# MOSAIC-HF 0.1.0a7 — Alpha Capability Matrix

This matrix records what has been demonstrated honestly. Claims without evidence
are explicitly marked. "YES" means a reproducible test or on-air log exists.

| Capability | Status | Evidence / Notes |
|---|---|---|
| UHF multi-user on-air (12 stations) | **YES** | `onair-12station-random` campaign: 12/12 stations decoded by two independent transmitters across repeated runs |
| UHF adaptive Basic/Fast/Turbo on air | **YES** | Two independent transmit radios plus HackRF observer: **3/3 CRC-valid**, 100% measured active-span occupancy, 53.4% two-or-more-user overlap. Evidence: `adaptive-mcs/uhf-three-radio-final-20260815/`. |
| HF adaptive Basic/Chat/Fast over UHF carrier | **YES, bounded** | Same three-radio topology at 431.200 MHz: **3/3 CRC-valid**, 100% measured occupancy, 51.3% two-or-more-user overlap. This validates the HF waveform family and receiver, not HF-band propagation. |
| UHF live keyboard QSO | **YES** | Fresh 2026-08-11 bidirectional: VE6SLP-1 (bladeRF/HackRF) ↔ VA6GA-2 (B210) at 431.200 MHz; `run-artifacts/alpha-uhf-qso/qso-manifest.json` |
| Realtime browser-controlled UHF QSO | **YES** | Fresh 2026-08-14 bidirectional exchange at 431.200 MHz: B210 burst decoded CRC-valid in the VE6SLP-1 browser console; console-queued bladeRF reply decoded CRC-valid on VA6GA-2. Includes live spectrum/waterfall, station selection, QSO lifecycle, actual RF completion events, and fail-closed TX. Evidence: `run-artifacts/operator-ui-rf-20260814/` and site screenshots. |
| HF audio path, single-user | **YES** | Byte-exact end-to-end through all audio converters; deterministic unit tests pass |
| HF continuous browser/KISS sound-card service | **YES, implementation bounded** | Packaged `mosaic-hf-audio serve` reuses the operator console and KISS TCP service, continuously captures sound-card audio, runs bounded peer-catalogue HF SIC, and keeps audio/PTT fail-closed until armed. Service lifecycle and CRC-valid peer decode are deterministic tests; no on-air HF QSO is claimed. |
| HF realtime decode (single user) | **YES** | 0.10× wall-clock with `hf-audio` profile on reference hardware |
| HF ITU F.1487 single-user, good SNR | **YES** | Passes to SNR cliff; graceful degradation below cliff is expected |
| HF multi-user under Watterson fading (sim) | **LIMITED → improved** | Simulation only. `hf-congested` + multi-order known-payload residual rebuild: **4/4**, **6-user mean 5.83/6 (full 83%)**, **8-user mean 6.50/8 (full 50% / 6 seeds)** at 2 ms / 0.5 Hz / 6 dB. Hard seed 19 now 8/8 via rebuild. Seed 21 pure-MAI: **6/8** at 1.5 s arrival span, **8/8** at 40 s span (protocol stagger). Multi-copy same-signature peel after CRC for retransmits. Live modem default HF TX access jitter ≈ one packet duration. Not on-air HF. Artifacts: `hf-fade-harness-*.json`, `hf-8user-rebuild-6seeds.json`, `seed21-span-probe.json` |
| HF bounded CW QRM + impulsive QRN (sim) | **YES, bounded simulation** | Four users, two Watterson-style seeds: robust front end **48/48** versus baseline **30/48**, clean unchanged at 8/8, zero false deliveries. Cells cover persistent 1500 Hz CW, one declared three-window keyed schedule, +0.75 Hz/s linear drift, impulses, and persistent CW plus impulses, all at the declared levels. Arbitrary keyed patterns, nonlinear chirps, digital/speech QRM and on-air HF are unmeasured. Preprocessor worst timing 1.026% of observation time; full four-user SIC is not realtime. Evidence: `hf-qrm-qrn-congested.json`, `hf-qrm-qrn-evidence.png`. |
| UHF multipath FIR proxy (lab) | **YES (cabled + radiated)** | Static two-path FIR (`apply_two_path_iq_fir`, 2 ms / 0.7). **Cabled** pad path and **radiated** antenna path at 431.200 MHz: 2-user HF profile, both CRC-valid (`VE6SLP`/`VA6GA`, MPATH1/MPATH2), false cancels 0. Evidence: `uhf-multipath-cabled-2user-evidence.json`, `uhf-multipath-radiated-2user-evidence.json`. Offline Watterson+FIR audio 4-user: **5/5 full**. Profile `uhf-multipath` enables multi-tap SIC on UHF waveform. |
| HF direct SDR IQ path | **YES (cabled + radiated)** | HF profile (16-FSK, 80 ms) over free-running UHF SDR LO at 431.200 MHz. **Cabled** pad path and **radiated** antenna path both CRC-valid after local trajectory refine + residual-aware header (`run-artifacts/hf-iq-cabled-crc/`, `run-artifacts/hf-iq-radiated-crc/`). Sound-card audio remains the intended HF deployment model; this closes the historical header-reject gap on the IQ path. |
| On-air HF contact | **NO** | No confirmed two-way HF QSO; UHF only to date |

| UHF multi-user under HF-equivalent fade (sim) | **LIMITED → improved** | Offline Watterson scaled to UHF symbol (`uhf-multipath` congested stack). At 1.5 s arrival span: **2/2×4**, **4-user mean 2.5/4 (10/16)**, 8-user seed0 **3/8**. At **8 s** arrival span: **4-user 15/16 (mean 3.75/4, full 75%)**. Same protocol lesson as HF seed-21: stagger matters. Artifacts: `run-artifacts/uhf-hf-emu/uhf-multipath-hf-equiv-remeasure.json`, `uhf-multipath-hf-equiv-span8.json` |

## Interpretation

- "YES" entries are safe to cite in release notes.
- "LIMITED / NO" and "PARTIAL" entries describe active development areas; they
  are not blockers for alpha distribution but must not be overstated.
- This matrix is the authoritative claim surface for v0.1.0a7. Update it before
  any subsequent release.