Skip to content

Decoder workspace

The signal-intelligence workspace separates lane navigation, pinned target, decoder controls, decoded activity, protocol-specific evidence, and product workspaces. Panels can be moved/resized on a grid; layout is persisted locally. The default places the pinned-signal panel beneath the receiver lane so the decoder content can use more horizontal space.

Decoder controls

The Decoder selector shows compatible manifests and the recommendation/manual source. Input target identifies manual VFO or a detector track. Auto Start sits beside its label and status. Start/Stop, Pin job, bounded media, and export actions remain lane/job scoped.

Changing a decoder workspace does not retune the receiver, stop another lane, or start a second application instance. A pop-out is a same-origin React portal; the main window remains the sole owner of API clients and workers.

Start is explicit by default. Auto Start is a persisted opt-in, not the same control as Follow live in the event table. Suggestions can use a documented service frequency as a prior; a trial needs manifest admission, timeout and cooldown, and actual decoder evidence to become a protocol claim. Manual selection, operator-started/pinned jobs, and validated protocol holds take precedence over new suggestions.

Input and job state

The input readout should establish the target, reference frequency, input domain/rate/bandwidth, clock quality/uncertainty, and continuity. Listening volume, mute, squelch, AGC, de-emphasis, and audible-filter settings do not change a decoder-only input unless the manifest requires that listening domain. Manual-VFO Start remains available with signal detection disabled.

Preparing, synchronizing, running, holding, complete, stopped, cancelled, and fault describe the job, not interchangeable validation levels. A retune, incompatible manual adjustment, target loss, route handoff or receiver release stops or safely resynchronizes the affected job; another receiver lane must remain independent. Pinning and assisted-activation undo preserve operator authority. A fault report stays inspectable after processing stops.

Decoded activity

The event table is presentation-model aware. Shared storage keeps normalized events; each decoder presentation defines stable useful columns, units, summary, search fields, and evidence groups. Unknown decoders get a generic fallback.

For GOES HRIT/EMWIN, useful rows distinguish telemetry, validated CADUs/frames, packets, and product/file progress where the source event actually supplies them. Virtual channel, application/packet identity, sequence, product, and validation are shown only when meaningful. Fields such as callsign, grid, audio offset, or FT decode pass are omitted rather than displayed as green N/A.

Read the validation label beside the event summary. For example, an HRIT frame can pass Reed–Solomon correction while the image it belongs to is still incomplete. Validated only uses the same evidence shown in the row.

Timing and frequency

Rows preserve event-specific timestamps and offsets. A session's current metric is not stamped onto old rows. Absolute RF appears only when compatible capture/VFO metadata establishes it. Carrier/baseband and audio offsets retain units and reference. Mode-specific DT or dB fields disappear when they do not apply.

For example, a GOES Residual carrier offset of −980 Hz is relative to the decoder reference. Use RF reference to see the tuned frequency in MHz.

Throughput

Ingestion retains every event up to the existing bounded policy. UI publication is coalesced independently, and visible rows are windowed/virtualized. Concise summaries/search tokens are derived once per stable event identity. Selecting a row formats detailed evidence on demand; the render path does not repeatedly serialize raw payloads or scan 100,000 events at 100 events/s.

Follow live, Freeze view, selection, order, counters, search, and validation filtering remain stable. JSON/CSV export uses retained normalized events, not only the columns currently visible.

Freezing a docked or popped-out review surface does not pause decoding. Each surface can keep its own cursor, filters, scroll and selection while sharing retained events. New-result counts and Jump to latest expose incoming history without restarting the job. Event-history playback reads stored normalized events; IQ DVR replay is a separate operation that can run the decoder again on retained samples.

Audio, media, and diagnostics

Decoder audio uses the existing audio engine with an exclusive source handoff; unaligned digital and analog audio are never implicitly mixed. Bounded session media retains only validated types, lengths, dimensions, filenames and fragments; downloads are explicit. No media entry is a legitimate state, not permission to remove shared diagnostic or export controls. NRSC-5 has additional public-service and content rules.

The project-owned diagnostic-iq monitor reports unvalidated ABI/timing/continuity evidence, not protocol identification. NRSC-5 sanitizer jobs are a separate explicit operator action and never an Auto Start candidate.

Ledger and remaining proof

docs/roadmap/decoders.md Phase 6A — Decoder UI and Operator Control and 6B.1B — FT8 and FT4 Decoder Pack preserve the review/control contracts. The live tune-to-suggestion/opt-in Auto Start matrix across independent lanes and the extended FT8/FT4 workspace matrix remain open; a documentation build or one accepted protocol slice does not close them.

SATCOM products

GOES validated CADUs also feed the dedicated SATCOM product worker. Imagery, messages, environmental records, and acquisition activity live in their purpose-built workspace; the generic activity table remains a compact protocol diagnostic, not a file-drop-zone replacement.