From TESS Sector 94, QLP eclipse depths of 49 southern CVZ binaries are 11% smaller in the 3x3 aperture and 36% smaller in the 1x1 aperture than before, while SPOC depths do not change

The TESS Quick-Look Pipeline (QLP) changed from FIPHOT circular-aperture photometry to TGLC photometry with 1x1, 3x3 and 5x5 pixel square apertures at Sector 94 (S94). We measured the primary eclipse depth of 49 detached eclipsing binaries in the southern continuous viewing zone (ecliptic latitude below -78 deg, T = 6.6-13.0) in QLP light curves of S87-S93 and S94-S98. SPOC 2-minute light curves of the same stars and sectors are the control. The median depth ratio (S94-98 / S87-93) is 1.004 for SPOC. For QLP it is 0.888 in the 3x3 aperture (45 of 49 stars below 1, sign test p = 8e-10), 0.635 in the 1x1 aperture (16th-84th percentile 0.43-0.86, p = 4e-12) and 1.006 in the 5x5 aperture (p = 0.15). The step occurs at S94 and stays in all five later sectors. Before S94, the QLP 3x3 depth agrees with SPOC to within 1%. The raw QLP flux has the same step (0.897), so the detrending does not cause it. Four known hot Jupiters have 3x3 ratios of 0.87-0.91. Without a correction, their radii from S94+ QLP depths are about 6% too small. After S94, the 1x1/5x5 depth ratio is below 0.5 for 43% of the on-target eclipses, so this ratio is not a valid on-target test. Our own box photometry on TESScut images of S93-S95 gives depths within 3.1% of SPOC in all three apertures (3x3: 1.001-1.007). Thus the image pixels did not change, and the cause is in the QLP processing. The QLP depth factor is not equal to the fraction of the target light in each aperture (p < 1e-9). The data agree with a constant extra flux in each aperture after the background subtraction: +17% (1x1), +10% (3x3) and -3% (5x5) of the target flux. For the 1x1 aperture this extra flux increases with the distance of the star from the pixel centre. The TGLC code sets the median aperture flux to a model value, and this agrees with the data. We did not run TGLC, so we do not identify which model term is wrong. Limits: one sky region, stars with T < 13, five sectors on each side of the change, and QLP v01 files as downloaded in October 2026. For depths from S94 onward, use the QLP 5x5 aperture (DET_FLUX_LAG) or SPOC data. The package has the code, the measured depths of each star and sector, the file lists and the reproduction instructions.

TESS · Quick-Look Pipeline · QLP · TGLC · eclipsing binaries · eclipse depth · transit depth · photometric bias · aperture photometry · southern continuous viewing zone · SPOC · TESScut · data validation · hot Jupiters · planet radius