Skip to content

Decoder roadmap

Status: WebSDR 6A, 6B.0 and 6B.1A–C have accepted slices. Phase 6 remains incomplete: extended FT8/FT4 workspace coverage, deferred WSPR, character, packet/paging, RDS, ISM and full NRSC-5 acceptance remain open. WebSDR Phase 7 is the future on-demand native broker, not the separate Server Phase 7.

Scope and dependencies

Use the shared bounded lane input, rational sample clock, isolated WASM ABI and explicit decoder start/stop. The catalog is the implemented inventory; the requested modes are a wishlist. Automatic scan integration depends on deferred WebSDR Phase 4.

Phase 6A was operator-accepted 2026-08-27; exact timing and FT8/FT4 slices on 2026-08-28. The live FT8 checkpoint retained 55 CRC-valid messages with no reported gap/drop/reset; FT4 passed its pinned reference fixture. These are bounded observations, not the extended multi-rate/dual-lane/live-FT4 matrix. NRSC-5 FM checkpoints through 2026-08-31 established sustained stereo, program selection and received artwork/traffic/radar objects. Positive AM, diagnostic fault replay/isolation, upstream comparison and formal full acceptance remain open. See NRSC-5 limits.

Acceptance criteria and open gates

These are retained acceptance records, not qualification of the current source. Checked items keep their recorded scope; implementation, qualification and operator acceptance are distinct. Later work does not close unchecked gates.

Phase 6 — Decoder Core Pack

Phase 6A — Substrate and Contracts

  • [x] ~~Version and extend the decoder-manifest contract for complex IQ, linear audio, FM discriminator, FM multiplex, envelope/pulse, and protocol-bit inputs; exact sample rates and bandwidth; tuning reference; timing needs; output audio and data services; execution ABI; automatic-launch policy; cost; limits; and hardware/path compatibility.~~
  • [x] ~~Version and extend normalized decoder events with lane, receiver, path, job, track or manual-VFO target, capture epoch, RF and decoder reference, sample time, synchronization, CRC/FEC, BER/MER, frequency/time offset, service/program identity, structured payload, and bounded media references.~~
  • [x] ~~Build an allowlisted, build-time-validated decoder catalog with pinned source revisions, content-addressed lazy-loaded assets, and an explicit ABI handshake. A manifest describes data and must never load arbitrary code.~~
  • [x] ~~Implement a lane-scoped decoder manager and explicit job state machine: suggested, preparing, synchronizing, running, holding, complete, stopped, cancelled, and fault. Keep bounded status and result history after a job stops.~~
  • [x] ~~Support decoder targets from either a stable detected parent emission or the operator's current manual VFO/passband. Manual decoding must remain usable when signal detection is disabled.~~
  • [x] ~~Implement a continuous decoder-specific channelizer/resampler downstream of the existing selected-signal IQ bridge. Decoder input must be independent of speaker volume, mute, squelch, browser AGC, de-emphasis, and the audible filter unless the decoder manifest explicitly requires that audio domain.~~
  • [x] ~~Add a backward-compatible session clock anchor pairing server real and monotonic time, map IQ sample counters to UTC, expose clock quality and uncertainty, and use recording metadata for replay timing.~~
  • [x] ~~Route only bounded low-rate decoder input to dedicated Web Workers/WASM; never duplicate full-rate IQ through React or the main thread.~~
  • [x] ~~Keep every decoder module separate from the core receiver WASM, validate its exported ABI by instantiating the shipped browser artifact, and ensure a missing, stale, or failed decoder cannot prevent receiver, audio, spectrum, waterfall, detection, or candidate analysis startup.~~
  • [x] ~~Bound decoder input queues, memory, wall time, output rate, retained events, binary objects, retries, and CPU admission by performance profile. Mark every discontinuity and reset only protocol state invalidated by it.~~
  • [x] ~~Stop or safely resynchronize decoder jobs on retune, incompatible manual adjustment, target loss, receiver release, route handoff, session close, browser worker failure, or cancellation without disturbing another lane.~~
  • [x] ~~Permit independent compatible decoder jobs on the two physical receiver lanes and preserve manual control, setting pins, receiver leases, and clean per-receiver demand shutdown.~~
  • [x] ~~Extend intelligent-activation snapshots and undo so a decoder start, replacement, stop, and prior decoder state are visible and reversible. Decoders may request refinement only through the shared activation policy and must never retune or change gain directly.~~

Phase 6A — Decoder UI and Operator Control

  • [ ] Live-accept tune-to-suggestion and opt-in Auto Start on independent receiver lanes across representative FT8/FT4, NRSC-5, 1090ES, and 978 UAT frequencies, including manual override and incompatible-rate diagnostics.

