Quadratic background fits give false asymmetry changes in symmetric eclipse tests at DI Herculis sample times

We test one fixed odd flux component in four TESS secondary eclipses of DI Herculis. We use public 120-second data from 2020 and 2024. Each fit has a local background, an even spline, a free center, and one analytic odd function. The main cubic fit gives a 2024 minus 2020 contrast of +203 parts per million (ppm). Archive errors give a local standard error of 90 ppm. A residual scale correction increases this to 144 ppm. Cubic and quartic fits give contrasts from -63 to +479 ppm across the tested windows and flux columns. We also test a symmetric profile on the actual sample times. The data do not define this profile. A sine background with 1% amplitude and a 1.07-day period gives false contrasts up to 2,799 ppm with quadratic fits. The eclipse profile is even. We add no noise. The largest false contrast falls below 175 ppm with cubic or quartic backgrounds in the tested phase pairs. Tests include signal recovery, a separate center search, and 1,280 simulated four-event data sets with five noise rules. We also save local response weights for other flux variation periods. The observed data do not establish a stable change in this component. The tests show a background fit failure. They do not identify the cause of the observed background. The package contains source FITS files, public URLs and hashes, code, selected flux, all results, simulation draws, figures, and reproduction commands.

DI Herculis · TESS · secondary eclipse · eclipse asymmetry · stellar variability · background model · signal recovery · public data · linear response · model error