The Wizard Nebula (NGC 7380), HOO with true-colour stars

The night: a wrapped power cable

The Origin runs one-hour sessions so it re-centres, and between sessions it swings back toward the pole and out again. On this target that swing wound the power cable round the mount, and three sessions near the end were dragged off target. Every sub was graded (tools/origin_sub_qc.py): star shape, star count against the whole night’s median (two sessions were bad from start to finish, so a per-session median would have hidden them) and the pointing in each header. 49 culled — 34 of them more than 10′ off target, and the rest with no stars, streaked, or thin. The culls are kept, not deleted.

Why HOO, not SHO

SII (672 nm) and Hα (656 nm) land in the same red band on the same red pixels, so they can’t be separated. HOO is the most this filter can honestly give.

Stacking in Siril

1. Each session calibrated on its own with the matching 15 s dark (20.8 °C, against 21–24 °C lights) and a location-based bad-pixel map.
2. Split before any debayer: red pixels → Hα, green + blue → OIII (seqextract_HaOIII). Half-size frames, each pixel a 2×2 block — 4.0 µm effective, 2.46″/px.
3. The same calibrated subs split a second way for the stars: R, G, G, B as four separate planes on the same 2×2 grid, stacked as a true-colour star master. HOO stars can only be red, cyan or white; these keep their real colours.
4. Six sessions merged with Siril star registration, one reference for all channels. Row and column banding from the Origin’s small dithers removed on the linear masters (9%/8% → 4%/3% of the noise). Cropped to the area every session covered, 1910 × 1061.
5. A one-pass stack of all 777 subs was tried against the per-session merge: no sharper (4.51 vs 4.53 px) and noisier, so the merge stayed. The softness is in the data, not the stacking.

PixInsight — linear

ChannelCombination R = Hα, G = B = OIII; ImageSolver at 4 µm / 335 mm; SPFC narrowband; MultiscaleGradientCorrection with the MARS survey (it works for HOO — MARS has no SII, so SHO couldn’t use it); SPCC narrowband 656.3 / 500.7 / 500.7 nm at 25 nm; BlurXTerminator correct-only then stars 0.50; MLDenoise 0.90; and an OIII lift while linear, ($T-med($T))*1.3+med($T) on green and blue — the med() form lifts the nebula without tinting the sky.

PixInsight — non-linear

GHS stretch (factor 6.27, local intensity 0.05), black point 0.058; BlurX non-stellar 0.80; StarXTerminator to split the stars off; HDRMultiscaleTransform (6 layers) on the starless nebula; three curves; a second BlurX non-stellar pass at 0.80.

The stars

From the true-colour stack: StarXTerminator, BlurX on the stars (0.50), a PixelMath stretch ((3^5.04)*$T)/((3^5.04-1)*$T+1), ColorSaturation, and MultiscaleMedianTransform with layers 1–2 off to trim the swarm of small stars the Origin’s wide field produces. A HistogramTransformation black point then removed faint coloured “worms” — noise that StarX had carved into squiggles. Screened back onto the starless image: ~(~starless*~stars). Cropped to 1655 × 1031.

Result

Highlights at 194 of 255, sky 12/7/7, 0.04 % clipped; true-colour stars (orange, white, blue) over an HOO nebula.