  • [x] ~~Add a stable decoder section to the pinned inspector with suggested and active decoder, input reference/rate/bandwidth, timing quality, job state, synchronization, validation evidence, signal quality, progress, warnings, start/stop/retry, decoder pin, and compatible manual choices.~~

  • [x] ~~Present decoded text, messages, telemetry, services, positions, media, and protocol-specific fields in a bounded session history that does not jump as status changes and can export normalized JSON/CSV where applicable.~~
  • [x] ~~Distinguish band-plan suggestion, attempted synchronization, protocol confirmation, validated message/frame, and unvalidated diagnostic output. Never promote a frequency prior or decoder attempt into a protocol claim.~~
  • [x] ~~Keep decoded data browser-local by default. Escape all text, validate MIME types, dimensions, lengths, and filenames, and never automatically upload to APRS-IS, PSK Reporter, MQTT, lookup services, or other third parties.~~

Phase 6B — Initial Browser Decoder Pack

6B.0 — Shared Foundation

Protocol Substrate Hardening
  • [x] ~~Extend the accepted WASM worker ABI with generic bounded protocol-event, status, audio, and binary-media drains while retaining the Phase 6A diagnostic ABI as a compatibility fixture.~~
  • [x] ~~Replace diagnostic integrate-and-dump rate conversion with a properly anti-aliased polyphase or equivalent multistage rational resampler. Preserve exact sample counts, phase continuity, capture epochs, and fractional decoder rates without weakening discontinuity reporting.~~
  • [x] ~~Implement the manifested linear-audio, FM-discriminator, FM-multiplex, envelope/pulse, and protocol-bit producers in addition to complex IQ. Keep decoder input independent of speaker mute, volume, squelch, de-emphasis, and the audible filter unless the manifest explicitly selects that domain.~~
  • [x] ~~Resolve decoder search span, channel-grid center, carrier/edge reference, input rate, and guard bandwidth from protocol policy rather than generically copying a detected rectangle or VFO reference.~~
  • [x] ~~Add a bounded stereo decoder-audio path with exact 44.1-to-48 kHz conversion where required, low-jitter buffering, source switching, and continuity telemetry separate from analog demodulator audio.~~
  • [x] ~~Implement a bounded session-local media store and transfer contract for broadcast images and binary objects, with validated type, dimensions, filename, length, fragments, lifetime, and explicit operator download.~~
  • [x] ~~Replace the single diagnostic build script with a deterministic decoder build matrix supporting pinned sources, readable patches, source and artifact hashes, explicit linker exports, scalar compatibility, and WASM SIMD where it provides measured benefit.~~
Detachable Signal-Intelligence Workspace
  • [x] ~~Add an explicit Pop out control to open the pinned signal inspector, assisted analysis, decoder controls, decoded content, media, and session history as a same-origin multi-monitor workspace.~~
  • [x] ~~Keep the tuner window authoritative for receiver leases, IQ WebSockets, DSP and decoder workers, audio engines, and jobs. The pop-out must synchronize state and explicit control actions without opening duplicate receiver paths, streams, workers, or leases.~~
  • [x] ~~Synchronize lane selection, signal pins, decoder suggestions and jobs, validation state, program selection, results, warnings, and exports between windows. Support dock/reattach, clean close/reload recovery, independent scrolling, stable reserved layout, and both simultaneously active receivers.~~

