RexySDR daily RF report — 2026-10-05

Rolling 24-hour evidence window; see source limitations

Historical report for 2026-10-05; this page is not a live observation. Values are relative and uncalibrated. Source limitations remain authoritative; complete capture coverage is not established.

Observation status

- **P19 eight-band series:** 247 points at 5-minute cadence vs 289 expected for a contiguous window — 42 samples absent, all from one 215-minute gap (last point 2026-10-03T23:10Z, next 2026-10-04T02:45Z). Latest observation id 18638 at 20:35Z, completeness 8/8 bands, freshness 117 s.

- **P20 panorama:** one observation in the current block — id 261004_203230, observed 2026-10-04T20:32:30Z, completeness 1.0, 12,500 points, 13.1 s scan, 5 kHz spacing over 5 kHz–62.5 MHz; 12,431 valid bins, 69 DC-hole bins excluded; all artefact hashes verified; A and B share identical valid-bin masks.

- **Historical P20 evidence:** present but sourced from an *earlier* observation, id 261004_200231 (20:02:31Z), with only 5 history observations and `history_maturity=INSUFFICIENT`.

- **FT8 logger:** 1,077 recorded minutes, 1,073 active, 16,089 declared decodes, spanning the full window.

- **Environment context:** solar geometry (elevation +8.3°, azimuth 92°, 20:35Z), BOM space-weather (status OK, cache hit), Open-Meteo grid-cell weather — all labelled by the contract as context, not measurements at the site or proof of anything.

Eight-band RF behaviour

Per-band raw levels across the window (measured; medians of 5-min samples, dB relative):

|---|---|---|---|---|

| 160m | -80.0 | -65.5 | -54.9 | -54.9 → -67.5 |

| 80m | -80.7 | -78.7 | -62.3 | -63.1 → -80.5 |

| 40m | -73.8 | -71.9 | -62.0 | -63.5 → -72.6 |

| 30m | -79.6 | -78.6 | -68.1 | -75.2 → -79.0 |

| 20m | -80.1 | -78.8 | -68.2 | -79.1 → -78.6 |

| 17m | -76.6 | -75.3 | -59.5 | -61.6 → -73.7 |

| 15m | -84.3 | -82.3 | -63.7 | -65.3 → -83.6 |

| 10m | -91.8 | -90.8 | -80.2 | -80.2 → -91.7 |

Deterministic segment medians (dB relative):

| Segment (UTC) | 160 | 80 | 40 | 30 | 20 | 17 | 15 | 10 |

|---|---|---|---|---|---|---|---|---|

| Oct 3 20:35–23:10 (before gap) | -57.0 | -63.1 | -63.5 | -75.1 | -78.8 | -61.8 | -66.0 | -88.5 |

| Oct 4 02:45–05:50 (after gap) | -79.5 | -80.2 | -72.4 | -78.6 | -78.3 | -76.1 | -82.4 | -90.8 |

| Oct 4 05:55–06:55 | -66.0 | -79.3 | -72.4 | -78.4 | -78.2 | -73.4 | -83.8 | -90.5 |

| Oct 4 07:00–10:10 | -65.0 | -74.7 | -63.2 | -68.7 | -68.5 | -66.5 | -79.6 | -86.3 |

| Oct 4 10:15–13:00 | -65.7 | -78.7 | -70.3 | -78.6 | -79.1 | -75.6 | -82.3 | -90.8 |

| Oct 4 13:05–20:35 | -65.5 | -78.8 | -72.1 | -79.0 | -79.0 | -75.5 | -82.4 | -90.9 |

Descriptive reading (measured facts + deterministic aggregation): the evening of Oct 3 carried the window's highest 160/80/40/17m levels; samples after the gap sit 10–22 dB lower on most bands; a 40/30/20/17/15m elevated interval spans 07:00–10:10Z with a stepped return to baseline at 10:15Z; 10m stayed at or near the bottom of its range through the last ~15 h. The before/after-gap contrast is a transition **across missing data**, not a measured 5-minute step. Model interpretation only: the elevated 07:00–10:10Z interval is where several mid-bands moved up by roughly 6–11 dB relative to adjacent plateaus; the contract does not permit attributing these movements to any mechanism.

