Observation 14745664

PE0SAT-31OBJECT JNORAD 63218BPSKlabel: with-signalqueue rank 16

the fitted corridor, the path the matched filter scored, with its half width shaded. the same pass geometry at zero frequency offset. the commanded receive frequency, where a Doppler-corrected capture would sit. The gap between the first two is the measurement. Scale: 127.7 Hz per pixel across, 0.316 seconds per pixel down, drawn one screen pixel to one measured pixel.

Black point 0.05White point 0.85Gamma 1.00

The black and white points open at a fixed pair, not one fitted to this image. Across the 25 committed waterfalls the noise floor runs from 0.20 to 0.38, so on most observations the ramp starts well below the first real sample and a weak pass reads flat until the black point is raised toward the floor.

PaletteOverlay

Reviewer note

deterministicUNGROUNDED_NUMBER · a number not in the evidence packet

Ranked 16 with score 0.566, flagged NO_REASON. The network label is with-signal and the model probability is 0.438. The corridor sits -7149 Hz from the catalogue centre, -16.4 ppm, with a half width of 2000 Hz; the curved fit scores 1.5 against 0.5 for a vertical line. Peak elevation was 30.3 degrees at 0.50 of the pass.

A generated note was refused for UNGROUNDED_NUMBER, so the deterministic summary is what is shown. The refusal is the measurement, not a fallback that failed quietly.

Check a claim about this observation

The grounding checker from the pipeline, ported to TypeScript and running here. Edit the sentence, or start from one of the drafts below, and the verdict changes as you type. Nothing is sent anywhere: the whole rule set is in the page.

The sentence to check

322 of 700 characters

GROUNDED

Every number in this sentence is a token of the packet below, every code and label in it is this observation’s, and no permission rule fired. That is all the checker claims: the sentence is consistent with what was measured, not that the measurement is right.

The evidence packet this is checked against, exactly as the model was shown it

observation_id             : 14745664
ground_station_id          : 1696
ground_station_name        : PE0SAT-31
norad_catalogue_id         : 63218
transmitter_mode           : BPSK
receiver_frequency_hz      : 436400000
network_label              : with-signal
pass_duration_s            : 489
max_elevation_deg          : 30.3
closest_approach_fraction  : 0.50
fitted_offset_hz           : -7149
fitted_offset_ppm          : -16.4
corridor_half_width_hz     : 2000
sigma_curved               : 1.5
sigma_vertical             : 0.5
hz_per_pixel               : 127.7
seconds_per_pixel          : 0.32
axis_derivation            : axis_ticks_ocr
axis_derivation_confidence : 0.92
model_probability          : 0.438
ensemble_uncertainty       : 0.0091
flat_row_fraction          : 0.00
queue_rank                 : 16
queue_score                : 0.566
queue_reason_codes         : NO_REASON
offset_at_bound            : false

The closed world. A number that is not a token of this text is not grounded, which is why 6490 is refused although those digits appear inside 436490000: containment is not membership, and a transposed offset passed the old check for exactly that reason.

The pass

The same propagation that produced the corridor above, drawn from the station's point of view and from orbit. Elevation is measured from the WGS-84 geodetic normal, which is the reference the corridor was scored against.

Sky track

Azimuth clockwise from true north, elevation from the horizon at the rim to the zenith at the centre. The open circle is the rise, the square is the set, and the filled mark is closest approach at 30.3°.

Ground track

The subsatellite point, the station as a cross, and the satellite’s horizon circle at closest approach from 468 km. There is no basemap: this console ships no coastline data and will not draw one it cannot cite.

Elevation and Doppler against time

The same pass on a time axis. Two stacked panels rather than one panel with two vertical scales, because a twin-axis chart lets the author choose where two curves appear to cross. This window spans closest approach, and the Doppler zero crossing lands on the sample where elevation peaks, because both happen at the same instant.

