6.6 Dealing With CTE Losses in WFC3 UVIS Images

There are many ways of dealing with imperfect charge-transfer efficiency in WFC3/UVIS images. The White Paper by MacKenty and Smith (2012), provides a detailed discussion of the many issues and trade-offs. Here, we give an overview of their conclusions.
The best way to deal with CTE losses is by minimizing them in the first place. Neither the reconstruction algorithm nor the empirical correction is perfect, so minimizing the need for these corrections is the best way to combat CTE losses. This can be done in several ways. The easiest method is to place the source close to the readout register, thereby minimizing the number of readout transfers during which CTE losses occur. This is, of course, possible only for fields-of-interest that are relatively small.
Another way to minimize CTE losses is to ensure that the background has at least 12 electrons per pixel. This puts the marginal electrons into the "sweet spot" of the mini-channel (see the right half of Figure 6.1), so that a maximum fraction of them will survive the parallel-transfer process (about 80% as of Aug 2012). Users can consult the graphs found in WFC3 ISR 2012-12 to predict the number of electrons they might expect in a typical background pixel for empty fields observed through various filters for a given exposure time. The WFC3/UVIS imaging exposure time calculator provides an estimate of the predicted total background and will provide a warning if it falls below the recommendation. In images where the predicted background is less than 12 electrons, using the POST-FLASH option is recommended to add enough electrons to increase the background counts up to this level. Note that exceeding 12 electrons has not been shown to improve the charge-transfer efficiency, but it does increase the noise in each pixel. Therefore, it is best to use the minimum POST-FLASH necessary.

NOTE: The post-flash levels in APT and the ETC represent the average across the FOV. The PF illumination pattern varies by ~ +/-20% across the FOV, brightest in B/D quadrants and falling off towards A&C corners, see: 
As a consequence, observers desiring a specific e/pixel level across the entire FOV will want to increase the requested levels by ~20%.

For exposures that were treated with POST-FLASH to raise the background level for CTE loss mitigation, the background noise is naturally increased and the signal to noise ratio of the science targets correspondingly lowered. While long exposures should always be broken up into dithered sub-exposures to improve PSF sampling and remove detector artifacts and cosmic rays, exposures requiring POST-FLASH should be broken into fewer, longer sub-exposures if possible. This way, the POST-FLASH induced signal to noise reduction will be alleviated as the science source signal increases linearly with exposure time, while the background noise is kept constant at approximately 12 electrons per pixel through a suitable choice of added POST-FLASH background. Once CTE losses have been minimized in a data set, it will still likely be important to correct the observations for whatever losses have occurred, or at the very least obtain an estimate of how the readout process may have degraded the original image. One can use either the pixel-based correction for general scenes of point and extended sources ( Section 6.4) or the empirical formula correction (see Section 6.5) for point sources.