P20 wideband evidence

Single 13.1 s snapshot at 20:32:30Z; no temporal-change claims are possible from one observation. Power A and Power B are treated strictly as two parallel measurement channels — their physical meanings are not defined by this package.

Global statistics (deterministic):

- **Channel A** (12,431 valid bins): median -70.8, p90 -54.4, p99 -43.3, peak -26.9 dB_rel at 5.985 MHz (43.9 dB above global median); MAD 3.4 dB; 4,353 bins (35.0%) above median+6 dB. Twelve strongest peaks cluster at 1.06–1.325 MHz, 5.935–6.1 MHz, and 9.48–9.525 MHz.

- **Channel B**: median -72.3, p90 -63.3, p99 -48.2, peak -33.9 dB_rel at 85 kHz (38.4 dB above median); MAD 1.9 dB; 2,225 bins (17.9%) above median+6 dB. Its strongest peaks sit below 300 kHz, with additional peaks near 14.9–15.1 MHz and 35.905 MHz.

- **A−B (deterministic):** median +1.04 dB, p10 -3.9, p90 +16.6, MAD 3.0 dB.

FT8-neighbourhood windows (±100 kHz analysis windows, not allocations):

|---|---|---|---|---|

| 160m | -49.7 | -70.9 | -40.1 @ 1.840 MHz | 5 / 2 |

| 80m | -59.9 | -73.5 | -56.1 @ 3.595 MHz | 0 / 0 |

| 40m | -52.2 | -74.0 | -50.3 @ 7.055 MHz | 0 / 0 |

| 30m | -58.7 | -74.1 | -57.1 @ 10.225 MHz | 0 / 0 |

| 20m | -55.7 | -57.9 | -52.3 @ 14.095 MHz | 0 / 0 |

| 17m | -53.6 | -62.5 | -51.7 @ 18.165 MHz | 0 / 0 |

| 15m | -63.4 | -67.2 | -61.0 @ 21.055 MHz | 0 / 0 |

| 10m | -71.3 | -73.3 | -69.7 @ 28.125 MHz | 0 / 0 |

Measured fact: the 160m window is the only window with any bins above local median+6 dB in either channel (5 in A, 2 in B). Deterministic observation: in the 20m and 10m windows channels A and B sit within ~2 dB of each other, versus ~9–22 dB separation elsewhere; the contract assigns no meaning to this.

Historical P20 evidence

`historical_p20` is available (archive `p20-historical-evidence.json`, `rexysdr-ai2-historical-evidence-v1`, generated 20:07:31Z) and is anchored on observation **261004_200231 (20:02:31Z)** — a different, 30-min-earlier panorama observation than the current-window P20 block above.

Reference availability: previous (19:32:31Z) available; ~one-hour-earlier (19:02:30Z, ±40 min tolerance) available; **previous-day same time unavailable** (±90 min tolerance, target 2026-10-03T20:02:31Z). Cadence `valid_for_sequence=true` for every band with a 30.0-minute previous gap, so short-sequence descriptions are permitted; day-over-day description is **not** — the prior-day reference is missing and history holds only 5 observations with maturity `INSUFFICIENT`.

Deterministic band deltas (current 20:02Z obs vs previous / vs one-hour; dB):

|---|---|---|---|

| 160m | -0.78 / +0.77 | -0.83 / +0.26 | A p99-above-median **-7.08 / -5.68 dB**; A active-fraction +0.049 (≈2 bins of 41), B -0.024 |

| 80m | -0.95 / -0.59 | -0.76 / -0.46 | A p99 excess -2.70 dB vs prev |

| 40m | -1.32 / +1.06 | -0.33 / -0.32 | — |

| 30m | -0.78 / -0.16 | -0.10 / -0.06 | — |

| 20m | -0.03 / -0.20 | +0.45 / +1.89 | — |

| 17m | +1.92 / +1.23 | +1.31 / **+5.32** | largest B-median movement in the table |

| 15m | -1.91 / -1.52 | -2.13 / -2.16 | both channels' medians down vs both refs |

| 10m | -1.60 / -1.14 | +0.01 / +0.18 | — |

Cross-observation deterministic calculation (two supplied observations, 30 min apart): 160m window A median moved from -48.87 dB_rel (20:02Z) to -49.71 dB_rel (20:32Z), and the A peak frequency from 1.760 MHz to 1.840 MHz. Persistence: `bands_with_active_persistence=[]`; every band's persistence record is `active=false, count=0, reason=no_ge_6db_step_in_latest_sequence`. Screening flags: empty for all bands.

