Add set_named_gain to the IqSource trait and implement it in
RealIqSource via soapysdr Device::set_gain_element. Wire a
lna_gain_cmd channel through SdrPipeline so the IQ read loop applies
LNA gain changes on the next iteration. Add set_sdr_lna_gain to
SoapySdrRig and expose the current value via filter_state.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Map ClientCommand::SetSdrLnaGain { gain_db } to and from
RigCommand::SetSdrLnaGain in both directions.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Add RigCommand::SetSdrLnaGain(f64), the corresponding default RigCat
trait method set_sdr_lna_gain, and an sdr_lna_gain_db: Option<f64>
field to RigFilterState for backends that expose a named LNA gain
element.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Replace vol-label with wfm-control on the Hardware AGC checkbox so its
label typography and alignment match the RF Gain control. Change the
outer container to align-items: flex-end so controls bottom-align.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
The bidirectional check required both A→B and B→A directed messages to
draw a contact line, which was too strict — the receiver may not hear
both sides of a QSO. Now a decode path is drawn whenever a directed
message is decoded and the target's locator is known from any message
in the 24 h history window.
Also rename "Longest QSOs" → "Longest decode paths" and update
related UI labels to better reflect what is actually shown.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
A QSO requires both parties to hear each other. Previously the map drew
contact paths whenever A sent a directed message to B and B's locator
was known from any message. Now paths only appear when both A→B and B→A
directed messages are present in the decoded history.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Add hardware AGC on/off control for SoapySDR backend, wired through the
full stack from RigCommand to the web UI:
- RigCommand::SetSdrAgc(bool) + ClientCommand::SetSdrAgc in protocol
- set_sdr_agc() on RigCat trait (not-supported default)
- SoapySdrRig: agc_enabled field, set_sdr_agc() via pipeline agc_cmd,
sdr_agc_enabled in filter_state(); removes the "not yet implemented"
warning — gain_mode="auto" now properly enables hardware AGC via
SoapySDR set_gain_mode()
- IqSource::set_gain_mode() trait method; RealIqSource implements it
- SdrPipeline: agc_cmd channel, read loop applies it each iteration
- POST /set_sdr_agc endpoint in trx-frontend-http
- New "SDR settings" full-row in index.html with Hardware AGC checkbox
and RF Gain (moved out of WFM controls); row hidden when
show_sdr_gain_control is false
- app.js: AGC checkbox handler, disables RF gain input while AGC is on,
syncs checkbox state from filter.sdr_agc_enabled
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Three issues caused audible distortion on AM reception:
1. DC blocker shared r=0.9999 (τ≈1.25 s at 8 kHz) across all modes.
For AM the envelope detector outputs A_c+m(t) — always positive —
so the blocker needs to track the carrier bias quickly. AM now uses
r=0.999 (τ≈125 ms), 10× faster, while keeping the highpass cutoff
below 2 Hz so speech is unaffected.
2. Audio AGC time constants were inverted relative to good AM AGC design:
attack=200 ms (should be fast to prevent overload) and
release=3500 ms (unreasonably sluggish). Changed to attack=5 ms /
release=200 ms, target=0.5, max_gain=36 dB.
3. No IQ AGC before envelope detection meant carrier amplitude variation
went directly into the audio chain, forcing the slow audio AGC to
handle both RF level and audio level simultaneously. Added an AM IQ
AGC (attack=0.5 ms, release=50 ms, target=0.7, max=30 dB) that
normalizes carrier power before demod_am, so the DC blocker always
sees the same steady-state bias regardless of signal strength.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Apply #ft8-messages container style (border, rounded corners, monospace
font, max-height with scroll) to #ft4-messages, #ft2-messages, and
#wspr-messages which were missing it.
Add #ft4-decode-toggle-btn and #ft2-decode-toggle-btn to the narrow-
screen white-space:nowrap media query rule alongside FT8/WSPR.
Cap DOM rows rendered per history view to 200 (FT8_MAX_DOM_ROWS,
FT4_MAX_DOM_ROWS, FT2_MAX_DOM_ROWS). Full history is retained in
memory; only the DOM representation is bounded. This prevents tab
switching from becoming sluggish after a long decode session where
thousands of rows accumulate in the DOM.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Three bugs caused USB/LSB to sound like AM:
1. The IQ low-pass filter was symmetric (passband ±BW/2), so both
sidebands were passed equally — taking .re then produced DSB-SC
rather than SSB audio.
2. cutoff_hz was computed as bandwidth_hz/2, halving the usable audio
bandwidth (1500 Hz for a 3 kHz USB channel).
3. demod_lsb claimed spectrum inversion was "handled upstream by
negating channel_if_hz", but that negation was never applied; USB
and LSB were functionally identical.
Fix: add a shift_norm parameter to build_fir_kernel / BlockFirFilterPair
that complex-modulates the time-domain FIR coefficients by
e^{j·2π·shift_norm·n}, shifting the passband in the frequency domain.
A new ssb_shift_norm() helper returns +cutoff_norm for USB/CW/DIG
([0, BW] Hz passband) and -cutoff_norm for LSB/CWR ([-BW, 0] Hz
passband); all other modes get 0.0 (symmetric LPF unchanged).
After the one-sided filter, taking .re correctly reconstructs the
selected sideband. No IF negation is needed for LSB.
