Pixel-Based CTE Corrections for ACS/WFC
A detailed description of CALACS with pixel-based CTE corrections is also available at http://www.stsci.edu/hst/acs/performance/calacs_cte/calacs_cte.html
The ACS Team has released a new version of CALACS containing significant improvements: a pixel-based CTE correction, and corrections for all the electronic artifacts introduced by the repair in SM4 (bias stripes, bias shift, and cross-talk).
A comprehensive analysis to quantify the accuracy of pixel-based CTE corrections, and efforts to monitor the evolution of transfer losses are presented in this web page.
The structure of a single isolated CTE trail produced by a hot pixel with ~1000 electrons is showed in Figure 1. The schematic on the right shows the position of the hot pixel with the label HP, followed by the first, second, and third pixels in the trail, marked as P1, P2, and P3, respectively.
Figure 1: A detail of the WFC2 chip of image jbmncoakq_flt.fits, a 1040 sec dark frame taken on December 27, 2010. The vertical trails extending upward from the hot pixels are indicative of imperfect CTE. The color schematic zooms into a single hot pixel to show the CTE trail structure: HP is the position of the hot pixel, followed by the first, second, and third pixels in the trail marked as P1, P2, and P3, respectively. The amplifier for this image is located towards the bottom of the image.
The evolution of the assigned number of electrons in P1 for each ACS/WFC anneal cycle is presented in Figure 2. The maximum of the distribution for two types of images are represented, those not corrected for imperfect CTE (FLT images) and those corrected for imperfect CTE (FLC images). The red points represent the FLT data obtained while ACS was operating at a temperature of -77℃. Blue points correspond to FLT data obtained at a temperature of -81℃. Green points represent data from CTE-corrected images (FLC) regardless of the instrument temperature.
Figure 2: Evolution of number of electrons in P1 as a function of time. This plot corresponds to hot pixels with 1000±50 electrons. The black lines represent linear fits to the empirical data. The data obtained after the temperature switch have been scaled and aligned to data taken at a temperature of -77℃.
The following section presents a detailed explanation of the method used for this study. Additional information can be found in ACS ISR 2012-03 (Ubeda & Anderson 2012)
Detailed Discussion of the Procedure
The ACS Team has released a new version of CALACS containing significant improvements: a pixel-based CTE correction and corrections for all the electronic artifacts introduced by the repair in SM4 (bias stripes, bias shift, and cross-talk). This new version of CALACS requires CTE-corrected dark reference files for applying CTE corrections to raw science images.
CTE losses are created during the CCD readout process--it is an issue found in several HST instruments, such as the Wide Field and Planetary Camera 2, Wide Field Camera 3 UVIS channel, Space Telescope Imaging Spectrograph CCD, and the Advanced Camera for Survey's Wide Field Channel (ACS/WFC).
Standard data products are: CRJ files (CR-rejected images), FLT files (flat fielded images), and DRZ files (geometrically corrected images).
There are now three new types of files delivered by the Archive that contain CTE-corrected data products: CRC, FLC and DRC files. These are the CTE-corrected counterparts to the CRJ, FLT, and DRZ files. Users can choose to use either standard or CTE-corrected products. Both sets of data are corrected for bias striping, bias shift, and cross-talk.
In an effort to quantify the accuracy of pixel-based CTE corrections and to monitor the evolution of the transfer losses, a comprehensive analysis was developed, presented below.
The new CALACS incorporates a refined version of the Anderson & Bedin (2010) code. This algorithm first develops a model to reproduce the observed trails, then inverts the model to convert observed pixel values in any image to an estimate of the original pixel values.
The original version of the code worked very well for intermediate to high flux levels (> 200 electrons). Improvements in the new version of the code have made corrections more effective at low flux levels (< 100 electrons), and employs more accurate time- and temperature dependent corrections for CTE over ACS's lifetime.
The Dark Current Data Set
Most of HST's orbits have an Earth occultation. The observatory regularly uses this time to collect calibration images such as dark frames, bias frames, and flats. As a result, the Archive contains an enormous number of such calibration data. Usually, four ACS/WFC dark frames, each with an exposure time of ~1000 seconds, are obtained four times a week.
The analysis starts with RAW dark current observations. These are 4144 X 4136 2-byte integer images (34.3 Mb) that contain the raw count number recorded by the analog-to-digital converter for each pixel. Typical post-SM4 dark current in the ACS/WFC is 0.006 electron/pixel/sec, but like all CCDs operated in the Low Earth Orbit radiation environment, ACS CCDs are subjected to radiation damage by energetic particles trapped in the radiation belts. As defined in the ACS Instrument Handbook, hot pixels are all pixels with a discernible dark current excess of over 0.08 electron/sec/pixel.
Figure 3 shows a zoomed section of a single dark exposure, jbmncoakq_flt.fits. It covers a region near the top of WFC2, about 2000 pixels away from the readout amplifier. Hot pixels trails are quite obvious in that image.
Figure 3: Detail of dark frame showing CTE trails.
The study presented in this web page covers the entire history of the ACS/WFC, starting from December 2002 to the present time. Dark frames were sorted according to their observation dates, then grouped by their anneal dates. The histogram in Figure 4 shows the number of dark frames that were considered in each anneal cycle.
Figure 4: Distribution of number of dark frames used for each anneal cycle. During an anneal cycle, the CCDs and the thermal electric coolers are turned off and the heaters are turned on to warm the CCDs to ~19℃.
The Calibration Process
In order to test the pixel-based CTE correction, all RAW dark frames were calibrated using the latest version of CALACS. The automated pipeline performs a bias correction, removal of bias stripes, the pixel-based CTE correction, and a transformation from counts (DN) to units of electrons using the appropriate gain for each of the four amplifiers.
