A browser-based astronomy tool agreed with reference rotation periods for 1,228 of 1,423 TESS light curves it classified as periodic, a recovery fraction of 86.3%, according to a preprint. Across the full comparison, however, the agreement rate was 3,200 of 5,439 sources, or 58.8%. The first figure comes from a classifier-defined periodic subset, while the second comes from the full heterogeneous sample.
The software, called TESSExtractor, is presented as an open-source web application for extracting, displaying and analysing TESS light curves directly from full-frame images. It combines aperture photometry, which measures the light collected within a chosen area around a target, with optional cotrending corrections, period searches and interactive plots. The online workflow automates retrieval of TESSCut data, background subtraction and systematic correction, and offers three user modes, including bulk processing for up to 100 targets.
A simple test for repeating patterns
To estimate a period, the program uses a Lomb-Scargle periodogram, a way of looking for repeating patterns in a time series. It tests 500 trial periods, spaced logarithmically from 0.01 to 25 days, takes the highest peak and calculates an analytic Baluev false-alarm probability. That probability is an estimate of how likely such a peak could be produced by chance.
For the photometric validation, the researchers processed same-target light curves from TESSExtractor, SPOC, QLP, Eleanor and TGLC in a homogeneous comparison. A detailed check used Brun 691, a young eclipsing binary observed in Sector 6 at 30-minute cadence. All of the compared pipelines recovered its dominant variability and eclipse morphology, with the durations, depths and overall timing preserved. The authors reported precision and baseline stability for TESSExtractor comparable to Eleanor and SPOC in that detailed comparison. The detailed morphology comparison used one target.
Precision stayed close in a larger sample
A broader precision test compared approximately 34,088 TESSExtractor light curves with 7,150 from TGLC, using the study's one-hour CDPP precision measure and excluding sources above 105 parts per million. The median TESSExtractor CDPP tracked TGLC within the 16th to 84th percentile spread, and the study described the two as essentially equivalent for catalogue TESS magnitudes of about 11 and higher. The bright-end comparison included relatively few sources, and no confidence interval was reported.
Filtering changed the period result
Period validation used a cross-match with the Kounkel datasets. It contained 16,936 common sources, but 5,439 had periods in both datasets. The study counted periods as agreeing when they met its stated 10% relative-agreement rule. On that basis, 3,200 sources, or 58.8%, agreed in the full sample.
The result changed sharply after morphology screening. TESSExtractor represented each light curve with 28 variability measures, then fed the 18 most informative features into a pre-trained classifier that assigned morphology labels. Among the 1,423 light curves labelled periodic, 1,228 met the same period-agreement rule, producing the 86.3% figure. That higher rate applies to a classifier-defined subgroup, not automatically to the full set of light curves.
Two independent catalogue checks were less favourable overall but followed the same pattern. In the Oelkers and Stassun comparison, 120 of 322 sources agreed, or 37.3%; in the Fetherolf comparison, 115 of 326 agreed, or 35.3%. Restricting those comparisons to periodic light curves raised the agreement fractions to 62.9% and 58.6%, respectively. The supplied analysis reports no confidence intervals for these fractions.
Useful for exploration, with limits
The study flags a familiar photometry problem: crowding and contamination. Its warning uses Gaia DR2 G-band photometry and a fixed circular aperture to approximate contamination. Caution is recommended when the contamination metric exceeds 15%, but the estimate does not model the actual TESS pixel response function, meaning how a star's light is spread over the detector, or changes from sector to sector.
Nor should the reported magnitudes be read as absolute brightness measurements. Default and bulk outputs anchor the median instrumental magnitude to catalogue Tmag, which may differ from the star's true brightness during the observed sector. That limitation makes the current outputs unsuitable for treating sector-to-sector magnitude changes as reliable long-term brightness measurements.
The document is labelled a preprint and retains unresolved acceptance and receipt placeholders. The TESS full-frame-image data are publicly available through the MAST archive, and processed products can be reproduced through the public TESSExtractor web interface. The acknowledgements report institutional, project, fellowship and mission support.
Paper data and sources
Original title: TESSExtractor: A web-based interactive tool for light-curve visualisation and period estimation using TESS full-frame images
Authors: Javier Serna, Jesús Hernández, Sean P. Matt et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text