Two bugs introduced by 60697bb:
1. dsp.rs passed `channel_idx == 0` as force_mono_pcm, which forced the
primary pipeline channel to output mono samples. The Opus encoder was
configured for stereo, so it received half the expected frame data,
causing distortion for all connected audio clients.
Fixed by passing `false` — hidden virtual channels already set
force_mono_pcm=true via set_force_mono_pcm() in vchan_impl.rs.
2. main.rs short-circuited channel conversion when no audio clients were
connected, sending raw frames to pcm_tx (decoders). When clients then
connected, decoders switched to receiving stereo-interleaved frames,
making decoder input format dependent on client presence.
Fixed by always performing the channel conversion before sending to
pcm_tx; the no-client skip now only bypasses Opus encode + rx_audio_tx.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Skip APRS packets whose ts_ms is older than 120 seconds. Live RF-decoded
packets arrive within milliseconds; history replay items can be up to 24
hours old and must not be re-uploaded to APRS-IS as live traffic.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Rewrite the PSKReporter uplink to match the protocol spec exactly:
- Fix template FlowSetIDs: receiver uses 0x0003 (Options Template Set),
sender uses 0x0002 (Template Set); previously both used 0x9992/0x9993
- Add missing enterprise numbers (0x0000768F = 30351) to all enterprise
field specifiers in both template blocks
- Fix sender template field IDs: use correct attributes (senderCallsign
30351.1, frequency 30351.5, sNR 30351.6, iMD 30351.7, mode 30351.10,
informationSource 30351.11, senderLocator 30351.3, flowStartSeconds 150)
- Fix sender data field order to match the template declaration
- Add iMD byte (0) required by the 8-field template
- Add 4-byte null padding on receiver and sender data records
- Batch spots into one UDP packet per 5-minute window (spec requirement)
- Deduplicate by callsign within each window (keep most-recent spot)
- Send template descriptors only in first 3 packets then once per hour
- Increment sequence number by report count, not packet count
- Guard against history replays: drop any spot older than the flush
window (live FT8/WSPR is seconds old; history can be 24 h old)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Rebuild the visible-source chips when live APRS, AIS, or VDES markers are first added so the map filter list updates without a page refresh.\n\nCo-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Use an unbounded virtual-channel command queue so background decode and scheduler transitions do not silently drop subscribe or remove commands.\n\nCo-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Remove hidden background decode channels when the owning audio client disconnects to avoid stale DSP and decoder buildup.\n\nCo-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Run FT8 and WSPR decode steps in blocking sections so the server listener stays responsive under decode load.\n\nCo-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Replace the misleading scheduler task countdown with the actual time-span interleave switch timing in the main controls row.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Raise remote spectrum polling from 100 ms to 50 ms while keeping the relaxed timeout and subscriber gating.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Relax the remote spectrum timeout, poll at the backend update cadence, and stop polling when no spectrum subscribers are connected.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Add separate map path toggles, move scheduler handoff into the channels row, and show a live countdown to the next scheduler cycle.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Remove settings rig pickers, restore the last scheduler cycle on release, fix FT8 locator role parsing, and add toggleable decode contact paths on the map.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Rewrite the README, remove AI-generated planning docs, and regenerate the combined example config.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Move Scheduler under a new Settings tab in the HTTP frontend.
Add the virtual-channel audio implementation plan document.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Use the active channel frequency for spectrum bandwidth edge hit-testing.
Co-authored-by: OpenAI Codex <codex@openai.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Each RDS PS overlay item (position: absolute within the shared #rds-ps-overlay
container) now receives a z-index derived from its channel frequency: items are
sorted by freq_hz ascending so higher-frequency layers sit on top of
lower-frequency ones by default.
Hovering any layer temporarily assigns it the maximum z-index (entry count + 10)
to bring it to the front; mouseleave restores the frequency-derived default
stored in data-default-z.
Also reverts the incorrectly applied vchan picker layer changes from the
previous commit.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Render virtual channels as absolutely-positioned layer strips inside a
shared relative container (#vchan-freq-layers). Layers are sorted by
frequency ascending so higher-frequency channels receive a higher z-index
and sit on top by default. Hovering any layer temporarily assigns it the
maximum z-index to bring it to the front; leaving restores the original
stacking order. Each layer is offset by 11 px vertically so all channels
remain visible as a staggered card stack.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
sjg
