.
- Per-line deviation (measured sync position minus expected) is recorded
in self._sync_deviations throughout the decode.
- get_image() fits a line through the deviations (linear regression) to
estimate the per-line SDR clock drift rate, then applies a horizontal
shear to the assembled PIL image: each row is shifted by
-round(row × drift_rate × width / channel_samples) pixels.
- Worst case (all measurements fail): no correction applied, image
quality identical to the pre-change baseline.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The step-49 coarse scan introduced up to ±24 sample uncertainty in R
channel placement. When accumulated SDR clock drift pushed the actual
sync 35+ samples early in the search region, the step-49 windows could
land on the B-pixel tail and return position 0, misplacing R by ~50
samples (~16 pixel colour shift) — worse than no correction at all.
Replace with a vectorised goertzel_batch sliding-window scan at step=1
over a short window (sync_duration / 3 ≈ 3 ms), giving single-sample
accuracy. Use consumed=pos (instead of max(pos,line_samples)) when the
sync is found, so the next line starts at its correct separator and
per-line timing errors stop accumulating entirely.
Falls back to the fixed-offset path whenever the sync is not found
(e.g. noisy signal), preserving the pre-change baseline quality.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
