LIGHT IS A MEASUREMENT, NOT A BRIGHTNESS.
One star. A crowded sky.
Measure the crosshair-marked reference star within 7% of its known flux and signal-to-noise ≥70. Center the aperture, capture its faint wings, and choose clean background pixels. Neighboring stars can spoil the measurement.
Detector frame
How the pixels become a measurement
The GPT-generated starless nebula supplies the background pixel intensities. Known synthetic Gaussian stars are added to that raster. The image is not a real observation. Source and annulus weights use Photutils’ 4×4 subpixel circle-overlap method. Masked pixels are excluded from weighted flux, area and squared-weight variance sums. Sky mean is subtracted from source flux; its uncertainty is propagated. Model variance is photon counts plus a 4-count read-noise standard deviation; noise is not randomly sampled in these teaching frames.
Aperture and sky radii must not overlap. Missing wings, a masked target, or a contaminated sky region can bias flux even with a large signal-to-noise value. Real calibration also needs detector corrections, PSF characterization and uncertainty validation.
Source ports: astropy/photutils, BSD-3-Clause, commit e2a3c8d9, October 29 2026; 315 observed stars. Original exposure generator, subtraction composition and interactive game. License.