6B.1 — Timed Weak-Signal Modes

6B.1A — Exact Timing and Reference Contracts
  • [x] ~~Move decoder ingress to a continuous capture-order tap before the lossy fine-analysis and candidate queues. Channelize at the trusted realtime DSP boundary, transfer only bounded decoder-rate samples to the isolated module, and prove that analysis backpressure cannot reset an otherwise continuous decoder stream or disturb audio.~~
  • [x] ~~Account separately for source-IQ gaps, decoder-queue drops, timing-epoch changes, and decoder resets. Report contiguous and total processed samples, never claim continuous flow across an undisclosed reset, and recover only the affected lane-local decoder job.~~
  • [x] ~~Extend decoder-rate chunks and the backward-compatible WASM ABI with the first output sample's exact rational source position, decoder sample counter, UTC time, clock epoch, and discontinuity state so a module never has to infer timing from WebSocket or worker-message boundaries.~~
  • [x] ~~Account explicitly for multistage resampler phase and FIR group delay, preserving exact input-to-output sample mapping across transport frames and proving it to within one decoder-rate sample.~~
  • [x] ~~Refresh the server realtime/monotonic clock anchor during active sessions, distinguish normal bounded correction from a material clock step, and start a new timing epoch when continued slot assembly would be unsafe.~~
  • [x] ~~Separate VFO or suppressed-carrier RF/tuner references from occupied signal bounds, decoder search bounds, and protocol grid/dial references. Express asymmetric relative search windows and validate sample-rate feasibility according to the selected real or complex input domain.~~
  • [x] ~~Cover live and replay clocks, mid-stream anchor refresh, clock steps, retunes, gaps, fractional decoder rates, frame-boundary invariance, and every configured hardware sample rate without weakening the accepted diagnostic decoder or receiver/DSP isolation.~~
6B.1B — FT8 and FT4 Decoder Pack
  • [x] ~~Qualify and pin the FT8/FT4 browser engine revisions against current official WSJT-X reference output. Record source revisions and hashes, readable patches, deterministic build inputs, supported 77-bit message families, decode depths, calibrated and uncalibrated measurements, and known recall limitations rather than assuming parity from a successful build.~~
  • [x] ~~Add allowlisted FT8 and FT4 manifests and browser-WASM artifacts using the accepted ABI and exact input-domain, sample-rate, bandwidth, dial/carrier reference, UTC uncertainty, search-window, output, cost, and retry contracts. The shipped browser artifact, not merely a native build, must pass the ABI and reference fixtures.~~
  • [x] ~~Implement integer-sample-clock UTC-aligned FT8 and FT4 slot assembly with full compatible decode windows independent of the audible filter. Expose waiting, collecting, decoding, incomplete, discontinuous, timing-uncertain, complete, and next-slot recovery states without allowing a damaged slot to contaminate the next valid one.~~
  • [x] ~~Search multiple simultaneous signals and emit structured normalized events containing UTC slot, RF and decoder frequency, time offset, SNR, decoded and raw message, message type/subtype, callsigns, CQ modifier, grid, report or acknowledgement, contest/free-text/telemetry fields, hash-resolution state, synchronization evidence, LDPC state, CRC result, and decode pass. Deduplicate repeated candidates without collapsing the same message received at materially different frequencies.~~
  • [x] ~~Replace the short raw-JSON preview with a protocol-agnostic, lane/job scoped append-only decoder event store indexed by time, slot, frequency, protocol, validation, and identity. Use chunked retention and range queries outside the large React lane snapshot so long histories do not copy, sort, or render in full on every decode; report any eviction explicitly.~~
  • [x] ~~Add an allowlisted protocol presentation registry and an FT8/FT4 activity workspace with All activity and selected-target scopes; UTC, mode, SNR, DT, absolute RF, decoder offset, message, parsed meaning, and validation columns; slot grouping; sticky headings; virtualization; search and filters; readable highlighting; a structured evidence inspector; copy; and normalized JSON/CSV export. Keep all content escaped and browser-local.~~
  • [x] ~~Give docked and popped-out windows shared decoder history but independent review state. Follow live defaults on; Freeze view holds only that surface's scroll position, time cursor, filters, and selected row while decoding and history ingestion continue. Show a new-result count and Jump to latest without stopping, resetting, or restarting the decoder.~~
  • [x] ~~When a view is frozen, expose a retained-history time slider with UTC endpoints, protocol-slot ticks, decode/discontinuity/retune/job markers, snap-to-slot behavior, previous/next slot and result controls, direct UTC entry, clear out-of-range feedback, Go live, and bounded real-time or accelerated event replay. Replay stored normalized events only; historical raw-IQ retention or re-decoding is separate future work.~~
  • [x] ~~Bound candidates, LDPC iterations, optional deeper passes, interference subtraction or OSD, decoder deadlines, event retention, indexes, memory, and wall time by measured performance policy. Permit generous configurable desktop history while preserving audio, waterfall, spectrum, detection, UI response, decoder slot collection, and the other receiver lane.~~
  • [x] ~~Validate the exact browser modules against pinned official/reference WAVs, browser-instantiated ABI gates, partial-slot and jittered-clock tests, and live over-the-air FT8. Confirm CRC-valid structured output, bounded decode latency, and uninterrupted receiver, audio, spectrum, waterfall, and decoder input.~~
  • [ ] Complete the extended operator-workspace matrix against controlled mixtures spanning message families, weak SNR, time/frequency offset, adjacent and overlapping signals, duplicate messages at same and distinct frequencies, gaps, clock steps, retunes, job boundaries, every configured capture rate, both receiver lanes, long histories, freeze/scrub/jump/replay, independent pop-out review, text sizes, memory/latency, and live FT4. Record recall, false output, timing error, decode latency, and UI/audio continuity as the matrix is exercised.
6B.1C — Explicit Intelligent Activation
  • [x] ~~Let explicit intelligent activation attempt FT8 or FT4 only when a live, persistent candidate, stable or held structured-digital evidence, artifact counterevidence, and compatible timing/capture state support the hypothesis. Frequency and service context may only reorder compatible trials after that measured evidence exists. Treat all pre-decode inputs as permission to try, not proof of protocol identity.~~
  • [x] ~~Derive the USB/LSB/symmetric tuning reference and decoder search extent from the stabilized RF energy actually observed. Never replace that reference with a customary working frequency. Keep known working frequencies as a low-weight, operator-visible protocol-selection prior only; receiver error or nonstandard operation must not prevent a measured candidate from being tried.~~
  • [x] ~~Bound ambiguous FT8/FT4 attempts by evidence, channel, slot cadence, resource policy, and retry state. Do not launch an indefinite competing- decoder loop or silently replace a pinned decoder choice.~~

