Sky transparency — night of 31 August – 1 September 2026

Normalised transparency index, by filter

100 = that target and filter's own clear‑sky median (this night's own pre-cloud median). Shaded columns are autofocus runs; the first target is named at the top left and dashed vertical rules mark each target change; ticks along the base are errors.

All‑sky background brightness

Mean luminance of the upper sky per all‑sky still — independent of anything N.I.N.A. recorded. The camera auto‑exposes, so this is an instrumental number, not photometry.

Roof shut 01:21–05:04, 06:46–06:55. The all-sky camera is inside the observatory, so those 659 stills measured the roof, not the sky, and are left out of this chart and the cloud detection.

PHD2 guiding

RMS tracking error per minute from the PHD2 guide log, taken about zero rather than about each minute's mean, so slow drift counts as error. 2870 guide frames over 5.2 h of guiding in 14 sessions; the gaps are slews, dither settles and autofocus runs, when PHD2 was not guiding. PHD2 discarded 175 further frames — usually a lost star. The worst minute (3.27″ at 00:54) is clamped at the top of the plot.

All‑sky frames at the key moments

Times come from the still filenames, so they line up with the charts exactly.

Clear‑sky reference medians

The divisors behind the index — median detected stars per light sub before 01:12.

Event timeline

Only the post-slew event predates the cloud — PHD2 failing to settle on a new guide star after a slew, not a sky problem.