Also fix two unit tests missing the force_mono_pcm argument introduced
after they were last updated.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
The OSD-3 (triples) path over 5 LLR passes was doing ~11,600 CRC checks
per candidate. With a 14-bit CRC this gives ~0.7 expected false positives
per candidate — far too high.
Remove OSD-3 entirely. Cap max_candidates at 16 for OSD-1/OSD-2, giving
136 CRC checks per pass (680 total). Gate OSD-lite behind a check that
LDPC reached within 6 parity errors of converging, so it only fires when
the LLRs are already trustworthy. Combined false-positive rate drops to
~0.04 per near-miss candidate.
Also remove the now-unused ft2_osd_decode and ft2_codeword_distance
functions.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Increase BP/SP iteration count from 30 to 50 to match WSJT-X reference
and give belief propagation more opportunities to converge near-threshold
candidates.
Replace the parity-based OSD-1/OSD-2 fallback (which required LDPC to
have nearly converged) with ft2_osd_lite_decode applied to all five LLR
combination passes. The CRC-based decoder works directly from raw LLRs
without depending on LDPC convergence, searching the 24 least-reliable
systematic bits for up to three bit errors via OSD-3.
Also increase max_candidates in ft2_osd_lite_decode from 12 to 24 for
broader coverage of likely error positions.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Diagnostic logging showed the FT2 BP/SP decoders consistently reach
1-8 residual parity errors rather than zero — the LLRs are correct
in direction but LDPC belief propagation stalls just short of
convergence.
Add ft2_osd_decode() implementing Ordered Statistics Decoding orders
1 and 2: after the five-pass BP/SP loop fails, sort the 174 codeword
bits by |LLR| ascending and trial-flip single bits (OSD-1, always)
or all pairs of the 50 least-reliable bits (OSD-2, when the remaining
error count is <= 4). Each trial costs one O(83) parity check;
worst-case overhead is ~1300 checks per candidate, negligible next to
the 5 x 30-iteration BP/SP passes already performed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Add err=N/N/N/N/N to the FT2 window diagnostic log line, showing the
minimum number of unsatisfied parity equations across all candidates
for each of the five LLR passes. This makes it possible to distinguish
between a signal-quality-limited failure (small error count) and a
systematic decoder bug (large error count), which is the key unknown
in diagnosing the current FT2 LDPC non-convergence.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Normalize FT2 log-likelihoods before LDPC and fall back to\nthe standard waterfall candidate decoder when the raw FT2\npath produces no decodes.\n\nCo-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Use a wider FT4 time-offset search range in candidate acquisition\nso decode remains robust with current pipeline latency.\n\nCo-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Use millisecond-based slot indexing for FT4 decode windows in\nforeground and background decoders to avoid premature 7s resets.\n\nCo-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Use an FT2-specific raw sample window and candidate acquisition path.
Keep the build clean by removing the stale monitor warning.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Use FT2-specific analysis settings and a wider receive span.
Switch server-side FT2 decoding to a rolling async window.
Widen FT2 candidate timing search in the vendored decoder.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Implement a distinct FT2 protocol path in the decoder stack and align\nits timing with the confirmed FT2 framing used by Decodium.\n\nCo-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
The function handles FT8, FT4, and WSPR locators so the ft8 prefix
was misleading. Updated all call sites in app.js, ft8.js, ft4.js,
and wspr.js.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
- Add ft4 to DEFAULT_MAP_SOURCE_FILTER (visible by default)
- Add ft4 hue to locatorThemeHues (peakHue + 30° offset from FT8)
- Propagate ft4 through locatorSourceLabel, locatorFilterColor,
locatorHueForEntry, isLocatorOverlay, applyMapFilter,
ensureDecodeLocatorMarker, pruneLocatorEntry, rebuildDecodeContactPaths,
clearMapMarkersByType, markerSearchText, navigateToMapLocator, and
the source items chip row
- ft8MapAddLocator now maps type="ft4" to markerType="ft4" so FT4
locators get their own distinct marker colour and filter chip
- ft4.js: pass "ft4" (not "ft8") to ft8MapAddLocator
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
bmReadDecoders, bmWriteDecoders, and bmApply were all missing FT4,
so the decoder checkbox was never saved and tuning a bookmark never
toggled the FT4 decoder state.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Decouple the longest-QSO summary from the contact path render toggle.
Keep the summary filtered by current map visibility while the toggle only
controls whether path overlays are drawn.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stanislaw Grams <stanislawgrams@gmail.com>
Serve grouped decode history payloads and restore each decoder through
explicit history restore hooks instead of replaying a mixed message stream.
This reduces replay overhead further by removing type regrouping and keeping
history restoration on decoder-specific bulk paths.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Serve a dedicated decode-history worker and move compressed history fetch
and CBOR parsing into that worker.
The main thread now drains ready-made decode batches within a frame budget,
which further reduces UI disruption during large history restores.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Override the global button chrome on longest-QSO cards so they keep
the intended card layout instead of inheriting fixed control sizing.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Replay decode history in decoder-specific batches instead of feeding every
message through the single-message path.
This reduces per-message array churn and UI scheduling during large history
loads while keeping the existing live decode behavior unchanged.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Serve decode history as gzipped CBOR and decode it in the frontend.
Defer map materialization until replay completes to avoid replay-time stutter,
and include the pending longest-QSO style adjustment.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
sjg
