Hubble Space Telescope
Post-Flashed Darks

Overview

As of January 15th 2015 the ACS team has begun to produce post-flashed dark reference files for the Wide Field Channel (WFC). With this new method of dark imaging we hope to combat the charge transfer efficiency (CTE) losses the two WFC CCDs have suffered since being put into orbit.

CTE losses become significantly worse when there is a low background in the image (Anderson & Bedin, 2010). Low level signal against a low background (20 electrons as of 2015) can be irretrievably CTE-trailed in the background during WFC readout, particularly for signal furthest from the serial registers. The dark exposures face this issue when trying to image dark artifacts. After more than 13 years in orbit, WFC has a large population of warm pixels. Because of their low contrast relative to the global dark current (less than 10 electrons per pixel), some of these pixels are completely trailed into the background during readout. This artificially suppresses the incidence of warm pixels in dark exposures, and artificially boosts the global dark current. In addition, the pixel-based CTE-correction (Anderson & Bedin, 2010) used during dark processing is imperfect, and not optimized for exposures with extremely low background. This can result in spurious warm pixels in the remaining trail of a CTE-corrected hot pixel. The LED post-flash allows us to fill a majority of the WFC charge-traps by providing a higher background level. This prevents the irretrievable CTE-trailing of warm pixels and allows the CTE-correction code to provide a more accurate reconstruction of the true dark current.

The new calibration darks

The new calibration darks were taken with a flash level of 65 electrons (~4.6 second flash duration). The types of calibration exposures we have taken have also changed with this new method. We have reduced the number of traditional non-flashed bias frames in half, and replaced those exposures with a flashed bias frame (due to engineering requirements this has to be taken as a 0.5 second dark exposure with the same 65 electron flash). These bias frames are used to subtract the flash signal from the dark frames (along with 0.5 seconds of dark exposure), leaving the calibration dark with an exposure time of 1,000 seconds. The exposure time is then divided out so we are left with units of electrons/second. The regular calibration bias files are unchanged with the exception of the number of exposures taken per anneal cycle. One daily set (taken every 2-3 days) of the new reference file cadence consists of:

4 x 1,000.5 second flashed dark exposures
2 x 0.5 second flashed dark exposures
2 x 0 second non-flashed bias exposures

With the addition of flash, we are also increasing the noise in the dark exposures. Furthermore, the flash signal is not consistent across the frame. For more details please see ISR 14-01. This noise information is propagated to the ERR array of the final calibration dark, which will then be propagated into any science image during CALACS processing. Figure 1 shows a measurement of the per-pixel noise, measured from a p pixel histogram of the difference between consecutive bi-weeks of stacked darks. The two curves represent the new flashed dark stack from Jan-Feb 2015 (in green), and the non-flashed dark stack from Nov-Dec 2014 (in blue). After dividing the widths of the histograms by sqrt(2) to account for the image differencing, we find noise levels of 0.0016 electrons/second for the non-flashed darks, and 0.0026 electrons/second for the flashed darks. Even in the case of a long WFC exposure of 1200 seconds, the dark-subtraction noise would be ~3 electrons, below the 4-5 electron WFC read noise. Figure 2 shows a set of four superdark files (the final dark calibration file), CTE-corrected at the top, non-CTE-corrected at the bottom, non-flash on the left, flashed on the right. Figure 3 shows the histograms for these four superdark files. All four images show a different pixel distribution. After fitting the peaks of these histograms with a guassian function, we find the peak x value (an estimation of the dark current) to be 0.0078 electrons/second for the non-flashed and non-CTE-corrected image, 0.0072 electrons/second for the flashed and non-CTE-corrected image, 0.0050 for the non-flashed and CTE-corrected image, and 0.0055 for the flashed and CTE-corrected image. We see in Figure 2 that the CTE-corrected,flashed dark has the least CTE trailing, while still maintaining the signal in the hot and warm pixels.

Flag Changes

Exposure to radiation over time continues to deteriorate the health of the WFC CCDs. This includes an increase in malfunctioning pixels, such as hot and warm pixels. We have also seen a steady rise in the global WFC dark current rate, with the exception of the decrease in 2006 when the ACS operating temperature was lowered. There have been two past changes in the definition of hot and warm pixels in the lifetime of ACS. Table 1 summarizes these changes. With the switch to post-flashed calibration darks, we have chosen to further updated the flag definitions used in the DQ arrays. The new value range for warm pixels will change from 0.04 - 0.08 e/s to 0.06 - 0.14 e/s. The hot pixels will likewise change from > 0.08 e/s to > 0.14 e/s. Table 2 shows the percentage of pixels for both WFC CCDs that have been flagged as warm and hot (this value is the average taken from a set of 14 calibration darks taken within a single WFC anneal cycle) for different years. The 2015 values were calculated using the new hot and warm pixel definitions.

More Information

References: Anderson & Bedin, 2010, PASP, Volume 122, Issue 895

Created by: Sara Ogaz. Last updated: April 7, 2015