TESS provides several data levels, from raw pixels to highly detrended light curve. Users should chose which they desire depending on their scientific needs.
Short Cadence Data
- Pixel-Level time series data are available as both un-calibrated and calibrated data.
- Simple Aperture Photometry (SAP_FLUX in light curve files) light curves are available for the short cadence targets. These are produced by summing the flux in specified apertures around the specified stars and correcting for known light from other stars using the TIC.
- Systematics removed (PDCSAP_FLUX in light curve files) light curve are produced by the Presearch Data Conditioning (PDC) algorithm of the TESS pipeline. The intent is to remove common systematics across each CCD, leaving behind the astrophysical signals. The algorithm uses the co-trending basis vectors an baysian priors to determine what signals to remove from the original light curves. This method of detrending is not optimal for all types of science and so users should evaluate its effectiveness (Smith et al. 2012).
- Do-It-Yourself systemtics removed light curves are possible by using a byproduct of the PDC algorithm. PDC produces basis vectors that can be fit and removed from the time-series. Tools are available to allow users to adjust which basis vectors they want to fit and remove from the data to create their own custom co-trended light curve similar to that provided by the pipeline. <A link to such a tool would be nice here>
- Median Detrended, harmonics removed, outlier removed light curves are produced by the Data Validation (DV) module of the TESS pipeline. Here the intent is to remove all signals besides those of exoplanets.
- Whitened, harmonics removed, outlier removed light curves produced by the Data Validation (DV) module of the TESS pipeline. Here the intent is to create a light curve that can easily be searched using a wavelet-based adaptive matched filter (Jenkins et al. 2018). However astrophysical signals, including transits, get distorted in this process, and so users need to be careful not to over-interpret these time series.