  • [x] ~~Separate instantaneous boundary evidence, the published clickable signal rectangle, the held activation/control envelope, and the protocol decoder search window in state, policy, diagnostics, and tests.~~

  • [x] ~~Stabilize the published rectangle on an absolute RF grid with independently justified edge evidence: discount overlapping windows, allow sustained substantial energy to pass a false-alarm-oriented sequential test and expand an edge promptly, require stronger sustained absence before contraction, retain sufficient statistics through ownership-window changes, and make accuracy and confidence improve with genuinely new observations.~~
  • [x] ~~Snapshot an evidence-backed control envelope at activation. Evolve it only under explicit follow policy, preserving a stable signal through geometry jitter without allowing a nearby intermittent emission to widen, pull, merge with, or remain attached to the primary target.~~
  • [x] ~~Preserve parent identity through compatible recurrence while ending a genuinely absent neighboring candidate. Reset held geometry only for sustained target loss, incompatible signal nature, route/capture change, or explicit operator action.~~

  • [x] ~~Track a robust carrier, dial, or paired-edge translation estimate rather than the latest rectangle center or either edge alone. Require sustained consistent motion, retain cadence/deadband/correction bounds and an activation anchor, and prove that width or edge breathing cannot move a stationary VFO.~~

  • [x] ~~Fit the initial receive passband from the held activation envelope using the selected mode's real geometry: lower reference for USB, upper reference for LSB, robust center for symmetric modes, and explicit CW/digital rules. Keep the occupied envelope contained with only resolution-appropriate guard room and without widening the audible filter to the decoder search width.~~
  • [x] ~~While following, permit evidence-backed outward containment adjustment only when bandwidth is unpinned; make contraction deliberately slower than expansion and never resize from a transient frame. If a pin prevents containment, report it rather than overriding the pin or silently chasing the signal with frequency.~~
  • [x] ~~Keep the tuning/passband policy registry extensible. Each later Phase 6 mode is incomplete until its manifest or allowlisted adapter supplies the mode-integration contract above and passes stationary-envelope, real-drift, asymmetric-passband, pin, and manual-takeover fixtures.~~

  • [x] ~~Present candidate, attempting, synchronizing, confirmed, unconfirmed, holding, and lost states distinctly. Confirm FT8 or FT4 only through protocol synchronization plus valid decoder evidence; absence of a decode in a bounded weak-signal interval is not proof that the hypothesis was false. Hold a validated protocol decision through ordinary fades on a minutes-scale protocol-aware timer, extending it on each new valid message.~~

  • [x] ~~Make decoder retention and loss handling protocol-slot aware. Brief detector contraction, fade, or off interval must not reset a contiguous job; valid decoder evidence may outweigh a momentary detector miss, while sustained loss, incompatible change, explicit retune, route release, or operator cancellation ends or resynchronizes only the affected job.~~
  • [x] ~~Keep manual-VFO decoder startup available with detection disabled or no band-plan match. Preserve all setting and decoder pins, cancellation, snapshot undo, and immediate manual takeover; automatic authority must never fight a subsequent operator adjustment.~~