Every row is a field of the exported geometry, not a figure read off the plots.
Quantity	Value	Where it comes from
Maximum elevation	30.286°	Highest sample of the propagated pass, from the geodetic normal
Azimuth at closest approach	292.19°	Clockwise from true north
Slant range at closest approach	850.1 km	Station to satellite, the minimum over the pass
Closest approach, as a fraction of the pass	0.5029	0 is the start of the recording window, 1 the end
Samples propagated	512	An error count above zero means the orbit model refused part of the window, and the track is drawn only where it did not

The fit

Frequency offset

-7149 Hz

-16.4 ppm of the catalogue downlink

Corridor half width

2000 Hz

15.7 pixels at this image scale

Max elevation

30.3°

pass geometry from the TLE that was current at capture

Closest approach

50%

through the observation window, where the Doppler curve is steepest

Matched-filter sigma at every whole-pixel offset the fit was allowed, subsampled for transport with the peak and both endpoints always kept. The peak is the fitted offset: the same function produces both, so they cannot disagree. The curve is what makes the fit a measurement rather than a choice, because a corridor that were merely near the trace would give a flat sweep with no peak to find. Scored at 341 offsets, 171 plotted.

The fitted curve is the predicted Doppler curve shifted by the fitted frequency offset, which is the path the matched filter scored.

The predicted curve is the same shape at zero offset, so the gap between them is the measurement.

Time runs bottom to top: the first row is the end of the pass.

The frequency axis runs against the Doppler sign.

Why it is in the queue

The composite score and the criteria it met, from the queue receipt.

Field	Value	What it is
Rank	16	inside the review budget of 50
Composite score	0.565976	rank-normalised blend of model disagreement, uncertainty and physics
Model p(signal)	0.437500	calibrated probability from the shipped image and corridor arm
Ensemble spread	0.009058	disagreement between seeds; a proxy for how sure the arm is
Flat rows	0.0000	fraction of the waterfall with no luminance variation
Episode	1696:63218:11	station, satellite and orbital revolution: the unit the interval is grouped on
Conflict	false	No conflict criterion

Precedent

The five nearest passes in the snapshot under the Granite embedding of a text card built from this pass's numbers.

Two conditions: any station, and then a condition that excludes this station and this satellite so a neighbour cannot be a near-duplicate of the query.

The measured agreement is on the precedent page, and under the second condition it does not beat chance.

Condition	Neighbour	Station	Satellite (NORAD)	Start	Network label
Any station	14745663	1696	63237	2026-08-09T23:20:32Z	with-signal
	14745659	1696	63215	2026-08-09T22:33:46Z	with-signal
	14745657	1696	63217	2026-08-09T22:10:36Z	with-signal
	14745929	1696	63217	2026-08-09T23:43:40Z	with-signal
	14745661	1696	58340	2026-08-09T22:59:49Z	with-signal
Other station, other satellite	14743674	431	60240	2026-08-09T15:15:12Z	without-signal
	14743911	431	98323	2026-08-09T16:21:57Z	with-signal
	14745607	431	63238	2026-08-09T23:01:25Z	with-signal
	14744128	431	52899	2026-08-09T16:31:50Z	without-signal
	14744444	431	27848	2026-08-09T17:27:00Z	with-signal

This pass carries the network label with-signal, station 1696, satellite OBJECT J (NORAD 63218). Counting how many neighbours match that label is one draw and settles nothing on its own. The neighbours link to the network rather than to this console, because most of them are outside the shipped set and the console carries no imagery for them.

Provenance

Enough to fetch the original and check it.

Field	Value
Source	network.satnogs.org/observations/14745664
Original waterfall SHA-256	625c26871c577683a16570f01bdc07d4bd8e7c2d602bc44af625f7a7718a463e
Window	2026-08-09T23:32:49Z to 2026-08-09T23:40:58Z
Receive frequency	436,400,000 Hz
Axis derivation	read off the spectrogram's own tick labels (confidence 0.92)
Displayed image	unweighted mean of the R, G and B channels, 8-bit