Shadow attention scores (operator-screening values only — not probabilities, not anomaly scores, not RF classifications, per the contract guardrails): 160m 22.82 (**MEDIUM**, highest); 15m 14.43, 80m 13.15, 17m 12.60, 10m 8.45, 40m 6.29, 30m 5.58, 20m 4.77 (all LOW). The 160m MEDIUM attention corresponds deterministically to its p99-excess reduction and small active-fraction change — but `anomaly_classification_permitted=false` and `source_identification_permitted=false`, so this remains a description of numbers, not a finding.

FT8 observational evidence

Contract role: observational evidence only; FT8 is not propagation and not a causal explanation. Per-band decode counts over the window (measured):

|---|---|---|---|---|---|

| 20m | 8,260 | 931 | 652 | -7.48 | -16 … +16 |

| 40m | 2,241 | 620 | 148 | -9.20 | -15 … +14 |

| 15m | 2,137 | 425 | 294 | -9.40 | -16 … +14 |

| 17m | 1,304 | 444 | 172 | -9.34 | -16 … +11 |

| 30m | 1,274 | 564 | 87 | -9.63 | -17 … +9 |

| 10m | 495 | 248 | 44 | -9.59 | -16 … +9 |

| 80m | 371 | 167 | 34 | -8.70 | -17 … +2 |

20m carried 51.3% of declared decodes (deterministic share). Count note (deterministic): the seven listed bands sum to 16,082 decodes vs the declared 16,089 — a 7-decode remainder not attributed in the supplied `band_activity` list (no 160m entry appears); no cause is asserted.

Recent decode sample (20:29–20:35Z): VK2LAW (grid QF56) repeatedly on 20m at -5…-13 dB; K8ST (EM79) 20m -11; EA8UP (IL18) 40m -11; YC2BUZ (OI43) 40m -6…-12; 3Z4DX (JO90) 30m -11…-12; VK4LW (QG62) 30m -13…-14. Reported grids are decoder-attached fields; per the rules, nothing about path quality, distance, or propagation mode is inferred from these SNR/grid values.

Historical-reference status

D1 (P19) historical reference for the 20:35Z anchor (±30 min half-window): 16 cycles found, **9 qualified** (2026-09-20, 09-22, 09-23, 09-25, 09-27, 09-28, 09-29, 09-30, 10-03), 7 rejected. Rejection reasons: `minute_bins_lt_45` + `centre_missing` + adjacent-window bin shortages in six cycles; `cycle-20261001-2035` failed only on `right_bins_lt_20` despite centre present. One cycle is far outside the recent block (2025-07-23, fully rejected).

Deterministic status per contract fields: `maturity_status=UNQUALIFIED` — the Q≥4 threshold "has not yet been qualified as sufficient"; `rf_state=INSUFFICIENT`; `anomaly_status=INSUFFICIENT`; `classification_permitted=false`. Location scope: history is keyed by station_id and time, **not** by test-site coordinates, so it is not location-stratified. Conclusion permitted by the contract: the reference set exists but cannot support any classified or mature historical judgement for today's bands; the cadence/references are insufficient, and this is stated explicitly rather than worked around.

Notable measured changes

Measured facts / deterministic calculations only (relative dB):

1. **Coverage gap dominates the series shape:** no P19 samples 23:10Z–02:45Z. The first post-gap samples are 10.1–18.1 dB lower than the last pre-gap samples on 160/80/40/15/17m — a before/after contrast across missing data, not an observed step.