  • [x] ~~Show published detector bounds, held control bounds, active passband, dial/carrier reference, protocol search window, containment state, follow state, and the reason a proposed correction was applied or withheld in the docked and popped-out signal-intelligence workspace without destabilizing layout or extending detection paint into the dB/VFO area.~~

  • [x] ~~Add deterministic regressions for stationary signals with breathing edges, sustained real drift, temporary edge loss and recurrence, widening and slow contraction, intermittent adjacent signals, FT8/FT4 frequency-offset operation and contextual-prior isolation, USB/LSB and symmetric containment, pins, manual takeover, undo, detection-off manual targets, lane isolation, and decoder failure isolation.~~
  • [x] ~~Validate against live FT8 and, when available, FT4 plus a stable broad symmetric signal such as 100.7 MHz WFM. Confirm accurate containment, no VFO motion from rectangle reshaping, bounded response to genuine drift, stable decoder collection through normal fades, and no regression to receiver startup, audio, spectrum, waterfall, detection, history, or the other lane.~~
6B.1D — WSPR and Final Weak-Signal Validation
  • [ ] Implement WSPR with its independent even two-minute UTC slot, explicit collection progress, bounded frequency/time/drift search, and deterministic recovery after a gap or partial slot.
  • [ ] Decode WSPR message types 1, 2, and 3 with callsign, locator, transmitter power, hash-resolution state, synchronization and FEC evidence, and an explicit CRC-not-applicable result. Never invent a callsign for an unresolved hash or upload reception reports automatically.
  • [ ] Bound Fano/Jelinek cycles, block demodulation, subtraction, duplicate handling, optional OSD, search width, passes, and latency by performance profile, with deeper work admitted only after measured benefit.
  • [ ] Validate FT8, FT4, and WSPR against pinned official/reference WAVs and controlled synthetic mixtures spanning SNR, time/frequency offset, drift, sample-rate error, adjacent and overlapping signals, mid-slot starts, gaps, retunes, both receiver lanes, every configured capture rate, pop-out state, and live over-the-air activity. Record recall, false output, timing error, latency, memory, and UI/audio continuity before operator acceptance.

6B.2 — Character Modes

  • [ ] Implement CW with adaptive tone, speed, spacing, polarity, and readable confidence/evidence rather than pretending unframed text has CRC certainty.
  • [ ] Implement RTTY/Baudot with common baud rates, shifts, polarity, stop-bit policy, AFC, and framing-quality telemetry.
  • [ ] Implement PSK31 plus common parameterized BPSK/QPSK variants where the shared carrier, symbol-timing, differential, and Varicode engine validates.

6B.3 — Packet and Paging Modes

  • [ ] Implement AX.25/APRS starting with AFSK1200, HDLC flags and bit stuffing, FCS validation, and structured APRS messages, positions, objects, weather, status, and telemetry.
  • [ ] Implement POCSAG 512/1200/2400 and FLEX with symbol timing, polarity, address/function data, BCH/FEC state, message type, and bounded handling of sensitive pager content.

6B.4 — Broadcast RDS/RBDS

  • [ ] Implement RDS/RBDS from the FM multiplex before audible low-pass and de-emphasis, including PI, PS, PTY, RadioText, RadioText+, alternative frequencies, traffic flags, clock, decoder state, and validated group data.

