The SDR audio bridge was forwarding PCM to server-side decoders but
never encoding it to Opus for rx_audio_tx, so TCP audio clients
(browser RX button) received nothing. Add Opus encoder initialised
from the rig's audio config and encode each PCM frame alongside the
pcm_tx broadcast.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Replace the BW slider + FIR taps filter panel with a visual bandwidth
bookmark drawn directly on the spectrum canvas:
- Semi-transparent amber gradient strip spanning dialFreq ± BW/2
- Rounded-top bookmark tab at the top of the strip showing the current BW
- Draggable left/right edge handles (cursor: ew-resize) that adjust bandwidth
live and send set_bandwidth on mouse-up; range clamped per-mode defaults
- Y-axis now labeled with dB values (floor to ceiling) drawn on canvas
- Configurable floor level via number input below spectrum (default -100 dB)
- Auto button fits floor/range to current noise floor and peak level
- Remove FIR taps selector (internal DSP implementation detail)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Add MODE_BW_DEFAULTS table mapping each mode to [default, min, max, step]
in Hz. When mode changes (via picker or SSE), applyBwDefaultForMode
updates the filter bandwidth slider range and sends set_bandwidth to the
server, so the DSP filter is rebuilt automatically for the new mode.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
- Compute hardware center as dial - center_offset_hz (was ignoring offset)
- Pass configured bandwidth_hz to device instead of hardcoded 1.5 MHz
- Add retune_cmd channel so set_freq repoints SDR hardware in real time
- Auto-add default channel with mode-appropriate bandwidth when [[sdr.channels]]
is empty, preventing silent audio with minimal config
- Add ChannelDsp::set_filter to rebuild FIR LPFs at runtime; wire
set_bandwidth and set_fir_taps to call it so UI filter changes take effect
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Spectrum panel is now placed above the freq/mode/jog row, spanning the
full card width. Key improvements:
- Scroll wheel zooms in/out at the cursor position (up to 64x); double-
click resets to full bandwidth view.
- Mouse drag pans the visible window; click-to-tune is suppressed when a
drag has occurred.
- Touch pinch-to-zoom and single-finger drag-to-pan supported.
- Hover tooltip shows the frequency under the cursor, formatted to the
currently selected unit (MHz/kHz/Hz, matching the jog-step selection).
- Frequency axis labels update to reflect the zoomed visible range.
- Canvas height increased to 160 px; axis bar styled with card bg.
- A small hint line below the panel explains the controls.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Poll GetSpectrum every 200 ms in remote_client via a dedicated timer that
bypasses the main state-watch channel (no SSE noise). The resulting
SpectrumData is stored in FrontendRuntimeContext::spectrum and served by
a new GET /spectrum endpoint (JSON or 204 when unavailable).
HTTP frontend shows a spectrum panel (canvas + frequency axis) only when
the rig reports filter_controls=true (i.e. SoapySDR). The canvas renders:
- dark background with dBFS grid lines
- green FFT spectrum line with semi-transparent fill
- red dashed vertical marker at the currently tuned frequency
- frequency axis labels (MHz/kHz) below the canvas
Clicking the canvas tunes the rig to the clicked frequency.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Add SpectrumData struct (bins, center_hz, sample_rate) to RigState and
RigSnapshot. Add GetSpectrum RigCommand and ClientCommand plumbed through
the protocol layer. SoapySDR DSP pipeline now computes a 1024-bin FFT
(Hann window, FFT-shifted, dBFS) every 4 IQ blocks (~10 Hz update rate)
and exposes it via RigCat::get_spectrum(). The rig_task handles
GetSpectrum without persisting spectrum data in ongoing state.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Three root causes prevented APRS decoding at 144.800 MHz with PKT/FM mode:
1. `RealIqSource::read_into` returned zeros — the SoapySDR streaming API
was never wired up. `RxStream<Complex<f32>>` is `Send` and
`StreamSample` is implemented for `num_complex::Complex<f32>` in the
soapysdr 0.3 crate, so the stream can read directly into the IQ buffer.
Now creates and activates an `RxStream` in `new()` and calls
`stream.read` in `read_into`.
2. PKT mode used `Passthrough` (take `.re`) demodulation. VHF/UHF packet
radio (APRS, AX.25) is FM-encoded AFSK — it must be FM-demodulated
before the APRS decoder sees the audio tones. Changed PKT to `Fm`.
3. `iq_read_loop` always slept `block_duration_ms` after each read. Real
hardware already blocks inside `read_into`; the extra sleep doubled
latency. Added `IqSource::is_blocking()` (default `false`; `true` for
`RealIqSource`) and skip the throttle sleep for blocking sources.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Pass `watch::Receiver<bool>` into `run_capture` and `run_playback` so
both threads check for shutdown at the top of their outer recovery loop
and inner monitoring loop. Without this, restarting the process (e.g.
via systemd after an ALSA error) left the old threads stalled forever
while new ones were created alongside them.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Replace the per-sample ring-buffer FIR convolution with block-level
overlap-save convolution using rustfft. For a block of M samples and
N taps the old approach costs O(N·M); the new one costs O(M log M),
with rustfft using SIMD (AVX2/SSE4) internally.
Key changes:
- Add rustfft = "6" dependency
- Add BlockFirFilter: overlap-save filter with pre-computed H(f) and
a single forward+inverse FFT pair per block (no per-sample multiply)
- ChannelDsp.process_block() now:
1. Batch-mixes entire block to baseband in one vectorisable loop
2. Applies BlockFirFilter to I and Q (one FFT pair each)
3. Decimates and demodulates as before
- Keep the old FirFilter for unit tests (sample-by-sample interface)
- Add BlockFirFilter unit tests (DC passthrough, length preservation)
- IQ_BLOCK_SIZE promoted to pub const for use in filter sizing
For the default config (4096-sample blocks, 64 taps, decim=40):
Old: ~262144 multiply-adds per FIR × 2 components = ~524k per block
New: ~2 × (3 × 8192 × log2(8192)) ops, all SIMD-vectorised by rustfft
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Fix compiler warning about discarded const qualifier in strchr() call.
The result of strchr(const char*, ...) should be assigned to const char*,
not char*.
This resolves:
warning: initialization discards 'const' qualifier from pointer target type
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
sjg