2. **07:00Z stepped rise in mid-bands (5-min steps):** 20m +8.6, 17m +9.2, 40m into a 30-min elevated interval (-63.2 median vs -72.4 prior), alongside 30m and 15m; sustained 07:00–10:10Z.

3. **10:15Z stepped fall:** 20m -9.5, 30m -8.4 in one sample interval; 80m rose ~4 dB to its -78.7 plateau at the same time.

4. **160m isolated excursions:** single-sample +11.0 dB rise at 04:00Z (reverted next sample); a +11.1 dB rise at 05:05Z sustained ~45 min at ≈-67 to -68 before settling to the -65 plateau.

5. **P20 160m tail reduction (historical obs):** A p99-above-median -7.1 dB vs previous and -5.7 dB vs one-hour, with A active-fraction +0.049 — opposing deterministic movements in the same window; no classification applies.

6. **17m B-channel median +5.3 dB vs one-hour-earlier reference** (P20 window statistics).

7. **15m both-channel median decrease of ~1.5–2.2 dB vs both references** — the only band moving consistently down in both channels.

8. **10m window level:** P19 10m median -90.8 for the window with late-window values at the series minimum (-91.0 to -91.8); this is a measured level, not a "quiet band" claim.

Cross-source observations

- **Timing:** P19 20:35Z sample, P20 scan 20:32:30Z (historical block: 20:02Z), and FT8 decodes to 20:35Z are near-simultaneous end-of-window snapshots; comparisons are cross-sectional, not longitudinal.

- **160m agreement (deterministic):** the band with the highest P19 window medians (evening -57) also holds the only P20 window with active bins and the largest A−B window separation (~21 dB); its shadow attention is the only MEDIUM. Coherence across sources, no classification implied.

- **20m contrast:** 20m shows the heaviest FT8 decode load (8,260 decodes, 652 unique calls, mean SNR -7.5) while P19 20m medians stayed near -79 outside the 07:00–10:10Z interval, and its P20 window had zero active bins and the smallest A/B median separation after 10m. High decode counts coexist with low window-aggregated levels; the contract forbids reading either as the other's cause.

- **Environment alignment (context only, interpretation):** the elevated mid-band interval (07:00–10:10Z) and the post-gap low plateau fall within the window of a BOM Australian-region magnetic **warning** (G1 range, Oct 3–4, cause "coronal hole"; Kp 4 at 18:00Z, Dst -42, A 5) — space-weather indices are regional context and explicitly not a local geomagnetic measurement or proof of RF causation.

- Solar elevation +8.3° (above geometric horizon, azimuth 92°) at 20:35Z is a calculated geometric value for the grid-centre coordinates; meteorology (16.6 °C, mainly clear, 29% cloud) is grid-cell model context.

What cannot yet be concluded

- **Cause of any level movement** — the gap-straddling contrast, the 07:00–10:10Z elevated interval, the 15m/10m declines, or the 160m tail change cannot be attributed to propagation, solar geometry, geomagnetic activity, local equipment, or interference. Environmental fields are context; FT8 is not a causal explanation; the contract's `environmental_role` and `rfi_rule` both forbid this.

- **Anomaly or RFI status of anything**, including the MEDIUM-screened 160m P20 window: classification permissions are all `false` in both `interpretation_constraints` and `classification_flags`, and no corroborating evidence is supplied in this package.

- **Day-over-day change** in either source: the previous-day same-time P20 reference is unavailable, D1 maturity is UNQUALIFIED, and P20 history holds 5 observations with maturity INSUFFICIENT.

- **Behaviour during the 215-minute P19 gap**, including whether the 23:10Z→02:45Z contrast was gradual or stepwise.

- **Temporal evolution from P20**: the current block contains one 13.1 s scan; the historical block covers 20:02–20:32Z only at 30-min cadence.

- **Physical meaning of Power A vs B**, absolute power, transmitter distance, station proximity, path quality, or source identity — none defined or permitted from these fields.

- **Nature of the 7-decode remainder** between the declared FT8 total and the band-list sum.

- **Whether the 160m p99 reduction or the 17m B-channel rise will persist** — each rests on two reference points within a two-hour span.