6B.5 — HD Radio / NRSC-5

6B.5A — Reproducible, Isolated Browser Module
  • [ ] Qualify and pin a compatible nrsc5 revision plus its exact FFT and HDC audio dependencies for a reproducible browser-WASM build; record exact source, patch, build, attribution, codec, fixture, and artifact provenance.
  • [ ] Build a content-addressed, lazy-loaded NRSC-5 module through the existing decoder allowlist and ABI handshake. Verify the shipped artifact rather than only a native build.
  • [ ] Compile only the IQ-fed decoder surface. Prove by source selection, WASM import/export inspection, and runtime tests that the module cannot open SDR hardware, rtl_tcp, files, sockets, network resources, or tuner controls.
  • [ ] Bound memory, callbacks, event/audio/media queues, malformed radio data, object sizes, processing time, retries, cancellation, and failure recovery so the receiver, audio, spectrum, waterfall, detection, and other lane remain independent.
6B.5B — FM and AM Acquisition
  • [ ] Implement FM and AM HD Radio/NRSC-5 synchronization and decoding from decoder-specific complex streams at their exact rational native rates. Preserve capture epochs, discontinuities, sample-clock timing, mixer reference, filter delay, and frame-boundary invariance.
  • [ ] Keep manual start available at any frequency. Intelligent activation may rank NRSC-5 from measured hybrid-sideband evidence and contextual priors, but a station-frequency convention alone must never tune, launch, or claim HD. Valid decoder synchronization is protocol evidence.
  • [ ] Surface acquisition, sync/loss, FM/AM service mode, frequency offset, MER/BER, block/CRC state, continuity, and actionable failure reasons without allowing brief fades to thrash an established session.
6B.5C — HD1–HD8 Audio and Program Control
  • [ ] Stream the selected public HD1–HD8 audio service through the existing browser audio engine with end-to-end left/right preservation, bounded low-jitter buffering, and exact 44.1 kHz to output-rate conversion.
  • [ ] Discover and present public programs dynamically; support clean program switching, reset stale audio at the switch boundary, preserve lane isolation, and distinguish unavailable, unknown, restricted, corrupt, and unsupported services without bypassing access controls.
  • [ ] Define explicit loss-of-sync and analog fallback/blend policy. Never mix two unaligned paths implicitly, and never let decoder audio disrupt ordinary analog reception when HD is absent or the decoder fails.
6B.5D — Metadata, Images, and Broadcast Objects
  • [ ] Expose all safely parsed NRSC-5 content provided by the signal: station ID/name/slogan/message and location, service guide, public audio/data programs, program type and audio details, title, artist, album, genre, comments, supported LOT/AAS/stream/packet data, broadcast time, and emergency alerts.
  • [ ] Display every supported station logo, album-art image, broadcast-carried photo, or program image actually transmitted with the selected service. Associate it with the correct service/program, object type, validation state, and broadcast time; retain it through brief metadata gaps; permit enlargement and explicit download; and clearly report when no image was transmitted. Never substitute an external artwork lookup for broadcast content.
  • [ ] Bound and validate binary-object reassembly, MIME, dimensions, fragments, strings, structured fields, and object lifetime. Treat every decoded field as hostile radio-supplied input and never present emergency content as safety-certified.
6B.5E — Operator Workspace and Acceptance Gates
  • [ ] Add a protocol-shaped docked and pop-out HD Radio workspace with stable station/sync/quality state, HD1–HD8 selection, now-playing content, received images, data objects, emergency notices, decoder status, and time-ordered history. Preserve pause/live review and history navigation while decoding continues in the main tuner process.
  • [ ] Validate deterministic FM and AM fixtures, all public program slots, program switching, audio continuity, metadata and image/object reassembly, corrupt/restricted/unsupported inputs, retunes, gaps, sync loss/reacquisition, cancellation, stale assets, worker failure, both receiver lanes, and every configured capture rate.
  • [ ] Compare normalized outputs against the pinned upstream decoder on the same provenance-known captures; record expected differences. Benchmark CPU, memory, queue depth, latency, event rate, and audio continuity while proving receiver/audio/waterfall/detection isolation.
  • [ ] Prove the operator-only sanitizer and fault-forensics path: reproduce a native memory fault with source-oriented runtime output, preserve an expandable and JSON-exportable report, export/replay the bounded pre-fault IQ against the exact content-addressed artifact, and keep normal decoder jobs free of diagnostic retention overhead.
  • [ ] Complete live operator acceptance of the deployed 6B.5 candidate before striking any 6B.5 item. Operator-requested checkpoint commits do not imply acceptance.

6B.6 — Curated ISM Decoder Pack

  • [ ] Implement selected lightweight rtl_433 paths or a browser adapter with a generic OOK/FSK pulse view, curated lazy-loaded protocol groups, structured fields, checksum evidence where available, and no hardware or network access.

Source Provenance and Reproducibility

  • [ ] Record the origin, exact revision, source hash, attribution, notice, local modifications, and transitive dependencies for every imported decoder, codec, FFT, fixture, and generated artifact in the third-party register.
  • [ ] Pin exact upstream commits or releases and hashes, keep readable patches and deterministic source build scripts in the repository, and do not check in an opaque decoder blob without its reproducible corresponding source.
  • [ ] Review the exact binary's source, notices, codec/patent considerations and fixture redistribution terms before publication. Record that review in the third-party register and revisit it when artifacts or distribution change.

