Superpose

Validated capability · Public corpus forthcoming

Communications v1

A 66-class communications capability for controlled single-signal examples and wideband multi-emitter scenes.

66signal classes
2 modessingle emitter + scenes
I/Q + truthlabels and provenance

Coverage

The catalog spans analog modulation, constellations, FSK, OFDM, chirps, tone, conformant resource-grid emitters, and coded constellations. In the single-emitter corpus, conditions sample SNR, bandwidth, channel behavior, transmitter effects, and receiver effects.

The scene configuration is a separate, narrower contract: it composes 1–6 emitters over a thermal receiver noise floor with co-channel interference. It does not currently apply a fading model or the single-emitter impairment chain. The measured scene pilot is sparse, with 4.1% median time-frequency occupancy.

Representative generated wideband multi-emitter scenes. Technical validation figure; not an over-the-air capture.

What validation supports

  • All 66 classes appear in the evaluated scene draw.
  • Validated structured provenance and time-frequency bounding boxes accompany the evaluated scene pipeline.
  • The single-emitter channel and impairment axes are tested against their declared contracts.

Limits and non-claims

The scene segmentation masks were checked for presence and configured resolution only. Their pixel-level energy alignment and per-plane class assignment were not independently validated, and detector-level validation has not been run for either masks or boxes.

This is a simulator-to-simulator capability. It does not establish real-receiver performance, protocol conformance for every surrogate class, synthetic-to-real transfer, or a publicly downloadable corpus today.

Read the technical validation report (PDF). View authoritative source at the pinned commit.