This process generates two calibrated FITS images with extensions FLT and FLC, each 167.8 Mb in size (Figure 5). A total of 240Gb of raw data were processed to deliver almost 2.0Tb of calibrated files.
Figure 5: (Left) A 1000 x 1000 pixel region at the top of the chip 1 extension in image jbmncoakq_flt.fits. CTE vertical trails are clearly visible. (Right) The reconstructed CTE-corrected FLC image after the execution of CALACS. (Center) An animation of the corrected and uncorrected images, blinking at a one second interval.
The Identification of Hot Pixels
In the presence of a high electric field, the dark current of a single pixel can be greatly enhanced. These hot pixels accumulate as a function of time on orbit; however, the reduction in operating temperature of the WFC CCDs has dramatically reduced the dark current of the hot pixels and, therefore, many pixels previously classified as hot may now be normal pixels. ACS CCDs undergo a monthly annealing process which greatly reduces the population of hot pixels and does not affect the normal pixels.
In order to identify most of the hot pixels in each anneal period and to avoid mistaking them for cosmic rays, pixels had to fall within the following conditions:
- pixel should have a dark current greater than 0.08 electrons/second
- pixel should be isolated within a 4-pixel radius
- pixel should appear consistently in at least 80% of the dark FLT frames in each anneal period.
This selection criteria allowed for the creation of a single hot pixel list for each anneal period by combining the results from individual dark frames.
The Empirical Trails
Figure 1 shows the structure of a single isolated CTE trail produced by a hot pixel with ~1000 electrons. The schematic shows the position of the hot pixel with the label HP and the first, second and third pixels in the trail are marked as P1, P2, and P3, respectively.
Using the list of hot pixels for each anneal period, every single FLT and FLC image was scanned to record the values of HP, P1, P2, and P3.
In order to analyze how well the pixel-based CTE correction performs, plans are in place to study the evolution of the number of electrons in pixels P1, P2, and P3 as a function of time, and as a function of number of electron in the hot pixels.
This can be done by studying the "behavior" of individual hot pixels, in subset ranges of 100±10, 1000±50, and 10000±1000 electrons, located at least 1500 pixels away from the readout amplifiers.
Most dark exposures are taken with 1000 seconds integration time, but there are some with longer or shorter exposure times. For these cases, pixel intensities were scaled to match a single 1000 seconds exposure. Figure 6 represents a typical RAW dark frame with the four readout amplifiers labeled in the corners as A, B, C, and D. During the CCD readout, charge is shifted in the direction of the white arrows. In order to maximize the CTE signature, hot pixels that were at least 1500 pixels from the serial registers were examined: y > 1500 in the WFC2 and y < 500 in the WFC1. Black rectangles are used to represent that space.
Figure 6: A typical processed FLT dark frame. The white arrows represent the readout direction.
Figure 7 shows a typical histogram distribution of the number of electrons in the first upstream pixel (P1) for 66 dark frames obtained around August 2009. They correspond to hot pixels with 1000±50 electrons.
These distributions always have the same shape: they present a steep increase, a defined maximum and a slower decline. This distribution is definitely not symmetrical. The assigned value to the most probable number of electrons in P1 is shown with the green vertical line. The dotted lines represent the adopted error in the average.
Figure 7: Histogram distribution of the number of electrons in the first upstream pixel (P1) for 66 dark frames obtained around August 2009. The average number of electrons in this distribution is 42.5±1.6 electrons.
In order to find the maximum, an iterative sigma-clipped mean was computed on the dataset. This yielded a value close to the mode of the distribution and provided an error to the measurement. This analysis was performed on both the FLT and FLC images.
The evolution of the assigned number of electrons in P1 for each anneal cycle is shown in Figure 8, where the maximum of the distribution is represented for both types of images, those that were corrected (FLC images), and not corrected (FLT images), for imperfect CTE. In this web page, only the results for hot pixels with 1000±50 electrons are presented. Please refer to ACS ISR 2012-03 (Ubeda & Anderson 2012) for other cases.
In Figure 8, red points represent FLT data obtained while ACS was operating at a temperature of -77℃. Blue points correspond to FLT data obtained with an instrument temperature of -81℃. Green points represent data from CTE-corrected images (FLC) regardless of the instrument temperature.
Figure 8: Evolution of number of electrons in P1 as a function of time. This plot corresponds to hot pixels with 1000±50 electrons. The black lines represent linear fits to the empirical data.
The data in the plot represents a comprehensive study of ~6000 dark frames obtained using the ACS/WFC from 2002 until 2012. The gap in the data represents the time between the ACS failure (January 2007) and the repair mission (May 2009).
Overall, there is a noticeable trend in the FLT data. A linear fit is provided for two segments: the pre-SM4 FLT data and the FLC data. The slope of the linear fits are 6.81 and 0.42 in units of electrons per year.
One significant feature of the plot is the drop observed around July 2006. This was the time in which the electronics on Side 1 failed and operations was switched to Side 2. At the same time, the ACS operating temperature was decreased from -77℃ to -81℃.
Data that was empirically corrected for charge transfer inefficiencies are shown as green points. This analysis shows that the flux from the first pixel in the trail (P1) is indeed being transferred back to the pixels where it originated.
This study also shows that the Anderson & Bedin algorithm performs the expected empirical correction, to the expected 10% accuracy, on ACS/WFC data obtained in a variety of circumstances such as (1) the newly installed camera, (2) electronics switch, (3) decrease in operating temperature, and (4) camera recovery. Also note the expected CTE loss at the time when the camera started operations.
Finally, this study demonstrates the successful integration and implementation of pixel-based CTE corrections in CALACS.