Regression and Acceptance Gates

  • [ ] Add synthetic generators and project-owned or otherwise provenance-known recorded CU8/WAV/SigMF golden fixtures for every decoder, including upstream reference captures and exact expected normalized events.
  • [ ] Test strong, weak, overlapping, drifting, fading, mistuned, clipped, time-offset, sample-rate-error, discontinuous, truncated, wrong-protocol, and noise-only inputs; require correct rejection as well as successful decode.
  • [ ] Compare messages, frames, CRC/FEC results, and supported metadata with authoritative upstream decoders on identical samples, and document expected differences rather than silently weakening fixtures.
  • [ ] Test every shipped decoder WASM artifact through the browser ABI, CSP, worker lifecycle, stale/mismatched asset, cancellation, timeout, malformed event, and bounded-output paths.
  • [ ] Give future native/WASM decoder imports the same opt-in sanitizer build, bounded runtime-output/stack capture, structured operator fault report, and exact input-replay methodology before calling their failure behavior scientifically inspectable.
  • [ ] Validate retune, gap, overflow, receiver release, route handoff, manual takeover, pin, undo, target fade/recurrence, browser disconnect, and decoder crash recovery on both independent receiver lanes.
  • [ ] Fuzz decoder input and all decoded strings, structured fields, and binary objects as hostile radio-supplied data; prove bounded allocation and safe rendering/download behavior.
  • [ ] Benchmark decode latency, CPU, memory, queue depth, event rate, and browser power by decoder and performance profile. Protect audio continuity, display responsiveness, waterfall history, detection, and manual tuning under the heaviest admitted decoder workload.
  • [ ] Live-validate available FT8/FT4 and broadcast signals, including RDS and HD Radio where locally receivable; use golden recordings as the acceptance authority when a protocol is not reliably on air.
  • [ ] Update AGENTS.md, README.md, manifests, license register, build and recovery instructions, and the completion ledger only with validated facts.
  • [ ] Validate all non-deferred Phase 6 deliverables and update the completion ledger. Phase 4 scan integration remains open until scanning resumes.

Reconciled Phase 6 Completion Ledger — 2026-08-31

Accepted slices are Phase 6A, 6B.0, 6B.1A, 6B.1B, and 6B.1C. The following work still prevents Phase 6 closeout:

  • Complete the extended 6B.1B FT8/FT4 operator-workspace matrix. Resume the deferred 6B.1D WSPR and final weak-signal matrix when the operator schedules it.
  • Implement 6B.2 character modes, 6B.3 packet/paging modes, 6B.4 RDS/RBDS, and 6B.6 curated ISM decoding.
  • Close the explicit 6B.5 HD Radio gates listed above, including positive AM, full rate/lane/failure coverage, diagnostic replay, comparative results, resource measurements, and formal operator acceptance.
  • Complete the cross-decoder source/provenance and regression/acceptance gates. The NRSC-5 and FT8/FT4 evidence already recorded is retained, but it does not prove decoders that have not yet been implemented.
  • Complete the remaining live acceleration matrix across modes, rates, FFT tiers, decoder paths, dual lanes, and renderer fallbacks. The accepted smooth 120/165 Hz Windows WebGPU checkpoints remain valid partial evidence.

Automatic decoder actions remain in deferred Phase 4 and are not a Phase 6 closeout dependency while that phase is deferred.

Phase 7 — On-Demand Decoder Broker

  • [ ] Define a safe native worker interface.
  • [ ] Add WSJT-X-family adapters.
  • [ ] Add fldigi-family adapters.
  • [ ] Add Dire Wolf.
  • [ ] Add multimon-ng.
  • [ ] Add native aviation adapters only where the Phase 10 browser/WASM qualification shows that an existing decoder cannot run safely or efficiently in the browser. Phase 10 remains the owner of aviation identity, history, workspace, metadata, selection, and map behavior.
  • [ ] Add digital voice.
  • [ ] Add DAB and broadcast decoders.
  • [ ] Add satellite and image pipelines.
  • [ ] Add CPU, memory, timeout, and output limits.
  • [ ] Make native workers selectable as automatic scan actions where safe.
  • [ ] Update AGENTS.md and README.md with validated decoder-worker operation and recovery.
  • [ ] Validate Phase 7 deliverables and update the completion ledger.