The Triangulum Galaxy

Photography notes

267 of 342 frames of 15 s, from two nights. 1.1 hours of light from two nights, the camera moved a little between frames.

Equipment

Telescope

Celestron 8SE Schmidt-Cassegrain: 203 mm aperture, 2,082.9 mm focal length by plate solve of this stack, f/10.3

Camera

Sony a6000, stock, at prime focus: APS-C 23.5 x 15.6 mm, 6000 x 4000, 3.9 micron pixels, RAW

Mount

Sky-Watcher EQ6-R, unguided. Set down about 10 degrees off the pole tonight; aligned from plate-solved photos, both axes driven from that model. Dithered between frames for most of tonight

Acquisition

Exposures

267 x 15 s at ISO 3200: 66.8 minutes. Tonight 208 frames, 52.0 minutes in three turns; 13 October 59 frames, 14.8 minutes

Frames

342 taken, 267 kept. Tonight, held still: 53 of 63; dithered: 26 of 30; after Saturn, dithered: 129 of 143. 13 October: 59 of 106. Dropped tonight: 21 smeared (every star stretched the same way; 7 of them the first frame after a dither move, one when someone was at the telescope), 2 with soft stars, 3 the first frame after a slew, 2 taken too soon after a move. Dropped on 13 October: 43 trailed or soft (the tracking let go now and then until it was fixed during that run), 3 the first frame after a slew, 1 too soon after a move

Dithering

Tonight's first turn held still (the field drifted about 2.5 arcmin). The second moved between frames in a spiral up to 1.5 arcmin. The turn after Saturn moved every 3 frames, small probe steps and then up to 4 arcmin from the start (the last place nearly 5): a bug in the sensor map chose the 4, and it was kept because it widened the field the frames cover. 13 October was not dithered

Sampling

0.776 arcsec per pixel: one pixel per 2 x 2 colour cell, never enlarged. Stars 4.5 arcsec wide (half-flux) in the stack, 4.9 in the reference frame; the frames' median 4.2 tonight, 3.6 on 13 October

Field

36.3 x 21.4 arcmin: the area that at least half the frames cover. Inside it 63 percent of the pixels have every frame and 94 percent at least three quarters of them; at the edges it falls to 50 to 53 percent. Not turned (north 64 degrees anticlockwise of up)

When

The night of 17 October 2026, in the evening and after midnight, either side of Saturn; and 13 October 2026, after midnight. About 50 degrees up in the evening, about 80 after midnight

Darks

None taken. Hot pixels found from the frames, one map per night (about 20,000 to 22,600 per colour plane tonight, 5,000 to 6,900 on 13 October), repaired by the 3 x 3 median of their colour plane

Flats

No flat frames. The hybrid flat of 13 October (large scale from that night's open sky, dust from its twilight frames), the flat tonight's Dumbbell Nebula run chose. On this run's dithered sky, the galaxy's middle masked, it leaves less than the twilight flat alone: 1.3 against 2.8 percent in green after a plane, galaxy light included. The galaxy fills the frame, so no large scale could be built or judged from these frames

Calibration

Dust that moved since 13 October, measured on each night's own frames (each frame against the first stack put back on its sensor grid, so the galaxy cancels) and divided out. Tonight 10 shadows, each seen in the first turn and the turn after Saturn alone, 0.77 percent of the sensor; among them the new crescent, 11.6 percent deep after the 13 October flat, left at 0.95 percent. 13 October: 19 shallow ones (1.5 to 3 percent), from its odd and even frames, a weaker test because those frames were not dithered

Processing

Registration

Each frame placed first by its own plate solve (all within 9 sensor px of the star solution), then rotation and shift on the stars, weighted rms at most 0.84 px. The field turned 10.2 degrees across all the frames, 8.7 of it over tonight's turns, because the mount stood about 10 degrees off the pole

Stacking

RAW colour planes (no demosaic), hot pixels by 3 x 3 median, registered with rotation, then a 3-sigma clipped mean weighted by each frame's clarity and noise. Not done, for the time limit: proper coaddition (1 to 4 percent more signal to noise on stars from 15 s frames), and lining red and blue up with green against the air's prism. One constant per colour per frame, set on the darkest 30 percent of the area every frame covers: the galaxy's faintest level, not the true sky; no surface fitted. Noise 13.7 times lower than one frame (green); with 13 October's frames 10 percent lower than tonight's alone

Colour

The camera's colour matrix with its white balance set by 44 Gaia stars in the field (red x 1.01, blue x 1.14). Colour only where the light is bright enough to carry it, 0.5 percent of the picture; the faint disc is grey

Finish

Star colours from Gaia, the camera's colour matrix. Arcsinh stretch; black at the faintest part's own level, gamma 0.9, S-curve 0.3, saturation x 1.25; the dark parts smoothed (bilateral 3 px). 0.2 percent of the picture is exactly black

Picture size

2804 x 1652 pixels

Software

Capture and mount: Observatory (Elixir on a Raspberry Pi 3). Stacking: numpy, OpenCV, rawpy. Plate solving: astrometry.net. Stacked from the camera's RAW frames. Nothing generated.

What the picture shows

What is real

The arms and clusters at every distance from the core: the odd and the even frames, each stacked alone, agree in 84 of 84 tiles of 3 arcmin, out to the corners 18 to 30 arcmin from the core; and tonight's frames against 13 October's, two nights with two dust corrections, agree in 84 of 84

Not seen

The galaxy's far outer arms: they lie beyond the frame. The true sky level: there is none in the frame

Honest limits

The tests compare structure at the scale of arms and clusters, not the smooth glow of the disc; with no blank sky in the frame they cannot say how far the faintest light stands above the true sky

Everything above as Markdown, ready to paste into a forum post.